Distance Controlled Rescue and Security

Mobile Robot
Golap Kanti Dey, Raqibul Hossen, [Link] Noor, Kazi Tanvir Ahmmed
Department of Applied Physics, Electronics and Communication Engineering,
University of Chittagong, Chittagong-4331, Bangladesh
[Link]@[Link], raqib_saikat@[Link], noorshafayet@[Link], tanvir@[Link]

Abstract——The intention of designing a robot is to facilitate the technology for wireless communication. Guangming’’s
human beings through giving security and working as a helping powerful home security robot could leap over particles more
hand through reprieving. The technology used in this rescue and than 4 times its own size. Reference [3] worked on the Dual
security robot has a number of important features such as functional reconfigurable mobile robot. They focused on
robotic car control by cell phone, automatically avoiding designing flexible mobile robot & various issues related to
obstacles in its way. It can sense gas, smoke and provide an alert interfacing various modules with CPLD (Complex
through text messaging. The robot can detect bombs, pick and Programmable Logic Devices). I. B Celik [4] worked on
place objects and also do distance measurements. A high quality development of a robotic arm controller by using hand gesture
wireless video camera equipped with a stepper motor for the
recognition. Sourangu Banerji [5] developed an unmanned
omni-directional view is used for the spying purpose which is
vehicle using a GSM network with microcontroller. Reference
mounted on the robo car to send real video images and audio
signals. These audio and video stream received by the receiver
[6] gave us an idea about real time human tracking using
unit can be used to take proper steps according to the received sensor for home security robot. They combined the data for
signals. This robot can also be used to reach places where both vision and depth sensor. The combat robot [7] which was
humans cannot reach like hidden places, small tunnels. The radio operated had all the controls like normal car. Through a
ultimate focus of this system is to offer far possible security for wireless camera this combat robot gave all the information
the human being. In this paper, we have illustrated our work by about the enemy. These types of applications are very much
using the entire sub modules. useful for the industries and places where the access of human
being is difficult, specifically, if the path of the robot is quite
Keywords-Objects pick and place module; Robotic car; Sensors; complicated and not suitable for laying cables. The only
Wireless Camera. solution exists is Wireless communication between the robot
and the base station.
I. INTRODUCTION
In our work we are controlling a robot through wireless
Robot is an electromechanical, reprogrammable, communication system, which is known as DTMF system.
multifunctional device which is capable of acting in the highly Microcontroller is used to control all operations. According to
varied environment, and takes autonomous actions. The word the motor operations the robot will operate in specified
““Robot”” was first used in a 1921 play titled R.U.R. Rossum’’s directions. If any obstacle is observed by the robot it will
Universal Robots, by Czechoslovakian writer Karel Capek. change its direction or it will stop. Our system is equipped
Now a day's every system is automated in order to face new with smoke sensor, gas sensor, object pick up and place
challenges. In the present days automated systems have less module for the security purpose of the human being. The
manual operations, flexibility, reliability and accuracy. Due to main aim of using these sensors is to monitor the smoke
this demand every field prefers automated control systems. through smoke sensor, leakage of gas using gas sensor in any
Especially in the field of electronics automated systems are industry or home. If input voltages of these sensors exceed the
giving good performance. threshold level, an alert is given through a text message. Here
In the present scenario of war like situations, unmanned we also used an Object pick and place module for the
systems plays very important role to minimize human losses. surveillance system.
So this kind of robot is very useful to do operations like
II. FRAMEWORK OF THE SYSTEM
detecting leakage of gases, obstacle, and object pick up and
place at the time of emergency as well as security purpose [1] The heart of our rescue and security robot is a
providing human-robot interaction in rescue robotics. Microcontroller, which is configured by programming in the
Micro C Pro, an advanced C compiler for the PIC
Wireless robotics is one of the evolving fields of Microcontroller Unit. Various sensory unit inputs, DTMF
application, where different technology is used. It is an decoder input are given to the microcontroller unit. After
application where the concept of embedded processor to applying the power supply [8], microcontroller takes action
wireless communication is being used for a practical through pre-defined instructions. Using the entire sensory unit
application. Guangming Song [2] developed surveillance robot like gas sensor, ultrasonic sensor for the obstacle detection,
with hopping capabilities which was adopted with the Zigbee

978-1-4799-0400-6/13/$31.00 ©2013 IEEE

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:17:18 UTC from IEEE Xplore. Restrictions apply.
 smoke sensor and giving an alert through SMS interfacing [9] is used here through DTMF sub module and wireless video
module we have offered a total package of security in the robo transmission unit. Wireless video camera is used for spying
car. Metal detector unit is also used for strengthen the security. purpose to reach place where it is difficult by a human being.
Total configuration of the rescue and security is given in Objects pick and place module is used as the helping hand for
Figure 1. the rescue purpose, controlled through a mobile phone from
remote place.
Now a days in modern world the best technology for
communication system, known as the wireless communication

Video camera unit

Object pick up and place unit

Wireless video
transmission unit

Stepper
Gas sensor unit
motor

Mobile phone
Obstacle detection and distance
measurement unit Wireless video
signal receiver
MICRO-
CONTROLLER DTMF decoder
unit

Smoke sensor unit

Display unit
SMS interfacing
unit

Power supply unit

H-bridge module
& speed control

Metal detector

Figure 1. Block diagram of the rescue and security mobile Robot

III. CONSTITUENT SUB MODULES

A. Object Pick and Place Module
Objects pick and place sub module is one of the major
parts of this rescue robot. For the up and down of the robotic
arm a servo motor with 18 degree step resolution is used. For
griping the object we used another servo motor. Maximum
opening of the grip is 6 cm and minimum is 0.5 cm. A 6V DC
servo motor is used for this purpose. Servo motors operate
differently from DC brush motors, which rotate when voltage
is applied to their terminals.

Figure 2. Servo motor driving circuit for object pick-up and

pick off

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:17:18 UTC from IEEE Xplore. Restrictions apply.
 According to the generated DTMF frequencies by pressing
the keypad on the mobile phone we can control the motors
behavior. When key ““7”” is pressed, the servo motor rotate to
upward position, when ““9”” is pressed the servo motor rotate to
downward position, key ““1”” is used for griping the object and
key ““3”” is used to release the object.

Start

DTMF signal

Micro-controller

Figure 4. Implemented DTMF circuit with power supply unit

Rotate H-bridge IC L298 is designed to control two gear DC

Rotate motors which take command as input from the DTMF decoder
servo
servo Object through a Microcontroller. There are 2 Input and 2 output pins
motor
motor pick-up for each motor. Two gear dc motor are connected to the H-
downward Bridge output pin. In order to activate the motor 1, the Enable
upward and
place A line must be HIGH.

Object
pick Object
up pick-
Rotate Rotate off
servo servo
motor motor
down- upwar
ward d

Rotate grip Rotate grip

motor inward motor outward

Figure 3. Flow chart for object pick-up and place

B. Robotic Car Movement

The main platform of our rescue and security robot is the
robotic car, which is controlled by a mobile using DTMF
circuitry. First of all a mobile phone kept on the desirable car,
receives call from the calling subscriber. If the DTMF decoder
is activated, the H-Bridge module begins functioning.
According to the generated DTMF frequencies from the
subscriber cell phone we can control the robot movement.
Figure 5. Robot movement controlling circuit using L298 H-bridge
IC

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:17:18 UTC from IEEE Xplore. Restrictions apply.
 The motor as well as its direction can be controlled by
applying a LOW or HIGH signal [3] to the Input1 and Input2
lines, as shown in Table I. When key ““2”” is pressed in the mobile phone, the
corresponding signal is detected by the decoder IC and both
gear dc motors rotate in the forward direction i.e. the robot
TABLE I. Motor behavior for different input moves forward. When key ““4”” is pressed the left gear dc
motor moves forward while the right motor remains idle, i.e.
Input 1 Input 2 Motor Behavior
the robotic car moves to the leftward direction. When ““6”” is
pressed the right gear dc motor, moves forward while the left
LOW LOW Break and stop motor remains idle i.e. the robot car moves to the rightward
direction.
HIGH LOW Turns forward
When key ““8”” is pressed both the motors move to the
backward direction. When key ““5”” is pressed both of the
LOW HIGH Turns backward motors are turned off which means that the robot car is now in
the breaking position and stop.
HIGH HIGH Break and stop

Figure 6. Object pick and place module on the robotic car

C. Metal Detector Unit

Metal detector [10] works on the principle of transmitting
a magnetic field and analyzing a return signal from the target
and environment. For the purpose of detecting metal , an
oscillator producing an alternating current that passes through
a coil. Here we used the CS209A proximity sensor IC, which
is the main component of our metal detection cicuit.
While power supply is on metal detector start its query
for metallic object. If it can detect any metallic object it sends
a signal to the microcontroller which is connected with a sms
interfacing unit. Due to this signal from the metal detector
unit, microcontroller sends a sms to a pre-defined cellphone
number. Meanwhile microcontroller keeps sms interfacing
unit inactive until the metal detectors output stage changes.
While that output goes low again metal detector starts
[Link] no metal is detected is keep its query unless its
power shuts down. Figure 7. Metal detecting circuit

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:17:18 UTC from IEEE Xplore. Restrictions apply.
 D. Gas Sensor
Gas sensor module is used for detecting the leakage of
Liquefied Petroleum Gas (LPG). When gas sensor senses gas
that is, when the sensor exceeds a level called threshold level
the output of the sensor is activated and it alerts through the
message by the sms interfacing unit to the subscriber. After
sensing the gas the sensor is kept inactive for 10 minutes.

Figure 9. Smoke sensing circuit

F. Ultrasonic Sensor
The ultrasonic sensor used here for the obstacle
detection and through proper programming and interfacing
module we had measured the distance of the obstacle. The
measurement distance of this module is within 0.03-3m. It
transforms the data into impulse with different width. The
advantage [11] of ultrasonic sensors is long-range detection,
the drawback of that is the unstable system and slow reaction
rate. The circuit connection for this unit is simple. Just have to
fed the output (pin 2) of the sensor to a pin of microcontroller
and connect the other two pins with the power rail
respectively.
Figure 8. Flow chart of the gas sensor G. Wireless Video Transmission Unit
E. Smoke Sensor Wireless video camera mounted on the robotic car is used
Apart from the gas sensor we have also used a smoke for the surveillance and spying purpose. We can make a large
sensor. So here we are monitoring the gas and smoke also. number of areas automated by controlling the direction of the
When smoke sensor senses smoke, that is, when the sensors camera by generating the DTMF frequencies. For this total
smoke sensing branch voltage exceeds the reference level action just install the wireless camera on the robotic car and
called threshold level the output of the sensor is activated and set the wireless receiver to a TV or monitor to watch the
action and take the necessary steps through recorded video
the microcontroller receives the sensor output. This sensory
footage.
element is very much compatiable with microcontroller and
having good sensitivity and rapid response time. LPG gas may A TV capture card is a computer component that allows
be harmful for the human being, so through the above process television signals to be received by a computer. It is a kind of
we can use it for security purpose of the mankind. Smoke television tuner. Most TV tuners also function as video
sensing circuit actually warns anyone against fire accidents. capture cards, allowing them to record television programs
Alert is given to the subscriber through the message by the onto a hard disk. The card contains a tuner and an analog-to-
sms interfacing unit. After sensing the smoke the sensor is digital converter along with demodulation and interface logic.
kept inactive for 5 minutes. After that it again starts sensing
the smoke. We can use it for both home and industrial security
purpose.

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:17:18 UTC from IEEE Xplore. Restrictions apply.
 IV. CONCLUSION
The distance controlled rescue and security mobile robot [3] Mr. Bhushan C. Patil, Mr. Rabinder Henry, ““Dual functional
is successfully designed and assembled. It is based on reconfigurable mobile robot,”” IEEE Region 10 Conference, pp.1-5,
microcontroller which is the brain of the complete system. TENCON 2008.
DTMF decoder decodes the received tone and gives binary [4] I.B. Celik, M. Kuntalp, ““Development of a robotic-arm controller by
equivalent of it to the microcontroller. The controller is using hand gesture recognition,”” International Symposium on
Innovations in Intelligent Systems and Applications (INISTA), pp. 1-5,
programmed such that appropriate output is given to motor IC July 2012.
L298 which will drive two DC motor connected with it. As [5] Sourangsu Banerji, ““Design and Implementation of developed an
rescue and security features are the most important factors of Unmanned Vehicle using a GSM Network with Microcontrollers,””
this mobile robot - object pick and place module, gas sensor, International Journal of Science, Engineering and Technology Research
metal detector, smoke sensor, distance measurement unit plays (IJSETR) Volume 2, Issue 2, pp.367-374, February 2013.
vital role to integrate these two aspects. In our work we have [6] June-Young Jung, Byoung-Kyu Dan, Kwang-Ho An, Seung-Won Jung,
deptict the modern application of cell phone. In the era of the and Sung-Jea Ko, ““Real-time Human Tracking Using Fusion Sensor for
industrial revolution, people tends to bulid big industries and Home Security Robot,”” 2012 IEEE International Conference on
consumer electronics (ICCE), pp.420-421, 2012.
high rise apartments to make profits. But many of them are
[7] [Link], [Link] ““Design of a intelligent combat robot for
very much unconscious about the secutiry of industry and high
war field,”” (IJACSA) International Journal of Advanced Computer
rise building as well as the human being who are working or Science and Applications, Vol. 2, No. 8, pp.64-70, 2011.
living their. After the disaster like earthquakes many people [8] Robert [Link] and Louis Nashelsky, ““Electronic Devices and
are embracing death due to the procrastination of the Circuit Theory,”” 8th Edition, 2006.
deliverance. Considering all the situations, the robot [9] Thomas Bräunl, ““EMBEDDED ROBOTICS-Mobile Robot Design and
integrated with different sub modules can be used for Applications with Embedded Systems,”” 2nd Edition © Springer-Verlag
redemption and security purpose. Berlin Heidelberg 2003, 2006, Printed in Germany.
[10] D. Sparsha, [Link], ““Remote Controlled Metal Detecting Robot
REFERENCES with Remote Image Transmission,”” Sridevi Women’’s Engineering
College [Link], Gandipet, [Link]-500 075, June 2010
[1] Robin R. Murphy, ““Human-Robot Interaction in Rescue Robotics,””
IEEE Transactions on Systems, Man, and Cybernetics, Part C: [11] Ren C. Luo, Tung Y. Lin, Te Y. Hsu, Po [Link], ““ Multisensor
Application and Reviews Vol. 34, No.2, pp.138-153, May 2004. Controlled Obstacle Avoidance and Navigation of Intelligent Security
Robot,”” 31st Annual Conference of IEEE Industrial Electronics Society,
[2] Guangming Song, Kaijian Yin, Yaoxin Zhou and Xiuzhen Cheng, ““A
pp.1827-1832, IECON 2005.
Surveillance Robot with Hopping Capabilities for Home Security,””
IEEE Transactions on Consumer Electronics, Volume 55, No.4,
pp.2034-2039, November 2009.

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:17:18 UTC from IEEE Xplore. Restrictions apply.
Habibian et al. Robomech J (2021) 8:1
[Link]

RESEARCH ARTICLE Open Access

Design and implementation of a maxi‑sized

mobile robot (Karo) for rescue missions
Soheil Habibian1*† , Mehdi Dadvar1† , Behzad Peykari1, Alireza Hosseini1, M. Hossein Salehzadeh1,
Alireza H. M. Hosseini1 and Farshid Najafi2,3

Abstract
Rescue robots are expected to carry out reconnaissance and dexterity operations in unknown environments compris-
ing unstructured obstacles. Although a wide variety of designs and implementations have been presented within the
field of rescue robotics, embedding all mobility, dexterity, and reconnaissance capabilities in a single robot remains
a challenging problem. This paper explains the design and implementation of Karo, a mobile robot that exhibits a
high degree of mobility at the side of maintaining required dexterity and exploration capabilities for urban search
and rescue (USAR) missions. We first elicit the system requirements of a standard rescue robot from the frameworks of
Rescue Robot League (RRL) of RoboCup and then, propose the conceptual design of Karo by drafting a locomotion
and manipulation system. Considering that, this work presents comprehensive design processes along with detail
mechanical design of the robot’s platform and its 7-DOF manipulator. Further, we present the design and implemen-
tation of the command and control system by discussing the robot’s power system, sensors, and hardware systems. In
conjunction with this, we elucidate the way that Karo’s software system and human–robot interface are implemented
and employed. Furthermore, we undertake extensive evaluations of Karo’s field performance to investigate whether
the principal objective of this work has been satisfied. We demonstrate that Karo has effectively accomplished
assigned standardized rescue operations by evaluating all aspects of its capabilities in both RRL’s test suites and train-
ing suites of a fire department. Finally, the comprehensiveness of Karo’s capabilities has been verified by drawing
quantitative comparisons between Karo’s performance and other leading robots participating in RRL.
Keywords: Field and service robotics, Telerobotics, Mobile robotics, Search and rescue robots, Mechanism design
of mobile robots, RoboCup rescue, Robot manipulation, Standard test methods, Response robots, Robot sensing
systems

Introduction abilities, they are mainly expected to play influential roles

Natural and manmade disasters have caused heavy casu- in urban search and rescue (USAR) [2], nuclear field
alties and significant economic damages over the past emergency operations [3], and mine rescue missions [4].
few decades. As the impacts of catastrophes are increas- Practically speaking, a rescue robot needs to demonstrate
ing, the necessity of rescue robotics, which explores mobility, dexterity, reconnaissance and exploration capa-
solutions to minimize the casualties for all phases of a bilities to efficaciously accomplish an USAR mission [5].
disaster, has increased as well [1]. Due to peerless capa- These principal capabilities practically empower rescue
bilities of rescue robots compared to human and canine robots to overcome obstacles, remove rubbles, explore
and map unknown environments, detect victims, and
deliver essential objects, such as bottle of water, food
*Correspondence: habibian@[Link] and drug, to victims’ locations. Although the minimum
†
Soheil Habibian and Mehdi Dadvar contributed equally in this paper
1
The Authors Were With Advanced Mobile Robotics Lab, Qazvin Azad required levels of those capabilities vary in a rescue robot
University, Qazvin, Iran depending on its application, embedding a standard level
Full list of author information is available at the end of the article of those complementary capabilities in a single system

© The Author(s) 2021, corrected publication 2021. This article is licensed under a Creative Commons Attribution 4.0 International License,
which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to
the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or
other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this
licence, visit [Link]
Habibian et al. Robomech J (2021) 8:1 Page 2 of 33

poses various critical challenges which is the primary are focused on mobility and maneuverability capa-
motivation of this work. bilities and neglect dexterity and manipulation skills
The very first academic attempts in the field of rescue which have resulted in the elimination of the systems’
robotics have been done by two groups at Kobe Univer- functionality.
sity in Japan [6] and the Colorado School of Mines in Mobile manipulators enhance the dexterity capabilities
the United States motivated by Kobe earthquake and of rescue robots by enabling them to open/close doors
the bombing of Murrah Federal Building in Oklahoma and valves, remove rubbles and obscurations, inspect
City respectively. However, the mobile platforms intro- the environment, and interact with victims. Embedding
duced in [7] and [8] can also be known as two early dexterity capabilities into a highly mobile rescue robot
experimental efforts in design and implementation of raises serious design and implementation challenges. In
mobile robot platforms which were potentially suited this regard, [23] presents a mobile robot design based on
for USAR, though the authors did not explicitly men- hybridization of the mobile platform and manipulator.
tion their USAR applications. Later on, rescue robot- Although this hybridization empowers the robot in term
ics started getting more attentions gradually among of mobility capabilities, the manipulator itself hardly sat-
the mobile robotics research community [9], which led isfies minimum dexterity requirements of a dexterous
to developing rescue robots with a variety of locomo- rescue robot. In another effort, a prototype of a 4-flip-
tion mechanisms such as spherical [10, 11], legged [12], per mobile platform is presented in [24] that has been
wheeled [13], and tracked [14] locomotion. integrated with a commercial 5-DOF (degrees of free-
Track-based locomotion mechanisms provide more dom) robot arm. Although this integration has resulted
versatilities with lower levels of complexities compared in standard dexterity capabilities of the robot, the non-
to other locomotion mechanisms [15]. In this regard, tracked design of the robot’s chassis is still a consider-
different attempts have been done on proposing vari- able issue while facing unstructured obstacles. Next, [25]
ous designs for tracked rescue robots in different sizes, introduces a ground mobile robot which has two paral-
as [5] categorizes them into man-packable, man-port- lel tracks and is equipped with a heavy-duty manipulator.
able and maxi-sized types. On this subject, [16] and Similarly, this work does not present a standard embed-
[17] present two man-packable tracked rescue robot ding of mobility and dexterity capabilities due to the lack
designs. The former develops a paradigmatic system of assistive mobility mechanisms such as flippers which
called inspection robotic system to demonstrate the restrains robot’s mobility skills. The problem gets even
feasibility of mechatronic solutions for inspection oper- more complicated when robots presented in [23–25]
ations. The latter proposes the design and development lack reconnaissance and exploration aspects required
of a novel reconfigurable hybrid wheel-track mobile for a rescue robot which can be a principal shortcoming
robot which is constructed based on a Watt II six-bar in situations where data acquisition is a prerequisite for
linkage. Furthermore, [18] demonstrate the application performing maneuvering or dexterity tasks.
of Quince, a man-packable rescue robot, in an emer- Although there are works addressing the exploration
gency response to the nuclear accident at the Fukush- and reconnaissance capabilities of rescue robots, little
ima Daiichi nuclear power plants. Although in [16–18] attention has devoted to embedding all three principal
different man-packable rescue robots with various lev- capabilities of rescue robots, i.e. mobility, dexterity, and
els of mobility are presented, man-portable and maxi- exploration. For instance, [26] presents a highly mobile
sized rescue robots perform higher degrees of mobility tracked rescue robot equipped with four flippers which
and stability while facing more complex obstacles and are empowered to accomplish exploration and dose
are well-suited to be equipped with a variety of assis- measurement tasks. However, the manipulation capa-
tive mechanisms and accessory devices. For instance, bilities of the robot are only limited to sampling func-
[19] develops a tracked man-portable rescue robot that tions by way of a simple 2-DOF manipulator which is
is not equipped with assistive mobility mechanisms an inadequacy for many dexterity operations. Similarly,
such as flippers, while the surveyor type robot imple- the rescue robot design introduced by [27] satisfies the
mented in [20] and the Packbot-like main rescue robot basic required mobility skills and also proposes a fron-
presented in are armed with two front flippers which tier-selection algorithm for robot’s exploration, while the
empower them to overcome more elevated and obsta- system is deficient in dexterity operations. Consequently,
cles stably. Moreover, there are some works introduc- there is a critical lack of attention paid to embedding all
ing 4-flipper mechanical designs such as [21] and [22] three complementary required capabilities in a single res-
which have heightened the robots’ mobility skills to cue robot for accomplishing an effectual USAR mission.
more sophisticated extent. All mentioned instances of The deep-seated objective of the Karo project is the
man-portable and maxi-sized tracked rescue robots design and implementation of a maxi-sized rescue robot
Habibian et al. Robomech J (2021) 8:1 Page 3 of 33

that meets sophisticated levels of all maneuvering, mobil- at RoboCup, demonstrate its comprehensive operation
ity, dexterity, and exploration requirements delineated in qualities that are expected from a response robot, which
the rulebook of RoboCup rescue robot league (RRL) [28]. confirms our initial design criteria and proposed analysis,
To that end, we first elicited the system requirements out (5) We ultimately show − based on Karo’s performance
of RRL’s framework inspired by standard test methods at the training field of a fire department − that neverthe-
developed by the U.S. Department of Homeland Secu- less standardized test methods are substantially useful for
rity, Science and Technology Directorate (DHS S&T) evaluating response robots and inspire developers, these
conjunct with the National Institute of Standard and methods should be extended more to real-life missions.
Technology (NIST). This discussion led to the concep- Overall, this work is a step towards preparing highly
tual design of Karo by drafting a locomotion and manip- capable response robots for USAR missions.
ulation system. Considering the proposed conceptual The remainder of this paper is organized as follows: the
design, this work presents comprehensive design pro- prior works of the authors are presented in "Prior works:
cesses along with detail mechanical design of the robot’s Karo’s ancestors" section. In "A discussion on conceptual
platform and its 7-DOF manipulator. Further, we present design" section, we provide a discussion on the concep-
the design and implementation of the command and tual design, followed by mechanical design in "Mechani-
control system by discussing the overview of the system, cal design" section. The elaborations of the command
robot’s power system, sensors, and hardware systems. In and control system and software system are discussed in
conjunction with this, this work elucidates the way that "Control and command system" and "Software systems"
Karo’s software system and human–robot interface (HRI) sections respectively. Then we propose the system speci-
are implemented and employed. Furthermore, we under- fications in "System specifications" section. Experimental
took extensive evaluations of Karo’s field performance to results and analyzing the robot’s field performance are
investigate whether the principal objective of this work carried out in "Experiments" section. Finally, a conclusion
has been satisfied. Experimental results of Karo’s evalu- is drawn from the paper’s discussions in "Conclusion"
ation in both RRL’s test suites and the training suites of section.
a fire department demonstrate that it has effectively
accomplished assigned rescue operations utilizing all Prior works: Karo’s ancestors
aspects of its capabilities. Besides, the comprehensive- During the last decade, several efforts have been made in
ness of Karo’s capabilities has been verified by drawing developing practical rescue robots at Advanced Mobile
quantitative comparisons between Karo’s performance Robotics Lab (AMRL). Lessons learned from the experi-
and other leading robots participating in RRL. mental and field performance of those developed systems
In this paper, we make the following contributions: (1) frame the roadmap to a more effectual and pragmatic
we propose system specifications for a response robot system design. Hence, we briefly review four distinct ver-
(Karo) that satisfies requirements and human opera- sions of rescue robots developed at AMRL.
tors’ expectations in USAR missions. These configura- Figure 1a demonstrates the most primitive rescue
tions are inspired by NIST standard test methods, and robot developed at AMRL in 2003 called “NAJI-I”, which
we here employ a holistic combination of their standards has been designed to perform basic maneuvering tasks.
when designing our system architecture, (2) we evaluate Its locomotion mechanism is track-based, and it is not
various conceptual hardware designs for locomotion and equipped with any assistive mechanisms such as flippers.
manipulation that were tested in our previous works, and This 35-kg robot can speed up to 15 cm/s at maximum,
point out properties of each design. We develop an opti- where it is actuated by two DC servomotors. According
mal conceptual design by incorporating the strengths of to the records of the robot’s field performance in Bam
the previous versions, which maximizes the functionality earthquake, its locomotion design was an inadequacy for
of the robot in field performances, (3) we propose a sim- overcoming the haphazardly formed obstacles in the dis-
plified and effective approach for design and analysis of aster site. Since maneuvering capabilities are prerequisite
the teleoperated system (i.e., electrical/mechanical hard- for all other capabilities of a rescue robot, upgrading the
ware, control and command, and human–robot interface) robot’s locomotion mechanism design was a crucial step
that leverages customized designs to meet the standard- to take.
ized requirements of response robots in one system, and Field performance results of “NAJI-I” led to an
(4) We empirically show that our approach in the design enhanced design for a second version of rescue robot at
and implementation of Karo, led to a mobile robot which AMRL in 2004, as shown in Fig. 1b. In the second ver-
exhibits a high degree of mobility at the side of main- sion, called “NAJI-II”, the robot’s chassis is equipped with
taining required dexterity and exploration capabilities two coupled flippers where their 360-degree rotation
for USAR missions. The results of Karo’s performance helps the robot to maintain its stability while climbing
Habibian et al. Robomech J (2021) 8:1 Page 4 of 33

Fig. 1 Two of Karo’s ancestors: a NAJI-I in a field performance in 2003, and b NAJI-II under test and development procedure in 2004

up inclined planes and stairs and enables robot to prevail obstacles to accomplish the assigned mission. And (2)
over the unstructured obstacles. Moreover, the robot’s the mobility itself does not suffice to accomplish a res-
maximum velocity and weight have increased to 25 cm/s cue mission successfully because a rescue robot basically
and 45 kg respectively, where three DC servomotors needs to maintain its stability while maneuvering in a dis-
function as robot’s actuators. Although “NAJI-II” could aster site to be able to perform other required operations.
accomplish a sort of basic mobility tasks such as climbing Besides, lack of a functional robotic arm for accomplish-
up the stairs and inclined planes, its stability and reliabil- ing manipulation and inspection tasks was a considerable
ity was still questionable while overcoming more compli- deficiency for accomplishing any assigned USAR mission.
cated obstacles. As a result, later designs evolved to 4-flipper style designs
According to the experimental results of the first two considering a robotic manipulator for robot’s dexterity
rescue robots developed at AMRL, the essence of a four- purposes.
flipper design for the rescue robot was justifiable to Figure 2a illustrates previously developed rescue
accomplish sophisticated rescue mission in an uneven robot NAJI-VII as it is overcoming a pipe step in Robo-
area. That is mainly because of two reasons: (1) the distri- Cup competition 2009. NAJI-VII is equipped with two
bution, complication and elevation of obstacles vary from front triangular flippers which can rotate 360 degrees.
one case to another. Therefore, a practical rescue robot Its mechanical design also consists of two regular rear
should be mobile enough to cope with various forms of flippers with a course of 90 degrees. The triangular

Fig. 2 Two more ancestors of Karo: a illustrates NAJI-VII while overcoming a pipe step in RoboCup competitions 2009, Austria, and b explains
Scorpion as inspecting inside a car in RoboCup competitions 2012, Mexico
Habibian et al. Robomech J (2021) 8:1 Page 5 of 33

design of front flippers helps them to leverage the A discussion on conceptual design
robot’s body by fewer degrees rotation since the flip- Mission description
pers inherently have three-point-contact structures. In All versions of rescue robots developed at AMRL has
despite of previous rescue robots developed at AMRL been tested in RRL which is held annually since 2001
prior to NAJI-VII, it benefits from a 2-DOF inspection inspired by the Kobe earthquake [29]. As a matter of fact,
arm. There are mainly two drawbacks associated with taking part in RRL and sticking to its developing frame-
this robot’s design: (1) although the triangular design works facilitated the advancement and improvement
of front flippers brings some advantages to robot’s of AMRL’s rescue robots. To investigate the rationale
mobility, the front flippers cannot be long enough behind that, we discuss three main reasons:
because of the inherence of triangular shapes and the
design constraints. This decreases robot’s competency 1) RRL basically provides objective performance evalu-
to overcome more elevated steps and obstacles. More- ations of rescue robots functioning in simulated
over, the asymmetric design of front and rare flippers earthquake environments. The frameworks of RRL
makes robot operator’s job more complicated. And (2) motivate participating rescue robots to demonstrate
rescue robots are required to inspect alcoves and holes their skills in maneuvering, mobility, dexterity and
for victim detection and object recognition. Therefore, exploration test suits [28], which necessarily enforce
the 2-DOF design of the robot’s arm does not suffice teams to focus on all aspects of required skills while
for the required inspection operations. Furthermore, developing rescue robots.
this arm is incapable of object manipulation and per- 2) Taking part in RRL gives the opportunity of compar-
forming a sort of dexterity tasks. ing our developed rescue robot with several other
Scorpion is the last rescue robot that we aim to por- works from all over the world. In fact, not only the
tray as Karo’s ancestor. It benefits from a 4-flipper inherence of the competition galvanizes the adequate
mechanical design where each of its flippers has two development of rescue robots, but also promotes col-
links, as shown in Fig. 2b. Considering, the second link laboration between researchers from different teams
of each flipper has a self-relative rotation with respect by sharing experiences during RRL’s annual events.
to the first link. This property helps robot to have more 3) The rules and structures of RRL follows the stand-
flexibility in rough terrain and enables it to overcome ard test methods developed by the DHS S&T con-
more elevated steps. Moreover, the flippers have been junct with the NIST. Thus, RRL’s framework is being
placed along with the robot’s chassis which facilitates updated regularly and its evaluations are getting
robot’s maneuvering in narrow corridors. However, more effective constantly as the NIST’s standard test
it made the robot more likable to overturn in uneven methods get developed. Having said that, the results
terrains. Furthermore, Scorpion derives profit from a of rescue robots taking part in RRL aptly reflect their
6-DOF manipulator for accomplishing inspection and actual potential skills in practice with respect to the
dexterity operations. Although its 6-DOF manipula- latest market demands.
tor was a decent step for getting inspection tasks done,
its limited payload (< 1 kg) restricts its dexterity skills The RRL as part of the International RoboCup com-
significantly. In addition, Scorpion can speed up to petitions provides a benchmark comparison for robot
50 cm/s which is more adequate compared to previous implementers and test administrators based on ASTM’s
designs, but it is still an operational deficiency. test methods. This benchmark is grouped into four major
Alongside mobility and dexterity discussions about categories: Maneuvering, Mobility, Dexterity, and Explo-
the robots mentioned above, their durability and reli- ration, that each includes various test scenarios (Fig. 3).
ability are another subject matter to analyze. According The Maneuvering suite includes apparatuses with simple
to the carried-out experiments, all these versions are terrains: flat surfaces (MAN 1: Center), bridges (MAN
short in reliability for fulfilling a USAR mission in which 2: Align), an inclined 30° surface (MAN 3: Traverse), 15°
the reliability is defined as steadiness functionality for continuous ramps (MAN 4: Crossover), movable verti-
repetitive operations. Moreover, the performances of cal and diagonal sticks (MAN 5: Negotiate), and ground
these robots have not been extensively investigated in a with bars (MAN 6: Curb), that are embedded in each test
real disaster field. Consequently, for instance, their slow for forward and reverse driving orientation. The Mobil-
set-up and break-down time exemplify a non-practical ity suite verifies the capability of robots to pass through
attribute. As a final point, not only we need to remedy apparatuses with medium to hard obstacles such as
the previous versions’ deficiencies for the new design, stacked rolling pipes (MOB 1: Hurdles), 15° surfaces with
but also, we must quantitively evaluate the new robot’s granular materials (MOB 2: sand/gravel hills), square step
field performance in a structured manner. field pallets (MOB 3: Stepfields), diagonal hills (MOB
Habibian et al. Robomech J (2021) 8:1 Page 6 of 33

satisfactory conceptual design on the robot’s mechanical

configuration.

System requirements
In this section, we aim to set an international standard as
benchmark to define the basic requirement for our sys-
tem. The development of robots has been always a labori-
ous task that forced many robot owners to retreat from
using them. In a market, filled with well-advanced robotic
platforms, exploring the best option is hardly feasible for
users. Thus, there is a big gap between the user’s demand
and developer’s approach in design. In 2005, a compre-
hensive suite of standard test methods was developed by
the U.S. DHS S&T conjunct with the NIST to compare
Fig. 3 Arena layout sample for Maneuvering (MAN), Mobility (MOB), response robots. The DHS-NIST-ASTM (American Soci-
Dexterity (DEX), and Exploration (EXP) test suites in RRL (figure
obtained from RRL’s rulebook 2019 [28])
ety for Testing and Materials) international standard test
methods describe the key features of a response robot
more broadly:
• Rapidly deployed.
4: Elevated Ramps), partly blocked stairs (MOB 5: Stair • Remotely operated from an appropriate standoff.
Debris). For Dexterity suite, objects such as pipes, doors, • Mobile in complex environments.
and wooden blocks are considered for manipulation and • Sufficiently hardened against harsh environments.
inspection tasks. The robot should inspect, touch, rotate, • Reliable and field serviceable.
and extract pipes that are placed in various arrangements • Durable or cost-effectively disposable.
namely DEX 1: Parallel Pipes, DEX 2: Omni-Directional • Equipped with operational safeguards.
Pipes, and DEX 3: Cylindrical Pipes. For the case of The above characteristics can be abstracted in maneu-
doors (DEX 4: Door Opening), sometimes equipped with vering, mobility, manipulation, sensing, endurance, radio
spring closures, the robot should open doors with lever communication, durability, logistics, and safety (see Fig-
handles and successfully enter the room. Additionally, ure S3 in [30]). The test methods designed by DHS-NIST-
this suite involves building vertical shoring structures ASTM seeks to address these capabilities for various
(DEX 5: Shoring) with wooden blocks on a flat surface. types of robots: Ground, Aquatic, and Aerial, which pro-
The Exploration suite has apparatuses for mapping a dark vide a quantitative method to evaluate the performance
labyrinth (EXP 1: Map on Continuous Ramps and EXP of robots in particular missions. The ASTM committee
2: Map on Crossing Ramps), recognition of objects (EXP has developed a test apparatus with terrains, targets, and
3: Recognize Objects), or detection of obstacles and ter- tasks for each mission (Figure S12 in [30]) that robots
rains (EXP 4: Avoid Holes and EXP 5: Avoid Terrains). are expected to operate safely in a limited time. The test
These apparatuses include terrains with continuous/ suites and evaluation metric get updated every year to
crossing ramps, amorphous negative obstacles, or ter- improve the evaluation quality, so robot developers and
rains that should be avoided. In each test, the operator’s users can compare the operational requirements of each
intervention is examined by the test administrator that robot in a specific capability spectrum. As a result, hav-
expects the driver to operate the robot remotely for as ing access to a concrete demonstration of operation for
many as iteration possible in a limited time. Lastly, classi- a response robot saves a substantial amount of resources
fied results are presented at the end of the competition to before selecting an operating system. Given this very
participants. By the means of these performance evalu- thorough description of the desired respond robot, we
ations, developers can compare their hardware and per- corroborated our system by setting a comprehensive
formance among other participants. operational requirement. Table 1 presents a detailed
Regarding the above justifications, we aimed to con- description of our robot configuration according to
sider RRL as the determined mission for the research ASTM’s benchmark.
and development phase of the proposed rescue robot.
Accordingly, we need to first scrutinize the system Conceptual design
requirements for success in the determined mission By holding our mission in "A discussion on conceptual
considering the RRL’s framework, which is rooted design" section on one hand, and the desired perfor-
in NIST standard test methods, and then propose a mance requirements in "System requirements" section
Habibian et al. Robomech J (2021) 8:1 Page 7 of 33

Table 1 System requirements of Karo

Power Endurance: continuous ramps Distance (m) per charge 2000

Time (min) per charge 60
Battery life cycle 300

Mobility Terrain: Flat Surface Capability to repeat a 100-m path for 10 times (Y/N) Y
Average time per repetition (s) 200
Obstacle: Inclined Plane Maximum incline (°) with 10 repetitions for vertical, 40
diagonal, and horizontal paths
Average time per repetition (s) 30
Gap Crossing Maximum gap (cm) traversed for 10 repetitions 45
Average time per repetition (s) 15
Stair Climbing Maximum successful incline (°) for 10 repetitions 45
Average time per repetition (s) 15
Manipulation Maximum height of reaching space (m) 1.7
Payload (kg) 8
Door opening capability (Y/N) Y
Sensor Visual Colored video (Y/N) Y
Near (40 cm) field acuity (Y/N) Y
Far (6 m) field acuity (Y/N) N
Field of view (°) 35*75
Resolution (ppi) 1024*768
Audio Full/half-duplex communication (Y/N) Y
Localization and Mapping Capability to generate map in a maze with flat ground Y
(Y/N)
Capability to generate map in a maze with uneven Y
ground (Y/N)
Radio Communication Maximum distance in Line-of-Sight (m) 1500
Maximum distance in Non-Line-of-Sight (m) 800
Human-System Interaction Interface Xbox 360 joystick
Control scheme Tele-operated

on the other hand, we discuss the design of our system the location of the center of mass (COM) that may cause
conceptually. This section focuses on drafting the initial backflipping during climbing. Besides, the triangular flip-
locomotion and manipulation system for Karo by ana- pers are aimed to work similar to wheels but with only
lyzing various cases. In the end, a practical concept is three points of contact, and usually lack enough length to
chosen by considering eliminated weaknesses of Karo’s work as a lever.
ancestors (see Sect. 2) and inspirations from outstanding Adopting two pairs of flippers as illustrated in Fig. 4d–f
features of contemporary similar platforms [19–22]. provides the operator with substantial flexibility to adjust
Figure 3 shows the side view of a conventional tracked the horizontal orientation to avoid backflipping. This
response robot and its variations of mobility with sev- becomes particularly crucial when the robot moves on
eral common locomotion configurations. The simplest uneven terrains and certain view angles are required for
tracked robot such NAJI-I without flippers (Fig. 4a) inspection or object manipulation. Additionally, employ-
barely overcomes obstacles that are higher than the ing more tracked linkages increases the traction capabili-
robot’s height. This issue led the designers to think of ties since the robot creates more contact with terrains.
embedding assistive arm mechanisms to improve the Although four flippers enhance the mobility of the robot,
maneuverability, NAJI-II was designed the same way. our testing proved that having intermediate linkages
However, using a pair of flippers helps the platform to without tracks [Fig. 4d] limits the locomotion. Mainly
conquer the steps in cases (b) and (c) of Fig. 4, keeping because of the gap between the main body and flippers
the horizontal balance of the robot remains still chal- which does not have any tracks. We frequently noticed
lenging, especially in the case with triangular flippers. this issue with Scorpion during tests, and sometimes
This issue makes the operator incapable of controlling small, trapped objects between flippers and the body
Habibian et al. Robomech J (2021) 8:1 Page 8 of 33

Fig. 4 Step climbing of a response robot with various locomotion mechanisms. a No flippers. b One pair of regular flippers. c One pair of triangular
flippers. d Two pairs of two-linkage flippers. e One pair of triangular flippers combined with one pair of regular flippers. f Two pairs of regular flippers

physically locked the system. On the other hand, asym- capable of overcoming obstacles. On the contrary, Scor-
metric flipper design of NAJI-VII that is shown in Fig. 4d pion experienced a lack of mobility in gap crossing due
did not provide all the benefits of the four-flipper design: to a smaller size. Accordingly, we selected the four-flipper
(1) dissimilar reachability of flippers, (2) uneven contact design [Fig. 4f ], with an optimal optimum robot size, as
points of flippers in a 360° of rotation, and (3) restricted an appropriate locomotion mechanism that efficiently
height adjustment capability of the robot. can satisfy the technical requirements of Table 1. The
Note that the functionality of each case depends on detailed description of the mechanical design and analy-
several factors such as size, weight, location of COM, flip- sis for the mobile platform is presented in "Locomotion
per’s geometry, and the obstacle. For example, the step in mechanism" section.
Fig. 4 is only one example of many complex barriers that Manipulation capability is also an essential element for
exist in the real world. Harsh environments may con- response robots which was absent in the primary versions.
tain gaps, irregularly shaped barriers, and more impor- Later, the reconnaissance and exploration needs grew over
tantly, various surface materials. In a similar manner, time and urged scientists to use camera arms to provide
the dimensions/shapes of linkages also play a major role more flexibility in the inspection. Nowadays, most of the
in the performance. The configuration with no flippers rescue robots are equipped with manipulators which are not
could perform better compared to the four-flipper design only capable of object manipulation but also have devices
which is smaller in size. It is also important to remem- such as camera, temperature sensor, carbon dioxide sensor,
ber that the size and the weight of a robot are directly etc. Thus, the new generation of response robots has com-
related and restrict designers to exceed a certain bench- plex manipulation systems that often makes the operation
mark. Imagine we increase the body size to improve more difficult. From DOF to joint arrangements, how can
obstacle negotiation: the larger the robot’s dimension is a manipulation system be flexible and dexterous [31, 32]
the more powerful actuators and power sources we need enough for the requirements of a response robot ("System
for the system, which not only makes it pricey but also requirements" section)? By defining applications and work
challenging for us to place parts. This leads to a trade- environments, how can we design the workspace sufficient
off between robot size and the overall locomotion per- enough for all tasks? We aim to create a conceptual design
formance. Our prior work with various robot’s overall to work towards answering these questions for the desired
dimensions indicates that larger robot size is not neces- manipulator of a rescue robot.
sarily the game changer in real field performance. For According to RRL’s Dexterity test scenarios, a response
instance, NAJI-VII was limited to entering narrow tun- robot should be able to search victims while performing
nels or turning in mazes because of the size, while it was dexterous manipulation tasks in various locations in the
Habibian et al. Robomech J (2021) 8:1 Page 9 of 33

field. These embeddable tests, usually made of PVC pipes Mechanical design
(see Dexterity section in RRL’s rulebook [28]), are repeat- Based on the former discussion on the conceptual design,
able and reproducible test packages. They can be mounted almost all components and equipment (915 total and 230
on the planer and cylindrical surfaces at different heights/ unique parts without counting fasteners) of the robot
angles, or even be placed in hardly accessible locations such were designed and modeled in SolidWorks (Fig. 6) for
as inside vehicles, packages, and narrow gaps. The tests initial assessments. This section elaborates on the proce-
can be categorized as four different manipulation tasks: dure employed to mechanically design the system, select
inspection, disruptor aiming, object insertion, and object the appropriate actuators, analyze the mechanical com-
retrieval, it may also include lifting heavy objects, open- ponents, and identify the manipulation workspace. As
ing doors, cutting, and unlatching. Thus, a minimum DOF shown in Fig. 6, all of the electrical devices are modeled
with a suitable joint arrangement is required for the manip- as well, and their location was adjusted by using the Col-
ulator. Considering various operational scenarios in a field lision Detection feature in SolidWorks. Modeling all of
of disaster "A discussion on conceptual design" section and the components in 3D not only provided accurate weight
performance requirements in "System requirements" sec- estimation but also enabled us to examine the mechanical
tion, we developed a conceptual design of a 7-DOF robotic
arm (Fig. 5) with a specific order of joints. Figure 5 shows
the kinematic design; the cylinders are rotational joints,
and the cubic is the only linear joint. This concept provides
enough resilience for the operator to perform various dex-
terity maneuvers in mentioned operational situations. The
first joint rotates the base of the arm relative to the robot’s
orientation. This feature is crucial in a situation that mov-
ing the mobile platform is difficult for the operator. The
second and third joints mainly adjust the position of the
end-effector in the plane of operation. Further, the linear
joint helps aiming, retrieval, and insertion tasks in any posi-
tion. Also, the last three joints resemble the human hand
to perform rotation and bending motions. This link/joint
arrangement satisfies the required resiliency in essence.
"Manipulation mechanism" section addresses the detailed
mechanical design and how the conceptual design delivers
insights about the workspace of the manipulator. Fig. 6. 3D model of Karo in SolidWorks

Fig. 5 Conceptual design of the 7-DOF arm, presented with kinematic parameters and frame assignments
Habibian et al. Robomech J (2021) 8:1 Page 10 of 33

design of the parts in various arrangements. Additionally, body (two flippers and one main track) is independently
it undoubtedly improved Karo’s manufacturing quality driven by one bevel gearbox.
done by computer-aided machining. As a result, Karo has Figure 7b depicts the cutaway diagram of the locomo-
better functionality, reliability, and component accessi- tion mechanism located at the front right of the body, it
bility compared to its previous versions, which makes it is part of the symmetric design, the same structure exists
convenient for the setup team to rapidly deploy it before on the left side. The bevel gearbox drives a hollow shaft
USAR missions. that is coupled to pulleys which drive the rotation of all
tracks on the right side. The smaller shaft (flipper shaft)
Locomotion mechanism that passes through the hollow shaft and coupled with
The mobile platform is designed into three different the worm gearbox, rotates the front flipper. One end of
modules: side chassis, middle chassis, and flipper, which the flipper shaft is coupled with a flange (fastened to the
are shown with orange, blue, and green dashed boxes in flipper’s plate) while the other end is linked to an Old-
Fig. 7a, respectively. This centrosymmetric configura- ham coupling to accommodate a small amount of axial
tion provides a convenient assembly of parts in addition misalignment caused by tension regulation in tracks.
to decreasing the costs of manufacturing. Each module The tension of each track is adjusted using two bolts that
can be taken apart or replaced individually without com- set the center distance between pulleys. All the bearing
plete disassembly if any maintenance required, which arrangements are designed according to the manufactur-
meets the field serviceable feature of DHS-NIST-ASTM er’s standards.
standards. The locomotion system is embodied in 4 DC
motors (two for flippers and two for traction), 4 custom- Torque requirements analysis
ized gearboxes (worm gears and bevel gears), 8 steel alloy Following the design of the locomotion mechanism, and
shafts, 16 polyamide pulleys, 6 polyurethane tracks, and after the initial estimation of dimensions and weight of
12 aluminum body plates. To increase the stability during the robot, we can select appropriate actuators and stand-
negotiation with obstacles, three out of four motor gear- ard components. Since the robot should reliably oper-
boxes (two traction and one flipper motor gearboxes) are ate in harsh environments and be able to climb stairs or
placed at front of the body to avoid backflipping. Note steep slopes, cross gaps, and even pass uneven terrains
that both traction motor gearboxes are located below the [15], it is always prone to experience unidentified exter-
main tracks and not shown in Fig. 7a. Each pair of front/ nal loads. Analyzing the unexpected circumstances is
rear flippers are driven by one worm gearbox (e.g., the pretty complicated and laborious. Hence, for input power
front worm gearbox controls the angle of front flippers at calculation, we only selected the two most challeng-
the same time). Similarly, the traction of each side of the ing movement scenarios that are frequently performed

Fig. 7 Module arrangement and detail design of Karo’s locomotion mechanism in CAD
Habibian et al. Robomech J (2021) 8:1 Page 11 of 33

in missions: climbing steep ramps and lifting the robot second law to describe motion in i direction Eq. (1). The
with flippers. Targeting those cases as the continuous exerted resistance forces are highly dependable on the
operation benchmark would guarantee that the robot can terrain types, so any consideration would not be account-
effortlessly perform regular movements. The following able for the actual power loss. Instead, we apply a margin
presents analyses with basic assumptions to determine in the calculation to compensate for the effects of all the
the minimum output torque and power for the system. frictions in the system.
The traction system should overcome various resist-
ant forces when passing haphazardly formed obstacles.
Ftr − Mgsinα = Mai (1)
Climbing a ramp is one of the most common and chal-
lenging mobility tasks that the robot has to overcome Ftr = Mgsinα (2)
during operations. Here, we consider the robot climbing
where the angular acceleration of the pulley is ω̇ and
a 40° ramp (left side of Fig. 8), determine the minimum
therefore, the linear acceleration of the body is ai . Writ-
output torque required by each actuator, and select a
ing the general form of Euler’s rotation equations with
DC motor available in the market based on the desired
respect to the center of the actuation pulley:
speed of the robot. Note that this analysis is intention-
ally simplified to avoid complicated calculations. For To − Ftr b = I ω̇ + ω × (I × ω) (3)
instance, friction between the robot and the environment
is neglected due to variations in friction force, the opera- where To is the motors’ applied torques,I is the inertia
tion site includes fields made by various materials. Addi- matrix, and ω is the angular velocity of the pulley. The
tionally, the contact area of tracks and the normal force first term on the right side of Eq. (3) is zero because the
exerted on the robot could exceptionally alter the friction system has very small acceleration. The inertia matrix
force, especially since Karo’s tracks are made of Polyu- only includes the principal moments of inertia since the
rethane. Special circumstances of the standard model of center of rotation is aligned with COM of the pulley, so
friction are usually considered for rubber materials. The the second term is also zero. Considering all of the actua-
friction force in the power transmission system is consid- tion torque to be merely transmitted through inextensi-
ered as the power loss in calculations. ble components, the following relationship in Eq. (4) can
Figure 8 (left) presents the free body diagram (FBD) be written:
of Karo on a slope where Ftr is the total traction force Ftr b
applied by motors, M is the weight of the robot, N is To = (4)
2
the normal force, and α is the slope angle. Considering
no terrain slippage between the tracks and the ramp, where b is the center distance of the pulley to the outer
the force balance can be written according to Newton’s surface of the track. For our system α = 40◦, m = 85 kg,

Fig. 8 Free body diagram of Karo climbing a 40° ramp (left) and lifting the robot on the gap by using flippers (right)
Habibian et al. Robomech J (2021) 8:1 Page 12 of 33

and b = 12 cm, which requires 64.25 Nm of output torque flippers, and the manipulator, respectively. Considering
to enable the robot to climb the ramp continuously. Since the accelerations to be equal to zero, the reaction forces
two DC motors are considered for the locomotion, 32.12 are found to be F1 = 392.8 and F2 = 441.2 N when β is
Nm should be provided by each side chassis motor. By con- zero. To calculate the maximum required torque, the
sidering the desired locomotion speed to be higher than larger reaction force ( F2 ) is selected and therefore, 136.7
0.5 m/s, the RE 50, 200 W Maxon DC motor with the GP Nm of torque is required to rotate the flippers. This case
52 C planetary gearhead (26:1 reduction) coupled to the occurs rapidly during the lifting process, and as the flip-
bevel gearbox (4:1 reduction) was selected for each side pers’ contact point with ground gets closer to the body,
chassis. The RE 50 motor generates a nominal speed of the calculated values decrease significantly. However,
5680 rpm and nominal continuous torque of 405 mNm at tests in conditions such as step field pallets show that
24 V. The GP 52 C planetary gearhead [33] and the bevel flippers experience greater loads. For this reason, we con-
gearbox [34] cause 17% and 4% power loss respectively, sidered a larger safety factor to choose the motors for the
which it makes only 33.6 Nm of the output torque account- flippers. The RE 40, 150 W Maxon DC motor [33] with
able. This amount of torque generated by each motor is the GP 42 C planetary gearhead (reduction: 43:1) coupled
sufficient for continuously climbing the 40° ramp with a to the worm gearbox (reduction: 30:1) was selected to
speed of 0.8 m/s. The operation range of the RE 50 motor actuate each pair of the flippers. The RE 40 motor gener-
suggests that it can provide higher rotation speed when ates a nominal speed of 6040 rpm and a nominal continu-
less output torque is required or conversely. Therefore, it ous torque of 177 mNm at 24 V. The GP 42 C planetary
ensures that the robot can easily move in complex environ- gearhead and the worm gearbox cause 28% and 15%
ments and the motors are still capable of providing more power loss respectively, which it makes only 151.4 Nm
power if any abrupt unexpected situation happened during of the output torque accountable. The rotational speed of
the operation. flippers is 32 deg/s when flippers are in contact with the
The flippers help the robot to pass obstacles by chang- ground, they rotate faster in the unloaded case.
ing the location of COM and also providing more con- The above analysis is aimed to obtain a reasonable ball-
tacts with terrains. As the angle of each flipper and the park of the power requirements and also verify the design
position of the manipulator changes, the location of of the locomotion system. We tested Karo for both of the
COM shifts proportionally. Hence, one flipper motor discussed movement scenarios. "Locomotion mechanism
could experience a larger external load than the other experimental results" section represents the experimen-
depending on the situation. Figure 8 (right) illustrates the tal measurements of motors’ current and torque and dis-
FBD of Karo on a gap when only flippers are in contact. cusses their alternations in several sequences of motion.
Separate COMs are assigned for the front/rear flippers,
main chassis, and the manipulator, to account for changes Manipulation mechanism
in reaction forces based on the position of each part. For The manipulator is an essential element of Karo that com-
this analysis, similar to the traction system, we consider pletes the mission when the robot successfully reached
the most challenging condition in which both flippers victim’s location. In this section, we explain the mechani-
are completely horizontal and lifting the robot (e.g., β cal design, actuator selection, and present the workspace
is zero). The position of the manipulator is usually fixed of the manipulator. As depicted in Fig. 9, it is a 7-DOF arm
during this motion. The location of the exerted forces are made of rotational and prismatic joints, including the grip-
l1 = 30, l2 = 150, l3 = 100, la = 260, and lb = 310 mm. per’s jaw (Link 7). For all of the joints, off-the-shelf Robotis
F1 and F2 are reaction forces applied at the tip of each Dynamixel servo motors were selected because of their
flipper. Writing the force balance according to Newton’s high resolution, output torque, control algorithm, and
second law in j direction Eq. (5) and the moment balance compact design. All of the links are made of carbon fiber
based on the general form of Euler’s rotation equations or Aluminum to provide a lightweight and rigid structure.
Eq. (6) with respect to center O: One noticeable feature of this arm is operating on a mobile
platform that makes the manipulation more challenging
F1 + F2 − (m1 + 2m2 + m3 )g = (m1 + 2m2 + m3 )ai
compared to the industrial counterparts, because the posi-
(5)
tion and orientation of the arm constantly change the posi-
m1 gl1 + m3 gl 3 + F1 (la + lb cosβ) − F2 (la + lb cosβ) − m2 d3 = I θ̈ (6)

where ai and θ̈ are the linear and angular acceleration of tioning of the end-effector. Link 1 sets the orientation of
the robot, I is the inertia matrix, and m1 = 50, m2 = 13.3, the whole arm relative to the robot, Links 2 and 3 mainly
and m3 = 8.5kg are the mass of the chassis, two of the adjust the height of the end-effector; and the rest of the
Habibian et al. Robomech J (2021) 8:1 Page 13 of 33

Fig. 9 Link arrangement and mechanism design of Karo’s manipulator in CAD

links provide more flexibility to object manipulation and coupled to a ball screw, as depicted in the orange box in
inspection. Additionally, Link 6 enhances the inspection Fig. 9. The ball nut moves across the ball screw while the
by rotating the sensor box independently. Similar to the flange is constrained between the internal walls of Link
mobile platform, this manipulator is designed in a modular 3 and the prismatic link sliding inside ABS bush. These
configuration to make the assembly and accessibility con- high-efficiency mechanisms combined with the high res-
venient for the setup team. olution of servo motors provide precise motion with the
The blue box in Fig. 9 illustrates the timing belt drive minimum backlash for the manipulator.
design in Link 1. The servo motor can be fixed at differ- Figure 9 also illustrates the detail design of Joints 1, 2,
ent positions relative to the output shaft to adjust back- and 4 in the green, blue, and orange box, respectively.
lash and tension in the timing belt mechanism. Two Since manipulation tasks in rescue missions tend to be
deep-groove bearings are mounted on the output shaft, slow-paced, we avoided the conventional dynamic anal-
apart from each other, to tolerate radial and bending ysis of robotic arms and simply used the static form of
forces on the second link. For Joints 1, 5, 6, and 7, the Euler’s equations to calculate the required torque of each
20 W Dynamixel Pro is employed, which have a smaller joint for motor selection. Considering the fully-extended
dimension and weight. Joint 1 is coupled to a spur scenario (Fig. 10) as the continuous operation bench-
gearset (5:2 reduction) which makes a compact design mark, which usually takes place briefly during motion,
for the base besides increasing the output torque of the ensures that the manipulator is capable of operating
first joint. As depicted in the green box in Fig. 9, a com- when links are positioned at other angles.
bination of bearings is embedded in the base to absorb Here, torque calculation of Joints 2 and 3 are only dis-
shocks, improve rigidity, and provide smooth motion. A cussed since they require higher torques, identical calcu-
needle-roller bearing is positioned apart from a regular lations were conducted for other joints. The parameters
deep-groove ball bearing to ensure capturing the radial of Fig. 10 are mt = 8.5 kg is the total mass, mc = 2.8 kg
exerted forces to the main shaft. Additionally, the axial is the mass of Links 3–7, and mex is the external mass.
forces are tolerated by the thrust bearing situated at the lt = 40 cm is the distance from Joint 1 to COM of the
bottom of the output shaft. Another deep-groove bearing whole arm, lc = 45 cm is the distance from Joint 2 to
located along the axis of the servo motor’s shaft to endure COM of Links 3–7, lex = 130 cm is the distance from
the radial forces produced in the spur gear set. The lin- Joint 1 to the external load, and lj is the length of Link 2.
ear motion of Link 4 is created by the MX-106 T that is Writing the moment balance based on the general form
Habibian et al. Robomech J (2021) 8:1 Page 14 of 33

Fig. 10 Free body diagram of manipulator when it is fully-extended

of Euler’s rotation equations with respect to center O1 for and position of every join in space with respect to an
Joint 2 Eq. (7) and center O2 for Joint 3 Eq. (8): inertial reference frame. Figure 5 presents the kinematic
diagram of the arm and frame selection according to the
T2 − mt glt − mex gl ex = I θ̈2 (7) D-H convention. By using the relationship between two
successive local coordinate frames, i and i + 1 we can
obtain the D-H kinematic parameters. According to this
 
T3 − mc glc − mex g lex − lj = Iθ3 (8)
method, ri , αi , di , and θi are required to define the trans-
where θ̈2 is the angular velocity of links 2–7, θ̈3 is the formation matrix between two coordinate frames next to
angular velocity of Links 3–7, and I is the inertia matrix. each other. The transformation matrix, which can calcu-
Without existence of an external load, Joint 2 requires late the position of a point in i + 1 coordinate frame with
3.4 Nm and Joint 3 requires 1.2 Nm of minimum torque respect to i coordinate frame is given as:
to support the weight of parts. We employed the 100 W
cosθi −sinθi cosαi sinθi sinαi ri cosθi
 
Dynamixel-P (PH54-100-S500-R) with the continuous
 sinθi cosθi cosαi −cosθi sinαi ri sinθi 
output torque of 25.4 Nm, the output speed of 29.2 rpm, Ai = 
0 sinαi cosαi di  (9)
and the resolution of 0.0004 deg/pulse for both Joints 0 0 0 1
2 and Joint 3. Due to the higher power requirements of
Joint 2, we added a timing belt drive (3:1 reduction) to For our system we need to calculate the above matrix
enhance the output torque. This mechanism with the for each degree of freedom to obtain a space-fixed matrix
high efficiency of 98% [35] enables the manipulator to lift of transformation as follow:
an external mass mex = 5.6 kg at the fully-extended posi-
tion by using Joint 2, the payload increases as lex becomes
Ti = A1 ∗ A2 ∗ A3 ∗ A4 ∗ A5 ∗ A6 (10)
smaller. The first six joints determine the location of the gripper
in space. Each joint has four kinematic parameters; there-
fore 24 total parameters should be defined to calculate
Achievable workspace
the position and the orientation of the end-effector. Fig-
Performing dexterity tasks is directly related to the loca-
ure 5 shows the coordinate frame assignments and D-H
tions that the end-effector can reach within an environ-
parameters of the arm, which are selected in a fashion
ment. For this manipulator, since it is mounted on a
to zero some of the parameter and make the computa-
mobile platform, there are some conflicts with the body
tion relatively faster. Table 2 lists these parameters of the
of the robot, which makes it more important to deter-
manipulator at the home position based on Fig. 5.
mine its workspace. In this section, we aim to simulate
By using the geometrical constant parameters and
the workspace of the manipulator by using the Denavit-
Eq. (8), the position of the end-effector is determined
Hartenberg method, based on the motion range of each
for various sets of joint positions and the workspace is
joint, to identify those conflicts and achieve the best
depicted in Fig. 11. The three view orientations of the
functionality for the operator. To calculate the transfor-
workspace provide useful information about the acces-
mation matrix and ultimately, positions of the gripper, we
sibility of the arm to the surroundings of the robot, it
need to define link kinematic parameters, known as D-H.
can reach points within a sphere with a radius of 130 cm
Using D_H parameters allows us to find the orientation
Habibian et al. Robomech J (2021) 8:1 Page 15 of 33

Table 2 D-H parameters of the first six degrees of freedom

Link Joint Variable ri αi di θi Range

1 θ1 0 − 90° d1 θ1 − 80° to 80°

2 θ3 d2 − 180° 0 θ3− 180° 0° to 180°
3 θ3 0 − 90° d3 θ3− 90° 0° to 180°
4 d4 0 0° d4 0 0 m to 0.4 m
5 θ5 0 − 90° d5 θ5 0° to 360°
6 θ6 d6 90° 0 θ6− 90° − 90° to 90°

which is centered at the base of the manipulator. The manipulation. Additionally, the front view shows that the
left view shows that the arm can access 48 cm below points within 100 cm of the left and right of the robot are
of the robot, however it is restricted to the front half of accessible. As the top view demonstrates, the operator is
the body. There is also an unreachable region close to able to reach points that are within 92 cm of front flip-
Link 1, which is practically not the area of interest for pers and 23 cm of rear flippers. Lastly, there are 10° and

Fig. 11 The workspace of manipulator from left, front, and top view orientation (all dimensions are in centimeters)
Habibian et al. Robomech J (2021) 8:1 Page 16 of 33

20° slices of the workspace due to the range of motion of design and analysis tool. As an example, Fig. 12a illus-
Joint 1. In conclusion, by using this method we are able to trates the geometry of the flipper’s shaft, imported from
control the workspace by adjusting the constant param- SolidWorks, under constraints and loadings in the mech-
eters or limiting the joints to achieve the desired work- anism. The AISI 1010 carbon steel is used as the material
space or avoid singular positions of the mechanism [36]. for this part, which is modeled with 13,315 quadratic tet-
Moreover, the operator can use the workspace plots gen- rahedron elements (type C3D10). As shown in Fig. 12c,
erated by forward kinematics study to get some insights 5.14 kPa of maximum principal stress is created by an
into positioning the robot while reaching objects within applied static torque of 152 Nm (calculated torque in
an environment. "Manipulation mechanism" section) creates, which does
not reach the yield strength of AISI 1010 steel (305 Mpa).
Component analysis The torsional displacement of the shaft is accordingly
The previous analysis helped determine the motors, gear negligible as shown in radians in Fig. 12b.
ratios, bearings, and revising dimensions of the mechani- A similar approach was taken for other critical compo-
cal parts. The next step is analyzing mechanical compo- nents to make the design more reliable and robust. How-
nents of the robot to ensure that all parts have enough ever not all of the loads on the system can be calculated
strength while the robot’s weight is optimized. Addition- due to the dynamics of the operation environment. Over-
ally, this analysis enables identifying the critical sections all, FEA assisted us to recognize the critical stresses for
of each part and redesigning it in such a way that stress each component and optimize the geometry, and there-
is not only reduced but also uniformly distributed. For fore, helped to reduce the overall weight of the robot,
power transmission components such as gears, bear- which essentially contributed to less power consumption.
ings, and shafts, the stress is calculated using standard
mechanism analysis [37] to avoid failure. Major compo- Control and command system
nents of the robot are either geometrically sophisticated Overview of the system
or they are prone to complex loading patterns, which We have proposed a mechanical design for robot’s plat-
makes solving stress–strain equations tedious. This mat- form and its manipulator in "Mechanical design" section.
ter becomes significant when studying fatigue failure due These designs took the first step towards the objective
to cycling loadings under varying circumstances [38, 39]. of this research, which is discussed in "Introduction"
In order to efficiently investigate those cases and opti- section. Thus, we discuss developing Karo from electri-
mize our design, we employed the finite element analysis cal system design point of view now on to take the sec-
(FEA) simulation, using the commercially available soft- ond needed step, which is the design of the control and
ware Abaqus from Dassault Systèmes, as an additional command system. Basically, we design a control and

Fig. 12 Finite element model of a flipper’s driving shaft in Abaqus

Habibian et al. Robomech J (2021) 8:1 Page 17 of 33

command system to (1) effectually control the actua- processes such as simultaneous localization and mapping
tors embedded in the mechanical structure, (2) derive (SLAM) and one ARM Cortex-m3-based board to handle
diverse sensors and perception devices to accurately per- low level peripherals and IOs. The manipulator module
ceive the environment and control the system itself, and also profits from two similar ARM-based boards acting
(3) build a sophisticated platform for high-level software as its controller and low-level peripherals handler. Basi-
systems and algorithms to be implemented. As Fig. 13 cally, in the explained distribution of computing units,
delineates, we designed a control and command system we tried to assign more expensive computations to mini
which mainly is comprised of three modules (1) mobile PCs, such as any kind of image processing, while the
platform, (2) manipulator, and (3) operator control unit hardware-involved computations, such as a position-con-
(OCU). Here, we aim to depict an overview of the system troller, are assigned to ARM-based microcontrollers.
from computation and communication perspectives. Communications: the communication scheme of the
Computations: Each module deals with several devices, system is highly critical for the stability and quality of
electronic or mechanical components and acquired the robot’s performance, since controlling tele-oper-
data by the way of computing units. As Fig. 13 illus- ated robot, video and sound streaming, system diag-
trates, OCU retains only one mini pc, that is Intel NUC nostics, sensors feedback, visualizing procedures and
kit, as its computing unit, while the robot platform SLAM in a remote station are all relying on the remote
module benefits from one similar mini pc for high level communication platform. In this regard, the OCU is

Fig. 13 The block diagram of the command and control system. For the sake of simplicity, all connections involving power distribution have been
excluded
Habibian et al. Robomech J (2021) 8:1 Page 18 of 33

equipped with UBIQUITI Networks 802.11a/b/g Bullet Thermal Camera: a thermal camera, Optris PI230, have
M5 Access Point/Bridge where it connects to the iden- been installed on the manipulator’s end effector which is
tical access point on the robot platform module wire- capable of synchronous capturing of visual and thermal
lessly. To ensure the communication stability and images. This camera is used for detection and position
robustness in long distances, the access points function estimation of victims during a mission as a part of the
on a 500 MW-power basis. All connections between the robot’s exploration operations. It also ensures robot’s vis-
robot’s access point and other computational nodes on ual acuity by providing thermographic information from
the mobile platform and the manipulator module are targeted objects.
established throughout Ethernet. Besides, we utilize CO2 sensor: we have prepared a gas sen-
the UDP communication protocol for transmitting the sor, MQ-9, which detects the presence of multiple types
control commands between OCU and the ARM-based of gases in an environment. While this sensor has sev-
control boards, since such command transmissions are eral applications for a rescue mission, we are more
time-sensitive. intended to verify a victim’s breathing as a vital sign
throughout this sensor. For instance, as illustrated in
Sensors Fig. 23, Karo uses its sensor module including the gas
To adequately perform the required tasks, Karo needs to sensor to explore the victim’s hole.
be equipped with a wide variety of sensors such as laser Analog Cameras: Four analog cameras have been
scanner, inertial measurement unit, RGB-D camera, shaft mounted on the robot’s platform and its manipulator
encoder, and thermal camera. This sensory system func- which provide a decent perspective for the robot’s opera-
tions to accurately perceive the environment, in terms of tor to monitor the robot, manipulator and surround-
obstacles, victims, signs, and objects. Here, we provide a ings by the installed cameras. Figure 16b demonstrates
brief description for each of them: an image capture of the robot’s graphical user interface
Shaft Encoder: as explained in "Locomotion mecha- (GUI) including the video streams of the analog cameras.
nism" section, the robot’s platform is equipped with IP Camera: when the manipulator approaches a
two brushed DC motors for its movement and two others detected victim, the operator needs a high-resolution
for its front and rear flippers. Accordingly, the robot has video of the location to complete the inspection task as
been set up with four incremental optical rotary shaft clear as possible. To that end, a Sony high-resolution IP
encoders, which make wheels odometry calculation pos- camera has been embodied into the manipulator’s end
sible. The system eventually benefits form this odometry effector and Fig. 16a illustrates how the IP camera pro-
for DC motor control purposes and the localization and vides a perceptive and broad sight of the car’s ceiling.
autonomous navigation methods [40]. Mono Microphone: two microphones have been
Laser Scanner: since the expected operation environ- installed on both the OCU and the manipulator’s end
ment of Karo includes slopes, steps, stairs, and unstruc- effector to facilitate a full-duplex audio connection.
tured obstacles, the odometry method alone does not
suffice for the robot’s localization. Thus, for both localiza- Power systems
tion and 2D-mapping of the system, a Hokuyo UTM30- According to the schematic design of Karo’s electrical
LX LIDAR is prepared for the robot, which is mounted on systems, there are two separated power sources to sup-
a stabilizer to be parallel with the ground on inclined sur- ply the consumption of signal devices and actuators. The
faces. As depicted in Fig. 13a, the laser scanner is con- rationale behind this separation is the inherent differ-
nected to the robot platform’s mini PC throughout the ence in their consumption patterns. The signal devices,
USB connection and in "Simultaneous localization and such as access points, cameras etc., have often a constant
mapping" section we will discuss how this data is used for consumption through the whole operating time. On the
Karo’s mapping. other hand, the robot’s actuators have significantly spo-
Inertial Measurement Unit: we utilize the measure- radic and dynamic consumption depending on the obsta-
ment regarding the changes in the attitude of the robot’s cles on the way or the operations in progress.
platform using a 6-DOF inertial sensor, Xsens MTI-100, To provide the power source for Karo’s devices and
to control the laser scanner’s stabilizer. The functional- actuators, we first need to determine the systems specifi-
ity of this stabilizer gets indispensable when the opera- cations from consumption point of view. The power sec-
tion terrain is not even. We will discuss in "Evaluation tion of electronic devices needs a 24-V power source with
of Karo" section regarding the highly uneven terrains on 4 A continuous current. For the sake of the actuators, we
which Karo operates and accomplishes the exploration similarly need a 24-V power source while its continuous
operations. current ranges from 0 to 20 A. The other factor to be con-
sidered is the system’s burst current, which determines
Habibian et al. Robomech J (2021) 8:1 Page 19 of 33

how quickly the batteries are going to be discharged. This The end-effector board: the end-effector board is
factor is only depended on the actuators’ behavior due to installed inside the sensor box of the manipulator. Basi-
their stall current when they are applying their maximum cally, it electrically handles all equipment and sensors in
torque. Accordingly, the c-rating of the power source the sensor box such as controlling the LEDs, controlling
must satisfy this characteristic of the system. the gripper’s DC motor, data acquisition from gas sensor,
Considering the above discussion, we provided two supplying the speaker and cameras etc.
24 V lithium polymer battery packs with 10,000 mAh Motor Drivers: one of the most serious challenges in
capacity and a c-rating of 10. Having said that, the power the hardware design of the control and command system
section of electronic devices can operate up to 2.5 h while is designing the motor drivers of the mobile platform’s
the provided power source for actuator suffices for half actuators. As elucidated in "Mechanical design" sec-
an hour on average for each a mission. We will later on tion, there are two Maxon RE 200 W DC motors for the
investigate the consumption pattern of the system and robot’s movement and two Maxon RE 150 W DC motors
sufficiency of the provided power sources in “Analysis of for the robot’s flippers. Thus, the main challenge would
Batteries’ Discharge” section. be handling of both continuous and stall currents of these
high-power actuators aligned with decent control and
Hardware design communication implementations in the motor drivers.
We need to design different ARM-based control- Accordingly, we designed the motor drivers based on the
ler boards for both robot’s platform and its manipu- H-bridge circuits for switching the motor’s polarity via
lator to fulfil the schematic design of the control and PWM signals. Practically, each of these custom-designed
command system, as illustrated in Fig. 13. All these motor drivers supplies a load with 24 V and up to 20 A
controller boards benefit from LPC1768 Cortex-M3 current, which confidently satisfies the system’s require-
microcontrollers with similar developed firmware. Since ments (Fig. 14).
there are numerous specific requirements, constraints,
and considerations regarding those controller boards, it Operator Control Unit (OCU)
was infeasible to choose a generic product in the market. The portability, reliability, and connectivity of an OCU
Accordingly, we aimed to design those controller boards in a rescue mission significantly impact the robot’s per-
from scratch to satisfy Karo’s system requirements as formance. By the way of the OCU, the operator should
effective as possible. Here, we explain the functions of be able to control the movement of robot and manipula-
each developed hardware for both robot’s platform and tor, observe the environment with broad and wide per-
manipulator modules: spective, monitor the robot’s internal and external states,
The main controller: the main controller board is and interact with the robot’s advanced functions such
installed on the mobile platform to (1) manage robot’s as mapping, exploration, object detection, etc. Accord-
power system, (2) communicate with motor drivers by ingly, we designed and implemented the OCU embedded
sending the commands received from the OCU, and (3) in waterproof and dust proof case, as shown in Fig. 15a,
control robot’s peripheral equipment. To that end, vari- which can be set-up and break down the robot operation
ous DC/DC convertors supply devices with different volt- system in less than 10 min. In this regard, Fig. 15b deline-
age levels such as 3.3, 5, 12, and 15 V. Besides, the board ates the operator remotely controlling Karo to climb up
controls robot’s front and rear lights, alarms, indicators, the stairs.
and the encloser’s fans by a series of relays. To commu- Besides, the OCU is facilitated by two packages of
nicate with the motor drivers, the main controller relies 10,000 mAh batteries that supply the whole system for
on Controller Area Network (CAN) protocol, which more than an hour of operation. All other the OCU’s
provides relatively a robust communication between devices and equipment have been depicted in Fig. 13c.
nodes on a mobile system. In addition, we implemented
the hardware requirements for Ethernet protocol on the Software systems
main controller board using a 10/100-Mbps Ethernet Human–robot interface
PHY to establish a reliable connection with the OCU. As the dexterity, mobility and exploration capabilities
The manipulator’s controller: the manipulator con- of the developed rescue robot presents higher complex-
trol board mainly functions to control the manipulator’s ity, operating the robot becomes more challenging and
servo DC motors. To that end, it is equipped with the complicated. Accordingly, the way that the HRI is imple-
same Ethernet hardware as the main controller board to mented can significantly affect the desirability of the
communicate with OCU. Further, it communicates with robot’s performance by providing an effective, conveni-
the servo motors through the RS-485 serial protocol. ent, and friendly interface for the operator. In the devel-
oped software system integrated with Robot Operating
Habibian et al. Robomech J (2021) 8:1 Page 20 of 33

Fig. 14 Hardware design of (left) the manipulator controller board, and (right) the end-effector board

Fig. 15 a The operator control unit of the robot, and b the operating controlling the robot

System (ROS) [41], we implemented the human super- movement, and joint states etc., as shown in Fig. 16.
vision and control of the robots as two independent but Moreover, the robot’s GUI provides a dynamic and inter-
closely packed software. Since the cores of these two active visualization of the 2D map generated by the robot
software are different and it is inadmissible in ROS for a by employing the RVIZ plug-ins [42] of ROS. Besides, by
single node to connect to more than one ROS core, we utilizing ROS’s RQT GUI plug-in, the interface’s features
tackled a software engineering challenge to create an can be evolved and customized as needed with no need
integrated GUI for the whole system. to major changes in the underlying deployed codes.
In the design of the HRI, we chose the colors of com- The control commands for robot’s mobile platform
ponents in the robot’s GUI in such a way that the color of and manipulator are sent throughout an Xbox 360
each section naturally reflects its application. As a result, wireless controller. The controller mainly serves in two
the operator can get use to the GUI’s feedbacks and con- mode (1) to control the robot’s movement and position
trols as effective and fast as possible. The GUI essentially of flippers, and (2) to controller the robot’s manipula-
serves for visualizing the sensory information regarding tor. The assignment of functions to buttons is a critical
the state of the environment and the robot itself including task to do since it must provide simultaneous control
video streams, gas sensor, thermal images, robot’s power, over multiple function of the robot for the operator as
Habibian et al. Robomech J (2021) 8:1 Page 21 of 33

Fig. 16 This figure illustrates the implemented HRI for Karo; a the video stream of the IP camera installed on the manipulator’s end-effector
inspecting the car’s ceiling, b an image capture of the GUI provided for the operator, c the operator while using the joystick to control the robot,
and d Karo in the operation field trying to inspect the targeted car

convenient and efficient as possible. Besides, we have To address the SLAM problem on mobile robots we
provided a safety feature for the robot called armed/ need to deal with some serious challenges: (1) the USAR
unarmed mode which prevents any unintentional com- fields are covered by unstructured obstacles, stairs and
mands from being sent to the robot. In addition to the steps, and inclined surfaces which makes normal planar
Xbox controller, the commands can be sent using the indoor solutions [45] insufficient and inapplicable; and
keyboard or other controllers utilizing an abstract ROS (2) most of the SLAM-based methods require accurate
node, which can be easily modified according to the odometry data as a part of the input, while odometry data
type of the controller. is inherently noisy and unreliable for localization pur-
poses. Hence, we utilize a flexible and robust SLAM sys-
Simultaneous localization and mapping tem with 6-DOF motion estimation [46] for the robot’s
Rescue robots are expected to effectively explore the 2D map generation. Furthermore, a precise LIDAR sys-
environment and accomplish the reconnaissance opera- tem with high update rate, as explained in “Sensors”
tions as well. Mapping is one the most imperative recon- section, has been installed on the robot to enhance the
naissance tasks assigned to rescue robots in a disastrous accuracy of the mapping. To cope with the disturbance of
situation. Principally, generating a 2D map of the envi- uneven terrains, a 2D stabilizer compensates the changes
ronment, in which the locations of victims have been in the attitude of the laser scanner and robot’s platform,
uncovered, provides a decent perspective for rescue per- caused by the slope gradient, throughout the read data
sonnel to efficaciously progress the rescue mission. In from the IMU. In this regard, Fig. 17 illustrates a 2D map
this regard, the occupancy grid maps [43] have received generated by the robot in an extremely rough environ-
adequate attentions throughout a wide variety of appli- ment. As a matter of fact, this map is a merged 2D map
cations such as localization, path planning, and collision by Karo and another complementary robot in the mis-
avoidance. Besides, there have been many works focus- sion. Map merging capability of Karo gives an overall
ing on the SLAM topic [44], which is based on occupancy and decent perspective of the environment to the rescue
grid maps, as an important concept of mapping in mobile personnel which is more effective and useful than giving
robotics. them multiple maps with partial observations. This has
Habibian et al. Robomech J (2021) 8:1 Page 22 of 33

objects, and other points of interest. To that end, we

developed a highly modular system which benefits
from a multi-layer architecture including behavior con-
trol, global path planner and trajectory generation lay-
ers. Here, we briefly describe each of these layers:
Behavior control: behavior control functions as a
high-level controller and determines the robot’s objec-
tive, which is approaching a potential victim or explor-
ing the environment.
Global path planner: the global path planner gener-
ates a path regarding the goal set by behavior control
layer in order to maximize the coverage over the envi-
ronment and minimize the distance travelled.
Trajectory generation: when goal point has been
determined and a path is generated for reaching that
goal, the trajectory planner generates trajectories that
can guide the robot to follow the generated path till it
gets to the goal point considering the current states of
the robot.
Fig. 17 This figure illustrates a merged 2D map by Karo and another
In practice, there are many complexities that makes
complementary robot in the mission which exemplifies a collective
behavior [47, 48]. In this map, the blue lines represent the obstacles the explained implementation challenging. For
where blue and red spots stand for decoded QR codes using instance, the environment is initially unknown for the
image caption generation method [49] and the detected victims, robot. Thus, it is infeasible for the global path planner
respectively. Moreover, the purple lines correspond to the paths to find the best path in a partially observable environ-
taken by Karo
ment. However, as the robot keeps exploring utiliz-
ing the frontier-based method [50], the environment
becomes more observable for the robot and the output
been done by finding transformation between two maps, of the global path planer converges to an optimal solu-
generated by two different robots, and merging them tion. In despite the abstractions and simplifications of
accordingly. the autonomous exploration function implemented for
the robot, its performance turned out to be an effec-
tive complement for the robot’s tele-operative mode
Autonomous exploration especially in flat and wide corridors. For instance, it is
To take a step towards autonomous behavior of the robot, assertively feasible to employ this autonomous func-
we implemented a simplified autonomous exploration tion in two of the Maneuvering test suites to enhance
utilizing the existing sensors and devices on the robot. the robot’s overall performance, as will be discussed in
By a “simplified” implementation we mean that the con- "Evaluation of Karo" section.
trol of flippers still needs to be tele-operative and the
algorithm controls only the mobile platforms movement.
Having said that, this function does not suffice for per- System specifications
forming autonomous exploration in excessive uneven Basically, this section mediates between the presented
terrains such as environments with stairs, ramps etc. design procedure of Karo and the experimental results.
However, the autonomous exploration function of the In other words, the robot’s specifications presented in
robot can enhance the robot’s performance in flat cor- this section stand for the theoretical expectations which
ridors or terrains with moderate obstacles such as the has to be practically investigated by a wide variety of
“Crossover” test of the RRL’s maneuvering test suites experiments and field evaluations. Table 3 illustrates the
or “Map on Continuous Ramps” from exploration test robot’s specifications mainly focused on robot’s move-
suites. ment, power system, manipulation, and application. In
The robot in the autonomous mode is expected to fact, these specifications are the building blocks of the
explore the unknow environment and perceive the sur- robot’s performance from mobility, dexterity, and explo-
roundings for detecting victims, QR codes, specific ration points of view. For instance, the set of sensors
that the robot is equipped with shapes its exploration
capabilities, which will be discussed in the next section.
Habibian et al. Robomech J (2021) 8:1 Page 23 of 33

Table 3 Karo’s Specifications

Robot’s specification Value

Name Karo
Typical operation size 0.8 * 0.6 * 0.6 m
Transportation size 0.9 * 0.8 * 0.7 m
System Weight 85 kg
Weight including transportation case 100 kg
Unpack and assembly time 210 min
Locomotion Tracked
Assistive mechanisms 4 flippers
Maximum speed (flat/ outdoor/ rubble pile) 0.8 / 0.6 / 0.3 m/s
Turning Diameter Zero
Payload (flat) 30 kg
Power source Lithium Polymer batteries
Battery Endurance (idle/ normal/ heavy load) 90 / 40 / 20 min
Batteries’ charge time (80%/ 100%) 45 / 60 min
Manipulator 2 link arm, DOF 7
Manipulator’s reach (vertical/ horizontal) 130 / 130 cm
Manipulator’s payload at full extend 5 kg
Sensors four analog cameras, one high resolution IP camera,
thermal camera, gas sensor, IMU, Laser Scanner, and
microphone
Communication IEEE 802.11a 5 GHz 500mW
Tether optional
Operation modes Tele-operative and semi-autonomous (flat ground)

However, these sets of specifications provide a compre- M

hensive comparison with robots investigated in DHS/ I= + I0 (11)
Km
NIST sponsored evaluation exercises [30].
Figure 18 illustrates the succession of Karo in climb-
Experiments ing a 40° ramp and drawn current/torque from the
Locomotion mechanism experimental results motor in each sequence. Sequences 1 and 4 both
In order to confirm that the purposed analysis of "Loco- include a period that the robot does not have any
motion mechanism" section assisted motor selection, motion (0–1 s and 9–10 s), and a period of moving on a
we tested Karo while performing both scenarios and flat surface (1–3 s and 7–9 s). Measured current values
measured the drawn current from each motor during during the motion periods are about 2 A, which is the
motion (Fig. 18). Particularly, these tests verify our prior indication of existing friction between the robot and the
considerations for power loss due to the undetermined surface. As the robot reaches the ramp (Sequence 2),
frictions. It should be mentioned that the robot tested the current increases until it is completely on the ramp
with full equipment, the manipulator is not shown in (Sequence 3), which draws the maximum current from
Fig. 18. Since each of DC motors has a unique mechani- the motor. For reasons unknown, there are irregular
cal specification (i.e., torque constant and nominal cur- spikes in measurements where the robot’s body enters
rent), the corresponding load torque at each measured the inclined surface and leaves the bottom landing.
current value is calculated by using Eq. (11), taken from However, there are several possibilities: the slippage
the Maxon Motor datasheet, to verify that both motors between tracks and the floor; uneven wear on tracks;
operate within the suggested continuous range by the and possible shocks during the transition between
manufacturer company. The torque constant Km is 38.5 surfaces. Although some of the measurements pass
mNm/A and 30.2 mNm/A, and the nominal current I0 the 10.8 A (405 mNm) maximum continuous current
is 236 mA and 137 mA for the 200 W and 150 W DC (torque) of the 200 W motor, the overall performance
motors, respectively.
Habibian et al. Robomech J (2021) 8:1 Page 24 of 33

Fig. 18 Feedback current and corresponding torque of the traction motor during ramp (40°) climbing (right) and the flipper motor during a course
of rotation (left)

does not exceed the continuous operation range, so the Analysis of batteries’ discharge
robot can smoothly operate in various conditions. This The power system of remotely operated robots plays an
can be seen clearly from the 8 A difference between the imperative role in their functionality and effectiveness
measured current on a flat surface and an inclined one. while completing an assigned mission, especially when
Additionally, climbing ramps or stairs at (40°) angle they are equipped with on-system power sources. Cor-
happens for a short period and it is considerably less respondingly, we need to validate the adequacy of Karo’s
frequent during operation. power system by investigating its durability in various
Similarly, Fig. 18 depicts Karo’s flippers lifting and standard missions (Center, Crossover, Travers, and Stair
putting down the robot in three sequences, starting at Debris). We selected these four different RRL’s tests as the
β = −45◦ and finishing at β = 45◦. The lower Greek testbench for measuring the batteries discharge pattern.
number indicates the starting position, and the higher We know that Karo would have the highest consump-
Greek number indicates the finishing position in each tion in Stair Debris test compared to other selected tests,
succession. There is no movement in the first and last since its movement actuators require higher torques to
seconds of motion. In Sequences 1 and 3, the robot/ overcome the obstacle. Pragmatically, we seek the durabil-
flippers’ weights help the motors to rotate effortlessly, ity of Karo consistent with the RRL’s 30-min missions for
so the demanding current or torque from the system is the worst-case scenario of the robot’s consumption (Stair
very low. The unknown frictions impose less than 1A to Debris test). As the robot is equipped with two battery
each motor. The highest current was measured when the packs for its actuators and electronic devices, we conduct
flippers contact the floor at the beginning of Sequence the experiment for both power resources separately.
2. There are similar unidentified spikes in this transi- Figure 18a illustrates the discharge measurements for
tion, which may be caused by the large contact surface or the actuators’ battery pack in 60 min. As we expected,
slippage between tracks and the floor. It is important to the batteries discharge in Stair Debris test earlier com-
know that the peak current happens in less than 1 s every pared to other tests, though they endure 42 min (i.e.,
time the flippers lift the robot. This indicates that flippers 12 min more than required durability for RRL’s missions)
are certainly capable of assisting the motion in various
conditions in which the robot overcoming obstacles or
trapped between objects.
Habibian et al. Robomech J (2021) 8:1 Page 25 of 33

and satisfy the primary requirements. Besides, since the exploration capabilities. As we stated in "Introduction"
Center test only includes flat terrain, the battery dis- section, the main motivation of this work is to embed all
charge rate is relatively slower which ensures more than four complementary capabilities in a single rescue robot
one hour (around 74 min) of continuous operation for for accomplishing an effectual USAR mission. In this
the robot. We observe that the durability of the robot’s regard, the RRL test suites inspired from DHS-NIST-
batteries decreases by increasing the obstacles’ complex- ASTM international standard test methods facilitates
ity in the selected tests, where the batteries’ durability this aim by offering methods to evaluate rescue robots
dropped 43% from Center test to Stair Debris test. quantitively. Accordingly, we tested Karo in all four test
The same measurement has been applied to the battery suites meticulously based on the rules and frameworks of
pack of the electronic devices, as shown in Fig. 18b. In RRL. To draw a conclusion about the results confidently,
despite of robot’s actuators, the electronic devices have we repeated each test 20 times where each test was con-
almost a constant power consumption regardless of the ducted in a 30-min time slot. Thus, we will discuss the
type of the mission. As a result, in all four tests the bat- results and their variation for each test first, and then we
tery’s discharge patterns are significantly analogous and will compare the Karo’s performance with other robots
demonstrate a drop of voltage from 12.6 V (fully charged taken part in RRL.
battery’s voltage) to 11.45 V after 60 min of operation, Maneuvering is the most basic capability required for
where the discharge voltage is 9 V (Fig. 19). accomplishment rescue missions. Figure 24a illustrates
the results of 20 tests for each Maneuvering test taken by
Karo’s performance in RRL the Karo. Although the Maneuvering tests have no major
Evaluation of Karo mobility challenge for the Karo, as shown in Fig. 20 for
In this section we aim to evaluate the Karo comprehen- the Traverse test, maneuvering in confined corridors is an
sively in terms of maneuvering, mobility, dexterity, and exacting task to accomplish for maxi-sized robots. Since

Fig. 19 The discharge pattern of the robot’s battery packs over a course of 60 min: a actuators’ battery pack, and b signal devices’ battery pack

Fig. 20 Karo is fulfilling the Traverse test of the Maneuvering test suite
Habibian et al. Robomech J (2021) 8:1 Page 26 of 33

the RRL’s rule allow five more minutes only for autono- Similarly, the dexterity skills of Karo have been evalu-
mous operations, we utilized the autonomous explora- ated quantitatively throughout five Dexterity tests. As
tion capability of the robot in Center, Continuous Ramps, illustrated in Fig. 24c, Karo has accomplished all Dexter-
and Crossover ramps of the Maneuvering test suite. As a ity tests fulfilling with small variations in the 20 repeti-
result, the robot completed up to two more repetitions in tions for each test. One explanation can be the similarity
those tests. of the test elements in parallel pip, omni-directional, and
The mobility capabilities of Karo have been evaluated cylindrical pipes tests. In this regard, Fig. 22 demonstrate
in five Mobility tests based on the RRL’s framework, as Karo’s performance while inspecting inside parallel pipes.
demonstrated in Fig. 24b. Statistically speaking, the robot Beyond the RRL’s dexterity test suit, Karo’s dexterity
performed the best on the Hurdles and Stair Debris, capabilities have also been evaluated in a complementary
which is one of the most challenging mobility tests. The testbench, where the robot must rely on both its mobil-
results of the mobility tests imply that the robot ben- ity and dexterity skills to inspect holes in a highly uneven
efits from an immense stability while overcoming diffi- environment. As shown in Fig. 23, Karo is inspecting a
cult obstacles, although its performance on Sand/Gravel hole on uneven Stepfields to detect a simulated victim.
Hills test is ordinary. Moreover, Karo also accomplished In contrast to all other evaluations, Karo’s exploration
a satisfying operation on the Stepfields test, as shown in capabilities have been examined via only three explora-
Fig. 21. tion tests. That is mainly because two tests of Explora-
tion test suite (i.e., Avoid Holes and Avoid Terrain tests)

Fig. 21 The robot is maneuvering on Stepfields test of Mobility test suite

Fig. 22 Karo is inspecting inside parallel pipes of Dexterity test suite

Fig. 23 Karo is inspecting a victim’s hole in highly uneven terrains

Habibian et al. Robomech J (2021) 8:1 Page 27 of 33

Fig. 24 The results of Karo’s performance in RRL’s test suits

are designated to fully autonomous robots. Since Karo’s the RRL’s mean performances, but also, we demonstrate
autonomous exploration function is not adequate for that all required capabilities have been embedded in Karo
those two tests, its exploration skills have been tested adequately with respect to the best RRL results.
only for Map on Continuous Ramps, Map on Cross- Figure 25 illustrate the comparison from the first
ing Ramps, and Recognize Objects tests, where Fig. 24d point of view, in which Karo have done far superior to
depicts the results. Considering the results, Karo’s per- the mean performance of all robots in RRL. The only
formances on the first two exploration tests are highly exception is in the exploration test suits where Karo has
identical. To elucidate, provided terrains in both tests are not participated in two tests since they are designated
sort of unproblematic for Karo’s mobility skills and as a to fully autonomous operations. However, the gap is
result, practically speaking, there is no noticeable differ- sizable in the rest of the exploration tests. In addition,
ence in these two tests. the RRL’s mean performance is closer to Karo’s perfor-
mance in the Maneuvering test suite than in any other
Comprehensiveness of Karo’s Capabilities test suite. That is mainly because the Maneuvering
In this section we aim to compare the performance of test suite is the most rudimentary evaluation of rescue
Karo with other robots participating in RRL. We will robots competing in RRL and all robots must take all
draw the comparison from two point of views: 1) compar- Maneuvering tests to demonstrate their basic capabili-
ing Karo with the average performance of participating ties. As a matter of fact, the results in Fig. 25 have not
teams in RRL, and 2) comparing Karo’s comprehensive been normalized and represent the raw score collected
performance with robots demonstrating the best perfor- by robots.
mance in each test suite. The former provides insights In Fig. 26, a comparison is drawn from the second
about Karo’s performance in each individual test and point of view, as explained above. In this compari-
compares its capabilities with the average results of all son Karo has been compared with three other res-
other teams participated in RRL based on the same test cue robots, each winner of one class such as Mobility,
frameworks. The latter examines the comprehensive- Dexterity and Exploration. In this comparison, we are
ness of Karo’s capabilities with respect to the best perfor- considering normalized scores of all five tests in a test
mances in RLL in each test suite. Altogether, not only we suite in order to investigate the robots’ comprehen-
ensure that Karo is functioning significantly superior to sive capabilities. In all three comparisons, Karo has
Habibian et al. Robomech J (2021) 8:1 Page 28 of 33

Fig. 25 Drawing a comparison between Karo’s performance and the RRL’s mean performance in all four test suits (based on the results from RRL
2016)

done comprehensively more satisfying performance training field of a fire department as its field performance.
compared to its opponents. For instance, concerning Generally speaking, standardized test methods including
Fig. 26c, although the winner of the Exploration has RRL put forward structured and reproducible test suites
accomplished a cut above Karo in one test suite (i.e., that cannot reflect the haphazard pattern of obstacles
exploration), in all other three test suites Karo has and corridors in the real world’s disaster sites. Besides,
demonstrated higher quality operations. Numerically the design of participating robots can be biased towards
speaking, the same discussion can be drawn for Fig. 26 the RRL’s test suites. As a consequence, the performance
a, b. of those robots remains uninvestigated in unprecedented
and less frequent scenarios in the real world’s problems
Field performance that has not been modeled by any of those standard test
Motivated to bridge the evaluation gaps of RRL’s frame- suites. By the same token, the performance of a robot’s
work inspired by NIST standard test methods, this sec- operator can be biased to the foreseeable structure of
tion discusses the test and evaluation of Karo at the the RRL’s test suites, while the operator has to make
Habibian et al. Robomech J (2021) 8:1 Page 29 of 33

Fig. 26 Comparing Karo’s overall performance with the RRL winners of Mobility, Dexterity, and Exploration test suites (based on the results from
RRL 2016).

unplanned and spontaneous decisions in unpredicted In mobility tests, each test item is considered completed/
scenarios in more realistic situations. Considering these passed when the robot successfully fulfills two repetitions
evaluation gaps, the field performance tests have been over the test terrain. Having said that, the robot effectively
conducted to validate Karo’s performance in mobility, performed multiple mobility tests such as maneuvering on
inspection, endurance, and communication in a more scaffolds, climbing up and down stairs with debris, travers-
generic and unrehearsed manner. These tests challeng- ing ramps and overcoming unstructured obstacles. Among
ing the robot’s practical capabilities help the developers all these mobility tests, stair with debris is the most chal-
to (1) challenge the robot’s capabilities beyond the scope lenging one which questions the robot’s mobility and sta-
of RRL, and (2) gain insights into the future designs and bility and the operator’s adeptness instantaneously. As
developments. Although the robot failed in three valida- Fig. 27 demonstrates, the robot successfully completed
tion tests as illustrated in Table 4, its promising compre- multiple trails on the stairs with debris where the width
hensive performance adequately satisfied the mission’s and slope of the stairs are almost similar to the RRL’s test
objectives. Here, we present a detailed elaboration of suite. However, the non-wooden material of the structure
each test item conducted in 30-min time slots. and haphazard distribution of debris along the way throws
up relatively more serious mobility challenges compared to

Table 4 Validation tests taken by the robot in the training field

Validation test Short description Passed/failed

Mobility Maneuvering over a flat surface structured by scaffolds Passed

Climbing up and down the stairs with harsh debris Failed
Climbing up and down the stairs with moderate debris Passed
Traversing ramps (< 45°) Passed
Overcoming unstructured obstacles Passed
Inspection Inspecting inside targeted cars Passed
Inspecting areas below the ground level Passed
Performing inspection in dark corridors Passed
Performing inspection in smoky corridors Failed
Communication Communicating in line-of-sight situation (> 800 m) Passed
Communicating in non-line-of-sight situation (> 100 m) Passed
Communicating in building through different stories Failed
Endurance Operating continuously (> 1 h) Passed
Habibian et al. Robomech J (2021) 8:1 Page 30 of 33

the RRL’s stair test suite. According to the results, the stair inspection, dark corridor exploration, smoky corridor
test with harsh debris aborted due to crossing the motor exploration, and inception of areas below the ground
drivers’ current limit. As a final note on the robot’s mobil- level. An inspection test is considered completed when
ity, its performance on different test items designed origi- the robot approaches the determined target closely
nally for human first responders was satisfactory, though and then transmits acute and perspicacious informa-
the robot’s depreciation was significantly higher compared tion of the target. In this regard, the robot successfully
to performing on the RRL’s test suites. inspected the space inside the car from distance enjoying
Next, we put the inspection capabilities of the robot the manipulator’s ample workspace (see Fig. 11), as illus-
to the test throughout four practical tests namely car trated in Fig. 28.

Fig. 27 The robot is completing repetitions on stairs with debris

Fig. 28 The robot is performing a car inspection

Fig. 29 The inspection of a dark corridor by the robot

Habibian et al. Robomech J (2021) 8:1 Page 31 of 33

Regarding the corridor exploration test, the robot Conclusion

effectually explored the dark corridor utilizing its night- This paper represents the methodology to design and
vision equipment, as shown in Fig. 29. However, it failed implement a maxi-sized rescue robot (Karo) that is capa-
to maneuver in the smoky corridor because of the high ble of carrying out the desired capabilities of response
density of the smoke which made the corridor almost robots in real-world missions: mobility, dexterity, and
unobservable for the robot. Relying on the laser scan- exploration. Our approach is based on RRL’s evaluation
ner’s multiecho function, we tried to utilize the online benchmark, obtained from the DHS-NIST-ASTM inter-
map obtained by the robot’s laser scanner to navigate the national standard test methods, to determine the system
robot through the corridor which failed similarly. requirements. We comparatively analyze the conceptual
Subsequently, the inspection test displayed in Fig. 30 design of our previous robots and therefore introduce an
exemplifies scenarios in which the manipulator must improved platform for Karo. We employ a streamlined
have sufficient reachability to inspect the areas beneath approach to implement Karo’s conceptual design that
its platform. Karo demonstrated its dexterity capabili- enables reliable operation during USAR missions. The
ties by accomplishing the test successfully. This capabil- experimental results and test evaluations of our work
ity of Karo’s manipulator has been completely analyzed in confirm that incorporating the knowledge about the
"Achievable workspace" section. real-world necessities into the design and implementa-
During all mobility and inspection tests, the quality tion process enables tailoring a comprehensive response
of wireless communication between the robot and the robot. We also highlighted that the important difficulties
OCU has been constantly monitored to corroborate its associated with real-life missions are not fully addressed
functionality. The communication maintained reliable in in the standard tests, but they are significantly beneficial
line-of-sight situation up to 900 m. In the case of non- for response robot developers to begin with. We envision
line-of-sight situation, the communication maintained that the presented approach lays a foundation method for
robust for a 120 m-distance. However, the reliability of the developers to strengthen the future generation of rescue
communication was not ensured when robot was operat- robots.
ing on the third floor of the building and the OCU was To identify Karo’s mechanical characteristics, we first
on the first floor. Although the OCU and the robot were developed a conceptual design for the locomotion system
still connected, the delay in the control commands had (tracked platform with four flippers) conjunct with a suit-
increased the risk of fault and failure of the operation. able joint arrangement (7-DOF) for the manipulator. Due
A screen capture video of Karo’s field performance to the unpredictable circumstances of rescue missions,
can be found as a supplementary material along we identified the operation range for motors accord-
with this paper, by using the YouTube link “[Link]/ ingly and considered the extreme payload condition as a
baseline for the continuous operation of our system. Our
approach is verified experimentally, through comprehen-
sive evaluation tests in RRL. Here, we employed a diverse
set of sensors and perception devices for vision, audio,
localization [51], and mapping, to control the robot from
a remote operator control unit. Furthermore, we imple-
mented a ROS-based software system for the GUI to
enable the operator to monitor live sensory information
in an environment besides providing real-time control of
UnbqkQWR7e4,” or by scanning the following QR code.
the robot (Fig. 16).

Fig. 30 The robot is inspecting areas beneath a scaffolded platform

Habibian et al. Robomech J (2021) 8:1 Page 32 of 33

In this paper, through various standard test evaluations, Acknowledgement

The authors would like to acknowledge the work of the team members at
we demonstrated that Karo’s overall performance is not AMRL who offered their enthusiasm and collaborative support to the Karo
only improved substantially compared to its counterparts project [54].
but also is superior to the average performance of RRL’s
Authors’ contributions
other participants. Furthermore, RRL evaluations verify SH, MD, BP, AH, MHS, and AHMH conducted research and experiments as
Karo’s extensive capacity in all test suits, however, the active members’ of AMRL. SH and MD also analyzed the data, developed the
top participants in RRL only performed well in some of key studies, and wrote the manuscript. FN advised the research and experi-
ments of this project. All authors read and approved the final manuscript.
the test suits. This highlights the importance of using the
benchmark since a real rescue mission is indeed a com- Funding
bination of mobility, dexterity, and exploration. Lacking This research was funded by Qazvin Azad University via internal grants to
assist the project at the Advanced Mobile Robotics Lab (AMRL) [54].
any of those capabilities could affect robots’ performance
drastically during missions. For instance, a robot with Availability of data and materials
inadequate dexterity cannot complete a mission that The data and source code used to support the findings of this study are avail-
able from the corresponding author upon request.
requires object manipulation such as opening a door.
Therefore, various capabilities should be considered in Competing interests
the development of a response robot. The authors declare that they have no competing interests.
As a final demonstration, we validated our approach Author details
more practically through real-world tests at the training 1
The Authors Were With Advanced Mobile Robotics Lab, Qazvin Azad
field of a fire department. The results of these valida- University, Qazvin, Iran. 2 Department of Electrical and Computer Engineering,
Qazvin Azad University, Qazvin, Iran. 3 University of Tehran, College of Engi-
tion tests closely followed Karo’s performance in RRL, neering, School of Mechanical Engineering, Tehran, Iran.
where about 77% of all 13 tests successfully met the
mission objectives. However, we observed a few short- Received: 30 July 2020 Accepted: 15 December 2020
comings in our system that captured useful informa-
tion for future improvements: (1) limited resiliency to
carry various devices that could be highly pragmatic in
a field operation, (2) insufficient robustness in climbing References
1. Murphy RR (2014) Disaster Robotics. The MIT Press, New York
stairs with harsh debris, and (3) inadequate non-line- 2. Kitano H, Tadokoro S, Noda I (1999) RoboCup Rescue: search and rescue
of-sight communication, which diminished the overall in large-scale disasters as a domain for autonomous agents research. IEEE
performance. SMC ’99 Conference Proceedings. IEEE, Tokyo, Japan, pp 739–743
3. Tsitsimpelis I et al (2019) A review of ground-based robotic systems for
Future work will aim to address the deployment and the characterization of nuclear environments. Progr Nuclear Energy 111:1
maintenance necessities and apply them thoroughly 4. Reddy AH, Kalyan B, Murthy CSN (2015) Mine rescue robot system – a
into the design process. The improvement of the review. Procedia Earth Planet Sci 11:457–462
5. Murphy RR et al (2008) Search and Rescue Robotics. In: Siciliano B, Khatib
manipulation system will continue to increase payload O (eds) Springer Handbook of Robotics. Springer, Berlin, Heidelberg, pp
and object handling. Another direction will be focused 1151–1173
on implementing the object recognition feature to clas- 6. Tadokoro S (1997) On robotic rescue facilities for disastrous earthquakes-
from the Great Hanshin-Awaji (Kobe) Earthquake. J Robotics Mechatr
sify the perceived information for the operator. Future 9(1):46–56
studies will also be focused on enhancing autono- 7. Iwamoto T, Yamamoto H (1990) Mechanical design of variable configura-
mous exploration by using the IMU orientation data tion tracked vehicle. J Mech Des 112(3):289–294
8. Hirose S, Sensu T. The TAQT Carrier: A Practical Terrain Adaptive Quadru-
to develop an advanced scheme to control the flippers. track Carrier Robot, in Proceedings of the IEEE/RSJ International Conference
To that end, the control scheme will be trained with a on Intelligent Robots and Systems. 1992, IEEE: Raleigh, NC, USA.
training data set obtained from an expert human–robot 9. McAndrew DR, Mangolds A, Deguire DR, Tactical mobile robotic search
and rescue mission modules. in Tactical mobile robotic search and rescue
operator by utilizing supervised machine learning tech- mission modules. (2000) Orlando. SPIE, FL, United States
niques, which effectively facilitates the robot’s naviga- 10. Moazami S, et al. (2020) Design, Modeling, and Control of Norma: a Slider &
tion in simple repeatable missions [52, 53]. Pendulum-Driven Spherical Robot. https​://arxiv​.org/abs/1908.02243​
11. Moazami S, Zargarzadeh H, Palanki S (2019) Kinematics of Spherical Robots
Rolling over 3D Terrains. Complexity, 2019
12. Roennau A, et al (2014) LAURON V: A versatile six-legged walking robot with
Abbreviations advanced maneuverability, In: International Conference on Advanced Intel-
USAR: Urban Search and Rescue; RRL: Rescue Robot League; DOF: Degrees ligent Mechatronics. IEEE/ASME: Besacon, France.
of Freedom; DHS S&T: U.S. Department of Homeland Security, Science and 13. Chiu Y, et al (2005) FUMA: environment information gathering wheeled
Technology Directorate; NIST: National Institute of Standard and Technology; rescue robot with one-DOF arm, in IEEE International Safety, Security and
HRI: Human–Robot Interface; AMRL: Advanced Mobile Robotics Lab; ASTM: Rescue Rototics. IEEE: Kobe, Japan
American Society for Testing and Materials; COM: Center of Mass; FBD: Free 14. Zhang H et al (2006) A novel reconfigurable robot for urban search and
Body Diagram; OCU: Operator Control Unit; SLAM: Simultaneous localization rescue. J Adv Robot Syst 3:4
and mapping; GUI: Graphical User Interface; CAN: Controller Area Network;
ROS: Robot Operating System.
Habibian et al. Robomech J (2021) 8:1 Page 33 of 33

15. Nie C, Corcho XP, Spenko M (2013) Robots on the Move: Versatility and 37. Shigley JE, Mischke CR (1989) Mechanical engineering design, 5th edn.
Complexity in Mobile Robot Locomotion, in Robotics & Automation Maga- McGraw-Hill, New York
zine. IEEE. p. 72 - 82 38. Jirandehi AP, Chakherlou T (2019) A fatigue crack initiation and growth
16. Rea P, Ottaviano E (2018) Design and development of an Inspection life estimation method in single-bolted connections. J Strain Analy Eng
Robotic System for indoor applications. Robot Comput Integrat Manuf Design 54(2):79–94
49:143–151 39. Jirandehi AP, Mehdizadeh M, Khonsari M (2020) Temperature-induced
17. Luo Z et al (2018) A reconfigurable hybrid wheel-track mobile robot buckling of ductile metals during cyclic loading and the subsequent
based on Watt II six-bar linkage. Mech Mach Theory 128:16–32 early fracture. Int J Mech Sci 176:105525
18. Nagatani K et al (2013) Emergency response to the nuclear accident at 40. Etemad M, et al. (2020) Using Deep Reinforcement Learning Methods for
the fukushima daiichi nuclear power plants using mobile rescue robots. J Autonomous Vessels in 2D Environments. In: Canadian Conference on Artifi-
Field Robot 30(1):44–63 cial Intelligence. 2020. Springer
19. Zang X et al (2016) Two multi-linked rescue robots: design, construction 41. Quigley M, et al (2009) ROS: an open-source Robot Operating System. in
and field tests. J Adv Mech Des Syst Manuf 10:6 ICRA Open-Source Softw. Workshop
20. Kamegawa T et al (2011) Design and implementation of grouped rescue 42. Kam HR et al (2015) RViz: a toolkit for real domain data visualization.
robot system using self-deploy networks. J Field Robot 28(6):977–988 Telecommun Syst 60(2):337–345
21. Enayati N, Najafi F (2011) Design and manufacturing of a tele-oper- 43. Thrun S (2001) Learning occupancy grids with forward models, in Proceed-
ative rescue robot with a novel track arrangement. Industrial Robot ings 2001 IEEE/RSJ International Conference on Intelligent Robots and
38(5):476–485 Systems. Expanding the Societal Role of Robotics in the the Next Millennium
22. Ma J, et al (2014) Design, simulation and manufacturing of a tracked robot (Cat. No.01CH37180). 2001, IEEE: Maui, HI, USA. p. 1676–1681
for nuclear accidents, in IEEE International Conference on Robotics and 44. Durrant-Whyte H, Bailey T (2006) Simultaneous localization and mapping:
Biomimetics. IEEE: Bali, Indonesia part I, in IEEE Robotics & Automation Magazine. 2006, IEEE. p. 99-110
23. Ben-Tzvi P, Goldenberg AA, Zu JW (2008) Design, Simulations and Optimi- 45. Grisetti G, Stachniss C, Burgard W (2007) Improved techniques for
zation of a Tracked Mobile Robot Manipulator with Hybrid Locomotion and grid mapping with rao-blackwellized particle filters. IEEE Trans Rob
Manipulation Capabilities, in IEEE International Conference on Robotics and 23(1):34–46
Automation. IEEE: Pasadena, CA, USA 46. Kohlbrecher S et al (2011) A flexible and scalable SLAM system with full
24. Guzman R et al (2016) RESCUER: development of a modular chemical, 3D motion estimation. 2011 IEEE International Symposium on Safety,
biological, radiological, and nuclear robot for intervention, sampling, and Security, and Rescue Robotics, Kyoto. IEEE, Kyoto, pp 155–160
situation awareness. J Field Robot 33(7):931–945 47. Haeri H, Jerath K, Leachman J (2020) Thermodynamics-inspired macro-
25. Ducros C et al (2016) RICA: a tracked robot for sampling and radiological scopic states of bounded swarms. ASME Lett Dyn Syst Control 1:1
characterization in the nuclear field. J Field Robot 34:3 48. Haeri H, Jerath K, Leachman J (2019) Thermodynamics-Inspired Modeling
26. Nagatani K, et al (2011) Redesign of rescue mobile robot Quince, in IEEE of Macroscopic Swarm States. In: ASME 2019 Dynamic Systems and Control
International Symposium on Safety, Security, and Rescue Robotics. IEEE: Conference
Kyoto, Japan. 49. Amirian S, et al (2019) A short review on image caption generation with
27. Liu Y, et al (2016) The design of a fully autonomous robot system for urban deep learning. in Proceedings of the International Conference on Image Pro-
search and rescue, in IEEE International Conference on Information and cessing, Computer Vision, and Pattern Recognition (IPCV). 2019. The Steering
Automation (ICIA). IEEE: Ningbo, China Committee of The World Congress in Computer Science, Computer ….
28. RoboCup Rescue Rulebook. 2019, RoboCup: retrieved from: https​://rrl. 50. Holz D et al (2010) Evaluating the Efficiency of Frontier-based Explora-
roboc​[Link]/rules​/ tion Strategies. ISR 2010 (41st International Symposium on Robotics)
29. Kitano H, Tadokoro S (2001) RoboCup Rescue: A Grand Challenge for Multia- and ROBOTIK 2010 (6th German Conference on Robotics). IEEE, Munich,
gent and Intelligent Systems, in AI Magazine. AAAI Germany, pp 1–8
30. Jacoff A, et al (2012) Guide for Evaluating, Purchasing, and Training with 51. Asali E, et al (2016) Shiraz soccer 2D simulation team description paper 2016.
Response Robots Using DHS-NIST-ASTM International Standard Test Methods, In: RoboCup 2016. 2016. Leipzig, Germany
in Standard Test Methods For Response Robots. 2012, NIST: retrieved from 52. Majd K, Razeghi-Jahromi M, Homaifar A (2019) A stable analytical solution
[Link] method for car-like robot trajectory tracking and optimization. IEEE/CAA J
31. Deemyad T, Hassanzadeh N, Perez-Gracia A (2018) Coupling Mechanisms Automatica Sinica 7(1):39–47
for Multi-Fingered Robotic Hands with Skew Axes. In: IFToMM Symposium on 53. Majd K, Razeghi-Jahromi M, Homaifar A (2018) Optimal Kinematic-based
Mechanism Design for Robotics. Springer Trajectory Planning and Tracking Control of Autonomous Ground Vehicle
32. Azizi V, et al (2017) Geometric reachability analysis for grasp planning in Using the Variational Approach. In: 2018 IEEE Intelligent Vehicles Symposium
cluttered scenes for varying end-effectors. In: 2017 13th IEEE Conference on (IV). 2018. IEEE
Automation Science and Engineering (CASE). 2017. Xi’an, China: IEEE 54. Najafi F, et al (2016) RoboCup Rescue 2016 Team Description Paper MRL. In:
33. Maxon High Precision Drives and Systems. 2019/2020: Switzerland Robocup Rescue 2016 TDP Collection. Leipzig, Germany
34. Bhatia A (2012) Basic Fundamentals of Gear drives. In: PDHonline Course
M229. [Link]
35. De Almeida A, Greenberg S (1995) Technology assessment: energy- Publisher’s Note
efficient belt transmissions. Energy Buildings 22(3):245–253 Springer Nature remains neutral with regard to jurisdictional claims in pub-
36. Deemyad T, Heidari O, Perez-Gracia A (2020) Singularity Design for RRSS lished maps and institutional affiliations.
Mechanisms. In: USCToMM Symposium on Mechanical Systems and Robot-
ics. Springer
 International Journal of Applied Mathematics, Electronics and Computers 8(4): 133-137, 2020

INTERNATIONAL JOURNAL OF APPLIED MATHEMATICS International

Open Access
ELECTRONICS AND COMPUTERS
Volume 08
Issue 04
[Link]/ijamec
e-ISSN: 2147-8228 December, 2020

Research Article

Design and Development of a Mobile Robot for Search and Rescue Operations
in Debris
Yakupcan Baldemira,* , Soner İyigüna , Oruj Musayeva , Cenk Ulua
aDepartment of Mechatronics Engineering, Yıldız Technical University, 34353 Besiktas, Istanbul, Turkey

ARTICLE INFO ABSTRACT

Article history: After an earthquake, critical tasks during the search and rescue operations are to quickly detect
Received 27 September 2020 the locations of victims and recover them. Additionally, effective communication between
Accepted 11 October 2020 responders in the field and the emergency management center is also important to provide
Keywords: successful operations. All these tasks are mostly performed by humans and trained dogs in
Communication Unit
dangerous and risky situations. In this study, the design and development of a mobile search and
Debris
Remote Control rescue robot are presented. The design criteria are determined based on operational requirements
Robot and stakeholder expectations. Considering the specified design criteria, the mechanical and
Search and Rescue electrical designs of the mobile robot are performed and corresponding control, communication,
and user interface software are developed. Designed parts of the robot are manufactured and the
system integration is done. Then robot functionality tests are performed. The test results show
that the performance of the developed search and rescue robot meets the desired performance
criteria. The developed search and rescue robot has great potential to provide an efficient way to
reduce losses of lives by decreasing operation time and for getting accurate information about
the situation of the disaster area.

This is an open access article under the CC BY-SA 4.0 license.

([Link]

1. Introduction from an inaccessible surrounding.

Today, with the improvement of sensor technology and
Today, mobile robots are widely used in many fields. a rising in the speed and capacity of microcontrollers,
One of the important application areas of mobile robots is robots that are small-scale enough to pass through rubble
the search and rescue operations. These robots continue piles can be evolved [2]. There are various research fields
to replace traditional search and rescue equipment in for search and rescue robots such as mapping, victim
various applications, especially in the last 20 years. detection [3], navigation, and exploration [4].
In history, unfortunate disasters comparable to the In literature, there are different types of search and
Mexico City earthquake in 1985, the capital of Oklahoma rescue robots. One of these has a double-track wheel
bombing in 1995, and therefore the World Trade Center system, RC remote control, and various sensors [5]. The
attacks on September 11, 2001 show that robots are other, known as SALVOR, has six wheels locomotion
required to help search and rescue teams in insecure and system and can operate in autonomous and semi-
out of reach places. 135 rescuers died within the Mexico autonomous modes [6]. The operation of SALVOR relies
City earthquake. 65 of these died as they struggled to on the data from its sensors about the environment and
save rescue workers trapped in a narrow space [1]. The directions from the human operators. Another robotic
first real test and application of search and rescue robots system is aimed to widen a narrow gap and open a new
are performed after the terrorist attack on the World space to unleash buried survivors from pressure as
Trade Center (WTC). In WTC, robots are used to explore temporary supports [7]. Also, the robot can create areas
these areas by transferring real-time display to saviors for operating and moving courses for other rescue robots

* Corresponding author. E-mail address: yakupcanbal@[Link]

DOI: 10.18100/ijamec.800840
 Yakupcan Baldemir et al., International Journal of Applied Mathematics Electronics and Computers 08(04): 133-137, 2020

and tools. This robot consists of a spreader tool and a Table 2. Functional requirements
mobile unit. The spreader tool driven by a hydraulic Function Requirement Covering the
actuator can spread the gap from 3 centimeters to 20 Requirement
centimeters with 1000kg payload. Function 1 Ability to operate in the dark Lighting

In this study, the development of a mobile robot for Function 2 Having a short turning Four-wheel
search and rescue operations in debris is presented. This radius drive system
robot provides direct intervention to the wreckage region Function 3 Detection of unconscious Thermal camera
victims
where the working of people and dogs is dangerous. In Function 4 Ability to communicate Microphone,
this way, the search operation time can be decreased. The with disaster victim speaker
robot is designed, manufactured, and integrated. Then, Function 5 Gas leak detection Gas sensor
Function 6 Image transfer, recording and User interface,
the functional tests are performed successfully and it is monitoring of system data camera
shown that the desired design criteria are met. The mobile
robot has a four-wheel drive system. The camera is
integrated into the system for navigation purpose and to 2.2. System Architecture
facilitate human detection in debris. Additionally, it is
equipped with temperature and gas sensors to measure The system architecture of the robot is determined as
temperature and detect gas leaks in the disaster area. The shown in Fig.1 considering the design and functional
requirements.
designed robot is remotely operated and able to move in
the narrow gaps. All sensor data and video stream can be
monitored via the user interface. Furthermore, the robot
provides voice communication between search and rescue
operator and victims. It can also be used to deliver water
and basic foodstuffs to the casualty when necessary.
Today, preparation and operations take a long time in
current search and rescue efforts. Special users and
training are required for the use of existing robots on the
market. Communication with the victim is not possible
with existing robots. Unlike the existing robots, our study
has the opportunity to communicate with the casualty.
Besides, the robot is able to continue its movement if it
encounters an obstacle while moving and turns over. The
robot has a thermal camera to locate the victims under the
Figure 1. System Architecture
wreckage and direct the search and rescue teams there
more quickly. The search and rescue operator can control the robot
with the RC controller unit and monitor all system data
2. System Design and the condition of the debris area via the user interface
of the robot. Wireless communication is preferred
2.1. Design Requirements
between the operator panel and the robot [8]. The
The design criteria are specified based on operational operator has the opportunity to talk to the victims in the
requirements and stakeholders' expectations such as wreckage region through the loudspeaker and
AFAD (Ministry of Interior Disaster and Emergency microphone on the robot. In addition to these, thanks to
Management Presidency). The determined design sensors on the robot, operational data such as gas density
requirements and functional requirements are shown in and temperature values in the debris region can be
Table 1 and Table 2, respectively. monitored on the user interface [9].
Table 1. Design requirements 2.3. Mechanical Design
Requirement Value
The system mechanical design is shown in Fig. 2. To
Maximum Weight 3 kg
Maximum 300x300x120 mm increase maneuverability, the robot uses a four-wheel
Dimensions drive system. DC motor is preferred as a drive motor. The
Minimum wireless 50 m robot chassis is positioned so as not to exceed the height
communication distance
Voltage level 12 VDC of the wheel. In this way, the robot can continue its
Minimum operation time 30 dk operation even when it turns upside down [10]. This also
contributes to the protection of the chassis from
collisions. ABS material is chosen as the chassis material.

- 134 -
 Yakupcan Baldemir et al., International Journal of Applied Mathematics Electronics and Computers 08(04): 133-137, 2020

The placement of the equipment in the robot body is orientation, heading angle, and 3 axis acceleration values
shown in Fig. 3. of the robot. The MQ-9 gas sensor board is used for the
measurement of carbon monoxide and combustible gases,
and the DHT11 sensor module for measuring the
temperature and humidity. In addition, D40D50R speaker
and MAX4466 Electret Microphone Module are
preferred in order to enable the operator to communicate
with the survivors in the wreckage area. The power
requirement of all subsystems of the robot is calculated
and 11.1V 3S 6000 mAh Li-po battery is used as the
power source.
2.5. User Interface Design
The user interface of the robot is shown in Fig. 4. The
user interface enables the operator to use the robot
effectively. Visual C# programming language is preferred
for the development of the user interface. In the
Figure 2. System Mechanical Design developed interface, the operator can switch between the
thermal camera and normal camera. By using this
interface, the lights on the robot can be controlled. The
operator is warned in case of any gas leakage hazard. It is
also possible to communicate with victims via the user
interface. In addition, all system data such as
temperature, gas density, and robot orientation data can
be monitored on the developed interface.

Figure 3. Internal placement of equipment

2.4. Electrical Design

Considering the robot's weight, acceleration value,
nominal speed, and sloping areas, a 12V 210 rpm DC
motor with gearbox is determined as the robot wheel
motor. A motor driver suitable for this motor is used.
Raspberry Pi 4 2GB - Model B is chosen as the main Figure 4. User Interface
processor of the robot. Transfer of images from the
cameras to the operator instantly is provided through this 3. System Integration and Functional Tests
processor. In addition, general data communication
between the robot and the operator computer is provided After completing the system design, the mechanical
through this card. Arduino Mega 2560 R3 is used for parts of the robot are produced. The chassis of the
controlling DC motors and collecting sensor data. robot is produced in 3D printer using ABS material.
Raspberry Pi Camera Module is preferred to enable the The produced chassis and the robot whose integration
robot to provide live video stream from the debris area. In is completed are shown in Fig. 5 and Fig. 6,
addition to this camera, the Adafruit AMG8833 IR respectively.
Thermal Camera card has been used to enable the robot
to detect victims in areas of debris that the operator
cannot recognize. USB Dongle is preferred in the robot’s
communication system to increase the communication
distance. 10DOF LSM303D sensor is used to measure the

- 135 -
 Yakupcan Baldemir et al., International Journal of Applied Mathematics Electronics and Computers 08(04): 133-137, 2020

Figure 5. Produced Robot Chassis

Figure 8. Sample Image from Functional Tests (2)

Table 3. Functional test results

Functional test Test result
Communication test Communication at a distance of 400m in the
open area is provided uninterruptedly.
In the building, communication is provided from
a distance of at least 30m.

Motion test The robot is successfully performed the given

motion commands.
The robot continues its movement when it turns
upside down.
User interface test The acquisition of images from the camera and
the transfer of the data received by the sensors to
Figure 6. Search and Rescue Robot the interface are successfully completed.

After the system integration is completed, the Min. operation 30 min.

time
functional tests of the robot are performed. Applied
functional tests and obtained results are given in Table 3.
4. Conclusions
As it is seen from the test results, the developed search
and rescue robot meets all design criteria. Sample images In this study, the design and development of a
from the functional tests are given in Fig. 7 and Fig. 8. mobile search and rescue robot are presented. The
robot searches the victims in the wreckage area after
the earthquake and also transfers the collected data
about the wreckage area to the rescue team in real-
time. The system has been designed, produced, and
integrated. The functional test results show that the
developed robot meets design criteria. Depending on
its features, the developed robot can effectively be
used in search and rescue operations. By using this
robot, the search operation time can be reduced.
Unlike the other existing search and rescue robots, the
developed robot provides the opportunity to communicate
Figure 7. Sample Image from Functional Tests (1)
with the victims. Additionally, if the robot encounters an
obstacle while moving and turns over, it is able to
continue its movement. Furthermore, thanks to its
thermal camera, victims in the wreckage can be located
and search and rescue teams can be directed to the
location faster by using this information.
In future studies, it is planned to add the features of
autonomous control and casualty detection by image
processing to the system in order to increase the

- 136 -
 Yakupcan Baldemir et al., International Journal of Applied Mathematics Electronics and Computers 08(04): 133-137, 2020

efficiency of the robot.

Author's Note
Abstract version of this paper was presented at 9th
International Conference on Advanced Technologies
(ICAT'20), 10-12 August 2020, Istanbul, Turkey with the
title of “Design and Development of a Mobile Robot for
Search and Rescue Operations in Debris”.

Acknowledgment
This work is supported by TÜBİTAK (Project No:
1139B411900495), Turkey.

References
[1] I. Casper, M. Micire and R Murphy, Issues in Intelligent
Robots for such and Rescue, SPIE Ground Vehicle
Technology II, 4: 41- 46, 2000.
[2] A. Davids, “Urban search and rescue robots: from tragedy to
technology,” Intelligent Systems, IEEE [see also IEEE
Intelligent Systems and Their Applications], vol. 17, no. 2, pp.
81–83, 2002.
[3] S. Bahadori, L. Iocchi, D. Nardi, and G. Settembre, “Stereo
vision based human body detection from a localized mobile
robot,” in Proceedings. IEEE Conference on Advanced Video
and Signal Based Surveillance, 2005., 2005, pp. 499–504
[4] D. Calisi, A. Farinelli, L. Iocchi, and D. Nardi, “Autonomous
navigation and exploration in a rescue environment,” in
Safety, Security and Rescue Robotics, Workshop, 2005 IEEE
International, 2005, pp. 54–59.
[5] Tunwannarux, A., ve S. Hirunyaphisutthikul. “Design features
and characteristics of a rescue robot”, 1083-87, 2005.
[Link]
[6] A. Denker, Baykal Sarıoǧlu, Yeşim Öniz, Murat Tümer, Emre
Can Akgül, Hilal Merve Altuğ, Kenan Kural, “DESIGN AND
IMPLEMENTATION OF A MOBILE SEARCH AND
RESCUE ROBOT ”, 2017.
[7] Zhang, Guowei, Li Bin, Zhiqiang Li, Wang Cong, Handuo
Zhang, Shang Hong, Hu Weijian, ve Zhang Tao.
“Development of Robotic Spreader for Earthquake
Rescue*”, 2014. [Link]
[8] [Link] and Henry Y. K. Lau, Robot Assisted Emergency
Search and Rescue System With a Wireless Sensor Network,
International Journal of Advanced Science and Technology,
Vol. 3, Pages 1-3, February, 2009.
[9] M. H. Zulkefli, K. A. Mohd Annuar, S. H. Johari, M. R.
Mohamad Sapiee and S. Ahmad, “Graphical User Interface
(GUI) Controlled Mobile Robot” in Journal of Advanced
Research in Computing and Applications, Vol. 1, No. 1, pp.
42-49, 2015.
[10] Siciliano, Bruno, and Oussama Khatib, ed. Springer
Handbook of Robotics. Cham: Springer International
Publishing, 2016. [Link]

- 137 -
 ISSN (Print) : 2320 – 3765
ISSN (Online): 2278 – 8875

International Journal of Advanced Research in Electrical,

Electronics and Instrumentation Engineering
( An I SO 3 2 9 7 : 2 0 0 7 Ce rt ifie d Org a n iz a t ion )

Vol. 3 , I ssue 6 , Ju n e 2 0 1 4

Mobile Rescue Robot for Human Body

Detection in Rescue Operation of Disaster
Trupti B. Bhondve1, Prof.R.Satyanarayan2 , Prof. Moresh Mukhedkar3
1
PG student[VLSI], Dept. of E&TC , [Link] College of Engineering, University of Pune, Ambi, Pune, India
2
Assistant Professor, Dept. of E&TC , [Link] College of Engineering, University of Pune, Ambi, Pune, India
3
Assistant Professor, Dept. of E&TC , [Link] College of Engineering, University of Pune, Ambi, Pune, India

ABSTRACT: Many areas of world are getting affected due to natural calamity. Disasters are exceptional &
unstoppable events that are either man made or natural, such as terrorist attacks, earthquakes, wildfires and floods etc.
[1]. Disasters create emergency situations to provide basic services to the victims must be coordinated quickly. Many
times we observe that many people dies by trapping in these disasters but the people also dies on large scale just
because they didn’t get help at instant time or the help provided to them is late. This paper propose an mobile robot
based on Wireless Sensor Network (WSN) which is designed for human existence & detection in an unmanned area
can be done only by an automated system. This system proposed a monitoring system using sensors unit and camera
module to recording, analyze conditions of human body and transmit data [3]. Mobile robots perform cooperative
Simultaneous human body localization function and communicate over the WSN. The main objective of this Paper is to
rescue more & more number of people from the adverse condition.

KEYWORDS: body detection, Natural calamity, Mobile rescue robot, Wireless sensor network, Disaster.

[Link]
This article design a mobile rescue robot system based on wireless sensor network technology to help the people on
time which are trapped in natural calamity like disaster, earthquake, floods etc. It gives timely & accurately reflect
dynamic situation of human in disaster region like in the underground regions to control room, so that rescue team of
Experts & doctors can be sending to the victim’s location for primary treatment and can be sent to the safe place or
hospital. Wireless sensor network can solve the key issue of communication bandwidth, data transmission, real-time
detection & so on. Wireless sensor network is composed of a large number of micro-sensors nodes which have small
volume, low cost, good compatibility & battery powered. At the same time due to the special nature of the wireless
network is that it can spread the wireless signal, we can easily locate human. By sensors like humidity, temperature,
visibility we will get all the weather reports
Existing System: Natural calamities do occur and they are exceptional and unstoppable event which disturb both
economic & social balance of society. Now awareness about concept of intelligent rescue operations in such calamities
are becoming increasingly in the people, so that precious life and material can be saved though calamities which are
cannot be stopped [1,3].Still there are lots of disasters occurs recently disaster occur at uttarakhand on date 16th July
2013. Disasters produce a devastating effect and they see no difference between human and material. This leads to
great loss of possession & life. Human’s detection is difficult when they are buried among the debris.

Copyright to IJAREEIE [Link] 9876

 ISSN (Print) : 2320 – 3765
ISSN (Online): 2278 – 8875

International Journal of Advanced Research in Electrical,

Electronics and Instrumentation Engineering
( An I SO 3 2 9 7 : 2 0 0 7 Ce rt ifie d Org a n iz a t ion )

Vol. 3 , I ssue 6 , Ju n e 2 0 1 4

Fig 1: Conventional Rescue System

A timely rescue can only save the people who are buried and wounded. Figure 1 shows conventional rescue
system, to identify live or injured human beings as quickly as possible to provide primary service & save its life. In
such situations, rescue system must take fast decisions under pressure, and try to get victims to safe location at their
own risk. The rescue system must collect the location information and status of victims, stability of the structures as
quickly as possible so that medics and firefighters can enter the disaster area and save people. All these works are
performed mostly in very dangerous and risky situations by human and trained dogs [5]. Detection by rescue workers
becomes time consuming and due to the vast area that gets affected it becomes more difficult. This is why since some
years mobile robots have been proposed to help them and to perform tasks that neither humans, dogs nor existing tools
can do. So the project proposes a mobile rescue robotic that moves in the Disasters area and helps in identifying the
alive people and rescue operations.

II. RELATED WORK

In this section, a brief discussion of some of the related work is presented; focusing on the used approach and its
advantages and disadvantages. Remote Operated and Controlled Hexapod (ROACH): ROACH is a design that provides
significant advantages in mobility over wheeled and tracked designs. It is equipped with cameras which transmit live
audio and videos of the disaster site, as well as information about locations of objects with respect to the robot’s
position to the interface on the laptop. Kohga: University of Tokyo - The most complicated task for most of the USAR
robots has been working on a rough terrain. Specialized robots have been designed for these types of environments
such as KOHGA the snake like [Link] robot is constructed by connecting multiple crawler vehicles serially,
resulting in a long and thin structure so that it can enter narrow space.
System for disasters made up of four parts sensor, mobile rescue robot, transmission network,
monitoring center. When victims are trapped inside debris, it will be difficult to search and rescue them quickly. At this
situation Mobile robots are use that can go inside the building and detect if any victim is present and then signal the
crewmembers for recovery. The purpose of using mobile rescue robots is to track the victims in disaster area. Disasters
come with many obstacles for the rescue team that makes it hard for them to reach the victims, for example rainstorms,
collapsed buildings, obstructions and dangerous substances. The rescue team must fast and securely find information of
the disaster areas, so disaster area is covered by mobile rescue robot to rescue people which are injured survivor,
unconscious survivor etc.
III. PROPOSED SYSTEM HARDWARE
The project proposes a mobile rescue robot that moves in the disaster, earthquake prone area and helps in
identifying the live people, injured people, location and rescue system operations. Hence due to the on timely detection
in natural calamities this can save precious life & great loss even without the help of large number of rescue operators.
The proposed system consists of a mobile rescue robot, PC control Module.

Copyright to IJAREEIE [Link] 9877

 ISSN (Print) : 2320 – 3765
ISSN (Online): 2278 – 8875

International Journal of Advanced Research in Electrical,

Electronics and Instrumentation Engineering
( An I SO 3 2 9 7 : 2 0 0 7 Ce rt ifie d Org a n iz a t ion )

Vol. 3 , I ssue 6 , Ju n e 2 0 1 4

Fig 2: Hardware Design of Mobile rescue robot & Control Room.

The mobile rescue robot consists of four units that are namely Sensor unit, Micro-controller, Camera unit, Motor driver
unit, Transmission unit [2]. The sensor unit must be directly interfaced to the micro-controller. The sensor devices
monitor current readings and sends data to the Micro-controller. The controller circuit is responsible for transmitting
this information. Controllers are designed at hardware level. The 2.4GHz RF module is interfaced with the micro-
controller via the standard serial port. These data’s are updated by the PC/SERVER So that rescue team can view the
readings real time and camera is used to detect motion & live visualization of critical situation of the disaster.
Microcontroller unit: The microcontroller that is been used is the PIC 16F877A controller. PICs are popular with both
industrial developers and hobbyists alike due to their cheap, easily available, large user base, easy to program according
to application, free development & low cost tools availability, and serial programming (and re-programming with flash
memory) capability. The microcontroller is used to gather the data from the sensor unit in real time and compare it with
the set point (safer level of temp) and transfer the corresponding information data to the CPU of control room. It also
receives commands from the CPU and transfers it to the robot unit for its movement. The microcontroller is the core of
the surveillance robot.
Sensor Unit: A sensor (also called detector) is a converter that measures a physical quantity and converts it into a
signal which can be read by an observer or by an instrument. Four sensors are used in the project. They are visibility
sensor, Human sensor, and Gas sensor, Temperature sensor & Bomb sensor for which LDR (Light Dependent
Register), PIR (Passive Infra-Red), MQ-7 and LM35 sensors, and A Bomb (Metal) sensor to detect the presence of
suspected material in Rescue operations. are used respectively. The LDR detect visibility if visibility is less after going
under the ground then lamp consist by mobile robot will glows. The LM 35 has a range of 0-110 degree Celsius. The
real-time values observe by the sensors are transmitted to a PIC 16F877A wherein it compares the measured value with
that of the set points. This data information provide by sensor helps in having good knowledge about the environment
in disaster area.
Camera module: The camera module consists of a web camera and it is mounted on the robot and the video signal
is transmitted to the receiver at control room. The camera module will transmit the video coverage of the paths and thus
helping in easier mapping of the path to be taken by the rescue team. For real time applications, camera of high range is
to be used to get good clarity and good coverage of area. The function of camera also help the robot from getting stuck
in a pit as the obstacles lying in path is foreseen and required action can be taken, thus improving the life of robot in the
disaster area. Due to which we can observe the robot & we can see live vision.
Robot driver unit: The robot driver unit is primarily concerned about the movement of the robot in x-axis and y-axis.
The robot is of conveyor belt type as it helps to maneuver over debris and rugged terrain. Two DC motors of 200rpm
Copyright to IJAREEIE [Link] 9878
 ISSN (Print) : 2320 – 3765
ISSN (Online): 2278 – 8875

International Journal of Advanced Research in Electrical,

Electronics and Instrumentation Engineering
( An I SO 3 2 9 7 : 2 0 0 7 Ce rt ifie d Org a n iz a t ion )

Vol. 3 , I ssue 6 , Ju n e 2 0 1 4

will run the wheels of mobile rescue robot. When both the wheels are given with positive pulse edge, then robot will
moves in forward direction. when the supply is reversed mean both the wheels are given with negative pulse edge, then
it goes in backward direction and similarly by varying the negative and positive edge, left and right turn can be achieve
successfully. The selection of supply given to each motor, L293D IC is used. This will drive the robot to move in
forward, reverse and turn left and right.
Transmission unit: The transmission unit is used to transmit data. For free bandwidth the 2.4GHZ RF modem is used.
To improve the security of system, other secure frequencies can be used. Transmitter is mounted on the mobile robot
unit; its function is gets information data from the micro-controller and transmits to the receiver placed in the control
room.
At the control room these weather parameters along with real time footage of concerned area is displayed on TV
screen. By using wireless communication exact location of human we will get from our mobile robot to control room.
The rescue team of Experts & doctors can be sending to the victim’s location for primary treatment and can be sent to
the safe place or hospital in case of emergency
IV. SOFTWARE IMPLEMENTATION
When the PIR sensor detects a motion signal in its surrounding, then control program will orders the camera to display
the surrounding area .If a human is detected in disaster area covered by mobile robot, then proposed system sends its
current location & shows live vision to the rescue team. Figure 6 shows the flowchart of the propose system operation.

Fig 3: Flowchart of Proposed System

Copyright to IJAREEIE [Link] 9879
 ISSN (Print) : 2320 – 3765
ISSN (Online): 2278 – 8875

International Journal of Advanced Research in Electrical,

Electronics and Instrumentation Engineering
( An I SO 3 2 9 7 : 2 0 0 7 Ce rt ifie d Org a n iz a t ion )

Vol. 3 , I ssue 6 , Ju n e 2 0 1 4

In the fig 3, it shows complete path of program & along with working of robot. When Mobile robot starts moving then
continuously PIR sensor check for motion if motion is detected then buzzer gives a beep & then all the other sensors
get ON & camera get activated else mobile rescue robot keep moving to check motion. If human is detected then its
live vision we can see in control room.

V. RESULT AND DISCUSSION

Fig 4: Mobile Rescue Robot

In the fig 4, it shows structure of mobile rescue robot & assembly of hardware on the top of conveyer belt mobile robot.
Hardware contains LCD which shows real time data of the entire sensor.

Fig 5: Video Footage

In the fig 5, it shows live video footage of the area covered by mobile rescue robot on laptop which is placed in control
room
Copyright to IJAREEIE [Link] 9880
 ISSN (Print) : 2320 – 3765
ISSN (Online): 2278 – 8875

International Journal of Advanced Research in Electrical,

Electronics and Instrumentation Engineering
( An I SO 3 2 9 7 : 2 0 0 7 Ce rt ifie d Org a n iz a t ion )

Vol. 3 , I ssue 6 , Ju n e 2 0 1 4

.
Fig 6: Real Time Sensor Data
In the fig.6, it shows the Real time data of all the sensor like temp(°c) ,Gas concentration, Light Intensity( %) ,
Obstacle, Bomb with information of Date & [Link] using Vb software Real time data is collected at control
room & according to real time data proper security action will be taken.

Fig 7: Temp & Gas Real Time Graph

In the fig 7, it shows the graph of Real time data of temperature & Gas sensor Vs Time in second. The real time data
of temperature & Gas sensor is converted into graphical format.
[Link]
The application of wireless sensor network will improve the saving of many lives by using mobile rescue robot in
disaster. In this paper we design an effective & safe system to ensure that there are no human left behind in rescue
Copyright to IJAREEIE [Link] 9881
 ISSN (Print) : 2320 – 3765
ISSN (Online): 2278 – 8875

International Journal of Advanced Research in Electrical,

Electronics and Instrumentation Engineering
( An I SO 3 2 9 7 : 2 0 0 7 Ce rt ifie d Org a n iz a t ion )

Vol. 3 , I ssue 6 , Ju n e 2 0 1 4

operation. This system can detect existence of human, temp, humidity, visibility in order to monitor weather report &
trace the location of victim in disaster area. It will greatly improve the performance & efficiency of data transmission. It
is convenient to expand & reduce the cost of extending system. The application of wireless sensor network can realize
the real-time monitoring of affected area by the natural calamities.

REFERENCES
[1] An autonomous wireless sensor network deployment system using mobile robots for human existence detection in case of disasters Ad Hoc
Networks 13 (2014) 54–68
[2] Sandeep Bhatia,“Alive Human Body Detection Using A Autonomous Mobile Rescue Robot”“[Link]/documents/Mr.
SandeepB [Link]”
[3] A joint network for disaster recovery and search and rescue operations journal homepage: [Link]/ locate/comet 2012
[4] Alive human body detection system using an autonomous mobile rescue robot India Conference (INDICON), 2011 Annual IEEE
[5] Research on human body detection methods based on the head features on the disaster scenes Systems and Control in Aeronautics and
Astronautics (ISSCAA), 2010 .
[6] Anjali Maggu1, Kiran Rana2, Mitesh Kumar3, Mohit Dahiya4& Soni Chaurasia5 “LIFE DETECTING ROBOT FOR RESCUE
OPERATIONS” International Journal of Information Technology and Knowledge Management July-December 2011, Volume 4, No. 2, pp.
653-657
[7] Mr. S.P Vijayaragavan Hardeep Pal Sharma,Guna sekar.C.H, [Link] Kumar “Live Human Detecting Robot for Earthquake Rescue
Operation” International Journal of Business Intelligents ISSN: 2278-2400 Vol 02, Issue 01, June 2013
[8] An Autonomous Mobile Robotic System for Surveillance of Indoor Environments International Journal of Advanced Robotic Systems, Vol. 7,
No. 1 (2010)
[9] Chou Wusheng. Wang Tianmiao, You Song, "Sensor-based autonomous Control for telerobotic system", Proceedings of the 4th World
Congress On Intelligent Control and Automation, 2002, vol.3, pp. 2430 - 2434.
[10] Z. Hu, “Localization system of rescue robot based on multi-sensor fusion”, IEEE Int. Conf. on Computing, Control and Industrial Engineering,
Aug. 2011.
[11] S. Miyama, M. Imai and Y. Anzai, “Rescue robot under disaster situation: Position acquisition with omni-directional sensor”,
[12] IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, Oct. 2003
[13] T. Suzuki, R. Sugizaki, K. Kawabata, Y. Hada, Y. Tobe, Autonomous Deployment and Restoration of Sensor Network using Mobile Robots,
International Journal of Advanced Robotic Systems (2010)
[14] [Link]
[15] J. Yick, B. Mukherjee, D. Ghosal, Wireless sensor network survey,
[16] Computer Networks 52 (2008) 2292–2330.
[17] E. Cayirci, T. Coplu, SENDROM: sensor networks for disaster relief
[18] operations management, Wireless Networks 13 (2007) 409–423.

Copyright to IJAREEIE [Link] 9882

 Proceedings of the 2005 IEEE
International Workshop on Safety, Security and Rescue Robotics
Kobe, Japan, June 2005

Mobile Robots for Search and Rescue

PeLoTe Session

Niramon Ruangpayoongsak Hubert Roth Jan Chudoba

Institute of Automatic Control Applied Computer and Software Gerstener Laboratory, Department of
Engineering Engineering Centre (ARS) Cybernetics
University of Siegen Wuerzburg, Germany Czech Technical University
Siegen, Germany Email: [Link]@[Link] Prague, Czech Republic
Email:niramon.r@[Link] Email:chudoba@[Link]

Abstract – Mobile robots are integrated into a search and in the firing scenario and the experimental results are also
rescue team as tools for searching victims in dangerous areas presented.
that is harmful for human, as to provide the perception data
for map building, and as to follow the human entity during II. MOBILE ROBOTS
the mission. The teleoperated control and the autonomous
path following is implemented on the robots for the semi- The PeLoTe project targets on the concept of creating the
autonomous navigation in the simulated firing scenario. This presence feature via remote control and navigation, as well
paper focuses on the robot entities in the PeLoTe project as knowledge sharing in combined communities of living
(Building Presence through Localization for Hybrid and nonliving entities. The foreseen applications include
Telematic Teams). The architecture of the software next generation teleoperation and telediagnosis systems
integration for mobile robots into the PeLoTe system and the incorporating both humans and semi-autonomous robots,
experimental results are presented. design of highly realistic man-machine interfaces,
allowing accomplished complex teleoperation tasks.
Index Terms – Mobile robot, Rescue team, Teleoperated
control, Semi-autonomous navigation. A. Robot tasks
In the present time, most of the rescue teams only require
I. INTRODUCTION teleoperated robots, exception are the military that claim
In the search and rescue operation, mobile robots are autonomous robots for map building or exploration of
presently developed in order to assist humans in dangerous environment. Fire fighters are in some cases cautious
and risky tasks during the mission. As a team, the about the reliability and adaptability of autonomy, for
corporation of human and robot can be managed by a these cases teleoperated robots are wanted. In any case, the
remote coordinator, who is located in a safe remote place robots need to provide autonomous low-level actions and
outside of the disaster area. This hybrid telematic team is emergency handling methods in order to meet the
presented in [1]. requirement of robustness. E.g. the robot finds the way
As a team member, the task of a mobile robot is to back in case of communication interruption. These
explore the unknown area and the dangerous area that is requirements will be addressed. Desired autonomies for
not accessible or risky for the human. By using the robots in a farer future are:
appropriate perception equipment, the mobile robot can x Provide data for map building
localize itself and also senses the surrounding in the dark x Explore the environment
area, in which the human has low visibility. Meanwhile, x Search for wounded people in wide areas
the human entity equips the sensors, a data processing unit x Carry equipments
and a small display. He can also localize himself and sense In the opinion of the fire fighters, more autonomous
the surrounding in the similar way as the robots. robots might be used in the farer future in order to handle
The mobile robots have two main functions during the tasks on their own. Typical tasks, which are desirable to be
mission; one is to provide the sensor data for the mapping executed autonomously are for examples, search for
module and another one is to follow the human entity. wounded people in wide areas and exploration of
These robots are based on the same control concept using dangerous areas, provide the updated map according to
on the architecture of the microcontroller and the PC104. detected dangerous areas from gas or fire or closed path of
Nevertheless, the software interface protocols of these building from collapses, and also to carry heavy
robots are not identical. Thus, the integration methodology equipments for the firemen during the rescue mission. In
of the robot software into the PeLoTe system is designed. addition, the robot might perform routine tasks for saving
This paper describes the tasks, features, functionalities, time for human.
and the integration methodology of the mobile robots in Telepresence techniques are thus necessary during the
the PeLoTe project (Building Presence through mission for mobile robots in receiving sensor data, video
Localization for Hybrid Telematic Teams) funded by the image and audio from the remote places. Under semi
European Community under the IST-program: Future and autonomous function, a fire fighter, who performs as an
Emerging Technologies. The experiments were performed operator can command the robot to reach a position on the

0-7803-8945-X/05/$20.00 ©2005 IEEE. 212

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:20:14 UTC from IEEE Xplore. Restrictions apply.
 Fig. 2. Human following robot

Furthermore, the user can also command by specifying

line paths or arc paths. This is called the autonomous path
Fig. 1. Mapping robot following control. In this mode, the robot automatically
controls its movement along the received path and stops at
the destination using self localization. These semi-
preliminary map by using the path planning and the path autonomous features are integrated into the PeLoTe
following control or by using the teleoperated control. system and are explained in the next section.
After a command is accomplished, the robots stand by and
wait for the next command, which can be either the path III. SOFTWARE INTEGRATION
control command or the teleoperated control command. The overall software architecture of PeLoTe system is
B. Robot descriptions described in [1]. The detail of the robot software
integration is described in this section. From the developed
There exists only few expensive rescue robots in the GUI and hardware interface of the robot as explained in
market and their features are not complete according to the the previous section, an interface layer is required to
project requirements. Their functionalities are limited mount two or more robots into the same data and
according to the mechanical structure and the sensors on command protocols. From different protocols into one
board. Thus, the exploited mobile robots in the PeLoTe common protocol that is recognizable for all robots, the
project are built based on the existing mobile robot Robot Entity Control Module (RECoM) is defined,
technologies. As shown in Figs. 1 and 2, the Mobile designed, and developed using java RMI for the server
Experimental Robot for Locomotion and Intelligent communication. The RECoM is a control system
Navigation (MERLIN) is controlled by 80C167 CR 16 bit- consisting of several modules. The data flow among these
processor and PC104 [2]. Two motors are the servo modules is shown in Fig. 3.
steering motor and dc driving motor for the robot
orientation control and the robot speed control, A. Server interface module
respectively. The equipments on board are encoders, It is necessary to keep the communication connection
ultrasonic sensors, infrared sensors, 3-axis magnetic during the mission. However, in some blind area, where
compass, gyroscope, laser scanner, beacon system, the WLAN signal is weak or hidden by walls, the
bumpers, web camera, WLAN, and white LED lamps. The communication could be lost. Thus, the server interface
mobile robots provide sensor data for the higher layer module responsible for establishing connection to the
software interface. These data are the driven distance on PeLoTe server and provides communication with server to
the left wheel and the right wheel, the driving speed, the other modules. In case of lost of connection, the server
angle of the robot heading, the absolute roll, pitch, yaw interface module informs all modules and tries to re-
rotational angles, the battery level detector and the establish the connection.
obstacle distance detection.
MERLIN is remotely controlled via the programmed B. Decision module
client and server using java software based on TCP/IP The interpretation of the path data from path planner and
communication. The wireless LAN is used for wireless notification of robot events are the functions of the
data communication between user GUI and the robot. The decision module. This module mainly downloads and
control commands and telemetry data from sensors are processes the global plan from the server that is generated
transmitted between the client and server continuously. by central planning module. The global plans are then sent
In semi-autonomous teleoperated control, the user to the planning module. Special procedure is needed for
commands the robot by using a joystick or arrow cursors the plan points marked as inspection points, where robot
as to steer the front wheels and to propel the rear wheels. stops and explores its visible vicinity. When an inspection

213

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:20:14 UTC from IEEE Xplore. Restrictions apply.
 SERVER INTERFACE
local localization module processes sensor data used for
the position estimation and sends the estimated position to
plan from map the server. One the other hand, the localization module
progress transfer to
server reporting also handles the corrections of robot position from the
server
operator or from the CoLo.
DECISION MAPPING
MODULE F. Direct movement control module
position In some cases, the operator may need to control the robot
global plan local position update
map correction manually using joystick or arrow cursors; the direct
PLANNING LOCALIZATION movement control module. The symbol 'direct TO conn.'
means direct teleoperator connection that is the direct
direct
trajectory messages for connection of sending and receiving the joystick control
TO.
DIRECT
teleoperator command over socket communication between GUI and
conn.
MOVEMENT DIAGNOSTICS the direct movement control module and the command is
CONTROL futher sent to the hardware abstraction layer.
G. Diagnostics module
obstacle BUMPER
MOVEMENT warning As its name, the module informs server and operators
CONTROL about robot status, e.g. battery level, and important sensor
movement sensor data data e.g. temperature, detected gas. These data are shown
commands on PeLoTe GUI as symbolic color tabs or numerical data
HARDWARE ABSTRACTION LAYER depended on the type of sensor data so that they are easily
and quickly recognizable by the operator.
Fig. 3. RECoM data flow
H. Movement control

point is reached, a message is sent to the server for Here, the trajectory plan from the planning module is
progress reporting. This is necessary for the global path translated into robot movement commands and sent to the
planning module on the server to know what places were robot via the hardware abstraction layer.
already visited, in case of the need for replanning. Another I. Bumper module
function of the decision module is to report operators via
server of messages about unexpected events and The bumper module raised a flag, when the obstacle is
notifications about performed goals e.g. explored areas, close to the robot. As to avoid collision, this information is
found victims, finished plan. sent to the movement control to block the forward
movement.
C. Mapping module
J. Hardware Abstract Layer (HAL)
To support the map updating based on entities perception,
the mapping module builds map from the data obtained The robot hardware abstract layer provides a single
from obstacle detection sensors and these maps are control system for all robots. The main function is to
transferred to the server where global map is built. The sendthe movement commands to the robot and to collect
map is then used by server for efficient replanning in case the sensor data back. During the navigation, the sensor
of unexpected situation like blocked corridor. In such a data are saved in the log file, are used for other modules on
situation, the updated map is very efficient for helping the RECoM internally, and are sent in parallel to the GUI via
fire fighter, who cannot find the way out from the sudden the PeLoTe server. Furthermore, HAL also provides
death end caused by the destruction of walls or parts of the possibilities to initialize and calibrate sensors, whenever
building. The local map is further sent to the planning the operator wants.
module. Even the above modules are defined for all entities, some
are used by an entity, might not be used for another
D. Planning module entities. Most of the described modules are used for the
The planning module takes the plan from the decision mapping robot. For the human following robot, some
module and the mapping module, generates the local plans modules are not fully used and in extra, a module
based on the obstacle detection, and sends these planned responsible for the human following is added. This module
trajectories to the movement control. checks the position of the human, the robot position, and
the obstacle positions. Then commands are generated for
E. Localization module the movement module on the robot, wrapped by the HAL,
in order to keep certain distance from the human and also
In the PeLoTe system, the global and local localization
to avoid collision to obstacles.
modules are integrated. The global localization module is
In parallel, the PeLoTe server dispatches all received
called the Cooperative Localization (CoLo). The CoLo
data, mainly position of all entities, to the connected
provides the correction of localization for all entities. The
teleoperator GUI stations, including the GUI used by the
CoLo process is not focused in this paper. On RECoM, the
man in action, giving good mission overview to all. On

214

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:20:14 UTC from IEEE Xplore. Restrictions apply.
 Fig.5. Robot founds a victim

Fig. 4. The PeLoTe experiment in the firefighter training house

of Wuerzburg, Germany

PeLoTe server, there are also some other modules running

internally such as the CoLo, the global map builder, and
the planning module.
IV. EXPERIMENTS
The PeLoTe experiments were performed in the
firefighter training house of Wuerzburg, and in the Juliux
Maxmilian University of Wuerzburg. The experiments are
done under different scenario. As shown in Fig. 4, the
search and rescue team members are in the living room of Fig. 6. Dangerouse zone (left) and dangerous exit (right)
the firefighter training house. The PeLoTe rescue team
consists of a fireman with the personal assistance and
navigation systems (PeNa) [3-5], a mapping robot, a
human following robot. In addition, a victim represented
by a doll is put on a chair in front of firing sofa. For the
experiment in the university, the simulated firing scenario
are designed and constructed.
A. Description of the PeLoTe system operation
Initially, the map of the house is obtained and created as
a PeLoTe global map. All entities are set up at a starting
position on the map and wait for commands from the
operator. At the beginning, the operator starts voice
communication with human entity and makes confirmation
for prompting. Then, the operator starts planning for the
Fig. 7. Robots and beacons
mission for each entity and sent the planned paths to each
entity. The human entity and mapping robot start
that the robot stops and waits for the next command from
following the individually obtained path, which could be
operator. The example of simulated dangerous area and
different from each other. The human entity and the
dangerous exit is shown in Fig. 6. The mapping robot was
mapping robot can also update the map, when they found
commanded to investigate in these area, which is harmful
victims or dangerous area. Meanwhile, the human
for human. During the mission, the sensor data and the
following robot receives no plan, instead, follows the
entity positions are displayed on the operator GUI whereas
human entities during the whole mission. The mission is
the voice communication is continuously hold between
ended when the time limitation is reached or when the area
human entity and teleoperator throughout the mission.
is completely investigated.
Furthermore, the robot localization is further improved
B. The robot operations by the additional ultrasonic beacon system described in [6,
7]. The beacon system corrects the entity positions and
As shown in Fig. 5, the doll represents the victim and the
headings in areas covered by beacons. The 4 beacon
wall is built using carpet for convenience in construction.
sensors are installed in the scenario area shown in Fig. 7
During the operation, when the robot found a victim, the
and a beacon sensor is mounted on each entity. This
positions of victims are updated on the global map. After
system provides measurement of distances between the

215

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:20:14 UTC from IEEE Xplore. Restrictions apply.
 fixed beacons and the moving beacons that are mounted on
entities. The position is corrected using the dynamic
multilateration algorithm.
C. The PeLoTe GUI
Fig. 8 shows the PeLoTe GUI described in [8, 9]. When
the operator sends new path to the robot, the generated
path is shown up as concatenated straight lines on the
global map. The small scale map with a zoom bar is
available at the bottom right of the panel. The entities
position and heading are represented by triangles. Each
entity is recognized by colors on the same global
coordinate. Their positions are always updated during the
mission. On the left hand side of the panel, there are
symbols of victim, gas, fire, and dangerous area symbols,
which could be added later on the map for map updating. Fig. 8. PeLoTe GUI with generated path and entities position
The message dialog box is also available at the bottom of
the monitor. This shows message from entities e.g. robot walls and collapsed sound. The path is closed behind the
founds victim, low battery. firemen after he has passed through the destructive point.
D. Experimental results and discussion They would hear the noise but they couldn’t see what was
happening. Finally, some firemen loosed the way they
1) The experiment in the firefighter training house: the have passed and loosed their orientations. Most of the
firemen observed the PeLoTe system during the operation firemen thought that the scenario was complicated and was
and they were interviewed after the demonstration. They hard for them to find the way out.
agreed that the PeLoTe system is useful in the search and In a team, one fireman is an operator and another goes
rescue mission. The robots have advantages in harmful into the firing scenario. After each experiment, the firemen
environment such as toxic gas, explosion, nuclear area, must fill the evaluation form about the experiment.
collapse. The human following robot could also carry the Without the PeLoTe system, some of them failed to find
equipment, taking over the work from human in the the way out before the time is over because they didn’t see
destructive area, and doing measurement. The most the scenario and they didn’t know where they were. In the
important information for search and rescue are, where I opposite, the team with PeLoTe system could know their
am, where the exit is, and where the way back is. In some position during the mission using PeLoTe system and
situation, someone would loose his way out from the finally could come back to the exit.
building. In this case, the localization is necessary. The experimental results related to questionnaire are
Furthermore, the robots are expected to close the valve and presented in the PeLoTe session, “Human-Computer
check the chemical substances in the future. However, the Interaction in the PeLoTe rescue system.” There, the
firemen believe that they would rely on themselves more comparisons between the experimental results with and
than robot decisions in the non harmful area. without the PeLoTo system are also provided. The
2) The final experiment at the university: The map of the operating time of each experiment, the coverage of area in
experimental area is presented in the PeLoTe session, percentage, the number for fire area found, the number of
“Human-Computer Interaction in the PeLoTe rescue dangerous area found, and the number of victim found are
system.” The map represents the firing scenario and shows compared. For the final experiment with PeLoTe system,
the location of the dangerous area, the firing area, the gas from six experiments, the average time of all team is 25.2
area, the activated and not activated fire alarm, the victims, minutes and the rescue team covered 98.3 % of the overall
and the exits. The simulated firing scenario in the building area. All victims are rescue where as the average of fire
of the university represents the gas, fire, dangerous zone area found is 3-4 from the total of 4 and dangerous area
and dangerous exit by the symbols printed on a white found is 3-5 from the total of 5.
paper that are placed on the ground. The experiment area
was closed and all lights are turned off during the V. CONCLUSIONS
experiment to lower the visibility of human. This paper presents the mobile robots used in search and
Before the experiments started, the firemen were trained rescue operations to support the hybrid telematic teams in
for understanding the mission and knowing how to equip PeLoTe project. As a member of the team, the robots have
the PeLoTe system. The mission has 25 minutes time functionalities for searching victims, investigating
limitation. Before the time is over, the firemen should dangerous area or low visibility area, providing sensor data
search and rescue all victims and also come out from the for mapping and following human. The experiment were
firing scenario. The scenario was setup as the complicated performed in the firefighter training house in Wuerzburg
and low visibility large area. During the mission, the and in the Juliux Maxmilian University of Wuerzburg. The
victim volunteers were crying for requesting help, at the firemen performed the test using the PeLoTe equipment. It
position where the doll is placed, until they are found and was demonstrated, that semi-autonomous robots may
rescued. The building collapse was also simulated using improve quality of whole missions, mainly in the terms of

216

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:20:14 UTC from IEEE Xplore. Restrictions apply.
 speed, coverage area exploration and lower danger to the
firefighters. The whole PeLoTe system showed to be
effective for the large-scale complicated search and rescue
missions.
ACKNOWLEDGMENT
The work has been supported under the “PeLoTe-Building
Present through Localization for Hybrid Telematic
Systems” project within the IST-2001-FET framework.
The authors would like to thank the contributions from our
consortium partners at the Julius-Maximilians University
Würzburg, the Czech Technical University in Prague,
Helsinki University of Technology, and Certicon a.s.
REFERENCES
[1] F. Driewer, H. Baier, K. Schilling, J. Pavlicek, L. Preucil, M. Kulich,
N. Ruangpayoongsak, H. Roth, J. Saarinen, J. Suomela, A. Halme,
“Hybrid Telematic Teams for Search and Rescue Operations,”
Proceedings of the SSRR'04, IEEE International Workshop on Safety,
Security, and Rescue Robotics, Bonn, Germany, ISBN: 3-8167-6556-
4, May 2004.
[2] J. Kuhle, H. Roth, and N. Ruangpayoongsak, “MOBILE ROBOTS
and airships in a multi-robot team,” The 1st IFAC Symposium on
Telematics Applications in Automation and Robotics, Helsinki
University of Technology, Finland, pp. 67-72, June 2004.
[3] J. Saarinen, R. Mazl, P. Ernest, J. Suomela, and L. Preucil, “Sensors
And Methods For Human Dead Reckoning,” Proceedings of the 8th
Conference on Intelligent Autonomous Systems, IAS-8, Netherlands,
Amsterdam March 10-13 2004.
[4] J. Saarinen, J. Suomela, A. Halme and J. Pavlícek, “Multientity
Rescue System,” Proceedings of the 1st IFAC Symposium on
Telematics Applications in Automation and Robotics, Helsinki, June
2004
[5] J. Saarinen, R. Mazl, M. Kulich, J. Suomela, L. Preucil and A.
Halme, “Methods For Personal Localisation And Mapping,”
Proceedings of the 5th IFAC symposium on Intelligent Autonomous
Vehicles, IAV2004. Portugal, Lisbon, July 5 – 7, 2004.
[6] M. Kulich, J. Pavlíþek, J. Chudoba, R. Mázl, L. PĜeuþil, “Hybrid
Telematic Systems for Rescue Missions,” Cybernetics and Systems
2004, Vienna: Austrian Society for Cybernetics Studies, 2004, vol.
1,2, p. 731-736. ISBN 3-85206-169-5.
[7] J. Chudoba L. Preucil, “Localization using ultrasonic beacons,” the
3rd International Congress on Mechatronics 2004, Prague, 2004.
[8] F. Driewer, H. Baier, K. Schilling, “Robot / Human Interfaces For
Rescue Teams,” Preprints IFAC Symposium on Telematics
Applications in Automation and Robotics, Helsinki 2004, p. 79-84.
[9] K. Schilling, F. Driewer, H. Baier, “User Interfaces for Robots in
Rescue Operations,” Proceedings IFAC/IFIP/IFORS/IEA Symposium
Analysis, Design and Evaluation of Human-Machine Systems,
Atlanta, USA, 2004.

217

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:20:14 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the 2011 IEEE International Symposium on
Safety,Security and Rescue Robotics
Kyoto,Japan,November 1-5 2011

Redesign of rescue mobile robot Quince

–Toward emergency response to the nuclear accident
at Fukushima Daiichi Nuclear Power Station on March 2011–
Keiji Nagatani, Seiga Kiribayashi, Takeshi Nishimura, Yasushi Hada
Yoshito Okada, Satoshi Tadokoro Tomoaki Yoshida,
Eiji Koyanagi

Tohoku University Chiba Institute of Technology Kogakuin University,

6-6-01, Aramaki-Aoba, Aoba-ku, 2-17-1, Tsudanuma, Narashino, 1-24-2, Nishi-shinjuku,
Sendai, Miyagi 980-8579, Japan Chiba, Japan Shinjuku-ku, Tokyo, Japan
keiji@[Link] koyanagi@[Link] had@[Link]

Abstract — On March 11, 2011, a massive earthquake and tsunami called Quince, are designed for practical use in search and res-
hit eastern Japan, particularly affecting the Tohoku area. Since then, cue missions. The Quince was waterproof and highly mobile
the Fukushima Daiichi Nuclear Power Station has been facing a crisis. over rough terrain. However, its hardware reliability, commu-
To respond to this crisis, we considered using our rescue robots for
surveillance missions. Before delivering a robot to TEPCO (Tokyo nication, and basic sensors were not sufficient to employ it
Electric Power Company), we needed to solve some technical issues as part of the disaster response in Fukushima. Therefore, we
and add some functions to respond to this crisis. Therefore, we began began a project to redesign the Quince robot.
a redesign project to equip the robot for disaster response missions.
TEPCO gave us two specific missions. One was to explore the inside
The first disaster response task given to us by TEPCO
and outside of the reactor buildings to perform dose measurements. (Tokyo Electric Power Company) was an exploration and dose
The other one was to sample contaminated water and install a water measurement mission inside and outside the reactor buildings
gauge in the basement of the reactor buildings. To succeed in the that were seriously damaged. In this mission, we needed to
above two missions, we redesigned our mobile robot, Quince, and consider the following issues:
performed repeated operational test to improve it. Finally, one of the
robots was delivered to the Fukushima Daiichi Nuclear Power Station 1) Hardware reliability
on June 20, 2011. In this paper, we will introduce the requirements
for the above two missions and report how we fulfilled them. It was impossible for us to control the robot ourselves.
Instead, operators from TEPCO needed to tele-operate
Keywords: Rescue robot, Disaster response it. This meant that the robot would be controlled by a
novice operator, who might damage it. Furthermore, once
I. I NTRODUCTION
the robot was given to TEPCO, it was impossible for us
On March 11, 2011, a massive earthquake and tsunami to maintain the robot. This was because the robot was
hit eastern Japan, particularly affecting the Tohoku area, going to be exposed to radiation, making it unacceptable
and claimed many lives. Furthermore, the Fukushima Daiichi for us to maintain the robot after delivery to TEPCO.
Nuclear Power Station was also damaged, and resulting in Considering the above situation, we needed to ensure that
meltdown accidents and the release of radioactive material. the hardware used for our robot was very reliable.
This emergency is still continuing (June 2011). 2) Communication reliability
In this emergency, the first disaster response mission was to In NEDO’s project, we developed a hybrid mesh net-
check on the damage to the target environment, including dose work that included a wired mesh network and wireless
measurements at the disaster site. However, the site, including mesh network for multi-robots tele-operation over a wide
the outside and inside of the nuclear reactor buildings, is very area [2]. In this mission, the target area was not very
dangerous for humans because of the potential for high radi- large. However, the reactor buildings contain very thick
ation exposure. Therefore, there is a great need to use mobile concrete walls that block gamma rays, making it highly
robotic technology for such exploration missions, instead of probable that they would block radio communication.
humans. Therefore, non-wireless communication was required to
Our joint research group with support from NEDO (New maintain the reliability in this mission.
Energy and Industrial Technology Development Organization) 3) Checking radiation hardness
has been researching and developing tracked robots to assist In comparison with other disaster situations, a feature of
rescue crews in search and rescue missions in dangerous this mission was high radiation exposure. Our robot was
environments [1]. Some of the robots that we have developed, composed of conventional electric devices, and we had

13
978-1-61284-769-6/11/$26.00 ©2011 IEEE
Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:13:09 UTC from IEEE Xplore. Restrictions apply.
 Wireless Com. Quince1:Cable robot
Quince6 Quince1
2.4GHz 13ch
(Exploration vehicle) (Cable vehicle)
Wireless Wireless
Wireless device
Wireless Wireless
slave unit base unit
slave unit base unit
Wireless Com. Main camera
2.4GHz 1ch

Twist-pair cable, Overhead camera

500m VDSL com.
Air lock
Inside of reactor building
Cable unreeling
mechanism Cable drum
Outside
LAN Cable

Twist pair cable,

200m VDSL com.
Metal cable reel
Control Control
desk 2 desk 1
Sub-tracks×4
Fig. 1. Overview of configuration proposal for dual-robot system.

Quince6:Exploration robot
very little information about how well it would survive
exposure to gamma rays. We will report the results of Wireless device
a gamma-ray irradiation test of the electric components Main camera
Dose measurement
that we performed in another study. device Overhead camera
4) Additional sensors
One of the important requirements in this mission was to
Laser range
equip Quince with a dose measurement function, which scanners
the original robot lacked. Furthermore, 3D mapping was
a very good tool for understanding the disaster environ-
ment. Therefore, some modifications of the sensors were
required.
5) Easy tele-operation for first responders Sub-tracks×4
In this mission, it would be impossible for skilled opera-
tors from our research group to operate the robot directly,
as described above. Therefore, a good user interface for
tele-operation was required for first responders. Fig. 2. Overview of cable robot (upper) and exploration robot (lower).

During our redesign of the robot ( called the Fukushima-

version Quince, in this paper), TEPCO gave us an additional II. A PPROACH FOR FIRST MISSION
specific urgent mission. It included (1) a contaminated water-
sampling task and (2) a water-gauge installation task in the As shown in the introduction, the first mission was to
basement of the reactor buildings. In this extra mission, we perform exploration and dose measurement inside and outside
needed to consider the following issues: of the damaged reactor buildings. For the operator’s safety, an
operation box for the operator controlling the robot needed to
6) Manipulation function
be located as far from the target area as possible. However,
The task required an object handling such as a scooping
the target environment included thick concrete walls, a reactor
motion for water and installing a water gauge. Therefore,
vessel, and higher elevations. In such an unfriendly environ-
an additional manipulator needed to be mounted on the
ment for transmitting and receiving radio waves, we judged
robot.
that conventional wireless communication was unsuitable for
7) Countermeasures against overweight
this mission.
It was necessary to mount additional equipment on the
robot, including the manipulator. Thus, we needed to To secure a reliable communication, we configured a com-
consider an overweight problem, because we did not munication system using a wired/wireless combination net-
assume the presence of such heavy functions/sensors on work, which was a simple version of the hybrid mesh network
the robot in its original design. that we developed in NEDO’s project. Figure 1 shows an
overview of our proposal at the beginning of May 2011.
Finally, we redesigned the robot and after training TEPCO’s (Note that our configuration proposal had been extensively
operators, it was delivered to the Fukushima Daiichi Nuclear revised because the situation had changed in Fukushima.) Two
Power Station on June 20, 2011. developed robots, a cable laying robot and exploration robot,
In this paper, we will introduce the requirements of the two are shown in Fig.2. The details for the items used in the
missions and report how we fulfilled them. redesign are shown in the next section.

14
Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:13:09 UTC from IEEE Xplore. Restrictions apply.
 III. R EDESIGN ITEMS FOR MISSION
The changes from the original Quince are itemized in the
following:
A) Additional hardware and sensors
B) Improvement in reliability of the power supply system
C) Wired/wireless combination network
D) Simplification of operating procedure
E) Improvement of the tele-operation system.
We will report on the above topics in detail, in the following.
A. Additional hardware and sensors
1) Dosimeter
Dose measurement was one of the basic tasks in this
mission. However, obviously, the original Quince did
not have a dose measurement function. Therefore, we
mounted a conventional digital dosimeter (CPXANRFA- Fig. 3. Cable reel-out mechanism.
30, Fuji Electric Co., Ltd.) at a height of 1.2m, which was
almost the same height as internal organs of a human. To
read the value displayed on this sensor, we used a CCD Furthermore, in rare cases, the PC was not booted, even
camera located very close to the dosimeter. if the power supply system worked correctly. If this occurred
2) Main camera during the above off/on action, the robot never worked. To
In an original Quince, we typically mounted an op- avoid this risk, we added another function to the small circuit.
tional pan-tilt-zoom camera, Axis 213 PTZ. However, In a case where the startup beep sound was not detected within
this camera has movable parts that may be broken easily. 10 s, it performed the off/on action for the main power switch.
Furthermore, it is not waterproof, and it would be difficult We installed the above small circuit in the Fukushima-
to make it waterproof because of the movable parts. version Quince and confirmed the rebooting function caused
Therefore, we replaced it with an AXIS212 PTZ camera. by a shutdown of the batteries.
This uses a wide-angle lens combined with a three mega C. Wired/wireless combination network
pixel sensor and realizes instant pan/tilt/zoom by clipping
an image from the high-resolution image without any In this project, we decided to basically use a wired network
mechanical motion. for the tele-operation of the robot. However, in the target
3) LED light environment, there are double-entry doors, called an air lock,
To obtain useful images when using the cameras in a to prevent internal air from flowing outside. When considering
dark environment, lighting was very important. At the this mission, the air lock should not be left open. Therefore, the
beginning of our implementation, we just attached a communication needed to be established through the air lock
light source that was purchased in a climbing gear shop. without a cable. Furthermore, the exploration robot needed to
However, it was not sufficient. Therefore, we mounted be capable of gathering information without a cable.
two LED lights (7 W each) in front and back, along with Based on the above, we proposed and configured a
extra IDX battery (IDX Endura 7, 14.8 V, 4.8 Ah). wired/wireless combination network, as shown in Fig.1. The
control boxes of the robots were supposed to be located
B. Improvement in reliability of power supply system outside the building, and a 200 m twisted-pair metal cable
In the power supply system of the original Quince, there and VDSL (Very high-bit-rate Digital Subscriber Line) devices
were safety measures to protect from overcurrent. When an (NVF-200LS and NVF-200R: NetSys) were used to establish
operator commanded the robot to perform an unreasonable communication between the control boxes and the device that
motion, the power supply system was stopped to protect the was located in front of the air lock (outside). To communicate
DC-DC converters and avoid motor malfunctions. The IDX through the air lock, we used 2.4GHz wireless communication
battery (IDX PowerCube 14.8V 5.7Ah) that the Quince used devices (FXDS540STDMS, Contec Co., Ltd.). Between the
also had a circuit breaker function that used a poli-switch. cable robot and the device located in front of the air lock
Unfortunately, it could not return to normal operation until (inside), we also used a 500 m twisted-pair metal cable
the main switch was turned off and then back on. This was a with the VDSL devices. The cable on the robot was reeled
very serious problem because the robot might stop and wait out when the robot moved forward, passively. Between the
for the operator’s off-and-on action in the reactor building. cable robot and exploration robot, we used 2.4 GHz wireless
To solve this problem, we added a small electric circuit that communication. The radio field strength of the communication
watched the battery voltage. When the voltage dropped below was beyond the limitation set by Japanese law (10 mW/MHz),
a threshold, the circuit performs an off/on action for the main by permission of the Japanese government. This enabled about
power switch every 2 s. 2 km tele-operation of the robot from an unobstructed view

15
Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:13:09 UTC from IEEE Xplore. Restrictions apply.
 outside. Figure 3 shows the cable reel-out mechanism mounted
on the cable robot.

D. Tele-operation
The operation box for the original Quince is composed of
two 15 inch touch-panel LCDs (liquid crystal displays), a game
pad, and one PC (Aspire Revo: Acer, Atom, ION), packed in
a pelican case (1700). The reference rotational velocities of
the mounted motors would be specified by the game pad and
sent to the robot. The battery voltage, internal temperatures of
the robot, and images obtained by four cameras were sent to
the operation box and displayed on the LCDs.
In this project, however, we could not use the above
hardware, because the box was too large to use in the target
field. Furthermore, one of our members would not be the
operator. Thus, the operating procedures for the robot needed
Fig. 4. Display layout of the operation software for exploration robot: (1)
to be simplified. Therefore, we developed a new operation box main camera view, (2) overhead camera view, (3) front camera view, (4) rear
for the two robots’ tele-operation, which was composed of two camera view, (5) 3D laser range scanner view, (6) dosimeter view (not installed
PCs (Toughbook CF-30 mk1, Panasonic) and two game pads yet), (7) battery indicator and temperature display, (8) pose of robot (pitch)
and sub-tracks, (9) pose of robot (roll), (10) LED light switch, (11) total data
packed in a pelican case (1700). Then, we installed additional acquisition button, (12) button for returning sub-tracks to prescribed-pose.
functions in our operation software, as follows:
1) An automatic recovery function in the case of instanta- IV. E XTRA MISSION STATEMENT
neous interruptions,
2) A motion function to return the sub-tracks to a prescribed By the end of the redesign project for the mission from
pose, TEPCO, the demand in the field had changed slightly, and an
3) An extra button to acquire all of the sensor data at once, extra mission was proposed by TEPCO that was urgent. This
4) A screen capture function for the whole mission, involved sampling contaminated water and installing a water
5) A function to display the poses of the robot and sub- gauge in the basement of the reactor buildings. The mission
tracks, was supposed to be performed by a single robot.
6) A real-time display of the scanned 3D range data. To complete this extra mission, we needed to consider an
additional manipulation function. Therefore, we mounted a
When we installed the above functions, display-layout became simple 2 DOF (degrees of freedom) manipulator on one of
one of the most important issues. For example, an additional the Fukushima-version Quinces. Additionally, we equipped the
function might cause a large change in the layout of the robot with a water-gauge/sampling-bottle handling function.
display. Therefore, we installed an adaptable layout-setting Furthermore, the target stairs to the basement of the reactor
function in the operation software. Figure 4 shows a display buildings posed a greater challenge. They were too narrow
layout for the exploration robot, which mounted a 3D laser to use passive cable-reel-out function with tangle-free motion.
range scanner [3]. Therefore, we added an active cable rewind function.
During the above development, we needed to consider coun-
E. Operation test termeasures against overweight. As shown in the introduction,
After the redesign of our mobile robots, we performed initial we did not consider the use of such heavy functions/sensors on
tests and repeated operational tests, and used feedback in our the robot in its original design. Finally, the Fukushima-version
system-design to improve our robot system. Quince with a manipulator reached a total weight of about 50
At the end of April 2011, the redesign project was almost kg.
completed. For example, two non-skilled operators from our V. R EDESIGN ISSUES FOR THE EXTRA - MISSION
group succeeded in carrying out a stable exploration task using
the dual-robot system shown in the previous section. The A. 2 DOF manipulator
target environment was the depot on the 1st floor of building In this project, we needed to develop a robot system rapidly
11 in Shibazono campus, Chiba Institute of Technology. The to respond to the given mission as soon as possible. Therefore,
environment was about 100 m long and was dark at night. we decided to develop a very simple 2 DOF manipulator
It included many racks, which formed a kind of maze, along just for this mission. Figure 5 shows an illustration of the
with some obstacles to be surmounted and steep stairs. mission scene, and Fig.6 shows a photograph of the developed
Furthermore, in May 2011, TEPCO’s staff joined in the manipulator. It had two actuators for pitch angle and yaw
tests to train for its operation. They also provided feedback angle. The top limb of the manipulator was supported by a
to improve the system. parallel linkage, and the tip of the manipulator was always

16
Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:13:09 UTC from IEEE Xplore. Restrictions apply.
 Cable

Operation
box

2 D.O.F Stairs
manipulator
Reel/unreel
mechanism

Sampling
container
Contaminated water

Fig. 5. Illustration of mission scene using 2 DOF manipulator.

Fig. 7. Display layout of operation software for cable robot with 2 DOF
manipulator: (1) overhead camera view, (2) rear camera view, (3) front camera
view, (4) pose of robot (pitch) and sub-tracks, (5) pose of robot (roll), (6)
battery indicator and temperature display, (7) pose of 2 DOF manipulator,
(8) dosimeter view, (9) crane camera view, (10) LED light switch, (11) total
data acquisition button, (12) mission timer, (13) button to return sub-tracks
to prescribed pose.

for us to realize a tension based automatic reel control for

the cable. However, we installed a teleoperated cable rewind
function because rapid development was required at this time.
When an operator sent a rewind command to the robot, the
rewind function began. The torque of the rewind mechanism
had a limitation, so that there was no risk of a tensile cut of
the cable.

D. Improvement in tele-operation software

Accompanying the additional mechanisms of the robot, the
tele-operation software was also changed. Figure 7 shows a
display layout for the cable robot that mounted the 2 DOF
manipulator.
Furthermore, to return up the stairs, we installed a reverse
Fig. 6. Photograph of developed 2 DOF manipulator with water gauge. mode that enabled the switching of the front and rear cam-
eras, flipping the pose-display of the robot between left and
positioned perpendicular to the bottom plane of the robot. A right, and sending inverted commands to the robot. Thus, the
CCD camera and a LED light (1 W) were mounted at the tip, operator could control the robot as if the rear of the robot was
and a sampling container (or a water gauge) was reeled out. its front.

B. Handling function for water gauge/sampling bottle E. Countermeasures against overweight

To install a water gauge in the contaminated water in the After development the above mechanisms, including the 2
basement of the reactor building, we needed to have a reel- DOF manipulator, the weight of the robot became about 50
out mechanism for the cable of the gauge. Furthermore, a kg. It was obviously overweight because the original Quince
sampling bottle to be moved up and down to sample the water. was 27 kg. To respond to the problems caused by this extra
Therefore, we installed a crane mechanism at the tip of the weight, we performed the following operations.
manipulator to realize both of these functions. 1) Load mitigation and heat check of motors
In the extra-mission, the robot was typically on the stairs,
C. Active cable-rewind-function and its weight was increased. Therefore, we increased the
To enable tangle-free motion of the communication cable reduction ratios of the locomotion motors to mitigate its
for the cable robot on the narrow stairs, we needed to add a load. Furthermore, we put heat sensors on the motors to
cable rewind function to go back up the stairs. It was possible check their temperatures and increase their reliability.

17
Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:13:09 UTC from IEEE Xplore. Restrictions apply.
 With the help of the TEPCO, we performed many tests.
Their feedback was also included into the system.
F. Operation test
After May 26, 2011, we began operational tests for the extra
mission. It was much more difficult than the first mission,
because it included many tasks. An operational test scene on
stairs is shown in Fig.9. One of the most difficult parts was
turning at the narrow staircase landings without hitting the
manipulator. The cable rewind mechanism worked effectively
to assist the operator. Although it was originally designed
Fig. 8. Impact absorption mechanism. The left side of the photograph shows just for stair climbing, it also worked fine during standard
the original one, which is broken. The absorption material is offset to the
corners, and a phase shifting occurs (about 45 degree). The right side of the explorations.
photograph shows the improved version.
VI. S UMMARY
In this paper, we described two missions given to us by
TEPCO as apart of the disaster response in the Fukushima
Daiichi Nuclear Power Station and reported on the redesign
of our rescue robot to enable it to perform these missions.
The modifications included the following:
A) Additional hardware and sensors,
B) Improvement in the reliability of the power supply system,
C) A wired/wireless combination network,
D) Simplification of operating procedures,
E) Improvement of the tele-operation system,
F) Construction of a 2 DOF manipulator,
G) An additional crane function for a water gauge/sampling
bottle,
H) An active cable rewind function,
I) Improvement of the tele-operation software,
J) Countermeasures against overweight.
After their redesign for these exploration missions, on June
20, 2011, one of the Quinces was offered to the Tokyo Electric
Fig. 9. Operational test scene of Fukushima-version Quince on stairs. Power Co., for use in the actual disaster field. We hope that
it will contribute to breaking away from this emergency.
2) Improvement of impact absorption mechanisms for sub- ACKNOWLEDGEMENT
tracks
We would like to thank NEDO and the Chiba Institute of
In an original Quince, the impact absorption mechanisms
Technology for the financial support needed to proceed with
were located at the joints of the sub-tracks to prevent
this research.
motor malfunctions. However, during the initial tests of
the Fukushima-version Quince, one of the mechanisms R EFERENCES
was dead, as shown in the left side of Fig.8. To avoid [1] Eric Rohmer, Tomoaki Yoshida, Kazunori 0hno, Keiji Nagatani, Satoshi
this situation, we replaced the nylon used as the impact Tadokoro, and Eiji Konayagi. Quince: A collaborative mobile robotic
absorption material with rigid urethane foam. However, in platform for rescue robots research and development. In Proceedings
of the 5th International Conference on the Advanced Mechatronics
the case of the destruction of the impact absorption mate- (ICAM2010), pages 225–230, 2010.
rial, the joint might still malfunction. That is because the [2] Keiji Nagatani et. al. Development of a networked robotic system for
center component would rotate freely after the destruction disaster mitigation -system description of multi-robot system and report
of performance tests-. In Proc. of the 6th Int’l Conf. on Field and Service
of the material. Therefore, we made the center component Robotics, pages 333–342, 2007.
slightly larger, as shown in the left side of Fig.8. [3] Tomoaki Yoshida, Kiyoshi Irie, Eiji Koyanagi, and Tomono Masahiro. 3d
3) Substitution of long sub-tracks laser scanner with gazing ability. In IEEE International Conference on
Robotics and Automation, WeA204.4, 2011.
When we performed steep stair-climbing tests using the
Quince with the 2 DOF manipulator, it slipped many
times. Therefore, we developed and installed long sub-
tracks on the robot for pressure dispersion.

18
Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:13:09 UTC from IEEE Xplore. Restrictions apply.
 2010 2nd International Asia Conference on Informatics in Control, Automation and Robotics

Design and Implementation of an Emergency Search

and Rescue System Based on Mobile Robot and WSN

Huiyong Wang 1,2 Minglu Zhang 1 Jingyang Wang 2

1 School of Mechanical Engineering 1 School of Mechanical Engineering 2 Institute of Information Science &
Hebei University of Technology Hebei University of Technology Engineering
Tianjin, China Tianjin, China Hebei University of Science and
wanghuiyong815@[Link] zhangm1@[Link] Technology
Shijiazhuang, Hebei, China
ever211@[Link]

Abstract-In order to monitor the leakage of hazardous complete solution yet, thus the interest of research on this
chemicals and guarantee the safety of people in hazardous areas, subject. We propose a solution for the chemical accident.
we developed RESRS (Robot Emergency Search and Rescue Sensors monitor the environment at normal time. When the
System). RESRS integrated technologies such as mobile robot sensors detect the emergency events, the mobile robots enter
and WSN (Wireless Sensor Network). The mobile robots in the into the accident location to carry out search and rescue tasks.
system act as mobile nodes ofWSN. We describe the components
of the generic RESRS's architecture in detail. A small-scale The rest of the paper is structured as follows. In section II
demonstration experiment was presented to highlight the concept we present related works. System architecture is discussed in
and application of the RESRS to an actual chemical accident. section III. Section IV describes the implementation of the
Although our proposed system architecture provides a versatile system. Experimental results provided in section V.
framework for integration, a number of new and interesting Conclusions and future works are discussed in section VI.
research problems need to be solved before actual deployment of
the system.
II. RELATED WORKS
Key words-RESRS, emergency search and rescue, mobile Related work includes the fire information and rescue
robot,WSN equipment project at Berkeley[l]; The Centibots project
examined the use of large-scale mobile robotic teams for
I. INTRODUCTION mapping and areas surveillance [2]; A method for the
transportation of resources by combining robots with sensor
Autonomous mobile robots have demonstrated their network services was suggested by Gupta et al.[3];wildlife
usefulness in supporting life-threatening tasks. They can be habitat monitoring[4]; precision agriculture[8]; intrusion
used and act as high-performance mobile sensor node in WSN detection and tracking [9].
because robots equipped with various sensors and
communication capabilities. If a robot can manipulate sensor
nodes, that robot can change the range and topology of its III. SYSTEM ARCHITECTURE
WSN according to the communication conditions, sensing and The system involves three main blocks are: the WSN fixed
adapting to the environmental situation. Radio signals emitted node, the WSN mobile node (mobile robot), the monitoring
by the WSN nodes might also be used for location and tracking center. Fig. 1 shows the architecture ofthe system.
purposes, thus allowing the robot locate itself even in
environments where GPS is not available [5], [6]. Mobile The task of the monitoring center is to prepare plans in a
robots can be considered to be mobile nodes that provide centralized way, and to monitor the state of mobile robots. It
additional sensorial information. The WSN can be considered uses feedback from the WSN to control the mobile robots for
as an extension of the sensorial capabilities of the mobile placing sensors, collecting data from sensors, locating targets
robots [7]. and delivering the information gathered from the WSN to
human users.
It is important to guarantee the safety of people in
hazardous areas. It is thus important to localize and monitor the It also encompasses the alarm monitoring station, which is
leak sources in an environment. The present paper describes in charge of performing cooperative perceptions processing,
RESRS, an emergency search and rescue system based on and specialized images processing activities such as leak
mobile robot and WSN which is oriented to monitoring the detection. MS SQL Server database located in the monitoring
leakage of hazardous chemicals. This integration between center was used to serve as a backend data store for the entire
mobile robots and a WSN is a problem which does not have a

206
978-1-4244-5194-4/10/$26.00 ©2010 IEEE CAR 2010

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:32:53 UTC from IEEE Xplore. Restrictions apply.
 system. It also stores the information coming from the The data rate is 250 kbps using a Direct Sequence Spread
infrastructure for longitudinal studies and offline analysis. Spectrum (DSSS) . We chose MICAz motes as WSN nodes.
We designed two working mode in the system. One is The MICAz specifications are as follows [10]:
normal mode and the other is rescue mode.
• IEEE 802.15.4, tiny, wireless measurement system
During normal mode the monitoring center analyses the designed specifically for deeply embedded sensor
site information and give the data to managers for making networks
decisions . Several fixed sensors were deployed at each
monitoring node, which is responsible for normal information • 250 kbps, high data rate radio
collection and alarm in accordance with the concentration of • wireless communications with every node as router
the dangerous gases. capability
When the accident occurs, the system will switch to rescue • expansion connector for light, temperature , RH,
mode. The mobile robots enter into the accident location to barometric pressure, acceleration/seismic, acoustic,
carry out search and rescue tasks. The monitoring center magnetic, and other Crossbow sensor boards
analyses the data sent back by the sensors and constructs the
leak model. The monitoring center sends the control TinyOS is an open-source operating system designed for
information to mobile robots. embedded systems with very limited resources, like the
Mica series of motes. Software on the motes runs under
TinyOS [11]. TinyOS uses the NesC language, an extension
to C, with similar syntax, that attempts to embody the
structuring concepts and execution model. As an embedded
operating system, it responds to hardware events with handlers,
while also allowing tasks, which are equivalent to

Qg' /-~ functions in other programming languages. TinyOS does

not implement object sharing.

n"E""':l ' I
I\kl1lt.:'r ing Center-....,
. \ \

\
1
Energy is the scarcest resource of WSN nodes, and it
determines the life time of WSN . WSN is meant to be
deployed in large numbers in various environments, including
~ I I remote and hostile regions, with ad-hoc communications as
W I I key. For this reason, algorithms and protocols need to address

~ ~ /
User the following issues: lifetime maximization, robustness, and
\ ~ I self-configuration.
', ~ ~ I

B. Mobile Robot System

Rescue robotics is an important stepping stone III the
scientific challenge to create autonomous systems. In
Q Chemica!Container
development of robots for search and rescue tasks it is
important to develop a robot which can actually work in a real
disaster site.
Figure I. System architecture
Robots have demonstrated their usefulness in undertaking
life-threatening human tasks in recent years. They are
IV. SYSTEM IMPLEMENTATION deployed at an earthquake site or other disaster and
autonomously search the area, co-ordinate with each other,
A. WSN System deliver assistance to those in need and assist in rescuing
The construction of low-cost and low-power sensor nodes survivors .
that are small in size and communicate in short distances has Challenges in these rescue operations are posed by factors
become feasible with the recent advances in MEMS (Micro such as the unstable nature of the collapsed structures, hard to
Electro-Mechanical Systems) technology , digital electronics, reach spaces, lack of oxygen, and hazards resulting from fire,
and wireless communications. The capabilities of these tiny toxic gases. In the case of rescue robots, currently they are
sensor nodes, which consist of sensing, data processing, and often remotely operated, resulting in a number of limitations,
communicating components, enable the realization of wireless such as the number of robotic devices and poor environmental
sensor network (WSN) based on the collaborative effort of a conditions, etc.
large number of nodes.
Robots can help in the overall search and rescue operation
The Crossbow Company produces several motes: MICA, by producing maps of how to reach a survivor's location,
MICA2, MICA2DOT, and MICAz . MICAz mote uses a helping in asserting survivors' conditions and existing hazards,
Chipcon CC2420 Processor/Radio Transceiver module and etc. A challenge is that search and rescue robots must become
follows the IEEE 802.15.4 and ZigBee Alliances at 2.4 GHz. more autonomous. The robots must have the ability to

207

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:32:53 UTC from IEEE Xplore. Restrictions apply.
 communicate with other members when far away or even if It is assumed that each mobile robot is able to manage
they are sporadically connected. Sparse connectivity may mean ordered sequences of elementary tasks, and to return execution
that frequently there may be no direct path between source and status of the tasks to the monitoring center. FigA shows data
destination. flow of mobile robot.
We used the Motes MICAz to implement our mobile robots
and sensor networks. MICAz is the main component
supporting computation and communication capabilities for the
mobile robots. We designed a single circuit board to drive the
motors. Fig.2 shows the block diagram of our system
architecture. It could process sensing data from the sensor
boards and control motors through a motor board . Separate
power units were used to supply the MICAz and the motors , in
order to increase the stability of the system.

Anl :n
r Boord Figure 3. Mobile robot

Figure 2. Hardware architecture of mobile robot

The low-cost autonomous robotic search and rescue system

(Fig. 3) was designed to cooperate in large quantity. The
architecture of the each mobile robot supports two different
levels of autonomy :
• Low autonomy mode : a supervisor module manages
individual elementary tasks and sequences of
elementary tasks, as they are requested from the Figure 4. Mobile robot data flow
monitoring center. This module also manages the
tasks and robot status returned during the mission
V. EXPERIMENTAL RESULTS
execution.
We have proposed the generic RESRS architecture and
• High autonomy mode: a CNP (Contract Net Protocol) defined its components in the previous sections . We describe a
[12] module allows to autonomously negotiate tasks case study in this section that illustrates the approach we have
allocation in a distributed way by using a variant of used for some of the major technological integrations. Let us
the CNP. In this mode, the monitoring center should consider an example in which leak accident occurs at A.
only provide a list of elementary tasks to be executed
by the mobile robot. Fig.5 shows the result. From the figure we draw a
conclusion that the mobile robot can move to leak source (A)
successfully to carry out search and rescue tasks.

208

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:32:53 UTC from IEEE Xplore. Restrictions apply.
 Second ACM conference on Embedded Networked Sensor Systems
(SerrSys), pp 214-226, 2004 .
[5) Vaidyanathan Ramadurai, Mihail L. Sichitiu, "Localization in
WirelessSensor Networks : A Probabilistic Approach ," Proceedings of
the 2003 International Conference on Wireless Networks ,pp 275-281,
2003.
[6) Maxim A. Batalin, Myron Hatting, Gaurav S. Sukhatme, "Mobile
Robot Navigation using a Sensor Network," Proc. of the 2004 IEEE IntI.
Conference on Robotics and Automation , pp 636- 641, 2004.
[7) Pablo Gil, Iv an Maza, An ' ibal Ollero, Pedro Jos ' e Marr on, "Data
Emergency escue Area centric middleware for the integration of wireless sensor networks and
mobile robots," ROBOTICA 7th Conferende on Mobile Robots and
Competitions, April 2007.
[8) J. Burrell, T. Brooke, R. Beckwith, "Vineyard computing : Sensor
networks in agricultural production," IEEE Pervasive computing,
38-45,2003 .
[9) A. Arora, P. Dutta, S. Bapat, V. Kulathumani, H. Zhang, V. Naik,
V. Mittal, H, Cao, M. Demirbas, M. Gouda, Y-R. Choi, "A wireless
sensor network for target detection , classification, and tracking,"
Computer Networks (Elsevier) , 46(5),pp 605-634, 2004.
Figure 5. Experimental results
[10) Crossbow, Inc., [Link] .[Link]/.
[II) Jason Hill, Robert Szewczyk, Alec Woo, Seth Hollar,David E. Culler,
VI. CONCLUSIONS AND FUTURE WORK Kristofer S. 1. Pister, "System architecture directions for networked
sensors," In Architectural Support for Programming Languages and
Emergency search and rescue system based on mobile Operating Systems,pp 93-104,2000
robot and WSN is an important domain from which we can [I2) R. G. Smith, "The contract net protocol : High-level communi cation and
expect sigficant scientific contributions toward the control in a distributed problem solver," IEEE Transaction on
development of intelligent systems . It features a tremendous Computers,ser.C-29,no.12,pp 1104-1113,1980
application potential while facilitating the investigation of
basic research topics .
A robot emergency search and rescue system is presented
in this paper . We outline some technologies such as mobile
robot and WSN that are available or require in further research
and development. Some of the technology integrations are still
in the development stage. However, we propose a generic
system architecture that would enable the different technology
integrations to be performed searnlessly . Finally we describe a
RESRS demonstration experiment that we have completed that
highlights a particular application for which RESRS could be
used.
The importance of emergency search and rescue system
based on mobile robot and WSN is expected to increase in the
near future, due to the technological advances in
miniaturization and communication. There are still many
issues that remain to be improved and resolved . One of the
most interesting challenges will be the capability to address
more tasks at the same time: for instance , the simultaneous
monitoring of multiple environmental phenomena and 3D map
building .

REFERENCES
[I) J. Wilson, V. Bhargava, A. Redfern, P. Wright, " A Wireless Sensor Net
work and Incident Command Interface for Urban Firefighting . Mobile
and Ubiquitous Systems," Networking & Services, Volume 00. 2007 :
IEEE Computer Society Washington, DC, USA.
[2) K. Konolige , C. Ortiz, and R. Vincent, "Centibots large scale robot
teams," In AAMAS, 2003.
[3) A. K. Gupta, S. Sekhar, and D. P. Agrawal, "Efficient event detection by
collaborative sensors and mobile robots," In First Annual Ohio Graduate
Student Symposium on Computer and Information Science and
Engineering, 2004 .
[4) R. Szewczyk, A. Mainwaring , J. Polastre, D. Culler," An analysis of
a large scale habitat monitoring applicat ion," Proceedings of the

209

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:32:53 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the 2006 IEEE/RSJ
International Conference on Intelligent Robots and Systems
October 9 - 15, 2006, Beijing, China

Design and Manufacturing of a Mobile Rescue Robot

S. Ali A. Moosavian' Hesam Semsarilar2 Arash Kalantari3

Department of Mechanical Engineering

K. N. Toosi Univ. of Technology, Tehran, Iran, P.O. Box 16765-3381
Email: moosav tu x r

Abstract. This paper presents design and manufacturing would be to find victims, determine their situation, and then
procedure of a tele-operative rescue robot. First, the general task to report back their findings based on a map of the building, [1].
be performed by such a robot is defined, and variant kinematic These will immediately be given to human rescue teams
mechanisms to form the basic structure of the robot will be preparing to extract all victims that are found. Further
discussed. Choosing an appropriate mechanism, geometric expectations of rescue robots such as being able to
dimensions, and mass properties will be detailed to develop a autonomously negotiate compromised and collapsed structures
dynamics model for the system. Next, the strength of each component and provide structural shoring, find victims and ascertain their
is analyzed to finalize its shape. To complete the design procedure, conditions, deliver sustenance and communications to victims,
Patran/Nastran was used to apply the finite element method for and emplace sensors (acoustic, thermal, seismic,...) are active
strength analysis of complicated parts. Also, ADAMS was used to research fields. Nevertheless, the basic capability of rescue
model the mechanisms, where 3D sketch of each component of the robots should be their maneuverability in destructed areas
robot was generated by means of Solidworks, and several sets of which thoroughly depends on their locomotion system and
equations governing the dimensions of system were solved using their dimensions. Various rescue robots have been designed
Matlab. Finally, the components are fabricated and assembled and manufactured. For instance, an autonomous Urban
together with controlling hardware. Two main processors are used Reconnaissance Robot, in the size of 20 Kg, was developed by
within the control system of the robot. The operator's PC as the a research group from University of Southern California, [2].
master processor and the laptop installed on the robot as the slave PolyBot is an example of modular reconfigurable robots,
processor. The performance of the system was demonstrated in which is a rather small and economic prototype, [3]. CEDRA
Rescue robot league of RoboCup 2005 in Osaka (Japan) and is also a rescue robot with the ability to adjust its locomotion
achieved the 2nd best design award. system with the terrain on which it performs, [4].
Index Terms - Robotics - Tele-operative - Locomotion - This paper presents an illustrative description of the
Mechanisms. Resquake project at KNTU. Innovative mechanisms and
software-based steps of the design procedure are highlighted
I. INTRODUCTION. here. First, the most successful projects in terms of their
mechanical structure and locomotion capabilities, regardless of
Mobile robots whether autonomous or tele-operative play their autonomy and the sensors mounted on them, are studied
an important role in different fields of human life. Mobile and concluded to develop new special mechanisms to enhance
robots are mainly operated for investigating areas in which the maneuverability of a mobile robot while trying to keep its
human health is endangered. Police robots, fire fighter robots dimensions relevant to the environment in which it performs.
and rescue robots are examples of such application. Mobile After selecting suitable mechanisms, dimensions and
robots are also used for assisting human forces for doing parameters of the system are defined. The system dynamics is
repeated works such as moving heavy boxes within a defined discussed and the sequence of stress analysis for each member
path in a factory or providing the patients with appropriate of the mechanism is addressed in order to finalize its shape and
medicine on time in a hospital. Earthquake is a natural to select suitable material for its fabrication. The last phase of
incident, which threatens human life. Aftershocks occurring a the project was to manufacture the parts and assemble the
while after the main earthquake cause secondary collapses and system. Electronic devices and control system of the robot is
take victims from the search and rescue personnel. In order to described in the last section. Resquake is a robot with great
minimize the risks for rescuers, while increasing victim capabilities in climbing obstacles in destructed areas, which
survival rates, fielding teams of collaborative robots is a good was participated in Rescue robot league of RoboCup 2005 in
alternative. The mission for the robots and their operators Osaka (Japan) and achieved the 2nd best design award.

1- Associate Professor
2- [Link]. Student
3- Graduate Student
1-4244-0259-X/06/$20.00 C)2006 IEEE
3982
Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:18:32 UTC from IEEE Xplore. Restrictions apply.
 environment. Whether destructed or not a rescue robot should
II. MECHANISM DESIGN have the ability to climb and move down stairways in order to
cover the whole area.
There are three major categories of rescue and search
robots in terms of locomotion system, i.e. wheeled, tracked, In order to compromise between the two contradictory
and legged robots as shown in Fig. 1. The simplest are wheeled aspects of providing a small robot with maneuverability of a
robots, while tracked robots are used because of their ability to larger locomotion system, an exceptional mechanism has been
move on uneven terrains and their inevitable traction. Legged developed. This mechanism, which includes a base with two
robots can be the most maneuverable, due to their high degree expandable tracks (arms), enables the robot to resize
of freedom. The number of actuators and sensors is relatively depending on the situation that encounters. Accordingly, these
high which makes their dynamic analysis and modeling more tracks should have a minimum length to prevent loosing its
complicated than the former types. Consequently, stable balance and having a steady movement without extra
control of such systems is more difficult, and eventually more vibrations, Fig. 3. However, such lengthy tracks will need a
expensive.
wide area for turning, which is rarely reachable in a destructed
environment, and a compromise between these two aspects is
also required.

Fig. 3. Minimum length for tracks of the robot

A. Expandable tracks (Arms)
The kinematic structure shown in fig. 1 (top right) enables the
robot to expand its tracks when it needs to bypass obstacles.
Fig. 1. Three major categories of rescue and search robots Like wise, when the robot is going through narrow passages
and needs to be rather small, the front tracks can be closed.
Wheeled robots could be considered as the lowest-priced This was the original idea, developed to overcome the
system to implement for searching flat areas, while developing contradiction.
the autonomy of such systems is easier due to its simple
dynamics. Of course wheeled robots are capable of climbing This concept is improved to a system with two pairs of arms at
up obstacles depending on the diameter of the wheels while a both sides of the vehicle, That would reduce the length of the
relatively small tracked robot has the same capability. The robot with closed arms while the expanded length remains
sketch shown in Fig. 2 compares the two systems encountering relevant. Another advantage would be the symmetry of the
the same obstacle to climb. plan, which enables the robot to approach in both directions.
This arrangement remains operative in turning in a confined
tracked platform space.
So far, the length of the robot with closed arms has become
/-
shorter than its width. The next improvement is to make the
f...
z 777 arms collinear where the main tracks are located at each side,.
The last improvement is adding another joint to each arm in
order to use an extra area between the arms when they are
Fig. 2. Two types of locomotion encountering the same obstacle closed. The tracks on each side of the robot are also separated
As well as the type of locomotion, the size of a mobile rescue into three parallel planes, which provides more efficient
robot is also an important issue. In a destructed indoor field traction. Top view of the designed locomotion system is shown
there may exist some obstacles that can not be passed by any in Fig. 4 . However, it should be noted that adding four
independent joints to the system would increase the complexity
system such as when the walls or the ceiling collapses. At this of its dynamic model as well as the number of actuators and
situation, the robot should search for a bypass or a way the total price of the system. Therefore, a planetary gear chain
between the obstacles rather than climbing over them, which is substituted to simply transmit the power of the main joint of
definitely depends on its size. A relatively small robot can each arm to its second joint. In this case, the rotation of the
easily pass a narrow aisle and continue its search. It should be two parts for each arm will not be independent. Thus, two
noted that stair way is an inseparable part of an indoor desirable positions of the arms are considered, and the gear

3983
Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:18:32 UTC from IEEE Xplore. Restrictions apply.
 ratio is obtained such that the arms move based on a path B. Tracks
between two main positions, Fig. 5. The traction of the locomotion system strongly depends on the
It should be noted that while the main part of the arm rotates friction between the track peaces and the surface on which the
21/2 rad, the second part should rotate more than 21 rad. The robot moves. That is why the material and the shape of the
gear chain with such performance should be a planetary track pieces are of great importance. On the other hand, the
gearbox. The main part of the first arm play the role of the arm tracks should also bear a reasonable tension. Thus, the tracks
in the planetary chain, which is directly powered by a motor. are designed to be assembled using two components. A basis
The sun gear should be attached to the body of the robot and of chain-sprocket provides the system with sufficient tensile
the planet gear should be a attached to the second part of the strength and tooth shaped pieces made of latex filled the gap
arm. A pair of medium gears are placed between the sun and between the chain and surface to create required friction.
the planet so that the diameter of gears does not exceed a Figure 8 shows how latex pieces are attached to the chains.
reasonable length (the diameter of the main wheels of the
tracks), Fig. 6. Another advantage of this mechanism is that the
distance between the centers of the two joints of the arm will Arm's motor
remain constant during its rotaion. This enables us to fill the
gap between the main track and the arm's with another track,
Fig. 4. This track can also be used to transmit power from the
main part of the tracks to the latest part of the arm.

Medium tracks

Fig. 4. Final mechanism chosen for the tracks

Fig. 8. Latex pieces fixed on the chain

C. Suspension system
Two major advantages are obtained by including a suspension
mechanism.
- Increasing the flexibility and stability of the system on
Fig. 5. The path for motion of the arms bumpy surfaces;
- Damping any shock to the system caused by collisions.
Gear4 Gear3&Gear2 Gear2
Planet Gear Medium Gears Sun Gear
The suspension system was designed by separating the two
sides from the main body and then attaching them by a
revolute joint.. A pair of linear springs limits the angle of
rotation and makes the system remain at a single position when
no extra forces are applied. It should be mentioned that using
dampers was not needed, because the friction of the plain
14 13 12 11
bearings used as the so-called joints was enough to limit any
Fig. 6. Planetary gear-chain extra shaking of the springs.
Helical gears are chosen for the planetary gears due to their Finishing the design of locomotion mechanisms, now it is time
small backlash and also higher strength of gear tooth to determine the dimentions. Some of the components for
comparing with spur gears, [5-6]. The angular velocity of the building the mechanisms are available as standard parts, so we
arm should be less than 2-4 rpm while the motor's output should select other dimensions to match their counterparts.
velocity is normally around 3000 rpm. Hence the transfer ratio Besides, the overall size of the robot and the formulas on the
between the arm motor and the link should be around 1000. gear chains must be considered in our calculations. Since there
Therefore, a combination of a 3 stage planetary gearbox, are numerous equations governing these factors, an optimised
(constructed right at the motor shaft where the angular velocity solution is not reachable by manual calculations. Thus, Matlab
is relatively high) with a ratio of 3:1 at each stage, and a is used to extract desired values from a set of equations.
wormgear set with ratio of 30:1 (gear is attached to the arm's
link) provides the desirable ratio within a compact space, Fig. III. DYNAMIcs Analysis
7. As shown in Fig. 7, the tracks at each side of the robot are
powered by a DC motor. A. Locomotion system

3984
Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:18:32 UTC from IEEE Xplore. Restrictions apply.
 There are some important points in robot motion: Finally, the appropriate speed, by operating the robot within a
1- Convenient control over the robot's motion by the variety of situations was obtained around 18 20 cm/s.
operator;
2- Stability of the robot while maneuvering; B. Arms
3- Reducing the risk of collisions. The rotational speed of arms is chosen in range of 3-5 rpm.
Using the expressions for the kinetic and potential energy, and The exact speed will be specified after evaluating the power of
applying Lagrange's equations for a constrained or actuators. In order to obtain the appropriate torque applied on
unconstrained mobile robotic system, the dynamics model can the arms, we should consider the most critical case that the
be obtained as presented in [7-8]. Here, avoiding a system may encounter, which is assumed as shown in Fig. 10,
cumbersome analysis, in order to determine the torques needed and the motor torque is determined such that the robot bears its
to actuate the locomotion system and arms, it is reasonable to own weight.
choose the input torques such that they suffice the highest
torques assumed to be applicable on the system. Thus,
considering a simplified model as shown in Fig. 15, the
minimum torque required for climbing a surface with slope of
35° is calculated. To this end, Newton's 2nd law can be written
for the motion direction: Fig. 10. Arms bearing the weight of the robot over a hole
YFX =ma (1) The torque of the arm's motors is calculated by considering the
free body diagram of the gear chain within each arm, Fig.6.
N Solving the moment equation yields:
(4)
8dG 'a.('I
Ft T =
4
+13 -12) +0.25W(ll +12 +13)
Mg where W=30 kg, la = 106 mm, 11 = 44 mm, 12= 43 mm, 13 =
40.5 mm, and dG4= 39 mm.
x By replacing appropriate values, the required torque on each
arm is obtained as:
Fig. 9. Free body diagram of the robot over a 35o slop
T= 26.5 N.m
Since the final speed of robot is to be rather small and
constant, we could assume the linear acceleration and thus the Simulating a model of the arms using ADAMS, the same result
rotational acceleration of the wheels to be zero: is abstained, which confirms the above approximations. By
ax O0 specifying the speed and power needed we can choose the
motors.
F, -mg sinO= O =FA = mg sinO (2)
Assuming the power losses due to friction in chains and other
So, an estimation of the robot's mass is considered as components to be less than 50% of the maximum power,
m=30 kg. This estimation should be made conservatively 7.5 Nm output torque would be sufficient for each side.
because the whole calculations should be repeated if the robot Specifications of the selected motors are listed in Table I.
appeared to be heavier when it was completely designed:
Table I
> Ft = 163.25N Specifications of motors
Motor Torque Rotary Nominal Nominal Gear
Writing the moment equation with respect to the center of one specification speed Voltage Current ratio
of the wheels, the desired torque is determined as follows:
Locomotion 10 N.m 14.75 24 V 1.92 A 16:12
>M =Ia
(3)
motor Rpm

> T =F .r=9.5N.m Arms motor 1.5 N.m 120 Rpm 27.5V 1.5 A 1:30
aoc O
Arm's motor is coupled to the arm of the planetary gear-set
As mentioned in section 1.3, the locomotion system consists of with a 1:30 worm gear set, resulting in output torque of 30
two separate sides connected to the main body while a motor N.m (considering losses) and maximum speed of 4 rpm.
independently actuates the track at each side. Consequently,
the desired torque of each motor is obtained by dividing the IV. STRESS ANALYSIS
T, by2:
Tmotor 5 N.m
; The aim of such analysis is a minimal weight design, to reduce
on-board energy consumption. The results lead us to redesign
critical parts such that undesired deformation and stresses are
prevented. There are two methods available:

3985
Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:18:32 UTC from IEEE Xplore. Restrictions apply.
 1. Classic methods; The hardware of the control system consists various devices:
2. Numerical methods.
1. An operator desktop PC (A P4 2.4 MHz with 256
Using classic methods, the stresses in all gears and bearings
can be calculated and properly designed. Using this method is Mgb sdRam)
not applicable to some parts, e.g. the motor base structure or 2. An on-board Laptop (HP Model: nc4000 with 12.1"
the gear-set basis, where equations became really complex and LCD,Wireless LAN 802.11 a/b/g, 512 MB Ram,
unsolvable. In such cases, the finite element method using CPU: 1.6 Mobile)
Patran/Nastran is applied.
3. Micro controllers (mostly ATMega 128 with 53 I/Os,
For instance, the part shown in fig. 11, is considered as the arm 8 A/D, 2 serial ports, Machine frequency of 16 MHz,
for the planetary gear-set, considering the following roles 128 KB of program memory and 4 KB Ram),
defined for it:
- This part acts as the arm of the planetary gearbox. 4. Sensors: (pyrometer, potentiometer and cameras),
- It must stand all the force and moments applied to electronic circuits (for Interface, power and drivers),
robot's arm, which are transferred to this part via 9 lithium-polymer on-board batteries.
Gear 4 of planetary system. As shown in Fig. 13, the operator's desktop PC is used as the
- It should form the cover of planetary gearbox. master processor and the laptop installed on robot as the slave
By applying the loads and constrains and assuming that this processor. The robot is remotely operated using either
part is made of aluminum the result of analysis was obtained as keyboard or joysticks, by taking advantage of the wireless
shown in fig. 11. According to this analysis, the maximum connection between the computers via LAN-Cards.
pitting stress is 78 Mpa. By ignoring the effect of pitting stress,
the maximum stress is 35 Mpa which comparing to the yield
strength of aluminum is acceptable.
[Link] 2003 09-Oct-0 17 094:7
SC! ODEFAUJLT, Al 'Static Subcase' Stress Ternsor, -(NON-LAYERED)(VONM9) ,807lllll
Sti-,DEFAULT. Al Ztatlc Subcase. Displacemnents, Translational-(NOWLIA ERED) &.44+001l ll

Fig. 11. Results of FEM analysis for the arm

Following a complete stress analysis of all different parts, and
redesigning of those required, the last step is to fabricate the
designed parts and assemble those together with the available
standard elements to achieve the robot as shown in Fig. 12.

Fig. 13. Correlation of the main parts

The data acquired by sensors and video streams captured from
the two cameras are collected in the laptop, and then sent to
the operator's PC. Operator commands received by the laptop
Fig. 12. Resquake on a sample wooden step field are transmitted to the actuators through a micro processor
(ATMEGA128L AVR). Angular positioning sensor
Next, a brief review of the control system and performance (potentiometer) of each arm is also connected to the micro
characteristics of the robot will be discussed. processor where its data is processed in order to set the arms at
V. CONTROL SYSTEM the desirable position. Power circuits consisted of

3986
Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:18:32 UTC from IEEE Xplore. Restrictions apply.
 LM2576HV-ADJ switching regulators; distribute the voltage
proportionally to driver circuits and other electrical devices. It
should be mentioned that powered by the battery cells with a
full charge (4.2 volts, 10 A.H.), the whole system operates
more than 4 hours.
Resquake works as a tele-operative mobile robot with the
following capabilities:
1. Surmounting uneven terrains with a relatively high
stability as shown in Fig. 12;
2. Climbing stairs and slopes up to 350 as shown in
Fig. 14;
3. Searching unreachable environments within a
maximum distance of 50 meters from the operator Fig. 14. Resquake climbing up the stairs
station;
4. Broadcasting thermal and visual data from the area;
5. High reliability of hardware and electronic devices. REFERENCES
Other specifications of the robot are listed in Table II. [1] Adam Jacoff, Elena Messina, Brian A. Weiss, Satoshi
Tadakoro, Yuki Nakagava, "Test Arenas and
Performance Metrics for Urban Search and Rescue
Table II Robots", IEEE/RSJ International Conference on
Main specifications of Resquake Intelligent Robots and systems, 2003.
Overall weight 25 Kg [2] L. Matthies, Y. Xiong, R. Hogg, . Zhu, A. Rankin, B.
Length with expanded arms 80 cm Kennedy, "A Portable, Autonomous, Urban
Reconnaissance Robot', California Institute of
Length with closed arms 41 cm Technology, Jet Propulsion Laboratory, Pasadena
Minimum height 26 cm CA, 91109, 2000.
Width 40 cm [3] Mark Yim, David G. Duff and Kimon Roufas,
Maximum velocity 19 cm/sec "Modular Reconfigurable Robots, An Approach To
Urban Search and Rescue", Xerox Palo Alto Research
Arms maximum angular velocity 4 rpm Center, 2000.
Number of ball bearings 82
[4] A. Meghdari, S. H. Mahboobi, and A. L.
Number of plain bearings 32 Gaskarimahalle, "Dynamics modeling of "cedra"
Total number of parts 688 rescue robot on uneven terrains", ASME International
Mechanical Engineering Congress, 2004.
[5] Joseoh E. Shigley, and Charles R. Mischke,
VI. CONCLUSION "Mechanical Engineering design", Sixth edition,
McGraw-Hill, 2003.
This paper presented an illustrative description of the [6] George Henry Martin, "Kinematics and dynamics of
Resquake project at KNTU. Innovative mechanisms and Machines", 1917.
software-based steps of the design procedure were highlighted
here. First, a new special mechanism was developed to [7] Moosavian, S. Ali. A. and Papadopoulos, E.,
enhance the maneuverability of a mobile robot while trying to "Explicit Dynamics of Space Free-Flyers with
keep its dimensions relevant to the environment in which it Multiple Manipulators via SPACEMAPL," Journal
performs. After selecting suitable mechanisms, dimensions and of Advanced Robotics, Vol. 18, No. 2, pp 223-244,
parameters of the system were defined. The system dynamics March 2004.
and the sequence of stress analysis for different parts were [8] Rastegari, R., and Moosavian, S. Ali. A., "Multiple
addressed. Electronic devices and control system of the robot Impedance Control of Mobile Robotic Systems,"
was briefly described in the last section. The outcome is Proc. of the ISME Int. Conf. On Mechanical
Resquake, which is a mobile robot with great capabilities in Engineering, Isfahan, May 2005.
climbing obstacles in destructed areas, and was participated in [9] Reutlingen and Ulrich Fischer, "Metals handbook"
Rescue robot league of RoboCup 2005 in Osaka (Japan) where Third edition, Tarrah, 2004.
achieved the 2nd best design award.

3987
Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:18:32 UTC from IEEE Xplore. Restrictions apply.
 Proceedings ofISCIT2005

Design Features and Characteristics of a Rescue

Robot
Amon Tunwannarux and Supanunt Hirunyaphisutthikul
School of Engineering, The University of The Thai Chamber of Commerce
126/1 Vibhavadee-Rangsit Rd., Dindaeng, Bangkok Thailand 10400
Tel: +66-2-697-6713, Fax: +66-2-275-4892
Email: amon_tungutcc.[Link], supanunt hir@[Link]
Abstract - This paper presents a design and joystick and head mounted display like virtual reality head
implementation of a rescue robot. It has two front arms with set [1]. The other is the rescue robot from CEDRA team,
track mechanism and tracks link between front and rear Iran, the first runner up. Their robot has six wheels
wheels. With the double track robot system, it is good for locomotion, which can operate well in unstructured
unstructured landscape and able to climb over the pile of
collapse. This robot is equipped with a lot of sensors such as environment [2]. Every team tries to put more reliability
IR temperature sensors, distance sensors, odometer sensors, and special features for their rescue robots to get the
pitch/roll, compass sensor, three pan/tilt CCD cameras and maximum points in the contest. So we purpose the track
voice sensor. For simple and practical concepts, hardware type rescue robot, which has two front arms with tracks,
design is divided into control/locomotion subsystem and and tracks link between front and rear wheels. With this
monitoring subsystem. The RC airplane remote control and tracked dual rotating arms feature, it allows the robot to
receiver are modified to use in control/locomotion subsystem. easily to climb over the pile of collapse. We divided
In monitoring subsystem, three 8 bit CPUs are used for hardware design into two parts as control/locomotion
sensor information managing and sending it to serial port of subsystem and monitoring subsystem for simple, reliable
a computer wirelessly. The operator station not only shows
all sensors information but also plots the route of victim and practical concepts. The plots of victim accessing route
access and locates victim positions based on simple triangular and victim location are shown at operator station.
calculation technique. Furthermore, for looking over Moreover, this rescue robot has the mast, which is
obstacles to rmd the victims from the top view, a pan /tilt equipped with a pan / tilt camera and able to stretch to 125
camera is placed at the top of the mast which can stretch to cm. height for looking over obstacles to find the victims
125 cm. height. Audio signal from voice sensor is filtered for from the top view. The parametric equalizer is also
selecting only wanted frequencies by the parametric developed and included in the operator suitcase for
equalizer. This rescue robot design was implemented and filtering the unwanted frequencies from victim voice. For
worked excellent in the competition, especially it can climb other features, our robot was designed and implemented to
up the stairs with 15 cm. high steps to find the victims on the
s e c o n d f 1 o o r correspond with the rules of the competition to identify the
Keywords: rescue robot, double track, monitoring system, detailed victim information accurately and access all
mast, victim access route victims as fast as possible.
I. INTRODUCTION II. CRITERIA OF DESIGN

There are some developing of rescue robots for Our criteria of design a rescue robot is based on
commercial purpose but still in a small group of research Thailand Rescue Robot Championship Competition rules
and very expensive. A lot of rescue robot competitions 2004. They are the international rules, which will be used
have been held lately with the main purpose of in the Robocup Rescue Robot Championship 2005, held in
encouraging researchers to develop their robots for Osaka, Japan. The size of robot is limited to 70x70x70 cm3,
practical usage in the real situations.. The most popular total weight should not be more than 70 kg. The
rescue robot contest is the RoboCup Rescue Robot League competition rules and scoring metric both focus on the
Competition, which started in 2001. For rescue robots, the basic Urban Search and Rescue (USAR) tasks of
basic objective is usually used to survey and locate the identifying live victims, determining victim condition,
victims of the disasters. They take the places of human providing accurate victim location, and enabling victim
who are sent to the risky area for finding the victims. recovery, all without causing damage to the environment.
There are many types of rescue robots for example, the All teams compete in several missions (three different
Orpheus mobile robot from RoBmo team, Czech Republic, arenas) lasting twenty minutes with the winner achieving
the winner of Rescue Robot Championship 2003 the highest cumulative score from all missions. The details
Competition, is the rescue robot which is controlled by

0-7803-9538-7/05/$20.00©2005 IEEE 1047

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:18:45 UTC from IEEE Xplore. Restrictions apply.
 K
So POINTS POSSLE PER IC LM FOUt PENALTIES PER Ef by serial transmitter. We separately send the audio and
video signal via an analog transmitter. With this concept,
three communication parts were set up. Although this
VICrvA iCriT.M, VICTIM - [Link]
+ VICTIM - ZCTI 1
concept may not be the best for robot communications, but
4 QUALITM LOCTJN TAG SrTTAIrN STATE _ WMflN U!IW
Vaoflm
it is the most suitable way to implement and finish in time.
Fia + NPUse F scong
I The control/locomotion subsystem is described first,
Fig. I Performance metric used for scoring [3]. then the details of mechanics and all sensors will be
of the performance metric used for scoring are shown in illustrated. Finally, the monitoring subsystem will be
F i g . I [ 3 ] explained.
Because of our time constraint and the arenas in the IV. CONTROL/LOCOMOTION SUBSYSTEM
contest are placed by a lot of unpredictable obstacles, we
decided to create the non-autonomous rescue robot and use Before describing the remote control unit and all sensors
one person only to control remotely. In this case, our score in this robot, the details of mechanics are illustrated. Fig. 3
will be divided by 4, however we hope that the penalties specifies the various parts of the rescue robot. The
should be reduced and our score should be greater than the hardware is separately explained as the following items.
autonomous robot. All sensors are set up to correspond
with the rules and in order to get the maximum score. The A. Body and locomotion driving system
robot needs multiple sensors for map creation, victim The 3.5 mm. thick aluminum sheet is folded to be the
location, victim situation, victim states and sensors for base frame and the locomotion driving system with all
penalty discouragement. motors are placed in this frame in order to have the low
level center of gravity. Our robot dimension is 50x60x60
III. RESCUE ROBOT HARDWARE CONCEPT cm3. Two 24 V DC motors are used for driving two front
wheels by driving each side separately. Because we need
Our main design concept is using the simplest method to drive both moving tracks (linked between front wheel
but highly effective and reliable. This robotic system and rear wheel) and arm tracks, two shafts at same rotating
composes of the control/locomotion subsystem and center are designed, as shown in Fig. 4. The double arm
monitoring subsystem, see Fig. 2. For control part, the tracks are very useful for climbing over the pile of
versatile multifunction 9-channel transmitter for airplane, c o 1 1 a p s e s
FF9, is used to control all functions of this robot. For
locomotion part, we design the double track type rescue
robot with multiple sensors as mentioned before. And for
monitoring subsystem, we made the sensor information
managing board, which interfaces to all sensors. This
board will transform analog/digital signal from sensors
into formatted information and ready to send. All sensor
information will be sent wirelessly to a notebook computer
Control / locon otion Subsystem

WirelessPCPA Receiver
X
:2
10o<.__._.____
Remote
Transmitter
72.710 MI
r i
~_ Fig. 3 The robot hardware details.
+---------

ireless senal WI Vss Wireless senaI Senso nformation

Receiver 400 rHz Transmifter Managing Board
ii j
l |Monitor v ~~~~~~~Sensor For driving
ann tack &
(~)vd VoidAudi Wir Iess Video/AudioC0 n ve wheel tack
Receiver 12 Hz Receiver

Equalizer Monitoring Subsystem

Fig. 2 Block diagram of rescue robot.

Fig. 4 Track driving system and Robot arms driving system.

1048

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:18:45 UTC from IEEE Xplore. Restrictions apply.
 B. Robot arms driving system
These arms are created in order to raise its body up for
better climbing. Because the rotating center of the front
arm is the same position of the front wheel centers, so
II
||| Axle
Fn

| )

another hollow shaft is required to be placed between two

shafts of track driving system (same rotating center), in III

Fig. 4. Two 24V dc motors and chains for power

t r a n s m i s s i o n a r e u s e d.
ssfvS

Fig. 7 The brake system.

l _~~~~~~ow

Fig. 8 Operation of weight transfer unit.

E. Brake system
Fig. 5 Therobot head mechanism Fig. 6Channel details in FF79remote This mechanism is seriously required when robot stay
and sensors control and compass module
on ramp or is climbing steps, because the robot needs to
C . H e ad of r ob ot park at the ramp for surveying and stop for climbing over
This is a very important part because it is the place on each step. The brake system is installed at the shaft of rear
which many sensors are equipped such as two color CCD wheels because there is no room at the front. Fig. 7 shows
camera, distance sensor, temperature sensor, laser pointer, our brake system with spring coil and two servos.
voice sensor. Its mechanism provides the pan and tilt
ability by using two RC servomotors. The panning F. W e i g h t t r a n s f e r u n i t
mechanism is placed under the tilt mechanism because we The robot centers of gravity (C.G.) need to be closed to
will use the laser pointer at the top of robot head for the front as much as possible, so the bottom of the robot
distance measuring calculation, which will be discussed can be raised up and climb over the step. Because the quite
later. The mechanism of robot head is shown on Fig. 5. high step (higher than diameter of the front wheels) and
the moment from brake system and camera mast, we need
D. The camera mast to install the weight transfer unit at the front in order to
As the obstruction in the competition field, a lot of 80 move its C.G. pass over the edge of step (fulcrum point).
cm. high partitions were placed for simulating the Thus a battery, 12V 7AH, 2.5 kg. and the sheet of 3 kg.
situations. The distance between partitions is quite narrow lead are attached in order to overcome this step as shown
(max =80 cm.), so we need to look over these partitions i n F i g . 8
for finding victims and for better-controlled movement
and not to touch the partitions. Thus the mast added on at G. Remote control unit
the rear part of robot body giving the bird's-eye view is The digital proportional radio control, T9CAP, FF9
necessary. This mast is modified from car antenna and Futaba is selected for controlling the robot because its high
equipped with a color CCD small camera and small reliability even the receiver is in the concrete area.
pan/tilt mechanism at the top of mast. It works very well Moreover there are up to nine channels and has many
because it can stretch up to 125 cm. high with lightweight operational functions in each channel and between
m e c h a n i s m channels such as inverting, mixing, end point setting, sub-
trim of position, fail safe, etc [4]. All channels are
assigned and used according to the channel controllable
ability and ergonomic. The details are in Fig. 6.

1049

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:18:45 UTC from IEEE Xplore. Restrictions apply.
 H Sensors compensated compass/magnetometer system. This module
The infrared (IR) temperature sensor is chosen to use not only outputs the azimuth but sends the dual axis tilt of
for measuring the victim temperatures remotely and pitch and roll and temperature data over the standard RS-
contactlessly. We use the Raytek MID module [5], which 232 interface [7]. Due to a strong metal object or magnetic
has the important specifications to be concerned as anomaly affects to this module so the module should be
- Optical Resolution = 10: 1 provides the ability to placed as far as possible from the source of that anomaly.
measure temperature of the pointed area. In this case we installed it at the top of robot body and
- The miniature sensing head can be separated from main placed in the plastic box, see Fig. 6.
unit and has small size and lightweight (see Fig. 5).
For visual system, two color CCD cameras are installed IV. MONITORING SUBSYSTEM
on the robot head. The I' one is for the front view and the
2nd is for the rear view as in Fig. 5. Additional color CCD A. Sensor infornation managing board
camera is installed at the top of mast as in Fig. 6. From block diagram in Fig. 2, there are two
As the voice sensor, a small conventional wide range communication paths in this subsystem. Video/audio
condenser microphone is equipped at robot head. It is signal are separated from all sensors and sent by 1 W, 1.2
installed at the center of parabolic cone for voice focusing GHz. analog transmitter. The rest of sensor signals used
as in Fig. 5. The sensing voice is sent together with the the 400 MHz. serial communication for sending all sensor
selected video signal via the 1.2GHz. 1 W wireless information back to operator. This "sensors information
audio/video transmitter. At the remote operator site, this managing board ", in Fig.9, collects all sensor signals and
voice will pass via the parametric equalizer for selecting transforms to formatted information. Three MCS-51, 8 bit
only the wanted frequencies. CPUs are used in this board. The CPUI, main CPU,
The infrared distance sensor, Sharp GP2YOA02YK, manages key switches, LCD display, wireless serial
with the detection range 20-150 cm is used to measure the transmitter and all sensors except odometer sensors
distance from the robot head to the pointed object (see Fig. handled by CPU2 and pan/tilt angle sensors handled by
5). It has less influence on the colors of reflected objects CPU3. All sensor information we get from this board is
and their reflectivity, due to optical triangle measuring sent wirelessly to serial port of a computer at the operator
m e t h o d [ 6 ] s i t e
Our odometer sensor is comprised of a disc with holes B. Operator monitoring software
along its perimeter and the infrared transmitter and At operator station, the operator monitoring software
receiver module. Two sets of this sensor are installed at was developed with visual basic in order to process the
each side of the shaft which drives the front wheels. The data from the sensor information managing board and plot
data of each wheel sensor can be used to plot the victim the victim access route and its location. The number of
a c c e s s r o u t e pulses counted by odometer sensors and the azimuth
Arm angle sensors are necessary for remote controlling. angles from compass sensor are used to plot the victim
This information is sent to remote computer for robot access route as shown in Fig. 10. However because of track
graphic creation. A 10 turns potentiometer resister is use wheels and movement slipping, the moving route of robot
as sensor for each side of arm. on the map may not be accurate. So we need to update the
The EZ-Compass-3 module is used for robot navigation. robot position from time to time. In the competition, we
It is a low cost advanced pitch/roll have the arena map and landmarks such as poles or stair
steps for updating robot position when we lost. At the top
of robot head, a laser pointer is attached for marking
center of camera and IR temperature sensor. Moreover we
can use it for distance calculation by simple triangular
method. Fig. 11(a) shows a side view of calculating
method. If we tilt and point the laser dot to the known
position, pole Ml, the distance from pole MI to robot is d
= h/tan(T). Then the coordinate of robot can be obtained
by x=xMi+dlcos(AJ+ Pp), y=yM -dlsin(AI+PB). In Fig.1 1
(b), the victim position can be found at xv1=x-d2cos(A -
PI), yv1=y-d2sin(A1 - P,). The arena map, the calculated
information and all sensor data are displayed on the
notebook computer and furthermore, audio/video signal
from robot will be monitored at the suitcase as shown in
Fig. 9 Block diagram of sensor information managing board. F i g 1 0

1050

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:18:45 UTC from IEEE Xplore. Restrictions apply.
 I
i.
-1 iA.
4"

ivI
, __ FU4"bJK.1
Wq-I--
j47IEIKV.,
; - LI
1 I- .
,_'_
I

I.. ^i 1.... ..... Fig. 12 The rescue robot in the arenas of the competition.
VI. FUTURE WORK
Fig. 10 Plotting of victim access route and locating the victim position on
the notebook computer, and A/V Suitcase at operator station. This is our first version for RoboCup competition. Our
robot can run with speed only 0.25 meters /second and not
practical in real situations. Many capabilities should be
Pole M1 d=
, wF
T T = TRIt angle enhanced, such as the dust and water spraying proof body,
the thermal resist wheel tracks, lightweight body material,
d. 9 using Ni-Cad instead of lead acid battery, using higher
efficiency dc motors, the controllable light filter for
P Pan angle
camera. All wireless communications should be integrated
to WiFi IEEE802.11 and the robot side view locomotion
a) Side View b) Top View graphics should be added.
Fig. 11 Self-locating and victim locating by triangular calculation. VII. CONCLUSIONS

V. TESTING RESULTS AND DiscussioNs The rescue robotics system with its arm track and wheel
track mechanism has been briefly described. The remote
A. Mechanic and movement testing monitoring system of robot was provided to manage a lot
This rescue robot works very well in all three arenas of of sensor data for victim identification, localization and
the competition. The camera on the top of 125-cm. high navigation. The robotic system has been tested in many
mast is used for looking over the 80-cm. high partitions areas and competitions. Its performance was observed to
and accurately mark the locations of victims in the first be excellent in unstructured environments and successfully
area. In the second arena, it can climb the 15-cm. high climb over the 15cm. high steps. Finally, it did make very
steps to find the victims with climbing up speed 7 steps in good score and got the I" runner up award with 130,000
2 minutes. At the most difficult arena, it climbs over the Baht prize from Thailand Rescue Robot Championship
pile of collapse well, see Fig. 12. The controller should be 2 0 0 4
more careful with obstacles tie or stuck in the wheel tracks
or fall over the robot. REFERENCES
[1] Ludek Zalud, "RoBrno Czech Republic", RoboCup Rescue
B. Sensor testing Robot League Competition, Italy, July 2003.
The obtained results have been satisfactory for all [2] Meghdari,etc.,"CEDRA Iran",RoboCup Rescue Robot
sensors. The IR temperature sensor can accurately measure League Competition, Italy, July 2003.
the temperature of victim. The data from compass sensor [3] RoboCub Rescue League,"USAR Robot Competition Rules
is quite fluctuate, but the programmed software can filter 2004", [Link]/projects/USAR /[Link]
this ripple and get quite satisfied results. The victim access [4] Futaba Corp., "Manual of T9CAP R/C system", 2004
routes are plotted with some errors but not more than 50 [5] Raytek Corp., "Thermalert manual", [Link]
[6] Sharp Corp., "Distance measuring sensors notes",2003.
cm./500 cm moving. With self-locating update and the [7] AOS Inc., "EZ Compass-3 manual", [Link]
route plotting, we succeed in the victim position locating.
Fortunately, the arena grid is 50x50 cm2, so we never
make a mistake in specifying the victim position.
Three cameras work excellent, especially in the area
beneath the stairs or in the low illumination places.
However the video picture will be so pale or too much
white when the robot works in the bright light area.

1051

Authorized licensed use limited to: St Petersburg Natl Uni of Info Tech Mech & Optics. Downloaded on January 20,2025 at 00:18:45 UTC from IEEE Xplore. Restrictions apply.