Technion Aerial Systems 2016

Journal Paper for AUVSI Student UAS Competition

The Technion – Israel Institute of Technology
Faculty of Aerospace Engineering
Faculty of Electrical Engineering

!!!‫תמונה זמנית‬

Figure 1: Athene

Abstract
This journal paper presents the design and development process of the Technion Aerial System (TAS) for the AUVSI
Student UAS Competition 2016. The team consists of undergraduate students from both the Faculty of Aerospace
Engineering and Electrical Engineering. The team has focused on improving and surpassing the capabilities of the
system presented last year. This year, the group presents the Athene system, a wing- fuselage plane with mounted
double electric-powered engines, improved aerodynamics, and avionics capable of performing autonomous flight. The
system is a result of intensive design, analysis, simulations and ground and flight tests. The system features innovative
use of image processing using GPGPU technology, 3D printing, CFD calculations, customized UI, a unique sense and
avoid algorithm and more.

-1-
TAS – Technion Aerial Systems Team
Table of Contents
1 System Engineering Approach......................................................................................................................... - 3 -
1.1 Mission Requirements Analysis .............................................................................................................. - 3 -
1.2 Design Rationale ..................................................................................................................................... - 3 -
1.3 Expected Task Performance .................................................................................................................... - 4 -
1.4 Programmatic Risks and Mitigation Methods ......................................................................................... - 4 -
2 UAS Design ..................................................................................................................................................... - 5 -
2.1 Aerodynamics ......................................................................................................................................... - 5 -
2.2 Airframe .................................................................................................................................................. - 5 -
2.3 Wing Design and Analysis ...................................................................................................................... - 6 -
2.4 Propulsion System................................................................................................................................... - 7 -
2.5 Autopilot (AP) & Flight Control System ................................................................................................ - 7 -
2.6 Payload System ....................................................................................................................................... - 8 -
2.7 Ground Control Station (GCS) .............................................................................................................. - 10 -
2.8 Data Link .............................................................................................................................................. - 11 -
2.9 Development Process of Specific Mission Tasks .................................................................................. - 12 -
2.10 Mission Planning................................................................................................................................... - 15 -
2.11 System Integration ................................................................................................................................ - 15 -
3 Test and Evaluation Results ........................................................................................................................... - 16 -
3.1 Aircraft Performance and Airworthiness .............................................................................................. - 16 -
3.2 Autopilot System Performance ............................................................................................................. - 17 -
3.3 Mission Task Performance .................................................................................................................... - 17 -
3.4 Payload System Performance ................................................................................................................ - 18 -
4 Safety Considerations .................................................................................................................................... - 20 -
4.1 Operational Safety Criteria.................................................................................................................... - 20 -
4.2 Design Safety Criteria ........................................................................................................................... - 20 -
4.3 Safety Risks and Mitigation Methods ................................................................................................... - 20 -
5 Conclusions .................................................................................................................................................... - 20 -

-2-
TAS – Technion Aerial Systems Team
1 System Engineering Approach
1.1 Mission Requirements Analysis
Athene was designed, developed and tested to accomplish all mission requirements while following the rules and time
limits, and minimizing development risks.

While taking into consideration the existence of the 2015 Tyto system, the team focused on the examination of the
system based on 2016 mission requirements and 2015 results and review.
To collect and gather crucial data and information regarding the system, the team implemented a “Red team”
approach.
A red team is an independent group that evaluates and challenges the last year Tyto system to identify its weak points
and improve its effectiveness. The "red team" sub-groups were assigned to investigate different aspects of the system:
aerodynamics, structure, propulsion, image processing and communication. The resulting work was presented in PDR
(Preliminary Design Review) and CDR (Critical Design Review) and the shape and configuration of the 2016 Athene
System was frozen.

Missions and tasks were prioritized according to the team’s capabilities and resources.
The factors determining which task to tackle were as followed:
 Grade- tasks were arranged according to the grade scale of the competition. The higher the grade - the higher
the priority.
 Previous Experience- Experience from last year’s competition and last years’ expertise was considered.
 Budget- Due to the project’s nature and the fact that it was mostly funded by donations, hardware component
costs were evaluated and selected in accordance.

Tasks were divided to two categories:

Will Accomplish- Capabilities developed and tested by last year’s team, including autonomous flight, search area,
actionable intelligence, emergent target and interoperability.
Will Attempt- Capabilities in various development stages, including: off axis target, ADLC, SRIC and SDA.

1.2 Design Rationale

A thorough investigation of the 2014- 2015 systems, combined with 2016 mission analysis has yielded several design
focus points:
 Reduction of development risks to maintain the main characteristics while further improving the
aerodynamics, structure and payload operation.
 Implementing methods of design and manufacturing similar to those used in the Aeronautical industry.
 Vast use of composite materials as a way to reduce weight and preserve structural strength.
 Improved image processing by replacing the camera.
 Introducing onboard image processing to reduce system response time.
 Closed loop gimbal system to improve off-axis target tracking while reducing the error to a minimum.
 New algorithm for the SDA and interoperability was developed and integrated into the system.

These focus points were addressed by the team through several changes in comparison to last year:
o Improved aerodynamics – Forward swept tapered wing with enlarged wingspan and aspect ratio. A new
gimbal camera turret with minimal drag was designed.
o Higher durability – While maintaining the durability and strength of the fuselage, the design and
manufacture of a hollow wing comprised of composite materials was implemented.
o On Board Image processing – A dedicated onboard computer performs the ADLC task on board the plane.
o New camera – A new camera was selected after a thorough testing process.
o New Antennas – New antennas were selected for imagery downlink and SRIC missions.
o Mapping GUI – Search area coverage was tracked using a custom mapping GUI.
o Software optimization – A complete re-writing of the algorithms to C++ enabled full operation directly
from the ground station while using mission planner as the only platform of operation.
o Improved airdrop mechanism – The airdrop mechanism was simplified to minimize odds of malfunction.
The mechanism was also redesigned to adapt for the 2016 mission change.
o Enhanced flight management software – Software modification and customization to the mission planner
to better handle the interoperability and obstacle avoidance tasks. An introduction of a GUI.
o Safety – Key emphasis was placed on safety through all design and development stages, resulting in
methodical, checklist-based, and operational procedures in addition to redundant safety mechanisms.
-3-
TAS – Technion Aerial Systems Team
 1.3 Expected Task Performance
Based on the experience of the past two years and the new competition standards, we conducted some aerodynamic,
structural, payload and other system changes to achieve a greater improvement in performing the competition’s
tasks – for instance, redesigning the wings improves the UAV performance and navigation area scan. Adding a
gimbal controller helps stabilize the camera during the flight and off-axis mission. Furthermore, the new air drop
mechanism features more reliable characteristics. Significant improvement was achieved using innovative image
processing algorithms. Custom mapping GUI was designed to track the search area. Antennas were redesigned and
relocated to increase communication bandwidth and robustness. A new algorithm was developed to perform
obstacle avoidance automatically. In addition, the Mission Planner GUI was upgraded to feature more robust and
efficient controls for the interior pilots. The Athene team is ready to perform all the main and secondary tasks.

1.4 Programmatic Risks and Mitigation Methods

During the beginning of the development process, we made a programmatic risk assessment. As this will be our third
time in the competition, we based our assumptions on the experience of the previous systems. During the year, we
updated the risk assessment table according to the problems we encountered.

Table 1: Risk methods

Risk Factor Description Impact Likelihood Mitigation Method

Design and/or Delays in the new, challenging wing High Medium o Use of 2015 system as a backup vehicle
manufacturing design or construction may shorten the to accumulate flight hours and improve
delays available time to fly, check and team practice.
evaluate the system performance.
Insufficient crew Schedule delays might result with Medi High o Early operation of the sub-system
training insufficient flight time of the system um component on lab and ground tests
for both debugging the system o Initial team training was done using
performance and training the team. smaller R/C Models and last year's
system.
Integration issues Since the system was developed by High Medium o Weekly design review meetings were
and cooperation of students from 2 Faculties - Aeronautics conducted.
two faculties and Electrical Engineering, it may o A single faculty advisor led the two
cause contradicting requirements and faculties combined team.
solutions which may endanger the sub- o Tasks assignments were not necessarily
system and system integration. assigned as per the faculty expertise.
Legal and safety The system should and must comply High Low o A dedicated team member assigned "as
issues with the competition rules. a lawyer" to be in charge of addressing
any legal issues and system safety.
o A thorough inspection of the rules and
regulations was made before airframe
design.
Crash during testing Air vehicle crash during evaluation High Medium o The 2015 aircraft was maintained as a
and system flight tests may endanger backup solution for training.
the team's ability to participate in the o Two air vehicles were constructed and
competition. tested to provide redundancy.

-4-
TAS – Technion Aerial Systems Team
2 UAS Design
2.1 Aerodynamics
In preparation for SUAS competition 2015 - TYTO’s aerodynamic
components were evaluated using a "red team" approach involving
analytic and numerical methods. The team tested and evaluated the
airfoils, wing planform, tail configuration, etc. This year, we
focused on optimizing the wing planform to maximize the aircraft's
flight performance in minimal development time.
The main design objectives involved in the design process were
𝐿
increasing , decreasing 𝑉𝑠𝑡𝑎𝑙𝑙 and simplifying production
𝐷
time. The team considered several planform options that Figure 2: Wing-Tail VLM Simulation (XFLR5)
feature better characteristics for the given flight
conditions. The options were evaluated and compared using self-develop multidisciplinary flight mechanic simulator
and a commercial 3D panel program (XFLR5). Last year, TYTO's wing was composed of the geometrical shapes
which hampered construction and
increased production costs. This year, a
tapered wing planform was chosen, e.g.
a trapezoidal wing. The new developed
wing features more endurance time, up
to 10 minutes more. In figure 3, it can
be seen that the wing L/D ratio was
increased by 25%. Furthermore, we
decreased the stall speed by 2-3 knots
that will reflect significantly in the
competition tasks. These improvements
were achieved by increasing the AR by
50% and increasing the wing area by
7%.
Taking into consideration the new wing
planform, we made some adjustments
in the aileron design. After several Figure 3: Wing Planform Evaluation
analyses, we received an aileron
pivoted at about 75% of the wing chord, length about 40% of half the wetted wingspan, and width about 20% of the
local chord. Following the new aileron design and reduction of wet surface, we reduced the aileron drag.
Furthermore, due to the larger camera, a new turret was designed. The design was focused on reducing pressure drag
using Sears-Haack body. Due to stability issues caused by the new turret, it was necessary to increase the horizontal
tail wet area by a dorsal fin. The fin also improves the UAS maneuverability.

2.2 Airframe
The airframe is comprised of five segments: the L/R wing, fuselage, tail assembly and top cover of the fuselage.
All segments were
designed to be
detachable and
allow for quick
access to the inner
components, easy
transportation and
maintenance.
The structure of the Figure 4: System inner
Athena system is components
made out of
lightweight
composite material (Graphite/Epoxy) and fabricated using methods widely used in the aviation industry. Fuselage
durability was achieved by tailoring the layers of carbon fiber in a direction to suit the stress the fuselage would
experience during flight. To reinforce the open cross section of the fuselage structure, we implemented four
longerons made of unidirectional graphite fiber and Rohacell core along the four corners of the fuselage cross
section. The manufacturing process also included designing a two-part mold for the fuselage, nose and wings using
-5-
TAS – Technion Aerial Systems Team
CAD (Computer Aided Design) software and milling from a solid piece of MDF wood using a CNC (Computer
Numerical Control) machine.
2.3 Wing Design and Analysis
In the Athena system, we presented a major change and improvement regarding the wing planform and structure.
The Athena wing features a hollow structure made of a carbon
fiber skin 1.6 mm thick in a layered structure of Gr/Ep and balsa
wood. The main spar is a U beam made of carbon fiber and epoxy
matrix with unidirectional fibers in the flanges for improved
endurance to bending moment and Rohacell foam core to increase
the beam moment of inertia. The aft spar is a U beam comprised of
195 carbon fiber in an epoxy matrix. Both the skins and the spars
were made as individual parts using MDF molds designed and
manufactured using CAD/CAM (Computer Aided Design /
Computer Aided Manufacturing) techniques. All parts were then
glued together using 460 structural adhesive. The wings were Figure 5: Wing skin & spars inner structure
connected to each other using a carbon fiber sleeve and tube
system. Both wings were attached to the fuselage through 3 mm thick carbon fiber lugs and carbon fiber tubes, and
aligned with a torque screw.
The current wing presents an improvement in
weight and strength in comparison to the
2015 Tyto wing. Careful analysis of flight
conditions and expected performance of the
airframe was performed to design a wing
strong enough to withstand predicted loads,
while keeping wing structure as light as
possible. The analysis of lift distribution and Figure 6: Ansys Analysis
subsequent loads were calculated using
analytical methods and validated using commercial Vortex Lattice Method (VLM) and Finite Element Method
(FEM) programs. In order to validate the structural strength and to make sure the wings will support the predicted

Figure 7: Athene Views

loads prior to production, we generated a Final Element model of the wing using a solid model of the wing and
considering the known material properties used to construct the wing. The model was tested and loaded in extreme
flight conditions on the edge of the flight envelope. The results were consistent with those acquired by a static wing
load test carried out postproduction. The static load test simulated the flight conditions in 3.5G with 45 Kg of
sandbags distributed atop the wing. The results correspond with the data calculated prior to production and
confirmed system airworthiness.

-6-
TAS – Technion Aerial Systems Team
 Table 2: Air vehicle performance and characteristics

Main Wing Vertical Stabilizer Horizontal Stabilizer Aircraft Dimensions

Airfoil Douglas Airfoil NACA Airfoil NACA 0012 Length 2.09 m
LA203A 0012
Span 2.9 m Span 0.35 m Span 0.75 m Width 2.9 m
Area 0.708 𝑚2 Area 0.077 𝑚2 Area 0.165 𝑚2 Height 0.686 m
Aspect 11.85 Aspect Ratio 1.75 Aspect 3.4 Weight 13 Kg
Ratio Ratio
Figures of merit Propulsion System Velocity Endurance 30min
Wing 15.5 Motor power 1650 W Stall 23 Kts Range 16.7 nmi
Loading Kg/𝑚2 Speed
Power 150 W/Kg Prop. Size 17X10 Cruise 32 Kts Rate Of Climb 600 ft/min
Loading Speed
Max Load 3.5 Batteries 14.7 V 27 Max 42 Kts Minimal turn 60 ft
Factor Ah Speed radius
2.4 Propulsion System
Following thorough evaluation of the last year Tyto propulsion system by implementing a "red-team" approach, we
chose to continue with the same propulsion system: a twin electrical motor configuration, i.e. one motor on each
wing. This configuration allowed us improved redundancy and enabled us to locate the camera turret (the main
component) in the nose. Furthermore, we focused this year on reducing the power consumption by means of
software, such as maintaining leveled flight and avoidance of accelerations and maneuvers during the competition
scenarios.
This year, we also focused on reducing the weight of the batteries by acquiring batteries with higher power capacity
per cell. Thus, we were able to maintain leveled flight of 30 minutes with a lighter set of batteries.
2.5 Autopilot (AP) & Flight Control System
This year, after
accumulating many flight
hours with last year -
Tyto’s AP performance, it Figure 8: Flight Control System
was decided to keep the
current AP system,
Pixhawk. The Pixhawk
provides fully autonomous
capabilities to the aircraft,
while maximizing safety
and minimizing
development risks. During
the evaluation, we took
into consideration the
following Pixhawk profits:

o Pilot manual override – possible at all stages of flight, directly from the pilot's radio controller.
o Easy sensor integration – "plug & play" air data, altitude & heading reference system (ADAHRS), navigation
and propulsion monitoring sensors.
o Two-way long-range radio communication capable of delivering real time telemetry and receiving flight plan
and commands from the ground operator.
o Dedicated ground control application.
o Vast functional capabilities – waypoint navigation, midair re-tasking, autonomous takeoff & landing and
designated mission scripting.
o Small and lightweight airborne hardware, with low power consumption and no special cooling requirements.
o Affordable overall solution

-7-
TAS – Technion Aerial Systems Team
 Figure 9: Payload system architecture
2.6 Payload System
Last year's performance review revealed the need
for several improvements in the imaging payload
system. To this end, a new camera survey and a
series of field tests were conducted. The survey
showed that the Sony a6000 camera optics,
tethering capabilities and price has the best
performance and, therefore, was selected. The
Sony a6000 has a 24-megapixel APS-C CMOS
sensor and uses the Sony 16-50 F3.5-5.6 OSS
lens, which provides 3.1x zoom and focal range
of 24-75mm (35mm sensor equivalent). The new
camera is attached to a newly designed 2 axis
gimbal and is connected to the on-board
computer (OBC) via USB cable. Real time
positioning and orientation of the camera was Figure 10: Athena’s camera and gimbal
tracked using a dedicated AHRS. A custom
circuit uses the camera’s flash signal to achieve precise synchronization between images and the AHRS readings.
Precise camera stabilization and control are key features for meaningful image acquisition and target tracking. To this

Figure 11: Wiring scheme system

end, a closed loop system was designed for controlling the camera gimbal. The STorM32 control board has been
selected for its ease of configuration, relatively widespread use, built-in Pixhawk support and low price. The STorM32

-8-
TAS – Technion Aerial Systems Team
board utilizes separate PID controllers for each rotation axis and an IMU attached to the camera, ensuring outstanding
spatial orientation control. STorM32 is connected to the autopilot by Mavlink, a two-way communication protocol
that provides the ground station with the means to control the gimbal. Although preserving its basic principle of
operation, the turret assembly had to be redesigned in order to accommodate the new camera and control system. The
adjustments consisted of increasing the size of the camera pod and balancing the loads and integration of stronger
brushless motors. A dedicated onboard computer (OBC), NVIDIA Jetson TK1, was introduced to handle image
processing tasks. An ODROID XU3 OBC handles camera control and communication with the imagery console. This
setup ensures quality and reliability of image acquisition and processing and minimizes the risk of performance loss.
System wiring was redesigned in accordance with the above modifications. The new wiring scheme provides an
excellent tool for health monitoring and ensuring safe operation and testing of the platform.
o Camera Selection
Assessment of previous year’s camera (Canon S110) performance revealed the following problems:

 Low tethering reliability inhibited the ability to change camera settings (i.e. zoom) during flight.
 Image quality was not good enough for decoding the QR code.
 Slow frame-rate (<1 FPS) limited search area coverage.

Due to the above problems, we decided to use the Sony Alpha 6000 camera. In the following section, we describe in
detail our camera selection process.

 Optical requirements
There are three missions in which the camera plays a significant role – Target recognition, QR code and OFF-AXIS
target recognition. Each of these missions has different optical requirements, detailed in the competition rules. Based

Figure 12: Camera Comparison

on these requirements, calculations were made to determine the minimal optical requirements for the camera. For
example, each pixel in the QR code is 2 squared inch. Based on this, the following graphs were calculated for
different camera types (assuming an ideal lens) at small focal length (wide FOV). According to the graphs above,
many consumer cameras can perform the mission. In reality, most cameras we tested could not produce a decodable
image of QR code from 200 ft. because the actual resolution of a camera is limited by the camera lens, and not by
the resolution of its sensor.

 Camera comparison tests

Although optical requirements were calculated, many lens
manufacturers do not publish MTF charts for their product
(especially for low-end cameras like P&S and cellphone-
cameras). Thus, several cameras of various types had to be tested
and compared.

Below are some of the results for several cameras that the team
tested in various flight and ground experiments. Note that the QR
code was cropped out of the images. Figure 13: QR Results

The cameras used (from left to right):

-9-
TAS – Technion Aerial Systems Team
  Sony Alpha 6000 with its kit lens 16-50mm. The image was taken at 16mm from a flight altitude of 200ft.
 Samsung Galaxy Note 4 camera – taken from a ground distance of 200ft.
 Samsung Galaxy Camera – taken at x1.2 zoom from a flight altitude of 200ft.

 Summary
The Sony Alpha 6000 mirrorless camera was chosen, along with its 16-50 kit lens which can give sharp images of
the QR code from 200ft and has an electronically controlled zoom. The following table compares this year’s camera
with last year’s camera:

Table 3: Camera comprasion

Parameter\Camera Canon S110 Sony Alpha 6000 with 16-50 lens

Camera Type Point and Shoot Mirrorless
Resolution 12MP 24MP
Optical Quality* Low Medium
Focal length (35mm eqiv.) 24-120mm (x5 zoom)** 24-75 (x3.1 zoom)
Pixel length (um)*** 2 4
Max capturing speed (fps) 0.7 11
Dimensions/weight 99[mm]x58[mm]x28[mm]/173[gr] 120[mm]x67[mm]x45[mm]/468[gr]
Interface CHDK Sony PTP
Price ($) 350 700
Missions it can perform Targets Targes, QR code, Off-axis, Mapping
*Based on our tests of QR code recognition from 200ft.
**As explained above, the zoom was not usable. ***Larger pixel =
less noise.
Fla Fl Image time
o Camera Interface Image
The official remote API released by Sony uses only the cameras
I D
WIFI adapter. Using the WIFI adapter would result in very slow Image + Data Data
FPS and complicate the RF design as it can interfere with the
FUTUBA antenna. By the use of packet sniffing on the USB
adapter of the camera and reverse engineering techniques, we were Figure 14: Camera Interface Diagram
able to write a custom PTP driver for the Sony over the USB adapter. Using this driver, we were able to reach 2 FPS
(X3 improvement over last year’s camera).

In order to considerably improve the accuracy of target geolocation and mapping algorithms, we developed a unique
system that synchronizes captured images with positioning data collected from the GPS, AHRS and compass:

1. An Arduino connected to the camera’s flash hotshoe records the exact time of the shot.
2. The OBC receives the time of the shot from the Arduino.
3. The OBC receives telemetry data from the pixhawk and camera AHRS, and syncs it with the image shot
time.

In this method, each time the camera takes a picture, the OBC attaches the data taken at the closest time to the shot
to the photograph. Timing error using this method is between 10-60 milliseconds.

2.7 Ground Control Station (GCS)

The GCS includes two main systems independent of each other – the Flight Management Console (FMC) and the
Imagery Console (IC). Both consoles are placed next to each other and use different communication protocols to
communicate with the UAV. Each system has been designed to allow a highly efficient and user-friendly operation.
The different systems control different aspects of the mission and communicate with each other through the GCS
commander. In addition, the GCS commander also maintains constant communication with the safety pilot using a
hand-held radio.

o Flight Management Console (FMC)

- 10 -
TAS – Technion Aerial Systems Team
The FMC allows the flight controllers to monitors the current UAV status as well as communicate different tasks to
the UAV. The FMC runs the command and control software (APM Mission Planner), which has been modified by
the team to perform several tasks which are not native to the platform:
 Interoperability task - Communication with the judges’ server.
 Mission Display GUI – Display of the current mission status, UAV telemetry and flight log in accordance
with competition rules.
 SDA task – Algorithm for obstacle avoidance integrated into the software.
 General changes – Minor changes to allow the flight operators a better user experience.
As a measure of safety, a second computer is preloaded with the same software and missions as the FMC. The
second computer is also used to decrease workload on the main operator by allowing paralleling mission planning.

o Imagery Console (IC)

Last year the Technion team developed a custom Imagery
Console (IC) from scratch. This enabled designing a user
interface (UI) tailored to fit the competition tasks. This year
the IC was further adapted to handle the higher image
resolution and FPS of the new Sony camera. A new mapping
GUI was added for assuring the coverage of the search area,
and a separate GUI was added to handle the ADLC task. Key
features of the IC:
Figure 15: Imagery Console
 Intuitive user interface.
 Full control of camera parameters.
 Easy access to geolocation data.
 Easily scales to multiple
users/workstations.

An important feature of the IC is

the mapping console, which builds
a real-time map of the terrain below
the UAV. With this console, the
pilot can assess the real coverage of
the search area, and change the
UAV route in order to locate more
targets. Figure 16: GUI Interface
o Safety Pilot
Integrated within the UAV is a separated communication module capable of bypassing the autopilot. This
communication module allows the safety pilot to take complete control over the UAV in any case of emergency.

2.8 Data Link

This year we focused our efforts on improving the
communication between the plane and the ground
station. To this end, we used RF analysis, simulations
and ground and flight tests.

In accordance with the system from previous years, it

was decided to keep the separation between
communication systems to four sub-systems. This was Figure 17: Antennas architecture
done to achieve a more robust solution and improve the
design process. The sub-systems were divided by the requirements of the tasks: autopilot control link (900 [MHz]),
manual control link for the safety-pilot (2.4 [GHz] - should be active only for very short periods), main imagery
data-link (5.8 [GHz]) and the link for the SRIC mission (2.4 [GHz]). The first two sub-systems require much smaller
bandwidth than the others require, and performed successfully last year. Thus, it was decided to focus on the other
two subsystems.

- 11 -
TAS – Technion Aerial Systems Team
 o Main Imagery Data Link
 Requirements
The main imagery data link was used for online downloading of pictures during the search area and off-axis tasks,
allowing the operators to manually inspect for targets. Moreover, this link was used for controlling the on-board
camera and monitoring the OBCs. The data link must be able to work well with the graphite-epoxy fuselage, remain
robust during the aircraft maneuvers and have a broad enough bandwidth for smooth operation. Thus, it was decided
to use the 5.8 [GHz] band with the intention to reduce interference and allow a good bandwidth.

 Transceivers
The chosen transceivers, ‘UBIQUITI Bullet M5-HP’ and ‘UBIQUITI NanoStation M5’ were chosen for four main
reasons: past experience, fit required band, standard connection compatible with many antennas, and use of a
specialized, stripped down Linux distribution. The last reason, together with a built-in spectrum analyzer, enabled
easy development of inspection tools. Through ground and aerial tests, the chosen transceivers worked well together
and achieved long-range stable signal.

 Chosen Antenna
The search area task requires high transmission rate at long range. Through research, simulations and testing in an
𝜆
antenna range, we concluded that the best data link would be created by a monopole antenna. This decision was
4
later verified thoroughly in flight tests.

 Automated Antenna Tracker

This year it was decided to use an automatic ground antenna
tracker. The reasons included risk of RF radiation, low accuracy of
manual tracking, and waste of flight line crew member. The
majority of tracking systems were designed for FPV (First Person
View) flights and use additional onboard components. To avoid
unnecessary components, the antenna tracker was reverse-
engineered to enable ground control from the ground station. We
chose the ‘My Fly Dream’ tracking system since it has a very short
response time and good accuracy.

o SRIC Link
 Requirements
The link system should be compatible with the judges' channel at
2.412 [GHz]. The judges' antenna uses a vertically linear polarization Figure 18: SRIC Scheme
and is connected to a 10 [dB] attenuator. The aircraft should be able to
obtain good connection when flying over the antenna beam.

 Transceivers
The chosen aerial transceiver is “UBIQUITI Bullet M2-HP”, which has the proper band for the channel. Moreover,
it has the same beneficial qualities mentioned above for the Bullet M5.

 Chosen Antenna
To meet the above requirements, we selected a circularly polarized “PATCH” antenna. This antenna enables good
signal downwards. The antenna was tested and verified in an antenna range (details are given in the Antenna-Range-
Measurements section). The circular antenna has 80̊ beam width, allowing both flying above the SRIC antenna, or at
a circular flying course above the antenna.

o Positions and Types of the Antennas

Since the plane airframe is made of conductive carbon , a dipole antenna would suffer from interference. Therefore,
λ
for the more demanding link (5.8 [GHz]) a length monopole antenna was chosen. A monopole antenna requires a
4
ground connection which the carbon airframe provides, and, more importantly, it does not create interference.
Nevertheless, the landing gear creates interference (as can be seen in the experiments section). Therefore, it was
decided to move the antenna to the tail. For the SRIC mission, the patch antenna was directed downwards.
o Link Budget
PT GT (θ1 ,ϕ1 )GR (θ2 ,ϕ2 )λ2
The receiving power is PR (θ1 , ϕ1 , θ2 , ϕ2 ) = (4πr)2
.

- 12 -
TAS – Technion Aerial Systems Team
For the 5.8 [GHz] system 𝑃𝑅 is the power received by the nanostation system, 𝑃𝑇 is the power at the bullet output,
𝐺𝑇 is the antenna gain, and 𝐺𝑅 is the nanostation antenna gain. The power and gain for each of the above for the 5.8
[GHz] system is:
𝑷𝑻 [𝒅𝑩𝒎] 𝑮𝒓 [𝒅𝑩𝒊] 𝑮𝑻 [𝒅𝑩𝒎]
Manufacturer Specs 19 to 25 16 n/a
Measured 8 to 9 n/a -4
Therefore at 1 [km] the reception power is 𝑃𝑅 ≅ −77.7[𝑑𝐵𝑚], meaning 48[Mbps].
For the SRIC system, for the downlink (in the case of transmission from the bullet to the judges’ antenna), 𝑃𝑅 is the
power received by the judges’ antenna, 𝑃𝑇 is the power at the bullet output, 𝐺𝑇 is the antenna gain (a circular patch
antenna) and 𝐺𝑅 is the judges’ antenna gain. The power and gain for each of the above for a 2.4[GHz] system is:
𝑷𝑻 [𝒅𝑩𝒎] 𝑮𝒓 [𝒅𝑩𝒊] 𝑮𝑻 [𝒅𝑩𝒎]
Manufacturer Specs 22 to 28 -6 n/a
Measured 14 n/a 0 to -2

Therefore at 200[m] the reception power is 𝑃𝑅 ≅ −80[𝑑𝐵𝑚], meaning 36[Mbps]. This shows that even at the
extreme conditions, the selected antennas should enable successful SRIC mission. For most of the flight, the aircraft
should be at a distance of less than 1 Km and lower than 200m.
2.9 Development Process of Specific Mission Tasks
o Autonomous Waypoint Navigation
After assessing the results obtained on last year’s platform that had difficulties in tracking the waypoints, this year
we designed the UAS to achieve greater maneuverability by enlarging the wings, optimizing the ailerons and adding
a dorsal fin. The fin helps to increase the stall angle of the vertical surface. Furthermore, we eliminated some GPS
electromagnetic interference by isolating it. Following the above, this year the system features more precise and
accurate waypoint navigation.

o Automatic Takeoff
Like last year, we based automatic take off on the built-in algorithm of the Mission Planner and Pixhawk. To
minimize risks and damage to the aircraft, this task development process included flight experiments using a safety
pilot. Automatic takeoff and landing was first tested on a small RC wooden plane, equipped with the Pixhawk
autopilot and all necessary sensors in order to fly autonomously. For safe and accurate autonomous landing, the
SF02 Laser Range finder was installed. Connected to the autopilot, this rangefinder was used as a laser altimeter,
yielding accurate AGL (Above Ground Level) altitude data.
o Image Processing (search area and ADLC tasks)
The ADLC system is based on state of the art Deep-Learning algorithms. The system is comprised of two
subsystems shown in figure 19. The first subsystem detects possible targets (candidates) in an image. The
“Maximally Stable External Regions”
(MSER) feature descriptor was used as blob
detector. The input to this subsystem is a
1600x1080px grey downscaled image from
the original 6000x4000px color image. The
reason the image is down-sampled is
because the MSER algorithm is both
computationally demanding and sequential
in nature (making it hard to accelerate using
GPUs); the smaller images allowed an 8
times faster processing while maintaining
stability. The outputs of the subsystem are a
set of square crops (coordinates) suspected
to contain target.

Shape Classification. The target candidates

from the MSER subsystem were fed to a Figure 19: ADLC system
shape classifier and classified as one of nine

- 13 -
TAS – Technion Aerial Systems Team
possible shapes (circle, half circle, cross, rectangle, etc.). The shape classification was done using a deep neural
network implemented using the Caffe framework. The network was trained on synthetic samples; the largest training
set comprises 500K samples (figure 20), half of them with targets of random shape, color, character, orientation and
scale. The other half are patches with no target. Patches classified as ‘no-target’ are dismissed.

Character Classification. The purpose of the character classification subsystem is to classify the character inside
the targets and its orientation. Its input are patches taken from the full resolution image classified as containing
target shapes. The first stage is segmentation of the pixels belonging to the character. This is done using the ‘k-
means’ clustering algorithm. A binary ‘character’ mask is formed by selecting pixels that belong to the ‘character’.
The ‘character’ mask is rotated at a angles 0, 15, 30… 345 and re-cropped
tightly. This procedure forms 12 copies of the binary mask at different
angles (see Figure 19: ADLC system). These masks are re-sampled at
28x28px resolution and fed into a CNN trained for classifying characters.
As mentioned before, the network is fed with 12 copies of the ‘character’
mask rotated at different angles. For each angle, the network either
discards the patch or selects for it some character class. The angle, whose
character class receives the highest probability, is selected as the shape
orientation and the corresponding class is used as the shape character. If
all the rotated patches are discarded by the network, then the patch itself is
Figure 20: Synthetic Targets (the only discarded as ‘no-target’. This way, the character classification subsystem
true target is the star at the top right is used also as a weak target classifier. The network was trained with
with character ‘C’) synthetic data, made of 300K samples evenly divided between straight
characters, rotated characters and false (non character) patches. The
ADLC algorithm is implemented onboard the drone using the Jeston TK1 computer.

o Interoperability
The algorithm for the interoperability task, developed by the students,
has been integrated into the main loop of the command and control
software. It uses an HTTP protocol, which operates on top of the TCP
protocol to send and receive information from the server. In case the
server returns an error, the software sends a message box containing
the error description and its cause, and the process stops. The
algorithm consists of an HTTP GET and POST requests: login,
download server time, download obstacle information and upload
UAS telemetry. The last three processes are executed in parallel using
multithreading to allow independent operation.
To accomplish the required task’s send and receive rates, a simple
proportional controller has been integrated to close the interoperability
communication protocol. Using the controller, the software introduces
a virtual delay to the communication rate and converges to the rate set
by the flight operator.

o Sense, Detect and Avoid

The obstacle avoidance algorithm is based on last year’s algorithm.
However, in contrast to the previous algorithm, the algorithm keeps Figure 21: Algorithm Scheme
“breathing” and being updated according to the current state. This
allows it to lose less time by staying as close as possible to the
obstacle instead of setting a constant waypoint in order to take a detour around the obstacle thereby improving last
year’s algorithm performance.
Because of camera resolution and image processing as well as climbing rate considerations, the option of bypassing
the obstacle from above had been rejected.
The new obstacle avoidance algorithm is based on the changing tangents to the obstacle or a safety radius around it.
By updating the tangent to the obstacle, the course will be converged to the arc around the obstacle.
The algorithm first seeks the first obstacle between the UAV current location and the target waypoint. To calculate
the two possible alternative courses, two candidate waypoints are defined by the intersections between four lines
connecting the UAV current position closest to the obstacle. To choose which one is better, we need to consider the
related position of the no-fly-zone border, related position of other neighbor obstacles and the length of each course,

- 14 -
TAS – Technion Aerial Systems Team
in this order of importance. After passing the obstacle, while heading the target waypoint, it is necessary to merge
again with the original course as soon as possible. Therefore, the next waypoint for the calculation will be a point
after the obstacle, added on the original course as close as possible to the obstacle. This process runs in a loop to
update the tangent lines and the current temporary waypoint. In order to deal with the dynamic obstacle, the
algorithm should calculate the location of the zero effort miss (ZEM) in case the UAV course intersects with the
obstacle course, both in time and space in addition to the considerations mentioned. It is also necessary to take into
consideration the aspect between the obstacle and the UAV velocities into account.
o Off Axis
To improve last year’s performance, a new refined algorithm has been written. Although preserving the core
technique of calculating the camera attitude required for proper tracking of the off axis target, the improved method
relies on the Mavlink connection protocol between Pixhawk and Storm32. This way the camera orientation angles
are adjusted relatively to its actual position in real time, as opposed to assuming a certain initial attitude.

o Air Drop System

Our Airdrop algorithm is based on real-time continuous calculation of the ballistic flight equations, in which it
calculates the optimal point of release while taking into consideration the following factors - full system lag, real
wind profile and drag force on the bottle. In order to develop the algorithm, we created a simulative version in
Matlab, which includes basic aircraft flight simulation and 3DOF bottle flight simulation. By using multiple point
data to forecast the future optimal point of impact and achieve minimal range from the bottle point of impact to the
target, we were able to achieve better results than the standard current-state calculations. It is important to note that
the full forces on the bottle and the wind speed were taken into consideration in our algorithm and simulation in
order to optimally simulate the conditions in the field. An airborne version has been developed in Python, which
interfaces with the Mission Planner, and relies on the simulative environment. In 27 out of 32 experiments, we were
able to reach the minimum range required for this mission and for several experiments we reached the bull’s eye
radius.

2.10 Mission Planning

The mission planning is built according to the main competition tasks. There are several mission plans that meet
different weather situations, for instance, wind regime. While building the mission plan, we take into account the
system constraints such as the camera focal length. The flight line team is comprised of 5 sub teams – main autopilot
operator, secondary autopilot operator, imagery operator, ground control station commander, team captain and
safety pilot. After the completion of each task, the team captain reevaluates the remaining time and decides whether
to land or to procced with the mission tasks

2.11 System Integration

The system engineering process was carried out in several phases. During the first phase, taking into account the
tight development schedule and keeping a low budget, we chose on shelf products which feature high reliability in
minimum risks. While choosing each product, we ensure that they can communicate and operate flawlessly with
each other. For instance, the chosen gimbal controller is relatively affordable. It was purchased only after a
comprehensive spec examination for compatibility. During the second phase, the products were combined and
organized into subsystems. Each subsystem was evaluated and tested separately. Moreover, the subsystem
prerequisites were matched by means of voltage, cooling, bandwidth and more. For example, a detailed testing plan
was designed to examine the imagery subsystem capabilities, which started in laboratory tests, continued in ground
tests and ended in flight tests. During the third and last phase, we bind the independent subsystem into an integrated
system – Athene. While assembling the subsystems together, several issues were detected which could not be
detected separately, such as overheating of the camera turret due to higher than expected current. The problem was
solved by isolating the turret engine from the other turret subsystems in an innovative manner.

- 15 -
TAS – Technion Aerial Systems Team
3 Test and Evaluation Results
3.1 Aircraft Performance and Airworthiness
o Wing Load Test
To evaluate and guarantee the airworthiness of the Athena system, structural static loading tests were performed. A
static wing loading test was used to evaluate whether the aircraft can withstand the loads predicted in the

Figure 22: Wing load test

preliminary stage of design. During the test, weights were distributed atop the wing in accordance to the lift
distribution calculated using commercial Vortex Lattice Method (VLM) software and the wingtip deflection was
measured. The results were compared to those acquired by using Finite Element Method (FEM) programs.

o Wing Flow Test

In order to understand the flow patterns that develop above the wing,
this year, we performed tuft tests to TYOT and Athenna. The motion of
the tufts during flight can be observed and recorded to locate flow
features such as boundary layer separation and reattachment. On both
wings, a GOPRO camera was mounted to record tuft motion. The results
on the TYTO platform demonstrated that the wing suffers from
transition and turbulence. Furthermore, the aileron dimensions and
location cause the flow to separate when performing turns due to the
aileron’s high angle of attack. After the experiment, we analyzed the Figure 23: Wing with tufts
results and used them to design the new wing.

o Visual Flight
Simulation
A simulator was written which
loads flight data from recorded
test flights and visualizes it using
Google Earth. This allows both
simulating first-person view of
these flights and better test flight
debriefing.

o Data Link
Figure 24: Radiation patterns (left) of the imagery and CST graph (right)
To significantly improve the
communication systems, a series of simulations and tests were performed, as described:

o Modeling the Antennas

An EM simulation software, CST Computer Simulation Technology, was used to simulate and test the radiation
pattern of different antennas at different locations on the aircraft. An example of such pattern is given in figure 24
(right).

- 16 -
TAS – Technion Aerial Systems Team
 o Antenna Range Measurements
Different antennas were considered for each of the links (imagery and SRIC).
Ground tests were performed, comparing the gain and effective bit rate for each
antenna at different distances. The top candidates were then tested in an antenna
range. One of the aircraft bodies was attached to the antenna range
measurement system at two different configurations in order to test two main
cuts (roll and pitch).

Figure 24 (left) shows examples of radiation patterns measured in the antenna

range. Besides verification of the simulations, the antenna range helps in
determining the preferred position for installing the antennas. For example, the
imagery antenna was positioned at the bottom of tail to avoid interference by
the conductive material of the landing gear. Measurements of the SRIC antenna
showed that it has about 80̊ of clear reception, which makes it adequate for both Figure 25: Antenna range test
a fly-by above or at a circular path around the SRIC antenna axis.

o Antenna gain and data rate measurements

Scripts were written to obtain data from the transceiver OS, allowing us to create logs of the transmitted and
received bit rate and the gains of the aerial
and ground units. Another test performed
took advantage of the NanoStation built-in
spectrum analyzer, with the purpose of
determining the noise level in each channel.

o Static Thrust Tests

Because of the change in the total weight of
the system, the cruising speed has been
increased. To evaluate the performance of
the propulsion system, static thrust tests were
Figure 26: 2.4[GHz] Channel gain and bit rate at 1[km]
conducted. During the test, several attributes
such as reliability, temperature and power consumption were tested and evaluated according to the system
requirements.

3.2 Autopilot System Performance

To investigate the autopilot performance and integration, ground tests regarding sensor communication with the
autopilot first took place. Valid and reasonable sensors values were confirmed to be used in the autopilot
calculations. After integrating the hardware, and due to lack of time and unsuccessful simulation connectivity, flight
tests and autopilot gain calibration took place in order to achieve satisfying maneuverability, altitude, speed tracking
and waypoint navigation, while a safety pilot was ready to take action in case of an emergency. Next, specific
abilities were evaluated according to competition tasks.
3.3 Mission Task Performance
o Sense, Detect and Avoid Task
After deciding that there will be no altitude change maneuvering to avoid obstacles, a 2D Matlab simulation was
developed to support the obstacle avoidance algorithm development. The Matlab simulation takes into consideration
a simplified UAV model based on UAV maneuvering performance. The Matlab simulation allowed a convenient
infrastructure for developing the algorithm and examining it at first without field tests. The simulation developed
was based on the competition rules in sense of obstacle radius and velocity, no-fly-zone, etc. The screenshot below
shows the simulation takes action with two obstacles with their safety ranges, close to the no-fly-zone borders. The
UAV trying to follow the original course is impossible due to the obstacles ahead.
Therefore, the algorithm builds a new course every iteration based on the relative location of the UAV, the
waypoints of the original track, and obstacle position. After working with the simulation, a conversion to C# was
made to integrate the algorithm into the ground station Mission Planner. The next step was ground tests during
which a person carried the Pixhawk with walked virtual obstacles on the course, and the new track was confirmed to
be built.

- 17 -
TAS – Technion Aerial Systems Team
 Figure 27: Obstacle Avoidance Simulation
The final test was on a real flight out in the field. Examining the algorithm can actually make the UAV avoid virtual
obstacles.
o Waypoint Navigation
Nearly all flight tests included waypoint navigation as a sub-test to ensure the system will comply with the objective
level criteria. The Athene system has been tested to reach the required waypoint accuracy of 40 [ft]. The rate of
climb of the aircraft is 11 [ft/s]. It is the maximum climbing between two points with 17 [kts] tailwind and 40 [kts]
speed.

o Autonomous Search Area

Like last year's Tyto system, the team decided on the 'double back' scan method. The system was able to scan the
area with minimum altitude lost on turns and faster scanning time. According to the camera, the turning radius is 70
[ft] and cruising speed does not exceed the UAV limits.

o Interoperability Task
As mentioned before, the interoperability system has been integrated within the command and control software. To
ensure a safe usage of the aforementioned software with the additional integrated code, we tested the interoperability
system extensively on different networks both in air and in ground tests. During these tests, a stable connection has
been achieved on different network load levels. Closing the loop on the data rate has proved to be efficient and an
average steady rate of 10Hz has been recorded in most flights.

o Off-Axis Target
To evaluate the new tracking algorithm, a series of ground tests were performed before taking the system into the
air. Varying target distances and altitudes were simulated, returning the success rate of 95%. Once the reliability of
the new method was established, flight testing began. The system was tried in 20 flights with the success rate of
85%.

o Air-Drop Task
During the flight tests, we evaluated the airdrop subsystem algorithm in many scenarios and flight conditions. We
managed to reach to accuracy of less the 1.5[m] when dropping the bottle using our algorithm.

3.4 Payload System Performance

o Camera Interface Testing
To allow capturing and transferring pictures at a rate of 2 FPS, several optimizations were applied. The camera has a
24.3MP sensor, resulting in 72MB (uncompressed) images in memory. Instead of manually resizing such images in
real-time (which takes valuable processing time) an already-resized version of the image was extracted from the
EXIF data of the image. Since the camera’s picture transfer rate is slower than the picture burst rate, an external
synchronization module was developed which allows accurate timestamping of the images using the camera’s

- 18 -
TAS – Technion Aerial Systems Team
external flash connector. The module consists of an Arduino Nano receiving the digital flash signal from the camera
and sending timestamps via UART to the OBC.

o Communication System Testing

Since the transmitters in use run a BusyBox system (a minimal Linux system) the data link could be measured by
reading a special file located at /proc/net/wireless. In order to measure the data link signal quality, a script was
written which connects to the transmitter via SSH and continuously reads this file, displaying the current signal
quality on-screen and recording timestamped samples for future data analysis. To measure the transfer rate of the
payload system, a script was written to simulate a continuously transmitting data server. The data is pseudorandom,
ensuring possible compression does not affect the results. Another client script was written which connects to this
server and continuously downloads the data, measuring the transfer rate, displaying it on the screen and recording
timestamped samples.

o Airborne System Performance

Because the airborne computer should operate in a non-standard environment (low airflow, high temperature, long
running time) it was decided to perform system robustness tests. Heating tests were performed to make sure the
system does not overheat while running for a long time; the system was left in operation at a high workload for a
long time. The on-board computer CPU temperature was monitored and recorded. The initial results have indeed
shown that the system overheats and crashes after 10 minutes of continuous work. As a result, the original on-board
computer (ODROID U3) was replaced with another version (ODROID XU3) which has better heat conduction.
When repeating the test, the system performed significantly better and did not overheat even after a long time
memory consumption. While initially testing the Sony A6000 camera with the airborne software platform, the
software platform consistently crashed due to high memory consumption. After testing the issue, it was discovered
that this high consumption happens due to the large raw images produced by the camera. Once more than a dozen
pictures were taken and stored in memory (waiting to be sent to the ground platform), the memory consumption was
high enough to cause the system to crash. To fix this issue, new pictures transferred from the camera are stored on
the permanent storage (memory card) and later fetched for sending only when requested to do so by the ground
platform. To minimize the dependency on field tests, the payload system includes a simulation mode. This mode
allows using images and flight data from earlier field tests, feeding them to the system instead of live images from
the camera. This allows performing robustness tests for almost the entire system, and allows testing and reproducing
issues easily. For example, geolocation was tested by feeding known satellite images with the relevant flight data
and checking geolocation in known coordinates in these images.

o Rotating Turret Performance

The redesigned turret and camera stabilization controller were tested and evaluated from the early stages of the
project. A laser pointer was installed onto the camera mount and its beam projected on the specially designed
measuring wall. A wide variety of angles was tried, each time measuring the laser movement on the wall and
comparing it to the analytical calculations. Stabilization was evaluated by increasing the distance between the wall
and a pointer and measuring the deviation of the laser beam while manually changing the attitude angles of the turret
base. The Storm32 demonstrated accurate and responsive leveling and pointing capabilities. Pointing precision was
found to be the hundredth of a degree. Stabilization was found satisfactory for distances up to 450 meters.

3.5 Flight Tests

Overall, 25 flight days were conducted with an average of 6 flights per day. We accumulated more than 100 Flight
hours in which we evaluated and checked all the system components and performance. During the beginning of the
development process, the flight tests were conducted using last year’s platform to evaluate specific subsystems.
Upon completion of the Athene UAS, we integrated all the subsystems and performed full competition simulations,
in which our Faculty Advisor served as a strict judge.

- 19 -
TAS – Technion Aerial Systems Team
4 Safety Considerations
Emphasis on safety was essential in successfully developing a fully functional unmanned system in the given short
time of the project. During the design process, manufacturing and verification of the aircraft, continuous risk
assessments were carried out to minimize possible harm to personnel and equipment. These assessments dictated
design restraints and guided team operational policies, which enabled safe operation of the system.
4.1 Operational Safety Criteria
Before each flight\experiment, a list of goals and guidelines were prepared. Prior to every flight test day, the flight
crew inspected the vehicle, with emphasis on airframe and autopilot systems. Checklists were developed and
expanded as needed throughout the year for safety protocols and pre-flight readiness procedures. Main sections of
several checklists include:
 Inspecting weight and balance of the vehicle to insure the aircraft CG location is within the required margin
 Inspecting the engines of the vehicle, carrying out a static thrust experiment.
 Inspecting the integrity of the vehicle structure.
 Inspecting all batteries for damage, and ensuring that they are all fully charged.
 Inspecting the various payload components of the vehicle and ensuring that they are connected and secured.
 Inspecting the servo linkages (pushrods, horns, etc.) and control surfaces.
4.2 Design Safety Criteria
During the design of the vehicle, a safety factor of at least 2 of the expected load was taken for all structural
elements of the airframe, and of a least 1.5 times of the expected load for electrical systems. Repeated testing was
performed to ensure structural and electrical performance behaved as expected.

4.3 Safety Risks and Mitigation Methods

The aircraft has a safety rely switch (“Battleswitch” single pole – double throw 10 A rely switch) in its control
system, enabling us to minimize risks related to loss of control during the flight. This safety switch separates the RC
receiver and autopilot on a hardware level, allowing manual control and servo operation in the event of an autopilot
failure. The safety switch also allows the external pilot to bypass the autopilot and take immediate control over the
vehicle in case of any unexpected aircraft maneuvers. In the event of losing both the RC and the autopilot
communications for more than 30 seconds, the system is programmed to enter RTL mode, in compliance to section
[Link] of the SUAS rule sheet. In case a loss of communication signal occurs for more than 3 minutes, the system is
programmed to enter a flight termination protocol, in compliance with section [Link] and section [Link] of the
SUAS rule sheet.
To minimize the risk of any unanticipated engine operation on the ground (during maintenance, preflight checks,
etc.), the aircraft has an integrated safety switch connected to the autopilot system which cuts all signals to ESCs on
a software level. In addition, a new physical plug was developed to replace last year's plug, which once removed,
disconnects the ESCs from their power source on a hardware level: last year's plug was heated quickly and was hard
to remove after the vehicle landing. The new one is easier to remove and does not heat as quickly. The UAV also
integrates a servo cutoff switch and a payload cutoff switch, which were moved this year outside the vehicle. This
allows the team members to develop and test specific components of the UAV without the unnecessary operation of
other components and systems that are not required at that moment, and without the need to open the fuselage.
Incorporating all of the aforementioned safety mechanisms ensures a minimal loss of components and a secure
flight.

5 Conclusions
During the past year, we managed to design, analyze, build, test and evaluate an innovative autonomous airborne
system. We implemented advanced design and analysis tools such as CFD, Ansys, Matlab simulations, antenna
range, wing flow testing and more. The system features novel approaches such as GPGPU for parallel classification
computations, 3D printing for lighter components and cheaper production, computer simulations using our own
algorithms or 3rd party applications and more. For the first time, we built a fully carbon fiber wing using this
innovative method. Using those approaches, we were able to perform all the competition tasks in the best possible
way. This year we have focused on the comprehensive process of system engineering and system integration, which
resulted in highly improved system performance and system integration.

- 20 -
TAS – Technion Aerial Systems Team
1 Appendix: Cyber Security
1.1 Definitions
o Definition: Encryption and Decryption
Encryption is the process of transforming data using a key such that the original data (“plaintext”) is unreadable from
the encrypted data (“ciphertext”). Decryption is the reverse process, which generates the plaintext using the decryption
key. Such algorithms can be symmetric where the encryption/decryption keys are identical, or asymmetric, where the
keys are different. In general, asymmetric encryption or decryption is much more computationally expensive. Both
symmetric and asymmetric encryptions are based on mathematical operations.

o Definition: Authentication
The verification of the message’s sender identity (usually achieved using mathematical operations).

1.2 Radio link protection

o Encryption of communications (Control and Payload)
To prevent possible communication eavesdropping of both control and payload data, encryption of communication
data is possible. If using symmetric encryption, the keys must be shared in advance between the ground station and
the airborne platform, or some kind of key exchange mechanism should be used.

o Authentication of communications (Control and Payload)

Since it is undesirable that some external party would gain control over the airborne platform or send manipulated
data back to the ground station, an authentication mechanism should be used, such as a Message Authentication Code
(MAC) - where each message sent is joined with a keyed hash of the data. The key should be kept secret.

o Frequency-Hopping Spread Spectrum

Frequency-Hopping is a technique of transmitting radio signals while constantly and quickly switching the
frequency of the carrier. The switching is done in a pseudorandom sequence, thus, making it much more difficult to
intercept the transmission during a mission. Furthermore, the spreading of the transmitted energy across a spread
spectrum decreases the narrowband interferences.

1.3 Ground Station protection

o Ground communication encryption and authentication
As in the previous section, ground communication data should also be encrypted and authenticated.

o Access control
In order to control access to the network, an access control mechanism should be used where only specific hardware
(MAC) addresses are allowed to connect to the network. Since it is extremely easy to spoof such addresses, this is
done only as initial filtering of connected devices.

o Network isolation
To minimize the risks of external threats, the network should be isolated as much as possible – that is, not connected
to the internet or any other network. If this is not possible, the number of external networks must be kept to a minimum
to be able to control the gateways more tightly.

o Firewall
In case full network isolation is not possible, and as a counter-measure even if the network is fully isolated, a firewall
should be employed which is configured to allow only specific defined communications needed for system operations .
Such communications should use unconventional ports.

1.4 Internal UAV protection

o System Communication
Encryption and authentication of data passing between different on-board systems should be used.

o EMI Protection

- 21 -
TAS – Technion Aerial Systems Team
Electro-Magnetic Pulse (EMP) is a high-energy electromagnetic pulse that can induce high currents in electrical
components and cause damage. Possible protective measure is placing the payload (except for the antennas) inside a
conductive enclosure which blocks electrostatic fields (Faraday’s cage).

o Physical Protection
1. Tamper Resistance: Several measurements that should be taken to prevent sabotaging includes:
a. Securing the Payload – to prevent access to the electrical components the fuselage casing can be closed by a
lock and the key must be kept safely.
b. Screws that secure sensitive systems must conation unique drives.
c. Power surge prevention circuits – needed especially next to the antennas that cannot be kept inside the
Faraday’s cage.
2. Tamper Detection: Might contain simple measurements, such as a seal that is broken once touched; or more
complex such as an alarm system.

1.5 GPS Spoofing Countermeasures

o INS
Inertial Navigation System (INS) is a navigation aid system that consists of inertial measurement units (IMUs, sensors
that monitor the velocity, acceleration and orientation) and a navigation computer. The computer uses the IMU sensor
readings to estimate the position of the vehicle. It is necessary to provide the system with its initial condition, but since
it does not rely on external measurements, it is more immune to jamming. These systems tend to accumulate error
over time.

o VAN
Visual Aided Navigation implements Simultaneous Localization And Mapping (SLAM) algorithm based on camera
images and IMU sensors. VAN systems are able to estimate the position of the plane to high accuracy and map its
surrounding. The passive nature of the measurements (camera and IMU) make it a good alternative to the GPS. Our
team already implemented a mapping algorithm, which is the base for a VAN system.

o SAASM
Selective Availability Anti-Spoofing Module – allowing satellite authentication, re-exchanging keys through air mid-
operation and contingency recovery.

o Other GNSS
Global Navigation Satellite System (GNSS) are receivers using the GPS, GLONASS, Galileo or BeiDou systems.
These are all satellite navigation systems that provide location using time signals. GLONASS receivers don’t rely
solely upon the US NAVSTAR GPS, and therefore might still operate in case of GPS spoofing.

o VOR/DME
3. VOR (VHF Omnidirectional Range) is a system which transmits two signals – an omnidirectional signal and a
rotating highly-directional signal, which allows the receiver to determine its azimuth to the transmitting VOR
station by measuring the phase shift of the two signals.
4. DME (Distance Measuring Equipment) is a system which includes a land-based transponder which responds
to signals received from aircraft. This allows the receiver to determine its distance from the DME station – which
is typically collocated with a VOR station to allow full localization using a single base station.
5. Use of VOR and DME for navigation: To use VOR/DME for navigation, the base station locations must be
known in advance. To have full localization, one of the following must be received – VOR azimuth and DME
distance, 2 different VOR azimuths or 3 DME distances.
6. Authentication of VOR/DME data: Since it is possible to spoof VOR/DME data, such systems should also be
authenticated in some way, such as an addition of an authenticated data signal on top the existing signals.

- 22 -
TAS – Technion Aerial Systems Team