Ing. Jiří Novák, Ph.D.
Cyber-Physical Systems Design Project Assignment
Welcome to the project assignment for Cyber-Physical Systems! This is your opportunity
to apply the concepts we’ve explored throughout the course in a hands-on, real-world setting.
You’ll work on designing, developing, and implementing a system that integrates computa-
tional and physical processes.
You will select one of the following four project options, each presenting unique challenges
and learning opportunities. You can use your creativity in solving the project and apply knowl-
edge acquired throughout the course. Your submissions must include technical report and
implemented code. Technical report has to be in PDF format and it will include explanation
of the methods you used, implementation details and achieved results. Your code must be
clearly structured and commented so it is clear for others. You can get maximum of 20 points
based on quality of your technical report and code. Minimum number of points is 10 in order
to get a credit and finish this course. Deadline for project submission is 9. May 2025, 23:59 in
E-learning. In case, you want to work on another relevant topic in CPS, please consult with
me, so I can approve your project proposal.
A: Drone Control
In this project, your task will be to design and implement a control system for a multirotor
drone. The drone must be able to takeoff, hover and land at different location. You should
have full attitude control (roll, pitch, yaw). Model of the drone is in the supplemented files.
You should first run initialization.m to load parameters into the working directory. You can
then open [Link] and run it. You should be able to see visualization of the drone
simulation with spinning propellers. For your control system, you can use any technique you
want like Proportional, Integral, Derivative (PID) control, Linear Quadratic Control (LQR)
or Model Predictive Control (MPC). You should evaluate your final design on the Simscape
model and report time-domain characteristics of your control system such as overshoot, rise-
time, settling-time and steady-state error. There are a lot of resources for drone modeling and
Figure 1: Virtual drone model in three views
control system design. You will find some of them in the project folder. Generally, you should
first use a mathematical representation (linear/nonlinear differential equations) of the drone
to design your controller. In case you are confident with your result, the code from MATLAB®
& Simulink® can be loaded on real Pixhawk controller.
Required software: MATLAB® & Simulink® ; SimscapeMultibodyTM
1
Ing. Jiří Novák, Ph.D.
B: Robotic Arm Control
In this project, your task will be to design and implement a control system for a robotic arm.
The robotic arm should be able to pick up a reference object (box) and move it to a different
location. Try to design your system such that the arm and the payload avoids collision with
itself. You should first run initialization.m to load parameters into the working directory. You
can then open [Link] and run it. You should be able to see visualization of the
arm and the payload. To make this project a little easier, Denavit-Hartneberg parameters are
provided in the initialization.m file named dhparams. This project was simplified such that the
joint motion behaves according to double integrator ordinary differential equations. The con-
trol inputs of this system are therefore the joint accelerations. You will first need to understand
inverse kinematics to design your end-effector trajectory. The end-effector is equipped with
four vacuum cups. The code in SimscapeMultibodyTM is designed, such that the vacuum cups
stick to the payload after they first come to contact.
Figure 2: Robotic arm model in three views
Afterwards, you will have to design a controller for the joint angle accelerations to con-
trol the joint angles. Your controller should be able to follow a specified trajectory with the
end-effector. At each point of the trajectory, you should be able to control the position and
orientation of the end effector. Please report your achieved Root Mean Squared error (RMSE)
accuracy in the trajectory following problem.
Required software: MATLAB® & Simulink® ; SimscapeMultibodyTM
2
Ing. Jiří Novák, Ph.D.
C: Vehicle Detection & Tracking
In this project, your task will be to implement vehicle detection and tracking algorithm in a
traffic scenario generated in CARLA driving simulator. The supplemented files include a one
minute recording sampled at 15 frames per second. Each frame has its unique ID correspond-
ing to the filename. Each frame has also associated ground truth 2D bounding box JSON
file with vertices of the bounding box in image coordinates and 3D bounding box JSON file
including vertices of the box in world coordinates and also projected vertices to the image co-
ordinates. The folder also contains camera intrinsic and extrinsic parameters. You can choose
to track vehicles in 2D image coordinates or 3D world coordinates. Try to address common is-
sues in tracking such as static and dynamic occlusion. You should be able to assess your object
detector and tracker using the supplemented ground truth bounding boxes. Typical metric for
object localization is the Intersection over Union (IoU) metric. Please report the IoU in your
technical report. You can use any technique you will find suitable. However, neural network
based detector such as You Only Look Once (YOLO) will be necessary for robust detection
of the vehicles. One of the most used tracking algorithm is the Simple Online and Realtime
Tracking (SORT) which is very fast even in presence of many objects.
Figure 3: Vehicle detection and tracking in CARLA simulator
You can work in any programming language you want but I can recommend a Stone Soup
library in Python which helps in designing the tracking algorithm. There is also a lot of exam-
ples in topics such as Kalman filtering and object association techniques.
Recommended software: Python; Stone-Soup
3
Ing. Jiří Novák, Ph.D.
D: Flight Simulation & Autopilot Modes
In this project, your task will be to implement a functioning flight simulation with two autopi-
lot modes, heading change (HDG) and altitude hold (ALT). When both autopilot modes are
active, the aircraft should not loose altitude when changing heading. You will first need to
setup a Flight Dynamic Model (FDM) based on Equations of Motion (EoM). You can choose
any aircraft model, but keep in mind that you will need to find the aerodynamic coefficients to
properly simulate motion of that aircraft. For reference, you can find derivation of linearized
EoM for a single flight condition and all the required parameters for a Boeing 747-200 in the
project folder. You can work in any programming language, but you should be able to visu-
alize the results of your flight simulation in software such as FlighGear. If you use Simulink®
and AerospaceBlocksetTM , you can do this directly from Simulink® . When designing your
Figure 4: Flight simulator
control system, you should first focus on deriving the State-Space model (multidimensional
linear model) of longitudinal and lateral directional dynamics. For your control system, you
can use any technique you want such as Proportional, Integral, Derivative (PID) control, Lin-
ear Quadratic Control (LQR) or Model Predictive Control (MPC). After you are done, please
report time-domain characteristics of your HDG and ALT autopilot modes in your technical
report. These should include rise time, settling time, overshoot and steady-state error.
Recommended software: MATLAB® & Simulink® , AerospaceBlocksetTM , FlightGear