Student projects and final assignments

Final work under mentorship: dr. Dušan Gleich, dr. Andrej Sarjaš

Description: The master's thesis presents the development and testing of a rocket system for automatic vertical landing using electric fan propulsion (EDF) and thrust vectoring. It includes a review of existing technologies, control theory and practical prototyping. The emphasis is on the design of mechanical assemblies and the implementation of stabilization algorithms. The system is based on a cascade architecture of controllers. The result is a working prototype capable of automatic hovering and controlled landing. The work contributes to the development of affordable research platforms for testing vertical landing technologies.

Description: This master's thesis presents a hardware-in-the-loop (HIL) approach for implementing adaptive control of nonlinear systems. Various control methods are discussed, including backstepping control, sliding mode control (SMC), PD control, and a proposed hybrid approach. For simulation purposes, the BLDC motor is modeled as a second-order system in the Simulink environment, while the control algorithms are implemented on a Beckhoff PLC with communication via the ADS protocol. The results evaluate the performance and robustness of the methods and confirm the effectiveness of the HIL approach for real-time validation of advanced control strategies.

Description: The master's thesis presented a method for detecting unmanned aerial vehicles, which combines an optical camera and radar. In the optical subsystem, objects were detected using the YOLOv8 algorithm, implemented on the energy-efficient NVIDIA Jetson Orin Nano platform, which enables real-time processing. The radar subsystem is based on an FMCW radar that exploits the Doppler effect at a frequency of 10 GHz and a bandwidth of 500 MHz, which allows for accurate measurement of distance and speed. Experimental results showed that combining optical and radar data increases the robustness and reliability of detecting unmanned aerial vehicles in different environmental conditions.

Description: In this thesis, we have created a robotic arm. For this research, we developed algorithms for linear and point-to-point motions. These algorithms use quadratic, cubic and sine profiles for servo easing. We have upgraded the robot arm with a camera. Using a neural network and machine vision, we developed an application where the top of the robot follows the hand of the user. Before implementation, we tested all the motion algorithms using MSC Adams.

Description: The thesis deals with vector control of the thrust of an air fan, which affects the inclination of the entire mechanism. Vector control is achieved by a tilt mechanism (gimbal) and is used in aviation with an emphasis on military combat aircraft to increase the maneuverability of the aircraft. The described concepts can also be used as a basis for implementing vector control of rocket propulsion. The thesis describes the concept of vector control with examples of 3D modeling of the mechanism. Finally, it includes the synthesis of a closed-loop system and its implementation. The results show the effectiveness of vector control for increasing the maneuverability of the aircraft system.

Description: In our diploma thesis, we created a robotic manipulator. For this purpose, we developed algorithms for performing point-to-point and linear motion. When performing these motions, we used square, cubic, and sinusoidal path profiles. We created a graphical interface for controlling the robotic manipulator. We upgraded the robot with a camera. Using a neural network and machine vision, we created an application where the top of the robot follows the user's hand. We previously simulated the motion algorithms and path profiles using the MSC Adams program.

Description: The doctoral dissertation presents the development of an airborne coupled ground penetrating radar for the detection of explosives under the ground surface for primary use on an unmanned aerial vehicle (UAV). Since the weight and autonomy of such a UAV are extremely critical, this requires special care in the development of such a ground penetrating radar system. For this purpose, a ground penetrating radar that transmits continuous waves with a stepped frequency step (SFCW) has been developed. The use of this method and a carefully designed receiver based on a hybrid analog-digital super-heterodyne architecture ultimately eliminated the need for a power amplifier for the transmitted signal. Such an approach allowed us to design the entire system using integrated electronic components, thereby ensuring low power consumption and compact design. Despite all this, the system offers a resolution in the range of a few centimeters, which allows the detection of even the smallest explosives. The system has been tested in a laboratory environment, as well as on a test site in combination with a commercial unmanned aircraft. In order to ensure even greater detection success, we have also introduced the use of polarization, which expands the existing GPR image to a multi-channel one. In addition, we have also developed a system for eliminating unwanted reflections, which removes them at the very entrance of the receiver; thus, allowing targets of interest, which are usually weak, to be additionally amplified.

Description: The master's thesis describes the design of a two-aircraft system, the development of guidance and trajectory generation algorithms. The basic equations of geographic calculations are given, which represent the basic building blocks of the algorithms. Three guidance algorithms have been developed, namely the algorithm for flying between two defined coordinates, the algorithm for a circular turn maneuver, and the algorithm for generating the trajectory of the slave aircraft, defined based on the trajectory of the master aircraft. The proposed solution enables synchronous flight of aircraft intended for bistatic radar. The system composition, functionality of components, and interconnection are described. The evolving IEEE 802.11ah wireless network protocol for communication between aircraft has been implemented. The developed guidance algorithms have been tested in a simulation environment and on a real system.

Description: The purpose of this master's thesis is to establish an RTK system that will be able to operate in three different modes and develop a user interface that will mark the desired area, generate the shortest path and display the current coordinates of the RTK receiver in real time. The master's thesis is divided into two parts. The first part shows in full the process of configuring RTK devices, setting up a base station, transmitting RTK correction via radio communication and using the UBX protocol. The second part presents the process of creating a user interface with the most important sections of the algorithm, the implementation of the ROS system and the results of field measurements.

Description: The master's thesis is based on the processing of satellite images and the use of deep convolutional neural networks. The content of the final work describes research work in the field of polarimetric SAR. The purpose of the work is to design and manufacture a system that could be able to process a satellite image so that soil moisture can be determined from it. Deep convolutional neural networks were used to evaluate this, which proved to be very useful. Polarimetric programs such as PolSARpro and SNAP were used in the fabrication process. The Python programming language in the Visual Studio environment was used to further process the images and design the deep convolutional neural network.

Description: The master's thesis describes and presents the process of face detection and recognition using radar and laser technology. The first part describes the operation of the radar and presents the procedures for placing targets in the visible area of the radar and processing signals to obtain the final image. The second part shows the process of capturing images of detected faces with Kinect and learning the model with the Siamese Neural Network for face recognition. The conclusion presents the results of research and verification of the detection and recognition of the face with radar and Kinect.

Description: The master's thesis deals with the issue of finger recognition, with the help of which we can manage a variety of tasks and processes in the background. The work is designed as a presentation of solving the same problem using two different approaches and a presentation of their strengths and weaknesses. Using the technology of finding a pattern in the picture, we tackled the problem in a direct way and looked for the feature in the picture, with the help of which we also understood the desired gesture of hands and fingers from the picture. Using deep learning technology, we left the search for characteristics to artificial intelligence, but in the beginning we needed a large base of already solved recognition cases. The findings from this work provide good starting points for all researchers and engineers in the further research and implementation of image recognition systems based on machine vision technology or deep learning.

Description: The final paper describes the control of the float in the air levitation system with the implementation of two different control algorithms on a remote computer and the effects of network control on the implementation of the controller. The content of the thesis includes a description of the components of air levitation, mathematical modeling, implementation of simulation with Matlab / Simulink software package, implementation of PID controller and controller based on sliding mode. The work includes a comparison of management in network and non-network mode of closed-loop management. The aim of the task is to design the regulators so that they will be able to control the system via the UDP communication protocol with delays.

Description: The aim of the master's thesis is the perception of respiration and heart rate. We used radar technology in three different ways of transmitting and receiving EM waves. For all three modes of operation, we gave a theoretical starting point and performed measurements. The modes of operation we used are frequency modulated continuous, frequency stepped continuous and pulse. Finally, based on the measurements, we assessed the suitability of each mode of operation and the radar on which it is implemented. Signal processing for all systems was performed in the MATLAB software environment.

Description: This work describes the procedure for determining the external coordinates of an object using two cameras and machine vision methods. In this way, the system enables contactless determination of the external coordinates of an object or active marker. We produced two markers and tested their accuracy in determining external coordinates. The knowledge gained from the first part of the work was also used to determine the external coordinates of the head orientation. The initial chapters describe the basics of digital photography and the theoretical background of the distortions that are present when capturing an image with a camera. The process of modeling the aforementioned distortions is also described. The following chapters cover the approach and design of the marker system. The greatest challenge was the assessment of the position accuracy and the assessment of the marker tracking speed. The following describes the method for calculating the external coordinates of the head orientation, as well as the problems that arose during the work itself. Finally, the measurement results of the comparison of both markers and the success of determining the head position are given.