Cyber-Physical Systems are about the tight integration of computing, communication and control techniques for (autonomous) mechanical systems. We specifically focus on systems that are deployed in production, such as autonomous robotic systems, autonomous vehicles, UAVs, etc.

Our thesis covers both informatics and mechanical engineering disciplines and complement our research projects. Specifically, the macro-areas of interest are:

• Artificial intelligence
• Control theory
• Machine learning
• Operating systems
• Real-time and embedded systems
• Robotics

Note: our thesis topics are open for students in Informatics as well!

Check our our dedicated page or contact us to learn more about open topics.

The topics in these areas aim to combine high performance and predictability on complex heterogeneous platforms --i.e., platforms including multicores, accelerators such as GPUs, FPGAs, various level of caches and a shared interconnect. Additionally, we investigate the potential of HW/SW co-design for reprogrammable FPGAs, security aspects of cyber-physical systems and their impact on system's predictability.
Example topics are related to:

• autonomous driving
• hypervisors (e.g., Jailhouse, Xen)
• kernels (e.g., Linux, microkernels)
• memory management
• QoS regulation
• safety-critical system design
• sandboxing
• temporal logic
• Worst Case Execution Time (WCET)
• Xilinx UltraScale+
• zero-copy

These areas target safety, predictability, energy, and computational aspects at-large of artificial intelligence (AI) algorithms when applied and deployed for cyber-physical systems, especially on embedded devices and edge IoT devices. Among others, we focus on vision-based 6D-pose estimation for robot picking tasks, reinforcement learning for robotic path and motion planning (applied for example to quadcopters and UAVs), and integration of DNN and DL onto low-power embedded devices.
Other topics are related to:

• 3D positioning
• factory automation
• stochastic modelling
• Unity
• Tensorflow
• Volterra series

Our labs allow hands-on experience in autonomous driving (F1/10), drones (quadcopter), and industrial robots to bring control theory to life. An NVidia A100 cluster helps training AI models. Several ARM and RISC-V embedded platforms to evaluate predictable operating systems and middleware are available.