We supervise Bachelor and Master theses or student projects (e.g., project or research module in BSc and MSc study programmes of the university) in 3D computer vision, robot learning, and robot perception.

?

Please see further details on the topics below. If you find one or several of the topic areas interesting for your thesis/project, please do not hesitate to ask us about possible topics via email (see contact details below)! Please also include your BSc/MSc transcripts so we can assess your prior knowledge.

?

Some topics are announced in our Oberseminar course in Digicampus.

?

Examples of previously advised thesis topics:

?

• Learning Terrain-Aware Foot Contact Planning for Dynamic Quadruped Locomotion (MSc)
• Slot-Attention and Expectation-Maximization (MSc)
• Category-Level 6D Object Pose and Shape Tracking from RGB-D Image Data (MSc)
• Event-Based Non-Rigid 3D Tracking (MSc)
• Learning a Context-Conditional and Terrain-Aware Kino-Dynamics Model for Autonomous Mobile Robots (MSc)
• Learning Gaussian Process Dynamics Models from Visual Observations for Control (MSc)
• Probabilistic Semi-Dense Mapping for Stereo Visual-Inertial Odometry (BSc)
• Autonomous Flight of a Low-Cost Quadrocopter using a Semi-Dense Monocular SLAM System (BSc)

?

We develop approaches that learn 3D models of the environment from camera images by using deep learning and foundation models. Specific challenges are reconstrucing dynamic objects and understanding and simulating possible actions with objects.

?

Robots need the ability to perceive objects, create maps of the environment, and localize in them. In this topic area we are interested in methods for scene perception using sensors such as cameras in the context of robotic object manipulation and autonomous navigation.

?

Robots need the ability to learn their object manipulation and navigation capabilities. We develop approaches based on reinforcement learning, model-predictive control, and foundation models.