Perception of the environment is at the prerequisite for robots to interact with their surroundings, and enables the intelligence for high-level autonomy. The 3D geometry and semantic meaning of the environment, the pose of a robot, the understanding of human motion, etc., are among the core of robot perception. In this seminar, we will discuss artificial intelligence algorithms that endow robots with an accurate and keen perception of the real world as well as how to teach robots to act in it. Students are required to do a literature review on relevant topics and present their findings at a colloquium in an informative and understandable way. The related topics include but are not limited to: Learning-Based and classical state estimation and mapping, depth completion, neural rendering, human pose estimation and forecasting, navigation and control (classical and learning-based, e.g. with deep reinforcement learning).
Prerequisites:
We require some form of pre-registration (in addition to the registration via the TUM matching system). Thus, students that are not able to attend the preliminary meeting, please send us an e-mail that you are interested in the seminar such that we can provide you with the necessary information and material for the registration for the seminar
Course related content will be uploaded here.