Master Thesis Title: Learning-based World Representation

A robot that moves through the world has to remember what it has seen. Classical mapping stores this memory in hand-designed structures such as grids and graphs, which fix the size, the resolution and the kind of information in advance. Hand-designed maps have limits. A grid has a fixed resolution and extent, a scene graph holds only what someone decided to put in it, and combining geometry with semantics takes extra engineering. Recent models can infer geometry directly from images, and sequence models can carry information over time. This opens many directions to explore: can we learn from perceptive data streams, take the role of the map? In this thesis, you investigate how to encode a stream of perception sensor data (e.g. RGB images) into a compact latent representation, reconstruct features such as geometry or semantics from it, or use such embeddings directly for a downstream task. Environment: The thesis is conducted at the RAI Institute, a top-notch robotics research institute with representation in Zurich. It would happen as part of a collaboration agreement between RAI and ETH Zurich. A mature robotics stack running on a multitude of robots and enabling fast testing cycles, strong engineering support, and a best-in-class robotics research laboratory with state-of-the-art computing infrastructures led by distinguished scientists in the field make this opportunity second to none. For examples of the platforms and deployments developed at the institute, please have a look at the videos on our channel https://www.youtube.com/@rai- inst/videos Please append your CV, letter of motivation, and transcripts to the application. Selection is highly competitive. If considered, you will be invited to an on-site interview.