On March 19, 2021, Stefan Lüdtke defended his dissertation titled "Lifted Bayesian Filtering in Multi-Entity Systems".
The thesis investigates symmetry properties of multi-entity systems, i.e. systems that consist of multiple, interacting agents. Estimating the non-observable state of such systems is highly challenging. The thesis identifies symmetries in the underlying probabilistic model as a fundamental property of such systems and presents inference algorithms that can detect and make use of such symmetries. This way, computations can be carried out more efficiently than by existing algorithms, often by several orders of magnitude.
These methods are, for instance, highly relevant for sensor-based multi-agent activity recognition, which is a fundamental building block of ingelligent, situation-aware assistants. They allow to perform activity recognition in large multi-agent systems, for which activity recognition was previously infesible.
Due to the high methodological innovation and the impact for sensor-based human activity recognition, the reviewers and the Faculty of Computer Science and Electrical Engineering awarded the grade of "summa cum laude".