Seminar: Towards Probabilistic (Logic) Programming for Robotics

Recently, the field of AI has devoted a lot of attention to statistical relational learning and probabilistic programming, which provide rich representations for coping with uncertainty, with structure and for learning.  The reason for the current interest is that they promise to bridge the gap between high-level reasoning and low-level perception.

In this talk, I shall introduce the concepts underlying probabilistic logic programming, their semantics, as well as different inference and learning mechanisms, also for use in dynamic settings.  Most of all I shall sketch some emerging applications of these languages in
robotics, where probabilistic programming is used for tracking relational worlds in which objects or their properties are occluded in real time, and for planning in complex (relational) Markov Decision Processes.  We have also employed probabilistic programs to learn complex "relational" affordance models involving multiple objects and their context in a robotics setting.