Bayesian cognitive robotics is a novel paradigm for the knowledge-enabled control of autonomous robot. The paradigm presumes that one of the most powerful ideas to equip robots with comprehensive reasoning capabilities is the lifelong autonomous learning of joint probability distributions over robot control programs, the behaviour they generate and the situation-dependent effects they bring about. Having learned such probability distributions from experience, a robot can make predictions, diagnoses and perform other valuable inference tasks in order to improve its problem-solving performance.
In this project, we aim to advance autonomous learning in Bayesian cognitive robotics by (1) extending a plan language to allow the autonomous collection of semantically interpretable data, and (2) autonomously generating the structure and learning the parameters of probabilistic models and inducing probabilistic rules for planning. The resulting software components will be integrated into an autonomous robot control system and empirically investigated in the context of human-scale and real world robot manipulation tasks such as preparing a simple meal or cleaning up a table.