Interactive Learning of Probabilistic Decision Making by Service Robots with Multiple Skill Domains

This thesis makes a contribution to autonomous service robots, centered around two aspects. The first is modeling decision making in the face of incomplete information on top of diverse basic skills of a service robot. Second, based on such a model, it is investigated, how to transfer complex decision-making knowledge into the system. Interactive learning, naturally from both demonstrations of human teachers and in interaction with objects, yields decision-making models applicable by the robot.