Active Learning of Users Preferences on Robot Behaviour

An important challenge in human robot interaction (HRI) is enabling non-expert users to specify complex tasks for autonomous robots. Recently, active preference learning has been applied in HRI to interactively shape a robot's behaviour. In this project we study frameworks where users specify constraints on allowable robot movements on a graphical interface, yielding a robot task specification. However, users may not be able to accurately assess the impact of such constraints on the performance of a robot. Thus, specification are revised by iteratively presenting users with alternative solutions where some constraints might be violated, and learn about the importance of the constraints from the users' choices between these alternatives.

The practical validation of this project includes user studies with a material transport task in an industrial facility. In a first study nearly all users accepted alternative solutions and thus obtain a revised specification through the learning process, and that the revision leads to a substantial improvement in robot performance. Further, the users whose initial specifications had the largest impact on performance benefit the most from the interactive learning. Recently, the work has been extended to learning from corrections, where user's alter the robots behaviour by manipulating the current trajectory.