A knowledge-driven shared autonomy human-robot interface for tablet computers

Autonomous and teleoperated robots have been proven to be capable of solving complex manipulation tasks. However, autonomous robots can only be deployed in predefined domains and teleoperation requires full attention of a human operator. Therefore, combining autonomous capabilities of the robot with teleoperation is desirable, yet balancing the work load between robot and operator for intuitive human-robot interfaces is still an open issue. In this paper, we present a knowledge-driven tablet computer application for commanding a robot on a high level of abstraction. The application guides an operators decisions based on the actual world state of the robot and enables the operator to command object-centered actions, which are autonomously interpreted symbolically and geometrically by the robot. We evaluate our approach using the humanoid robot Rollin' Justin for an elaborate manipulation experiment and an user study. We show that our approach efficiently balances the work load between robot and operator and provides an intuitive interface for human-robot interaction.