Publication
Shared control strategies can improve task-performance in teleoperation. In such systems automation guides or corrects a human operator. The amount of correction or guidance that is provided is denoted the level-of-automation. As the variety of teleoperation tasks is large, manually specifying the underlying automation is time consuming. In this work we present an approach to program this automated system by demonstration. Our approach determines the level of automation online, by combining the confidence of automation and teleoperator. We present particular implementations of our approach for haptic shared control, and state shared control. The method is evaluated in a user study. Although the users indicated they preferred the learned shared control strategies, teleoperation performance did not improve our metric (task execution time).
Rachid Guerraoui, Alexandre David Olivier Maurer, Omid Ardakanian