Enhancing Human-Robot Connection in Space Teleoperation through Shared Mental Model Creation

Kurzfassung

Previous research in human-robot-interaction in teleoperation suggests that human-robot teams with a SMM are able to achieve better task performances and lower the mental workload of the user. To create a SMM, the robot must be able to estimate the task model of the user and incorporate knowledge about the user into its user model. Most existing concepts use probabilistic methods or symbolic reasoning frameworks, with predefined features for estimating the user's teleoperation goal. They are therefore highly dependent on a-priori knowledge about the task, thus limiting their flexibility. Only very few approaches mention or explicitly take into account knowledge about the individual user and lack adaptability to human behavior and user preferences. ML based concepts, on the other hand, are more flexible to adapt their SMM during the user interaction, but require a lot of data points for precise estimation. In this paper, we propose a new approach to address these limitations by combining the advantages of symbolic and subsymbolic methods, and dividing our estimation of the user's teleoperation goal into subgoals next to the overall task goal. Furthermore, by integrating the knowledge about the individual user and human decision-making into the prediction of the user's task goal, we can improve SMM creation of human-robot teams for teleoperation. Our approach is applied to an example space exploration task in a planetary surface habitat by a telerobotic team. The SMM implementation is described and analyzed, followed by an outlook on future development and validation.