Continued Advances in Supervised Autonomy User Interface Design for METERON SUPVIS Justin

Abstract

The exploration of the universe remains a challenging endeavor, constantly pushing the limits of technology. Of special interest is the investigation of the other planets of our solar system such as Mars, which has been examined by various tele-operated and (semi-) autonomous satellites and landers. But an important milestone that is needed for a deeper understanding of the planet is still missing: A crewed landing. In order to send humans to such a remote location, an infrastructure for the landing crew including an energy supply, a habitat, and a return vehicle needs to be provided on the surface of the planet. The construction and maintenance of these structures is envisioned to be done by semiautonomous robots that are commanded from orbiting spacecrafts. The teleoperation of such ground-based robots poses high demands on the capabilities of the system including robot autonomy, orbiter-robot communication, and human-robot interface design. The METERON SVPVIS Justin space telerobotics experiment suite has been initiated by the German Aerospace Center (DLR) together with the European Space Agency (ESA) to investigate the requirements for such a system and evaluate an approach. During the experiment sessions, astronauts onboard the International Space Station (ISS) command DLR's humanoid service robot Rollin' Justin on Earth to execute complex surveillance, service, and repair tasks in a simulated Martian solar farm. The robot uses its local intelligence to support the astronaut operator upon task completion allowing a simple intuitive command interface and lowering the requirements on the communication link. This work gives an overview of the developed robotic system, communication link, and tablet computer user interface (UI). In particular the tight coupling between the autonomy system of the robot and the UI, that allows the intuitive robot commanding including action parameterization, is described in detail. The first space-ground experiment sessions of METERON SUPVIS Justin were conducted in August 2017, and March 2018 by four astronauts in total. During the first session, three astronauts demonstrated the operational readiness of our system by commanding Rollin' Justin to perform surveillance and inspection tasks. The astronauts were even able to successfully command the robot in scenarios, which were not trained prior to their spaceflight. This was possible, because our astronaut-robot collaboration concept efficiently guides the operator towards task completion. We used this property in the second experiment session to evaluate our system in even more complex scenarios. While in the first session it was sufficient for the astronaut to select the correct commands, the operator was now required to manually parameterize some of the commands to optimize the task outcome. By that, the robot has been successfully commanded to perform complex maintenance and adjustment tasks in the simulated Martian solar farm. In this work, we evaluate the preliminary results of the space-ground experiments and discuss the feedback we received from the astronauts and its impact on future space telerobotics UI design.