Results in Sliding Autonomy for Multi-robot Spatial Assembly

We have developed a software architecture for teams of robots and humans to jointly perform tightly coordinated tasks, such as assembly of structures in orbit or on planetary surfaces. While we envision that robots will autonomously perform such work in the future, the state of the art falls short of the capabilities necessary to handle all possible contingencies. Our architecture provides a principled methodology for human involvement to optimize both task efficiency and robustness by combining robot capability with human intuition. We call such mixed control strategies "sliding autonomy". Robots accomplish as many of the tasks as they can autonomously, and human operators take over control to perform those that cannot be easily automated or to provide help when the robots fail. In this paper, we discuss results from recent experiments that quantify the effect of different levels of autonomy on the system's overall performance. By introducing two modes of sliding autonomy, we are able to achieve the high reliability of a teleoperated system combined with the high efficiency of autonomous operation. The incurred mental demand of the operator is directly proportional to the increase in system efficiency.