Peer Instruction to Assist a Teleautonomous USAR System

While the goal of Urban Search and Rescue is the autonomous mapping of a damaged building coupled with the location of human victims, artificial intelligence is not yet advanced enough to fully support such a rich and complex domain. Human teleoperation of robots is still relied upon extensively in this and many other domains, but suffers because of operator fatigue and problems with situational awareness. These two factors have led to recent research on combined teleautonomous approaches. The work described here is intended to supplement teleautonomous approaches in multi-robot settings by allowing robots to recognize specific situations in which they can assist their peers, thus allowing a teleoperator’s limited attention to be better spread around a team of agents. We overview the techniques we employ to assist others in the robotic rescue domain, and describe how they integrate into an existing teleautonomous robot control system for the robotic rescue. We also evaluate the performance of these facilities in a simulated robotic rescue domain.