Copilot MIKE: An Autonomous Assistant for Multi-Robot Operations in Cave Exploration

Abstract

Operating a team of robots under time and risk constraints can be challenging for a human operator. Environmental conditions, extrinsic risks, and accessibility might restrict humans from directly partaking in exploration tasks altogether. Hence, robotic systems with autonomous exploration and disaster response capabilities have evolved over the past years and help keep human explorers and emergency response teams from harm. In this work, we introduce Copilot MIKE, an autonomous assistant for human-in-the-loop multi-robot operations. Copilot MIKE assists a single operator in monitoring robot teams, strategic planning, and communicating high level commands to the robots. During complex and potentially stressful exploration missions, Copilot MIKE helps to maintain a bearable workload and high situational awareness. In this work, we mainly focus on cave exploration tasks in the context of the DARPA Subterranean Challenge (SubT), but we designed a generic assistant that can be used in other domains, such as search and rescue, science, and (space) exploration missions as well. Experimental mission runs were conducted in preparation for the SubT cave challenge and Copilot MIKE has been tested in realistic cave exploration simulations. We show that Copilot MIKE has the potential to reduce workload, while our operators place trust in the system. They report that they focused on important parts of a mission, rather than planning, adopting and memorizing a complete mission strategy themselves.