A unified approach integrating human shared mental models with intelligent autonomous team formation for crisis management

Autonomous systems are being exceedingly used to assist humans in various crisis responses scenarios such as earthquakes and nuclear disasters. Because they operate in highly unstructured and uncertain environments, failures are an inherent part of such autonomous systems, and, techniques for making these systems robust to failures arising from computer hardware, software or communication malfunctions are already integrated into their design. However, an important aspect while designing such systems is often times overlooked: how to better coordinate and communicate across distributed, possibly diverse human teams who are working in cooperation with autonomous systems into the design of the autonomous system itself. Unfortunately, this results in limited adoption of autonomous systems in real-life crisis scenarios. In this working paper, we describe ongoing work that attempts to address this deficit by integrating research on shared mental models between humans with techniques for autonomous agent team formation in the context of search and rescue scenarios.