Adopting pedestrian navigation techniques for multi-robot coordination

Even when walking in complex environments, human beings are able to easily share space and navigate comfortably past each other without explicit communication. Our recent work has sought to build models of this human navigation for the purposes of producing high-quality, cooperative, collision-free motion in multi-agent navigation tasks. We will first describe some recent approaches to modeling pedestrian interaction. Next, we discuss how these models can be adapted to automate mutual collision avoidance in multi-agent environments. Lastly, we briefly discuss some recent applications of these techniques to collaborative analysis of evacuation route planning.