Distributed multi-robot evacuation incorporating human behavior

It is well known that trained leaders have an important influence on saving human lives in emergency evacuation. In this paper, we present a novel distributed multi-robot system for guiding people in an emergency evacuation mission. A cooperative exit-seeking algorithm is designed for the robots to guide evacuees by online estimating the gradient and tracing gradient-descent while maintaining a predefined formation in movement. To better deal with evacuees' behaviors in emergency situations, two human panic behavior models are taken into account in the evacuation strategies. Guidance algorithms of single robot team and multi-team are introduced in the evacuation system. We evaluate the proposed method through comparing evacuation time with non-robot-assisted procedures in the static and dynamic routing simulations. The results show that our method can significantly shorten the evacuation time.