Near-optimal emergency evacuation with rescuer allocation

Emergency management systems are Cyber-Physical-Human Systems (CPHS) that use sensing, together with communications and control, to guide humans and physical systems such as vehicles, towards safe desirable outcomes in the shortest possible time. When human health and safety, and lives also, are at stake, it is important to take decisions in real-time with the best possible use of resources, including the critical resource of emergency personnel. Such distributed decision problems are so complex that the resulting optimisation and allocation problems can only be handled in a fast and timely manner using efficient heuristics. Thus in this paper we apply a recent resource allocation algorithm based on the Random Neural Network (RNN) to allocate rescuers to those evacuees whose health level has deteriorated beyond a certain level in the course of an evacuation. The approach is evaluated by simulating the evacuation of a three story building using the Distributed Building Evacuation Simulator (DBES) developed at Imperial College. The simulations show that the outcome of the evacuation can be significantly improved in this manner, in particular for larger numbers of evacuees.