Market Based Adaptive Resource Allocation for Distributed Rescue Teams

The dynamic nature of real-world rescue scenarios (e.g., military, emergency first response, hurricane relief) requires constant reevaluation of resource assignments. New events can trigger additional resource requirements generating conflicts about how to reassign resources across tasks in an emerging crisis. Reallocation is further complicated as some resources are synergistic (i.e., helicopter and pilot) and many distributed rescue teams have limited information about other teams' status. We show how integrating a team-based multi-agent planning system with standard combinatorial auction methods to dynamically re-allocate resources can maximize overall rescue utility while providing for graceful managed degradation under conditions of extreme stress. The key innovation of our approach is that we explicitly provide a framework that incorporates the costs involved in dynamically switching resources from one task to another. We compare our system's performance against two other approaches.