Distributed Scheduling Agents for Disaster Response

In this paper, we describe the application of a multi-agent framework for collaborative scheduling to a disaster response coordination problem. The target problem is a field exercise mockup of a natural disaster, where a team of human agents must rely on their respective automated scheduling agents to coordinate and accomplish various infra-structure repair and casualty transport tasks. Following the ground rules of the program sponsoring the experiment, our starting point is a collaborative scheduling framework developed under the strong assumption that no single agent has a global view of the overall problem. This peer-to-peer approach to multi-agent scheduling and coordination gives rise to a complex distributed search problem, and effective performance in the field exercise is found to depend heavily on the ability to provide the multi-agent system with strong initial strategic guidance. We describe the mechanisms developed for steering the multi-agent scheduling system to address specific disaster response scenarios and report performance results that were obtained in the field exercise test cases.