Improving tasks allocation and coordination in a rescue scenario

Abstract

This paper presents a proposal aimed at improving the task allocation and the coordination of rescue agents in a disaster scenario. Task allocation is related to allocating victims to be rescued by each agent, while coordination is related to how these agents interact among themselves to save a larger number of victims. Rescuing lives in disasters takes place in a dynamic environment in which new victims appear all the time. In addition, the victims' injuries aggravate over time. A due-date-based technique is presented to schedule the tasks in the rescue operations. In this sense, the victims are sequenced in a non-decreasing death time order. To do that, our rescue agents search for the victims and predict their death time to establish a victim priority order in our scheduling algorithm and coordinate the rescue agents. Past rescue experiences of our rescue team are used in the victims' death time estimations. Several experiments using the RoboCup Rescue simulator are presented to prove the feasibility of our approach to improving multiagent performance.