Assessing large scale emergency rescue plans: an agent based Approach
Résumé
In this paper we describe a multi-agent based approach for modeling and simulating the dynamics of large scale crisis situations. Our aim is to show how agent-based approach can benefit modeling-simulating a complex socio-technical collaborative complex system. The development of the model and simulator is based on data obtained from field studies of quasi-realistic non-computer based simulations and from official French rescue plans. Using our agentbased simulator we conduct samples of tests of various combinations of heuristics, both real and hypothetical in order to assess various potential rescue plans. To illustrate this, we focus on two organizational strategies combined with two communication modes to assess how they affect rescue process efficiency in terms of the global evacuation time and the rescue rate. Seven selected configurations have been simulated for each of the four typical scenarios ((centralized or distributed strategy) and (traditional paper or new electronic forms)). These first simulations show that: (1) electronic communication reduces delays and victims losses; (2) although distribution strategy improves rescue in general, centralized rescue process remains the most efficient in case of having many victims and few rescuers.