Semantic modeling and generative programming for a multi-agent simulation in the context of disaster management

Disaster management requires collaborative preparedness among the various stakeholders. Collaborative exercises aim to train stakeholders to apply the plans prepared and to identify potential problems and areas for improvement. As these exercises are costly, computer simulation is an interesting tool to optimize preparation through a wider variety of contexts. However, research on simulation and disaster management focuses on a particular problem rather than on the overall optimization of the plans prepared. This limitation is explained by the challenge of creating a simulation model that can represent and adapt to a wide variety of plans from various disciplines. The work presented in this paper addresses this challenge by adapting the simulation model based on disaster management information and plans integrated into a knowledge base. The simulation model created is then automatically programmed to perform simulation experiments to improve action plans. The results of the experiments are analyzed in order to generate new knowledge to enrich disaster management plans in a virtuous cycle. This paper presents a proof of concept on the French national NOVI plan, for which simulation experiments have made it possible to know the impact of the physicians distribution on the plan application as well as to identify their best distribution.