Title:  MASVERP: A Multi-agents System for Safety Interventions on SEVESO Sites

Authors: Lydie Edward, Domitile Lourdeaux, Jean-Marie Burkhardt and Jean-Paul Barthes

Abstract. Virtual reality and simulation give us today new tools for improving the training 
or making better decisions in the domain of risk prevention thanks to virtual autonomous 
characters. In our work, we develop such a tool allowing storyboarding hazardous working 
situations on Seveso-type sites. Our architecture is based on a multi-agents system MASVERP 
(Multi-agents System in Virtual Environments for Risk Prevention) including virtual operators 
represent by our agents (cognitive and reactive) and human operators. The system interprets 
a cognitive activity and a related risk model resulting from field analyses. In the proposed 
environment a manager can visualize the risks incurred during an intervention. The emergent 
risks depend on the cognitive characteristics of the operators (human factors). The difference 
with classic MAS is that our cognitive agents are enriched with a planner that decides the 
action to realize according to their objectives, the environment and their personal 
characteristics (temporal pressure, cautious, tiredness, hunger).