Industrial systems are getting increasingly complex, heterogeneous and decentralized. Consequently their functioning undergoes more and more constraints witch very often have temporal characteristics. Multi-agent techniques are well-adapted to this kind of environment. They are widely used in many applications such as (Control. Supervision. Simulation. Piloting Temporal representations have been studied for modeling and reasoning purposes but they still are at an early stage within a multi-agent context especially on the social level. The aim of this study is to define a multi-agent society where temporal aspects are explicitly taken into account within the individual as well as the social reasoning of an agent. The model we proposed focuses essentially on agent dependencies and interactions. It also permits to study their evolution in time. This generic model is applied to the supervision industrial systems through scenario recognition. In this domain, tasks are always executed under temporal constraints. The STARS system is an implementation of this model. It enables a group of agent-s to recognize scenarios in a distributed way using a temporal reasoning.
