Multi-agent systems and their application to control vehicle underwater

An autonomous mobile robot must perform nonrepetitive tasks in an imperfectly known environment and uncooperative and even hostile. In this context the missions assigned to the underwater vehicle can't be defined precisely, and this vehicle should have the capability to interpret, analyze the environment, decide on appropriate action and react to asynchronous events. It also must permanently reconfigure to adapt to external conditions and objectives. To fill the requirements and to harmonize decision, reaction and performance with distributed intelligence, the control architecture proposed it's an hybrid architecture, based on multi-agent systems, combines the benefits of reactive and deliberative architectures. In this work, we first studied the principle of the underwater vehicle; the analysis identified the desired characteristics. In the second part, we focused on multi-agent systems (MAS) in order to understand the link between the approach "Distributed Artificial Intelligence" and our project. After discussing the different control architectures in the third part, we finally treat the solution proposed in this article. The development of our architecture is based on multi-agent systems and the modeling of underwater vehicle which will be addressed in another paper. These developments are part of the overall project initiated by the team of the Computer Laboratory, systems and renewable energy (LISER) of the National School of Electrical and Mechanical (ENSEM).