Decentralized control in the pursuit domain
Résumé
We study the prey-predator problem as a multi-agent system (MAS). We consider an hierarchical approach to searching optimal policies for each agent to capture the prey. In particular, we imagine decomposition of the state-space, and of the "situation" space. In other words, we consider an hierarchy of policies, the agent following the one which is most appropriate for his current state-space and for the situation in that state-space.