Planning Cooperative Homogeneous Multiagent System Using Markov Decision Processes
Résumé
This paper proposes a decisiontheoric approach for designing a set of situated agents so that they can solve a cooperative problem. The design of each agent's behavior results from solving a decentralized partially observable markov decision process (DECPOMDP). But, as solving a DECPOMDP is NEX Pcomplete, we propose an approximate solution to this problem based on subjectivity and empathy.