Reinforcement social learning of coordination in cooperative multiagent systems
Jianye Hao, Ho-fung Leung · 2013
Coordination in cooperative multiagent systems is an im-portant problem and has received a lot of attention in mul-tiagent learning literature. Most of previous works study the problem of how two (or more) players can coordinate on Pareto-optimal Nash equilibrium(s) through fixed and repeated interactions in the context of cooperative games. However, in practical complex environments, the interac-tions between agents can be sparse, and each agent’s inter-acting partners may change frequently and randomly. To this end, in this paper, we investigate the multiagent co-ordination problems in cooperative environments under the social learning framework, in which there exists a large pop-ulation of agents and each agent interacts with another agent randomly in each round. Each agent learns its policy through repeated interactions with the rest of agents via social learn-ing. We distinguish two different types of learners depending on the amount of information each agent can perceive: indi-vidual action learner and joint action learner. The learning performance of both types of learners are evaluated under a number of challenging deterministic and stochastic cooper-ative games.