Multiagent reinforcement learning using function approximation

Learning in a partially observable and nonstationary environment is still o ne of the challenging problems In the area of multiagent (MA) learning. Rei nforcement learning is a generic method that suits the needs of MA learning in many aspects. This paper presents two new multiagent based domain indep endent coordination mechanisms for reinforcement learning; multiple agents do not require explicit communication among themselves to learn coordinated behavior. The first coordination mechanism Is perceptual coordination mech anism, where other agents are included in state descriptions and coordinati on information is Learned from state transitions. The second is observing c oordination mechanism, which also includes other agents in state descriptio ns and additionally the rewards of nearby agents are observed from the envi ronment. The observed rewards and agent's own reward are used to construct an optimal policy. This way, the latter mechanism tends to increase region- wide joint rewards. The selected experimented domain is adversarial food-co llecting world (AFCW), which can be configured both as single and multiagen t environments, Function approximation and generalization techniques are us ed because of the huge state space. Experimental results show the effective ness of these mechanisms.