Coordination of cooperation policies in a peer-to-peer system using swarm-based RL

The performance of a system of interacting peers depends strongly on their individual resource contributions. In this paper, we have devised a self-organized coordination mechanism for cooperation policy setting of rational peers that have only partial views of the whole peer-to-peer system in order to improve the overall welfare of the system. The proposed mechanism is based on a distributed Reinforcement Learning (RL) approach and sets cooperation policies of the peers through their self-organized interactions by exchanging the local value functions among the neighbors. We demonstrate that a Pareto optimal equilibrium emerges in the system from fair cooperation of the constituent peers.