Abstract: This paper presents an LR-I lagging anchor algorithm that combines a lagging anchor method with an LR-I learning algorithm. We prove that this decentralized learning algorithm converges in strategies to Nash equilibria in two-player two-action general-sum matrix games. A practical LR-I lagging anchor algorithm is introduced for players to learn their Nash equilibrium strategies in general-sum stochastic games. Simulation results show the performance of the proposed LR-I lagging anchor algorithm in both matrix games and stochastic games. PDF
Keywords: Multiagent learning, Matrix Games, Game Theory