Linear Convergence of Independent Natural Policy Gradient in Games With Entropy Regularization

This letter focuses on the entropy-regularized independent natural policy gradient (NPG) algorithm in multi-agent reinforcement learning. In this letter, agents are assumed to have access to an oracle with exact policy evaluation and seek to maximize their respective independent rewards. Each individual’s reward is assumed to depend on the actions of all agents in the multi-agent system, leading to a game between agents. All agents make decisions under a policy with bounded rationality, which is enforced by the introduction of entropy regularization. In practice, a smaller regularization implies that agents are more rational and behave closer to Nash policies. On the other hand, with larger regularization agents tend to act randomly, which ensures more exploration. We show that, under sufficient entropy regularization, the dynamics of this system converge at a linear rate to the quantal response equilibrium (QRE). Although regularization assumptions prevent the QRE from approximating a Nash equilibrium (NE), our findings apply to a wide range of games, including cooperative, potential, and two-player matrix games. We also provide extensive empirical results on multiple games (including Markov games) as a verification of our theoretical analysis.