One of the main challenges in offline Reinforcement Learning (RL) is the distribution shift that arises from the learned policy deviating from the data collection policy. This is often addressed by avoiding out-of-distribution (OOD) actions during policy improvement as their presence can lead to substantial performance degradation. This challenge is amplified in the offline Multi-Agent RL (MARL) setting since the joint action space grows exponentially with the number of agents. To avoid this curse of dimensionality, existing MARL methods adopt either value decomposition methods or fully decentralized training of individual agents. However, even when combined with standard conservatism principles, these methods can still result in the selection of OOD joint actions in offline MARL. To this end, we introduce AlberDICE, an offline MARL algorithm that alternatively performs centralized training of individual agents based on stationary distribution optimization. AlberDICE circumvents the exponential complexity of MARL by computing the best response of one agent at a time while effectively avoiding OOD joint action selection. Theoretically, we show that the alternating optimization procedure converges to Nash policies. In the experiments, we demonstrate that AlberDICE significantly outperforms baseline algorithms on a standard suite of MARL benchmarks.

Value decomposition methods have gained popularity in the field of cooperative multi-agent reinforcement learning. However, almost all existing methods follow the principle of Individual Global Max (IGM) or its variants, which limits their problem-solving capabilities. To address this, we propose a dual self-awareness value decomposition framework, inspired by the notion of dual self-awareness in psychology, that entirely rejects the IGM premise. Each agent consists of an ego policy for action selection and an alter ego value function to solve the credit assignment problem. The value function factorization can ignore the IGM assumption by utilizing an explicit search procedure. On the basis of the above, we also suggest a novel anti-ego exploration mechanism to avoid the algorithm becoming stuck in a local optimum. As the first fully IGM-free value decomposition method, our proposed framework achieves desirable performance in various cooperative tasks.

Let a two-player Markov game where both players affect the transition. We will effectively show that the problem of best-responding to a correlated policy σ is526 equivalent to best-responding to the marginal policy of σ for the opponent. The proof follows from527 the equivalence of the two MDPs.528 Before that, given a (possibly correlated) joint policy σ we define a nonlinear program, (PBR), whose539 optimal solutions are best-response policies of each agent k to σ k and the values for each state s540 and timestep h:541 A.2 Proof of Theorem 3.2542 The best-response program. First, we state the following lemma that will prove useful for several543 of our arguments,544 Lemma A.1 (Best-response LP).

Team-Nash equilibrium (TNE) predicts the outcomes of such coordinated interactions.

Let a two-player Markov game where both players affect the transition. As we have seen in Section 2.1, in the case of unilateral deviation from joint policy Let a (possibly correlated) joint policy ˆ σ . By Lemma A.1, we know that Where the equality holds due to the zero-sum property, (1). An approximate NE is an approximate global minimum. An approximate global minimum is an approximate NE.

One such example is two-player zero-sum Markov games, in which efficient ways to compute a Nash equilibrium are known. Inspired by zero-sum polymatrix normal-form games (Cai et al., 2016), we define a class of zero-sum

Insuch applications, a team of agents aim to maximize a joint expected utility through a single global reward.