MARL with General Utilities via Decentralized Shadow Reward Actor-Critic
This addresses the challenge of incorporating diverse objectives like risk and exploration in cooperative MARL, offering a novel decentralized approach with theoretical guarantees.
The paper tackles the problem of enabling cooperation in multi-agent reinforcement learning (MARL) using general utilities beyond cumulative return, such as risk-sensitivity and exploration, and proposes DSAC, which converges to ε-stationarity in O(1/ε^2.5) or O(1/ε^2) steps with high probability and finds globally optimal policies.
We posit a new mechanism for cooperation in multi-agent reinforcement learning (MARL) based upon any nonlinear function of the team's long-term state-action occupancy measure, i.e., a \emph{general utility}. This subsumes the cumulative return but also allows one to incorporate risk-sensitivity, exploration, and priors. % We derive the {\bf D}ecentralized {\bf S}hadow Reward {\bf A}ctor-{\bf C}ritic (DSAC) in which agents alternate between policy evaluation (critic), weighted averaging with neighbors (information mixing), and local gradient updates for their policy parameters (actor). DSAC augments the classic critic step by requiring agents to (i) estimate their local occupancy measure in order to (ii) estimate the derivative of the local utility with respect to their occupancy measure, i.e., the "shadow reward". DSAC converges to $ε$-stationarity in $\mathcal{O}(1/ε^{2.5})$ (Theorem \ref{theorem:final}) or faster $\mathcal{O}(1/ε^{2})$ (Corollary \ref{corollary:communication}) steps with high probability, depending on the amount of communications. We further establish the non-existence of spurious stationary points for this problem, that is, DSAC finds the globally optimal policy (Corollary \ref{corollary:global}). Experiments demonstrate the merits of goals beyond the cumulative return in cooperative MARL.