Global Convergence for Average Reward Constrained MDPs with Primal-Dual Actor Critic Algorithm

TL;DR

This paper introduces a primal-dual actor-critic algorithm for average reward constrained MDPs that guarantees global convergence and near-optimal constraint violation rates, advancing theoretical understanding in this domain.

Contribution

It presents a novel primal-dual natural actor-critic algorithm with proven global convergence and rate guarantees for average reward CMDPs, even without knowledge of mixing time.

Findings

Achieves $ ilde{O}(1/ oot{T}{})$ convergence rate with known mixing time

Maintains near-optimal rates without mixing time knowledge under certain conditions

Establishes new theoretical benchmarks matching lower bounds for average reward CMDPs

Abstract

This paper investigates infinite-horizon average reward Constrained Markov Decision Processes (CMDPs) with general parametrization. We propose a Primal-Dual Natural Actor-Critic algorithm that adeptly manages constraints while ensuring a high convergence rate. In particular, our algorithm achieves global convergence and constraint violation rates of $\tilde{\mathcal{O}}(1/\sqrt{T})$ over a horizon of length $T$ when the mixing time, $\tau_{\mathrm{mix}}$, is known to the learner. In absence of knowledge of $\tau_{\mathrm{mix}}$, the achievable rates change to $\tilde{\mathcal{O}}(1/T^{0.5-\epsilon})$ provided that $T \geq \tilde{\mathcal{O}}\left(\tau_{\mathrm{mix}}^{2/\epsilon}\right)$. Our results match the theoretical lower bound for Markov Decision Processes and establish a new benchmark in the theoretical exploration of average reward CMDPs.