English

Towards Painless Policy Optimization for Constrained MDPs

Machine Learning 2022-04-12 v1 Artificial Intelligence

Abstract

We study policy optimization in an infinite horizon, γ\gamma-discounted constrained Markov decision process (CMDP). Our objective is to return a policy that achieves large expected reward with a small constraint violation. We consider the online setting with linear function approximation and assume global access to the corresponding features. We propose a generic primal-dual framework that allows us to bound the reward sub-optimality and constraint violation for arbitrary algorithms in terms of their primal and dual regret on online linear optimization problems. We instantiate this framework to use coin-betting algorithms and propose the Coin Betting Politex (CBP) algorithm. Assuming that the action-value functions are εb\varepsilon_b-close to the span of the dd-dimensional state-action features and no sampling errors, we prove that TT iterations of CBP result in an O(1(1γ)3T+εbd(1γ)2)O\left(\frac{1}{(1 - \gamma)^3 \sqrt{T}} + \frac{\varepsilon_b\sqrt{d}}{(1 - \gamma)^2} \right) reward sub-optimality and an O(1(1γ)2T+εbd1γ)O\left(\frac{1}{(1 - \gamma)^2 \sqrt{T}} + \frac{\varepsilon_b \sqrt{d}}{1 - \gamma} \right) constraint violation. Importantly, unlike gradient descent-ascent and other recent methods, CBP does not require extensive hyperparameter tuning. Via experiments on synthetic and Cartpole environments, we demonstrate the effectiveness and robustness of CBP.

Keywords

Cite

@article{arxiv.2204.05176,
  title  = {Towards Painless Policy Optimization for Constrained MDPs},
  author = {Arushi Jain and Sharan Vaswani and Reza Babanezhad and Csaba Szepesvari and Doina Precup},
  journal= {arXiv preprint arXiv:2204.05176},
  year   = {2022}
}

Comments

Paper under submission. 27 pages, 12 figures

R2 v1 2026-06-24T10:44:38.193Z