English

Adversarial Reinforcement Learning: A Duality-Based Approach To Solving Optimal Control Problems

Optimization and Control 2025-07-03 v2

Abstract

We propose an adversarial deep reinforcement learning (ADRL) algorithm for high-dimensional stochastic control problems. Inspired by the information relaxation duality, ADRL reformulates the control problem as a min-max optimization between policies and adversarial penalties, enforcing non-anticipativity while preserving optimality. Numerical experiments demonstrate ADRL's superior performance to yield tight dual gaps. Our results highlight the potential of ADRL as a robust computational framework for high-dimensional stochastic control in simulation-based optimization contexts.

Keywords

Cite

@article{arxiv.2506.00801,
  title  = {Adversarial Reinforcement Learning: A Duality-Based Approach To Solving Optimal Control Problems},
  author = {Nan Chen and Mengzhou Liu and Xiaoyan Wang and Nanyi Zhang},
  journal= {arXiv preprint arXiv:2506.00801},
  year   = {2025}
}

Comments

Accepted by the 2025 Winter Simulation Conference

R2 v1 2026-07-01T02:52:46.768Z