Learning General Parameterized Policies for Infinite Horizon Average Reward Constrained MDPs via Primal-Dual Policy Gradient Algorithm

The framework of Reinforcement Learning (RL) is concerned with a class of problems where an agent learns to yield the maximum cumulative reward in an unknown environment via repeated interaction. RL finds applications in diverse areas, such as wireless communication, transportation, and epidemic control (Yang et al., 2020; Al-Abbasi et al., 2019; Ling et al., 2023). RL problems are mainly categorized into three setups: episodic, infinite horizon discounted reward, and infinite horizon average reward. Among them, the infinite horizon average reward setup is particularly significant for real-world applications.