Hypercube Policy Regularization Framework for Offline Reinforcement Learning
Yi Shen, Hanyan Huang
TL;DR
Offline reinforcement learning suffers from distribution shift when learning from fixed datasets. The authors propose a hypercube policy regularization framework that partitions the state space into cubes using a precision parameter and enables localized exploration within cubes, integrating with TD3-BC and Diffusion-QL to form TD3-BC-C and Diffusion-QL-C. The theoretical analysis under a Lipschitz Q-function supports that increasing hypercube granularity can preserve or improve performance, with the cube diameter controlling the trade-off. Empirically, integrating the framework with TD3-BC and Diffusion-QL yields state-of-the-art results among policy-constrained offline methods on many D4RL Gym tasks, especially in suboptimal-data regimes, and the code is released open-source.
Abstract
Offline reinforcement learning has received extensive attention from scholars because it avoids the interaction between the agent and the environment by learning a policy through a static dataset. However, general reinforcement learning methods cannot get satisfactory results in offline reinforcement learning due to the out-of-distribution state actions that the dataset cannot cover during training. To solve this problem, the policy regularization method that tries to directly clone policies used in static datasets has received numerous studies due to its simplicity and effectiveness. However, policy constraint methods make the agent choose the corresponding actions in the static dataset. This type of constraint is usually over-conservative, which results in suboptimal policies, especially in low-quality static datasets. In this paper, a hypercube policy regularization framework is proposed, this method alleviates the constraints of policy constraint methods by allowing the agent to explore the actions corresponding to similar states in the static dataset, which increases the effectiveness of algorithms in low-quality datasets. It was also theoretically demonstrated that the hypercube policy regularization framework can effectively improve the performance of original algorithms. In addition, the hypercube policy regularization framework is combined with TD3-BC and Diffusion-QL for experiments on D4RL datasets which are called TD3-BC-C and Diffusion-QL-C. The experimental results of the score demonstrate that TD3-BC-C and Diffusion-QL-C perform better than state-of-the-art algorithms like IQL, CQL, TD3-BC and Diffusion-QL in most D4RL environments in approximate time.