Prioritized Trajectory Replay: A Replay Memory for Data-driven Reinforcement Learning

TL;DR

This paper introduces a trajectory-based replay memory technique, Prioritized Trajectory Replay, that improves data sampling efficiency and performance in offline reinforcement learning by leveraging trajectory information and prioritized sampling.

Contribution

It proposes a novel trajectory sampling method, extending replay memory to trajectories, and introduces prioritized sampling to enhance offline RL performance.

Findings

Improved offline RL performance on D4RL benchmarks.

Enhanced data efficiency through backward trajectory sampling.

Effective avoidance of unseen actions during training.

Abstract

In recent years, data-driven reinforcement learning (RL), also known as offline RL, have gained significant attention. However, the role of data sampling techniques in offline RL has been overlooked despite its potential to enhance online RL performance. Recent research suggests applying sampling techniques directly to state-transitions does not consistently improve performance in offline RL. Therefore, in this study, we propose a memory technique, (Prioritized) Trajectory Replay (TR/PTR), which extends the sampling perspective to trajectories for more comprehensive information extraction from limited data. TR enhances learning efficiency by backward sampling of trajectories that optimizes the use of subsequent state information. Building on TR, we build the weighted critic target to avoid sampling unseen actions in offline training, and Prioritized Trajectory Replay (PTR) that enables more efficient trajectory sampling, prioritized by various trajectory priority metrics. We demonstrate the benefits of integrating TR and PTR with existing offline RL algorithms on D4RL. In summary, our research emphasizes the significance of trajectory-based data sampling techniques in enhancing the efficiency and performance of offline RL algorithms.