GCHR : Goal-Conditioned Hindsight Regularization for Sample-Efficient Reinforcement Learning

TL;DR

This paper introduces Hindsight Goal-conditioned Regularization (HGR), a novel method that enhances sample efficiency in goal-conditioned reinforcement learning by generating action priors based on hindsight goals, outperforming existing techniques.

Contribution

The paper proposes HGR, a new regularization technique that, when combined with hindsight self-imitation, significantly improves sample efficiency in off-policy GCRL methods.

Findings

HGR achieves more efficient sample reuse than existing methods.

The combined approach outperforms prior GCRL techniques on navigation and manipulation tasks.

Empirical results demonstrate state-of-the-art performance with the proposed method.

Abstract

Goal-conditioned reinforcement learning (GCRL) with sparse rewards remains a fundamental challenge in reinforcement learning. While hindsight experience replay (HER) has shown promise by relabeling collected trajectories with achieved goals, we argue that trajectory relabeling alone does not fully exploit the available experiences in off-policy GCRL methods, resulting in limited sample efficiency. In this paper, we propose Hindsight Goal-conditioned Regularization (HGR), a technique that generates action regularization priors based on hindsight goals. When combined with hindsight self-imitation regularization (HSR), our approach enables off-policy RL algorithms to maximize experience utilization. Compared to existing GCRL methods that employ HER and self-imitation techniques, our hindsight regularizations achieve substantially more efficient sample reuse and the best performances, which we empirically demonstrate on a suite of navigation and manipulation tasks.