ACE : Off-Policy Actor-Critic with Causality-Aware Entropy Regularization

TL;DR

This paper introduces ACE, a causality-aware entropy regularization method for off-policy actor-critic algorithms that improves exploration and performance in continuous control tasks by prioritizing impactful actions and preventing gradient dormancy.

Contribution

The paper proposes a novel causality-aware entropy term and a dormancy-guided reset mechanism, enhancing exploration and sample efficiency in off-policy RL algorithms.

Findings

Achieves superior performance on 29 continuous control tasks

Effectively identifies and prioritizes impactful primitive behaviors

Demonstrates versatility and sample efficiency across domains

Abstract

The varying significance of distinct primitive behaviors during the policy learning process has been overlooked by prior model-free RL algorithms. Leveraging this insight, we explore the causal relationship between different action dimensions and rewards to evaluate the significance of various primitive behaviors during training. We introduce a causality-aware entropy term that effectively identifies and prioritizes actions with high potential impacts for efficient exploration. Furthermore, to prevent excessive focus on specific primitive behaviors, we analyze the gradient dormancy phenomenon and introduce a dormancy-guided reset mechanism to further enhance the efficacy of our method. Our proposed algorithm, ACE: Off-policy Actor-critic with Causality-aware Entropy regularization, demonstrates a substantial performance advantage across 29 diverse continuous control tasks spanning 7 domains compared to model-free RL baselines, which underscores the effectiveness, versatility, and efficient sample efficiency of our approach. Benchmark results and videos are available at https://ace-rl.github.io/.