Select before Act: Spatially Decoupled Action Repetition for Continuous Control

TL;DR

This paper introduces SDAR, a novel reinforcement learning framework that performs spatially decoupled action repetition, allowing each action dimension to be repeated independently, which improves flexibility, sample efficiency, and overall policy performance in continuous control tasks.

Contribution

The paper proposes SDAR, a new method for action repetition that decouples repetitions across action dimensions, enhancing flexibility and efficiency over existing methods.

Findings

SDAR improves sample efficiency in continuous control tasks.

SDAR achieves higher policy performance compared to traditional repetition methods.

SDAR reduces action fluctuation, leading to more stable control policies.

Abstract

Reinforcement Learning (RL) has achieved remarkable success in various continuous control tasks, such as robot manipulation and locomotion. Different to mainstream RL which makes decisions at individual steps, recent studies have incorporated action repetition into RL, achieving enhanced action persistence with improved sample efficiency and superior performance. However, existing methods treat all action dimensions as a whole during repetition, ignoring variations among them. This constraint leads to inflexibility in decisions, which reduces policy agility with inferior effectiveness. In this work, we propose a novel repetition framework called SDAR, which implements Spatially Decoupled Action Repetition through performing closed-loop act-or-repeat selection for each action dimension individually. SDAR achieves more flexible repetition strategies, leading to an improved balance between action persistence and diversity. Compared to existing repetition frameworks, SDAR is more sample efficient with higher policy performance and reduced action fluctuation. Experiments are conducted on various continuous control scenarios, demonstrating the effectiveness of spatially decoupled repetition design proposed in this work.