Actor-Critic for Continuous Action Chunks: A Reinforcement Learning Framework for Long-Horizon Robotic Manipulation with Sparse Reward

TL;DR

AC3 is a reinforcement learning framework that learns continuous action chunks for long-horizon robotic tasks, improving stability and data efficiency through novel stabilization and reward mechanisms.

Contribution

Introduces AC3, a new RL method for stable, data-efficient learning of continuous action sequences in robotic manipulation with sparse rewards.

Findings

Achieves higher success rates on benchmark tasks.

Uses few demonstrations with simple models.

Effective stabilization mechanisms improve learning stability.

Abstract

Existing reinforcement learning (RL) methods struggle with long-horizon robotic manipulation tasks, particularly those involving sparse rewards. While action chunking is a promising paradigm for robotic manipulation, using RL to directly learn continuous action chunks in a stable and data-efficient manner remains a critical challenge. This paper introduces AC3 (Actor-Critic for Continuous Chunks), a novel RL framework that learns to generate high-dimensional, continuous action sequences. To make this learning process stable and data-efficient, AC3 incorporates targeted stabilization mechanisms for both the actor and the critic. First, to ensure reliable policy improvement, the actor is trained with an asymmetric update rule, learning exclusively from successful trajectories. Second, to enable effective value learning despite sparse rewards, the critic's update is stabilized using intra-chunk $n$-step returns and further enriched by a self-supervised module providing intrinsic rewards at anchor points aligned with each action chunk. We conducted extensive experiments on 25 tasks from the BiGym and RLBench benchmarks. Results show that by using only a few demonstrations and a simple model architecture, AC3 achieves superior success rates on most tasks, validating its effective design.