Closed-Loop Action Chunks with Dynamic Corrections for Training-Free Diffusion Policy

TL;DR

This paper introduces DCDP, a novel framework that enhances diffusion-based robotic policies with real-time closed-loop corrections, significantly improving adaptability in dynamic environments without retraining.

Contribution

DCDP combines chunk-based action generation with real-time environmental corrections using a self-supervised encoder and cross-attention, enabling rapid adaptation in dynamic scenarios.

Findings

DCDP improves adaptability by 19% in PushT simulations.

Requires only 5% additional computation for real-time correction.

Modular design allows easy integration into existing systems.

Abstract

Diffusion-based policies have achieved remarkable results in robotic manipulation but often struggle to adapt rapidly in dynamic scenarios, leading to delayed responses or task failures. We present DCDP, a Dynamic Closed-Loop Diffusion Policy framework that integrates chunk-based action generation with real-time correction. DCDP integrates a self-supervised dynamic feature encoder, cross-attention fusion, and an asymmetric action encoder-decoder to inject environmental dynamics before action execution, achieving real-time closed-loop action correction and enhancing the system's adaptability in dynamic scenarios. In dynamic PushT simulations, DCDP improves adaptability by 19\% without retraining while requiring only 5\% additional computation. Its modular design enables plug-and-play integration, achieving both temporal coherence and real-time responsiveness in dynamic robotic scenarios, including real-world manipulation tasks. The project page is at: https://github.com/wupengyuan/dcdp