PACE: Phase-Aware Chunk Execution for Robot Policies with Action Chunking

2026-05-30

AI Summary

This work addresses the limitation of existing action-chunking–based robotic policies, which rely on fixed execution horizons and struggle to accommodate dynamic replanning demands during task execution. The authors propose a training-free, test-time adaptive execution method that automatically identifies low-velocity transition points in predicted action chunks as replanning boundaries by analyzing their velocity profiles, thereby enabling phase-aware adjustment of the execution horizon. Leveraging the kinematic phase structure inherent in predicted trajectories, this approach achieves plug-and-play online adaptability without modifying the underlying policy or requiring retraining. Evaluated on RoboTwin2.0, the method improves task success rate from 57.8% to 64.2%; real-robot experiments further demonstrate a substantial performance gain, with task scores rising from 60.7 to 77.7 and success rates increasing significantly from 50.7% to 70.4%.

Abstract

Recent vision-language-action and diffusion-based robot policies often use action chunking, where each policy query predicts a sequence of future actions and the robot executes an open-loop prefix before re-querying. While this interface improves local motion continuity, deployment still requires choosing the execution horizon: how much of each predicted chunk should be executed before acquiring a new observation. However, our experiments show that success is strongly task-dependent and non-monotonic with respect to the execution horizon, making a single constant horizon an unreliable deployment rule. We propose PACE (Phase-Aware Chunk Execution), a training-free test-time execution method that selects the execution horizon online from the predicted chunk itself. PACE exploits the phase-dependent kinematic structure of manipulation trajectories by identifying low-speed transition points in the predicted speed profile and using them as candidate replanning boundaries. Because PACE uses only the predicted action chunk, it is plug-and-play and requires no retraining or access to policy internals. We validate PACE through large-scale evaluations in both simulation and real-robot settings. On 50 RoboTwin2.0 tasks, PACE raises the average success rate from 57.8% to 64.2%. In real-robot experiments on bimanual ALOHA and single-arm Franka platforms, PACE improves the average task score from 60.7 to 77.7 and the average success rate from 50.7% to 70.4%. Ablations and rollout-level analyses show that PACE adapts execution horizons across manipulation phases, shortening near transitions while preserving longer execution during coherent motion.

Problem

• action chunking
• execution horizon
• robot policies
• manipulation trajectories
• replanning boundaries

Innovation

• action chunking
• execution horizon
• phase-aware execution
• robot policy
• kinematic structure