Commit While Futures Agree: Consequence-Aware Adaptive Action Chunking for Robot Manipulation

2026-10-06

AI Summary

This study addresses the limitations of fixed execution horizons and action-similarity-based criteria in robotic action chunking by proposing the CA3C framework. This method introduces a novel adaptive commitment mechanism based on imagined future consistency. Specifically, it employs an action-conditioned world model to predict future states and dynamically determines replanning timings through Bayesian change-point detection combined with a future consensus selection algorithm, all without modifying or retraining the underlying policy. Experiments in both simulated and real-world tasks demonstrate that CA3C substantially enhances the performance of diverse policies, achieving a relative failure rate reduction of up to 71.8%.

Abstract

Action-chunking policies predict multi-step control sequences, but a fundamental question remains: how much of a predicted action chunk should be committed before replanning? Existing systems typically execute a fixed-length prefix, implicitly assuming that the same execution horizon remains trustworthy across states. Some adaptive methods estimate this horizon from the similarity or stability of predicted actions. However, different actions may lead to the same successful outcome, whereas similar actions can produce different futures, suggesting that commitment should be determined by agreement among imagined futures rather than by similarity in action space. To this end, we propose Consequence-Aware Adaptive Action Chunking (CA$^3$C), an inference-time framework built on a simple principle: commit while imagined futures agree, and replan when they diverge. Without modifying or retraining the base policy, CA$^3$C uses an action-conditioned world model to imagine the future consequences of multiple candidate action chunks under the same sampling noise. Using these imagined consequences, we formulate execution-horizon estimation as a Bayesian change-point inference problem and select the execution candidate through future consensus. Across multiple simulation benchmarks and real-world robot manipulation tasks, CA$^3$C consistently improves diverse action-chunking policies, achieving up to a 71.8% relative reduction in failure rate over the corresponding base policies.

Problem

• Action Chunking
• Robot Manipulation
• Execution Horizon
• Adaptive Replanning
• World Model

Innovation

• Adaptive Action Chunking
• World Model
• Bayesian Change-Point Inference
• Consequence-Aware
• Robot Manipulation