Implicit Action Chunking for Smooth Continuous Control

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Liang, Bosun, Pei, Shuo, Chen, Zirui, Fan, Chuanzhi, Sun, Chen, Wu, Yuankai, Tan, Huachun, Wang, Yong
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911698275794944
author Liang, Bosun
Pei, Shuo
Chen, Zirui
Fan, Chuanzhi
Sun, Chen
Wu, Yuankai
Tan, Huachun
Wang, Yong
author_facet Liang, Bosun
Pei, Shuo
Chen, Zirui
Fan, Chuanzhi
Sun, Chen
Wu, Yuankai
Tan, Huachun
Wang, Yong
contents Reinforcement learning often produces high-frequency oscillatory control signals that undermine the safety and stability required for physical deployment. Explicit action chunking addresses this by predicting fixed-horizon trajectories but scales the policy output dimension proportionally with the horizon length, leading to optimization difficulties and incompatibility with standard step-wise interaction. To overcome these challenges, this paper proposes Dual-Window Smoothing (DWS), an implicit action chunking framework for smooth continuous control. Unlike explicit methods, DWS enforces temporal coherence without expanding the action space. It uses a dual-window design: an execution window that ensures physical smoothness through deterministic modulation, and a value window that aligns temporal-difference targets over the horizon to correct critic bias caused by open-loop execution. DWS also includes a lightweight actor-side temporal regularizer based on first-order action differences to promote global continuity. This design effectively bridges the gap between temporal abstraction and reactive step-wise control. Experiments on benchmarks including the DeepMind Control Suite and industrial energy management tasks show that DWS outperforms state-of-the-art (SOTA) baselines. In complex vision-based autonomous driving tasks, DWS achieves smoother control, safer behavior with reduced jitter, and attains a 100% success rate.
format Preprint
id arxiv_https___arxiv_org_abs_2605_19592
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Implicit Action Chunking for Smooth Continuous Control
Liang, Bosun
Pei, Shuo
Chen, Zirui
Fan, Chuanzhi
Sun, Chen
Wu, Yuankai
Tan, Huachun
Wang, Yong
Robotics
Artificial Intelligence
Reinforcement learning often produces high-frequency oscillatory control signals that undermine the safety and stability required for physical deployment. Explicit action chunking addresses this by predicting fixed-horizon trajectories but scales the policy output dimension proportionally with the horizon length, leading to optimization difficulties and incompatibility with standard step-wise interaction. To overcome these challenges, this paper proposes Dual-Window Smoothing (DWS), an implicit action chunking framework for smooth continuous control. Unlike explicit methods, DWS enforces temporal coherence without expanding the action space. It uses a dual-window design: an execution window that ensures physical smoothness through deterministic modulation, and a value window that aligns temporal-difference targets over the horizon to correct critic bias caused by open-loop execution. DWS also includes a lightweight actor-side temporal regularizer based on first-order action differences to promote global continuity. This design effectively bridges the gap between temporal abstraction and reactive step-wise control. Experiments on benchmarks including the DeepMind Control Suite and industrial energy management tasks show that DWS outperforms state-of-the-art (SOTA) baselines. In complex vision-based autonomous driving tasks, DWS achieves smoother control, safer behavior with reduced jitter, and attains a 100% success rate.
title Implicit Action Chunking for Smooth Continuous Control
topic Robotics
Artificial Intelligence
url https://arxiv.org/abs/2605.19592