Enhancing Control Policy Smoothness by Aligning Actions with Predictions from Preceding States

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Kwak, Kyoleen, Hwang, Hyoseok
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866911399460995072
author Kwak, Kyoleen
Hwang, Hyoseok
author_facet Kwak, Kyoleen
Hwang, Hyoseok
contents Deep reinforcement learning has proven to be a powerful approach to solving control tasks, but its characteristic high-frequency oscillations make it difficult to apply in real-world environments. While prior methods have addressed action oscillations via architectural or loss-based methods, the latter typically depend on heuristic or synthetic definitions of state similarity to promote action consistency, which often fail to accurately reflect the underlying system dynamics. In this paper, we propose a novel loss-based method by introducing a transition-induced similar state. The transition-induced similar state is defined as the distribution of next states transitioned from the previous state. Since it utilizes only environmental feedback and actually collected data, it better captures system dynamics. Building upon this foundation, we introduce Action Smoothing by Aligning Actions with Predictions from Preceding States (ASAP), an action smoothing method that effectively mitigates action oscillations. ASAP enforces action smoothness by aligning the actions with those taken in transition-induced similar states and by penalizing second-order differences to suppress high-frequency oscillations. Experiments in Gymnasium and Isaac-Lab environments demonstrate that ASAP yields smoother control and improved policy performance over existing methods.
format Preprint
id arxiv_https___arxiv_org_abs_2601_18479
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Enhancing Control Policy Smoothness by Aligning Actions with Predictions from Preceding States
Kwak, Kyoleen
Hwang, Hyoseok
Machine Learning
Deep reinforcement learning has proven to be a powerful approach to solving control tasks, but its characteristic high-frequency oscillations make it difficult to apply in real-world environments. While prior methods have addressed action oscillations via architectural or loss-based methods, the latter typically depend on heuristic or synthetic definitions of state similarity to promote action consistency, which often fail to accurately reflect the underlying system dynamics. In this paper, we propose a novel loss-based method by introducing a transition-induced similar state. The transition-induced similar state is defined as the distribution of next states transitioned from the previous state. Since it utilizes only environmental feedback and actually collected data, it better captures system dynamics. Building upon this foundation, we introduce Action Smoothing by Aligning Actions with Predictions from Preceding States (ASAP), an action smoothing method that effectively mitigates action oscillations. ASAP enforces action smoothness by aligning the actions with those taken in transition-induced similar states and by penalizing second-order differences to suppress high-frequency oscillations. Experiments in Gymnasium and Isaac-Lab environments demonstrate that ASAP yields smoother control and improved policy performance over existing methods.
title Enhancing Control Policy Smoothness by Aligning Actions with Predictions from Preceding States
topic Machine Learning
url https://arxiv.org/abs/2601.18479