Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Chandra, Namai, Mohan, Liu, Gu, Zhihao, Wang, Lin
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2603.14469
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912977205067776
author Chandra, Namai
Mohan, Liu
Gu, Zhihao
Wang, Lin
author_facet Chandra, Namai
Mohan, Liu
Gu, Zhihao
Wang, Lin
contents Reinforcement learning (RL) has achieved strong performance in robotic control; however, state-of-the-art policy learning methods, such as actor-critic methods, still suffer from high sample complexity and often produce physically inconsistent actions. This limitation stems from neural policies implicitly rediscovering complex physics from data alone, despite accurate dynamics models being readily available in simulators. In this paper, we introduce a novel physics-informed RL framework, called PIPER, that seamlessly integrates physical constraints directly into neural policy optimization with analytical soft physics constraints. At the core of our method is the integration of a differentiable Lagrangian residual as a regularization term within the actor's objective. This residual, extracted from a robot's simulator description, subtly biases policy updates towards dynamically consistent solutions. Crucially, this physics integration is realized through an additional loss term during policy optimization, requiring no alterations to existing simulators or core RL algorithms. Extensive experiments demonstrate that our method significantly improves learning efficiency, stability, and control accuracy, establishing a new paradigm for efficient and physically consistent robotic control.
format Preprint
id arxiv_https___arxiv_org_abs_2603_14469
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Physics-Informed Policy Optimization via Analytic Dynamics Regularization
Chandra, Namai
Mohan, Liu
Gu, Zhihao
Wang, Lin
Robotics
Machine Learning
I.2.6; I.2.9
Reinforcement learning (RL) has achieved strong performance in robotic control; however, state-of-the-art policy learning methods, such as actor-critic methods, still suffer from high sample complexity and often produce physically inconsistent actions. This limitation stems from neural policies implicitly rediscovering complex physics from data alone, despite accurate dynamics models being readily available in simulators. In this paper, we introduce a novel physics-informed RL framework, called PIPER, that seamlessly integrates physical constraints directly into neural policy optimization with analytical soft physics constraints. At the core of our method is the integration of a differentiable Lagrangian residual as a regularization term within the actor's objective. This residual, extracted from a robot's simulator description, subtly biases policy updates towards dynamically consistent solutions. Crucially, this physics integration is realized through an additional loss term during policy optimization, requiring no alterations to existing simulators or core RL algorithms. Extensive experiments demonstrate that our method significantly improves learning efficiency, stability, and control accuracy, establishing a new paradigm for efficient and physically consistent robotic control.
title Physics-Informed Policy Optimization via Analytic Dynamics Regularization
topic Robotics
Machine Learning
I.2.6; I.2.9
url https://arxiv.org/abs/2603.14469