Flow Policy Gradients for Robot Control

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Yi, Brent, Choi, Hongsuk, Singh, Himanshu Gaurav, Huang, Xiaoyu, Truong, Takara E., Sferrazza, Carmelo, Ma, Yi, Duan, Rocky, Abbeel, Pieter, Shi, Guanya, Liu, Karen, Kanazawa, Angjoo
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866912869078007808
author Yi, Brent
Choi, Hongsuk
Singh, Himanshu Gaurav
Huang, Xiaoyu
Truong, Takara E.
Sferrazza, Carmelo
Ma, Yi
Duan, Rocky
Abbeel, Pieter
Shi, Guanya
Liu, Karen
Kanazawa, Angjoo
author_facet Yi, Brent
Choi, Hongsuk
Singh, Himanshu Gaurav
Huang, Xiaoyu
Truong, Takara E.
Sferrazza, Carmelo
Ma, Yi
Duan, Rocky
Abbeel, Pieter
Shi, Guanya
Liu, Karen
Kanazawa, Angjoo
contents Likelihood-based policy gradient methods are the dominant approach for training robot control policies from rewards. These methods rely on differentiable action likelihoods, which constrain policy outputs to simple distributions like Gaussians. In this work, we show how flow matching policy gradients -- a recent framework that bypasses likelihood computation -- can be made effective for training and fine-tuning more expressive policies in challenging robot control settings. We introduce an improved objective that enables success in legged locomotion, humanoid motion tracking, and manipulation tasks, as well as robust sim-to-real transfer on two humanoid robots. We then present ablations and analysis on training dynamics. Results show how policies can exploit the flow representation for exploration when training from scratch, as well as improved fine-tuning robustness over baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2602_02481
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Flow Policy Gradients for Robot Control
Yi, Brent
Choi, Hongsuk
Singh, Himanshu Gaurav
Huang, Xiaoyu
Truong, Takara E.
Sferrazza, Carmelo
Ma, Yi
Duan, Rocky
Abbeel, Pieter
Shi, Guanya
Liu, Karen
Kanazawa, Angjoo
Robotics
Artificial Intelligence
Likelihood-based policy gradient methods are the dominant approach for training robot control policies from rewards. These methods rely on differentiable action likelihoods, which constrain policy outputs to simple distributions like Gaussians. In this work, we show how flow matching policy gradients -- a recent framework that bypasses likelihood computation -- can be made effective for training and fine-tuning more expressive policies in challenging robot control settings. We introduce an improved objective that enables success in legged locomotion, humanoid motion tracking, and manipulation tasks, as well as robust sim-to-real transfer on two humanoid robots. We then present ablations and analysis on training dynamics. Results show how policies can exploit the flow representation for exploration when training from scratch, as well as improved fine-tuning robustness over baselines.
title Flow Policy Gradients for Robot Control
topic Robotics
Artificial Intelligence
url https://arxiv.org/abs/2602.02481