Generalizable Domain Adaptation for Sim-and-Real Policy Co-Training

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Cheng, Shuo, Ma, Liqian, Chen, Zhenyang, Mandlekar, Ajay, Garrett, Caelan, Xu, Danfei
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866914259217154048
author Cheng, Shuo
Ma, Liqian
Chen, Zhenyang
Mandlekar, Ajay
Garrett, Caelan
Xu, Danfei
author_facet Cheng, Shuo
Ma, Liqian
Chen, Zhenyang
Mandlekar, Ajay
Garrett, Caelan
Xu, Danfei
contents Behavior cloning has shown promise for robot manipulation, but real-world demonstrations are costly to acquire at scale. While simulated data offers a scalable alternative, particularly with advances in automated demonstration generation, transferring policies to the real world is hampered by various simulation and real domain gaps. In this work, we propose a unified sim-and-real co-training framework for learning generalizable manipulation policies that primarily leverages simulation and only requires a few real-world demonstrations. Central to our approach is learning a domain-invariant, task-relevant feature space. Our key insight is that aligning the joint distributions of observations and their corresponding actions across domains provides a richer signal than aligning observations (marginals) alone. We achieve this by embedding an Optimal Transport (OT)-inspired loss within the co-training framework, and extend this to an Unbalanced OT framework to handle the imbalance between abundant simulation data and limited real-world examples. We validate our method on challenging manipulation tasks, showing it can leverage abundant simulation data to achieve up to a 30% improvement in the real-world success rate and even generalize to scenarios seen only in simulation. Project webpage: https://ot-sim2real.github.io/.
format Preprint
id arxiv_https___arxiv_org_abs_2509_18631
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Generalizable Domain Adaptation for Sim-and-Real Policy Co-Training
Cheng, Shuo
Ma, Liqian
Chen, Zhenyang
Mandlekar, Ajay
Garrett, Caelan
Xu, Danfei
Robotics
Artificial Intelligence
Behavior cloning has shown promise for robot manipulation, but real-world demonstrations are costly to acquire at scale. While simulated data offers a scalable alternative, particularly with advances in automated demonstration generation, transferring policies to the real world is hampered by various simulation and real domain gaps. In this work, we propose a unified sim-and-real co-training framework for learning generalizable manipulation policies that primarily leverages simulation and only requires a few real-world demonstrations. Central to our approach is learning a domain-invariant, task-relevant feature space. Our key insight is that aligning the joint distributions of observations and their corresponding actions across domains provides a richer signal than aligning observations (marginals) alone. We achieve this by embedding an Optimal Transport (OT)-inspired loss within the co-training framework, and extend this to an Unbalanced OT framework to handle the imbalance between abundant simulation data and limited real-world examples. We validate our method on challenging manipulation tasks, showing it can leverage abundant simulation data to achieve up to a 30% improvement in the real-world success rate and even generalize to scenarios seen only in simulation. Project webpage: https://ot-sim2real.github.io/.
title Generalizable Domain Adaptation for Sim-and-Real Policy Co-Training
topic Robotics
Artificial Intelligence
url https://arxiv.org/abs/2509.18631