Flow with the Force Field

Learning 3D Compliant Flow Matching Policies
from Force & Demonstration-Guided Simulation Data

Tianyu Li* Yihan Li* Zizhe Zhang Nadia Figueroa

* Equal Contribution · University of Pennsylvania

ICRA 2026
Paper Code (Coming Soon)
Data Collection Deployment
1
Demo
∞
Sim Data
Zero-Shot Sim2Real
Vision
Force
Proprio
Compliant
Output
Scroll
01
Overview

What We Built

This framework uses a lightweight scheme to generate force-informed training data entirely in simulation from a single human demo. With this data, we train point cloud + force attending policies with passive impedance outputs. At execution, our passive impedance controller ensures compliant, safe interactions. Together, these components enable robots to generalize from simple geometry in sim to unseen real-world objects and spatial configurations.

02
Results

See It In Action

03
Abstract

The Problem & Our Approach

While visuomotor policy has made advancements in recent years, contact-rich tasks still remain a challenge. Robotic manipulation tasks that require continuous contact demand explicit handling of compliance and force. However, most visuomotor policies ignore compliance, overlooking the importance of physical interaction with the real world, often leading to excessive contact forces or fragile behavior under uncertainty.

Introducing force information into vision-based imitation learning could help improve awareness of contacts. However, current visuomotor policy approaches require a lot of data to perform well. One remedy for data scarcity is to generate data in simulation, yet computationally taxing processes are required to generate data good enough not to suffer from the Sim2Real gap.

In this work, we introduce a framework for generating force-informed data in simulation, instantiated by a single human demonstration, and show how coupling with a compliant policy improves the performance of a visuomotor policy learned from synthetic data. We validate our approach on real-robot tasks, including non-prehensile block flipping and a bi-manual object moving, where the learned policy exhibits reliable contact maintenance and adaptation to novel conditions.

04
Method

Framework Pipeline

3D Compliant Visuomotor Policy Learning Framework
3D Compliant Visuomotor Policy Learning Framework: Starting from a single simulation demonstration, we generate point cloud and force data by introducing virtual targets and applying Laplacian editing beyond the original demonstration. This augmented data is used to train a flow-matching policy that receives point cloud and force input and predicts actions, including an impedance parameter. At rollout time, the policy's output is synthesized into a state velocity field, which is then executed using a Passive Impedance Controller to ensure compliant behaviors. While our data is only generated with one simple geometry for both tasks, the trained policies produce generalizable capabilities beyond the original shapes by using our framework.
05
Acknowledgments

We thank our anonymous reviewers, who provided thorough and fair feedback that improved the quality of our paper. This work was supported by the National Science Foundation (NSF) Foundational Research in Robotics (FRR) program under NSF CAREER Award Grant No. FRR-2443721.

06
Citation

BibTeX

@article{li2025flow,
  title={Flow with the Force Field: Learning 3D Compliant Flow Matching Policies from Force and Demonstration-Guided Simulation Data},
  author={Li, Tianyu and Li, Yihan and Zhang, Zizhe and Figueroa, Nadia},
  journal={arXiv preprint arXiv:2510.02738},
  year={2025}
}