Residual Compliant Policy: Learning Compliant Policies from Few Demonstrations

Chunyu Jiang, Yichen Cai, Shuai Li, Wenbin Sun, Jo\~ao Carvalho, Kostas Daniilidis, Oleg Arenz, Cristiana de Farias, Jan Peters · N/A · 2026

Summary

Contact-rich manipulation requires a robot to be precise in free space to establish the intended contacts, and compliant once in contact to avoid excessive forces.

Abstract Summary

Contact-rich manipulation requires a robot to be precise in free space to establish the intended contacts, and compliant once in contact to avoid excessive forces. Imitation learning has proven effective for simpler manipulation tasks, but the compliance underlying a demonstration is not directly observable and is therefore difficult to infer from demonstration data alone. We present a framework that makes compliance explicit at every stage. During data collection, the operator teleoperates the real robot while a synchronized digital twin visualizes the error between the commanded reference and the achieved pose, making the current compliance visible, and selects between two discrete compliance modes. To generalize across object poses, we combine Laplacian editing, which retargets trajectories efficiently but is not physics-consistent during contact, with simulation-based trajectory optimization, which enforces physical plausibility but offers no manual guidance. Finally, a residual reinforcement learning policy refines the interaction forces by adjusting the reference pose of the underlying admittance controller. We evaluate on three flipping and pushing tasks on a Franka FR3 based on a total of 7 demonstrations across tasks. Our method is successful on all 40 out-of-distribution trials, compared to 33/40 and 24/40 for two baselines, with our base policy alone reaching 35/40.

Key Points

Contact-rich manipulation requires a robot to be precise in free space to establish the intended contacts, and compliant once in contact to avoid excessive forces.
Imitation learning has proven effective for simpler manipulation tasks, but the compliance underlying a demonstration is not directly observable and is therefore difficult to infer from demonstration data alone.
We present a framework that makes compliance explicit at every stage
During data collection, the operator teleoperates the real robot while a synchronized digital twin visualizes the error between the commanded reference and the achieved pose, making the current compliance visible, and selects between two discrete compliance modes.
To generalize across object poses, we combine Laplacian editing, which retargets trajectories efficiently but is not physics-consistent during contact, with simulation-based trajectory optimization, which enforces physical plausibility but offers no manual guidance.

Method

This section covers the main components of our method. We first discuss our admittance controller design in section III-A . Next, we introduce our proposed data collection pipeline in section III-B and data augmentation method in section III-C , which generate hundreds of high-quality trajectories from a few collected demonstrations to train the base policy. We then describe our base policy in section III-D .

Finally, we present our residual RL policy in section III-E .

Method — Residual Compliant Policy: Learning Compliant Policies from Few Demonstrations — Fig. 1: We propose a framework for learning compliant policies from few demonstrations. Trajectories are collected in the real world using a digital twin and subsequently augmented in simulation to train our base policy. We then apply residual RL in simulation and deploy the residual policy in the real world.

Experiments

In this section, we evaluate our framework through experiments designed to answer the following questions: (i) Can our data collection and augmentation pipeline provide sufficient high-quality trajectories for policy learning? (ii) Can a policy trained on the generated data successfully perform contact-rich manipulation on the real robot?

(iii) How does our compliance strategy compare with an ACP-style compliance policy? (iv) Does residual reinforcement learning improve the performance of the base policy? We evaluate our method on a Franka FR3...

Experiments — Residual Compliant Policy: Learning Compliant Policies from Few Demonstrations — Fig. 2: Proposed compliance policy learning pipeline. (a) The operator uses the haptic teleoperation device with our digital twin to collect $N_{h}$ human demonstrations. (b) The collected human demonstrations are augmented via free-space motion Laplacian editing and in-contact trajectory optimization given randomly sampled positions and poses of the object in simulation. (c) A consistency flow matching-based policy is trained on the augmented dataset. (d) A residual RL policy is trained upon the base policy in simulation. (e) The system takes in the robot’s proprioception, stiffness of the admittance controller, external force feedback, and the object pose estimated by the digital twin to predict expected stiffness level and reference poses for implementing the task.

Sources and Demonstrations

Figures are reproduced from Jiang et al. See the full paper for experimental details and the project page, when the paper links one, for demonstrations and videos.