Residual Compliant Policy: Learning Compliant Policies from Few Demonstrations
Residual Compliant Policy: Learning Compliant Policies from Few Demonstrations
This research presents a framework for learning compliant policies from few demonstrations, combining teleoperation, simulation-based trajectory optimization, and residual reinforcement learning to improve contact-rich manipulation tasks. The method is evaluated on three flipping and pushing tasks using a Franka FR3 robot, achieving success in 35 out of 40 out-of-distribution trials, outperforming two baselines.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
demonstrations
7 episodes
View original evidence
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.
Open source S5out_of_distribution_trials
40
View original evidence
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.
Open source S5tasks
3
View original evidence
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
Open source S4
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
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.
Open source S5
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
Open source S4
Source：arXiv Robotics — research abstracts · arxiv.org