aiqbench analysis
This research introduces ORPA, a framework that allows robot manipulation policies trained via imitation learning to adapt instantly to errors and distribution shifts without costly full retraining. By adding a lightweight feedback-conditioned module in joint space, the system achieves real-time correction on precision tasks like those performed by ALOHA platforms. The approach directly addresses critical limitations of current large-scale robotic models regarding robustness and deployment efficiency.
ORPA: Online Residual Policy Adaptation for Robot Manipulation Control with Human Feedback
arXiv:2608.17323v1 Announce Type: new Abstract: Robotic manipulation policies trained via imitation learning, such as Action Chunking with Transformers (ACT), can achieve strong performance under ideal conditions but often remain sensitive to small execution errors and distribution shifts. Correcting these failures typically requires dataset aggregation and full-policy retraining, which is computationally expensive and unsuitable for real-time deployment. In this work, we propose Online Residual Policy Adaptation (ORPA), a framework that enables immediate, feedback-driven correction of robot actions without modifying the underlying policy parameters. ORPA augments a pretrained control policy with a lightweight, feedback-conditioned module that predicts residual adjustments directly in joint space, allowing the system to adapt its behavior at runtime. We evaluate ORPA on a set of precision-sensitive manipulation tasks using the ALOHA platform, demonstrating improvements in success rate and recovery from small perturbations compared to baseline control policies and rule-based inverse kinematics corrections.
Abstract:Robotic manipulation policies trained via imitation learning, such as Action Chunking with Transformers (ACT), can achieve strong performance under ideal conditions but often remain sensitive to small execution errors and distribution shifts. Correcting these failures typically requires dataset aggregation and full-policy retraining, which is computationally expensive and unsuitable for real-time deployment. In this work, we propose Online Residual Policy Adaptation (ORPA), a framework that enables immediate, feedback-driven correction of robot actions without modifying the underlying policy parameters. ORPA augments a pretrained control policy with a lightweight, feedback-conditioned module that predicts residual adjustments directly in joint space, allowing the system to adapt its behavior at runtime. We evaluate ORPA on a set of precision-sensitive manipulation tasks using the ALOHA platform, demonstrating improvements in success rate and recovery from small perturbations compared to baseline control policies and rule-based inverse kinematics corrections.
Robotics (cs.RO); Artificial Intelligence (cs.AI)
arXiv:2608.17323 [cs.RO]
(or arXiv:2608.17323v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2608.17323 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 235 words
Related coverage
The Machines that Make the Machines
NVIDIA · Oct 7, 2026
Online Sim-to-Real Adaptation via Closed-Loop System Modeling
arXiv · Sep 25, 2026
TAM: Torque Adaptation Module for Robust Motion Transfer in Manipulation
arXiv · Jun 6, 2026
Stability-Aligned Residual Adaptation for Rapid Recovery from Dynamics Shifts
arXiv · Sep 22, 2026