Real-Time Operator Takeover for Visuomotor Diffusion Policy Training

TL;DR

This paper introduces a real-time operator takeover framework for visuomotor diffusion policies, allowing seamless control intervention to improve policy robustness and performance across diverse object manipulation tasks.

Contribution

The authors propose a novel real-time takeover paradigm that enhances visuomotor policy training with targeted demonstrations and automatic out-of-distribution state detection.

Findings

Targeted takeover demonstrations significantly improve policy performance.

The Mahalanobis distance effectively identifies undesirable states during execution.

The framework is validated on tasks involving rigid, deformable, and granular objects.

Abstract

We present a Real-Time Operator Takeover (RTOT) paradigm that enables operators to seamlessly take control of a live visuomotor diffusion policy, guiding the system back to desirable states or providing targeted corrective demonstrations. Within this framework, the operator can intervene to correct the robot's motion, after which control is smoothly returned to the policy until further intervention is needed. We evaluate the takeover framework on three tasks spanning rigid, deformable, and granular objects, and show that incorporating targeted takeover demonstrations significantly improves policy performance compared with training on an equivalent number of initial demonstrations alone. Additionally, we provide an in-depth analysis of the Mahalanobis distance as a signal for automatically identifying undesirable or out-of-distribution states during execution. Supporting materials, including videos of the initial and takeover demonstrations and all experiments, are available on the project website: https://operator-takeover.github.io/