Online Imitation Learning for Manipulation via Decaying Relative Correction through Teleoperation

2025-03-19

AI Summary

In teleoperated robotic arm online imitation learning, frequent expert interventions impose substantial cognitive burden and hinder scalability. To address this, we propose Decay-based Relative Correction (DRC): a lightweight, spatial intervention mechanism that applies transient, self-decaying trajectory corrections via 6-DoF displacement vectors—without persistent updates to policy parameters. Integrated into an online imitation learning framework and deployed on a cable-driven teleoperation system, DRC dynamically refines executed trajectories to enhance policy generalization and task adaptability. Experiments demonstrate that DRC reduces expert intervention frequency by 30%. On raspberry harvesting and fabric wiping tasks, it significantly improves task success rates and accelerates convergence. This work introduces the first formulation of spatial displacement as a decaying instantaneous correction signal—striking a principled balance between intervention efficacy and expert cognitive load. DRC establishes a novel paradigm for low-intervention, high-robustness online imitation learning.

Technology Category

• Intelligent Robots: Manipulation
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Teleoperated robotic manipulators enable the collection of demonstration data, which can be used to train control policies through imitation learning. However, such methods can require significant amounts of training data to develop robust policies or adapt them to new and unseen tasks. While expert feedback can significantly enhance policy performance, providing continuous feedback can be cognitively demanding and time-consuming for experts. To address this challenge, we propose to use a cable-driven teleoperation system which can provide spatial corrections with 6 degree of freedom to the trajectories generated by a policy model. Specifically, we propose a correction method termed Decaying Relative Correction (DRC) which is based upon the spatial offset vector provided by the expert and exists temporarily, and which reduces the intervention steps required by an expert. Our results demonstrate that DRC reduces the required expert intervention rate by 30% compared to a standard absolute corrective method. Furthermore, we show that integrating DRC within an online imitation learning framework rapidly increases the success rate of manipulation tasks such as raspberry harvesting and cloth wiping.

Problem

• Reduces expert intervention in robotic manipulation tasks
• Improves policy adaptation to new tasks with less data
• Enhances success rates in tasks like raspberry harvesting

Innovation

• Cable-driven teleoperation system for spatial corrections
• Decaying Relative Correction reduces expert intervention
• Online imitation learning enhances task success rates