Compliant Residual DAgger: Improving Real-World Contact-Rich Manipulation with Human Corrections

2025-06-20

AI Summary

To address the inefficiency of DAgger data collection and difficulty in policy updates for contact-intensive manipulation tasks, this paper proposes a human-robot collaborative framework with flexible intervention. It introduces a lightweight, in-execution human action correction interface, integrates force-sensing closed-loop control with compliance regulation, and develops a residual-policy learning mechanism for adaptive policy refinement. Crucially, it incorporates real-time force feedback into residual policy learning to enable online, non-interruptive human corrective interventions. The DAgger pipeline is further optimized for both data acquisition and policy iteration. Evaluated on a physical robot platform, the approach improves base-policy success rates by over 50% on book-flipping and belt-assembly tasks—substantially outperforming both from-scratch training and fine-tuning baselines—while requiring only minimal human correction samples.

Technology Category

• Intelligent Robots: Manipulation
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Multiagent Systems: Adversarial Agents

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

We address key challenges in Dataset Aggregation (DAgger) for real-world contact-rich manipulation: how to collect informative human correction data and how to effectively update policies with this new data. We introduce Compliant Residual DAgger (CR-DAgger), which contains two novel components: 1) a Compliant Intervention Interface that leverages compliance control, allowing humans to provide gentle, accurate delta action corrections without interrupting the ongoing robot policy execution; and 2) a Compliant Residual Policy formulation that learns from human corrections while incorporating force feedback and force control. Our system significantly enhances performance on precise contact-rich manipulation tasks using minimal correction data, improving base policy success rates by over 50% on two challenging tasks (book flipping and belt assembly) while outperforming both retraining-from-scratch and finetuning approaches. Through extensive real-world experiments, we provide practical guidance for implementing effective DAgger in real-world robot learning tasks. Result videos are available at: https://compliant-residual-dagger.github.io/

Problem

• Improving real-world contact-rich manipulation with human corrections
• Collecting informative human correction data for DAgger
• Updating policies effectively with new human correction data

Innovation

• Compliant Intervention Interface for gentle corrections
• Compliant Residual Policy with force feedback
• Enhances manipulation tasks using minimal correction data