From a Single Demonstration to a General Policy for Contact-Rich Manipulation

2026-05-17

AI Summary

This work addresses the challenge of efficiently generalizing multi-stage, contact-rich robotic manipulation tasks from a single demonstration. It proposes a constraint-aware imitation learning framework that, for the first time, leverages environmental constraints as an inductive bias to decouple task structure from instance-specific details. The approach employs a four-stage pipeline—comprising environmental constraint primitive abstraction, self-guided exploration, human correction fusion, and online compliant interaction recovery—to achieve strong generalization across variations in object pose, local geometry, and unmodeled contact dynamics. Evaluated on seven real-world tasks, the method achieves over 90% success rates, demonstrating its effectiveness and robustness in practical settings.

Abstract

We present a Learning from Demonstration (LfD) framework that achieves one-shot generalization in multi-stage, contact-rich manipulation tasks. Central to our approach is the utilization of environmental constraints as the inductive bias. By representing a demonstration as a sequence of behaviors that exploit environmental constraints, the robot separates task-general structure -- the constraint types and their transitions -- from instance-specific details such as exact demonstration trajectories, poses, and local geometries. Our four-stage pipeline builds a complete policy on this representation: the robot first abstracts a single demonstration into environmental-constraint primitives, then disambiguates them through self-guided exploration, next assimilates targeted human corrections that handle out-of-distribution variations, and finally recovers the abstracted-away details online through compliant interaction. Because the resulting policy follows constraints rather than mimics trajectories, it generalizes across object poses, local geometries, and unmodeled contact dynamics. We validate our approach on seven real-world multi-stage contact-rich manipulation tasks and achieve over 90% success. These extensive experimental results establish environmental constraints as fundamental building blocks for efficient generalization in learning from demonstration.

Problem

• one-shot generalization
• contact-rich manipulation
• learning from demonstration
• environmental constraints
• multi-stage tasks

Innovation

• Learning from Demonstration
• Environmental Constraints
• One-shot Generalization
• Contact-rich Manipulation
• Compliant Interaction