Learning Forward & Reverse Skills from a Single Unfinished Demonstration for Constrained Manipulation Tasks

2026-07-15

AI Summary

This work addresses the challenge in contact-rich manipulation tasks where existing methods rely on multiple complete demonstrations and struggle to execute reversal skills. The paper proposes a one-shot imitation learning framework that, for the first time, jointly learns forward and backward manipulation skills from a single, potentially incomplete demonstration. By integrating Dynamic Movement Primitives (DMPs) with Screw Motion Primitives, the approach introduces a geometry-driven screw-axis segmentation algorithm to enable trajectory extrapolation and reverse execution. During deployment, admittance control is fused into the execution phase to perform real-time pose correction and velocity modulation. Evaluated on tasks such as peg insertion, battery installation, lock unlocking, and screw driving, the method significantly outperforms baseline approaches in both success rate and robustness.

Abstract

Learning from demonstration (LfD) enables robots to learn manipulation skills directly from expert demonstrations but remains challenging for contact-rich tasks involving geometric constraints and force interaction. Existing approaches typically require multiple complete demonstrations and do not support reverse skill execution. In this paper, we present a unified one-shot framework for constrained manipulation that learns both forward and reverse execution from a single, possibly unfinished demonstration. Our method decomposes demonstrations into non-contact and contact phases, with non-contact motion encoded with dynamic movement primitives (DMP), and contact motion represented as a sequence of screw motion primitives segmented by our proposed geometry-driven twist-direction segmentation algorithm. During execution, screw primitives are executed sequentially under admittance-guided pose correction and speed regulation, enabling task completion beyond the demonstrated trajectory length as well as reverse skill execution without additional learning data. Experiments on peg insertion, battery insertion, lock opening, and screw driving tasks demonstrate improved success rates and robustness over segmentation and one-shot trajectory learning baselines. Details are available on the project website: https://tuwien-asl.github.io/LfD-Screw/.

Problem

• learning from demonstration
• constrained manipulation
• contact-rich tasks
• reverse skill execution
• one-shot learning

Innovation

• one-shot learning
• screw motion primitives
• geometry-driven segmentation
• admittance control
• bidirectional skill execution