English

Logic Learning from Demonstrations for Multi-step Manipulation Tasks in Dynamic Environments

Robotics 2024-06-25 v2

Abstract

Learning from Demonstration (LfD) stands as an efficient framework for imparting human-like skills to robots. Nevertheless, designing an LfD framework capable of seamlessly imitating, generalizing, and reacting to disturbances for long-horizon manipulation tasks in dynamic environments remains a challenge. To tackle this challenge, we present Logic Dynamic Movement Primitives (Logic-DMP), which combines Task and Motion Planning (TAMP) with an optimal control formulation of DMP, allowing us to incorporate motion-level via-point specifications and to handle task-level variations or disturbances in dynamic environments. We conduct a comparative analysis of our proposed approach against several baselines, evaluating its generalization ability and reactivity across three long-horizon manipulation tasks. Our experiment demonstrates the fast generalization and reactivity of Logic-DMP for handling task-level variants and disturbances in long-horizon manipulation tasks.

Keywords

Cite

@article{arxiv.2404.16138,
  title  = {Logic Learning from Demonstrations for Multi-step Manipulation Tasks in Dynamic Environments},
  author = {Yan Zhang and Teng Xue and Amirreza Razmjoo and Sylvain Calinon},
  journal= {arXiv preprint arXiv:2404.16138},
  year   = {2024}
}

Comments

Accepted by IEEE RA-L

R2 v1 2026-06-28T16:05:30.330Z