Learning Dynamical System for Grasping Motion

Abstract

Dynamical System has been widely used for encoding trajectories from human demonstration, which has the inherent adaptability to dynamically changing environments and robustness to perturbations. In this paper we propose a framework to learn a dynamical system that couples position and orientation based on a diffeomorphism. Different from other methods, it can realise the synchronization between positon and orientation during the whole trajectory. Online grasping experiments are carried out to prove its effectiveness and online adaptability.

IEEE ICRA 2021 Workshop on Bridging the Gap between Data-driven and Analytical Physics-based Grasping and Manipulation II