Automate Robot Reaching Task with Learning from Demonstration

Over the last decades, robots have been moved from industries to domestic environments. Robot Learning from Demonstration (LfD) is one of the most significant methods to facilitate this trend. In this work, we first discuss details about a efficient motion planning strategy, e.g., Stable Estimator of Dynamical Systems (SEDS). A human first demonstrates reaching tasks several times, and Gaussian Mixture Regression is used to roughly encode the demonstrations into a set of differential equations. Then based on Lyapunov Stability Theorem, a constrained nonlinear optimization problem is formulated to iteratively refine the previously learned differential equations and SEDS is thus obtained. Experiments have been conducted on a KUKA LBR iiwa robot to verify two properties of the proposed method, e.g., asymptotical stability and adaptation to spatial perturbations.