Robot Manipulation Skill Learning Based on Stable Sparse Gaussian Process

In robot manipulation skill learning, due to the complexity of the system, encoding the human demonstrations with a non-parametric learning method is always more effective. Gaussian Process is a data-driven approach that can be used to accurately learn dynamic systems encoding human motion rules, However, it suffers heavy computational costs. To address this issue, we introduce the Sparse Gaussian Processes to reduce computational costs. Additionally, the Lyapunov control theorems are utilized to ensure the stability of the SPGP dynamic system. Finally, we validate the effectiveness of the proposed method on a publicly available dataset.