Data-Based Design of Inverse Dynamics Using Gaussian Process

Model-based controller design has been widely utilized for various purposes, and many methodologies have been proposed to identify accurate models of the target plants. In this paper, a different methodology to design dynamics model, particularly inverse dynamics model is proposed using Gaussian process. The design process and selection of training input pattern for inverse dynamics learning Gaussian process are analyzed in this paper. The simulation results reveal the potential and limitation of the proposed Gaussian process based inverse dynamics learning.