Obtaining Models and Control for Rigid Body Motion Systems : 4th Report, Modeling of Inertia Matrix and Coriolis's Force Vector by Basis Function Networks

Abstract

This paper proposes a motion learning method for motion systems using basis function networks. Motion systems have unknown nonlinearities such as variation of inertia, Coriolis's force, nonlinear electromagnetic driving characteristics and gravity that are complicated to formulate. In this study, these nonlinearities are described by basis function networks. The updating rules of the weights in the network are derived by Popov's Hyper Stable Theory. However, there exist modeling errors and unmodeled forces which generate tracking errors. Therefore, H_∞control is introduced to attenuate the influence of modeling errors. This paper shows an analytical formulation of basis function networks for motion systems and effectiveness of the proposed motion control method by demonstrating an experiment of tracking desired trajectory in 2 types of 2-degree-of-freedom link motion system that have different motion equations.