LPV Gray Box Identification of Industrial Robots for Control

Abstract

This paper treats the linear parameter-varying (LPV) model identification of an industrial robot. Since the model is supposed to be used to design an LPV controller, it must simultaneously feature low complexity and adequate accuracy. As for most systems, a simplified analytical model structure can be derived for the robot based on the laws of physics. Some physical model parameters however must be experimentally determined. Due to the model simplifications, these physical parameters vary over the workspace. In order to capture this variation in an LPV model, the physical parameters are scheduled. Based on an understanding of the system, three different scheduling laws are designed and the resulting LPV models are compared to experimentally determined frequency response functions.