State Variable Dependent Disturbances Compensation Using Gaussian Process Regression: With Application to Core-type Linear Motor

The manufacturing industry is experiencing an increasing demand for improved production efficiency. Achieving precise control of high-performance, high-speed core-type permanent magnet linear motors is therefore essential to meet this need. The primary objective of this research is to address the challenges posed by complex disturbances, such as cogging force, which are influenced by both position and velocity and significantly degrade tracking performance. To overcome these problems, this research employs a two-step approach. Firstly, the Kalman smoother is used to accurately estimate the state variables and disturbances. Secondly, Gaussian Process Regression (GPR) is applied to establish a relationship between these state variables and disturbances, allowing the generation of disturbance compensation signals in arbitrary trajectories. A method introduced in this research streamlines the GPR computations, allowing efficient handling of large estimated data sets. The results of this study demonstrate a significant improvement in tracking performance compared to alternative data interpolation techniques such as smoothing splines. To validate this improvement, tracking control experiments were performed using a stage-type experimental machine equipped with core-type linear motors.