Determination of the End of Positioning Phase Using SVM: Kernel Choice and Parameter Tuning

Abstract

Abstract: In many positioning control systems, the completion of the positioning phase must be determined correctly in order to initiate the start of the next task. However, this is not always easy because the output may exceed the allowable error bound. To resolve this problem, we proposed a support vector machine (SVM-) based method, and we verified the effectiveness of the proposed method by performing simulations and experiments using simple plant models. In SVMs, the selection of a kernel and kernel parameters is critical to the performance of SVMs; however, we used only the polynomial kernel in the previous study. In this paper, we use not only the polynomial kernel but also a Gaussian kernel, their performances are compared. Further, we discuss the selection of kernel parameters, and we evaluate the performance of the proposed method using positioning data obtained from an actual galvano scanner control system.