Stochastic Learning Control Framework in the Integrated Frequency and Position Domain

Abstract

This paper proposes stochastic learning control with disturbance observer and Gaussian process to achieve the robust-ness against disturbance. Disturbance observer is designed from a frequency domain perspective to effectively remove disturbances within a desired range. The Gaussian process stochastically models disturbances that may have high frequency components in the time domain but have a definite period in the position domain. As a result, the synergy between DOB and Gaussian model improves control performance. In addition, the stability of the Gaussian model is theoretically verified and the proposed method is compared under various conditions. © 2022 IEEJ-IAS.