Gaussian Process and Disturbance Observer Based Control for Disturbance Rejection

Abstract

This paper proposes Gaussian process and disturbance observer-based (GP and DOB-based) control by stochastic learning of position-based disturbance, where patterned disturbance is effectively eliminated in positions such as the cogging force and ripple. The GP-based control properl removes the disturbance with a deterministic period in the position domain, and the DOB effecti el compensates for the unexpected disturbance that is no patterned in the osition domain ithin the de igned frequene range. Therefore GP and DOB-based control achie 'e the high- end performance of the control s stem. In addition, the reliabilit of the Gaussian process model is theoreticall erified and simulation results are alidated b comparing with con entional DOB-based control and proposed GP and DOBbased control in arious disturbance conditions.