A study of equivalence between adaptive learning and loop shaping methods for disturbance compensation

Abstract

AbstractThis paper introduces a theoretical analysis of the disturbance compensation methods in positioning systems. Thereare two major methods used to compensate for disturbances. The first one is based on a loop shaping method andthe second one is based on an adaptive learning method. We make a comparison between a resonant filter used asthe loop shaping method and an enhanced adaptive feed-forward cancellation (AFC) method used as the adaptivelearning method. To compensate disturbances, the resonant filter decreases the gain of the sensitivity function atthe disturbance frequency by using the vector locus. The adaptive learning method adjusts the coefficients of theenhanced AFC to compensate for a disturbance. We found that the LTI model of an enhanced AFC is the sameas that for a resonant filter when we compared their characteristics. A simulation of the disturbance compensationin a hard disk drive showed that these time responses and frequency responses coincided. Moreover, the resultsshowed advantages of each of the control methods with the exception of the performance. A resonant filter realizeslow cost implementation with compare to an enhanced AFC. That is, a resonant filter easier to implement than anenhanced AFC. An enhanced AFC has the utility to realize a high quality function control system simultaneouslyable to compensate and estimate for disturbances. The results of the evaluation in this study will provide designengineers with guidelines for selecting an optimum control strategy in actual applications.Key words : Disturbance compensation, Periodic disturbance, Non-periodic disturbance, Adaptive feed-forward