Position Control Using Acceleration-Based Identification and Feedback With Unknown Measurement Bias

Abstract

A position-command-following problem for asymptotically stable linear systems is considered. To account for modeling limitations, we assume that a model is not available. Instead, acceleration data are used to construct a compliance (position-output) model, which is subsequently used to design a position servo loop. Furthermore, we assume that the acceleration measurements obtained from inertial sensors are biased. A subspace identification algorithm is used to identify the inertance (acceleration-output) model, and the biased acceleration measurements are used by the position-command-following controller, which is constructed using linear quadratic Gaussian (LQG) techniques.