Predictive and linear quadratic methods for potential application to modelling driver steering control

Abstract

A brief review of the literature reveals that both predictive control theory and linear quadratic (LQ) control theory have been used to design path-following controllers with preview, but it is not clear how the controllers compare. This article derives optimal linear preview controllers using the two approaches starting from a common state-space description of the vehicle dynamics. The transformation of the controllers from ground-fixed axes to vehicle-fixed axes is discussed. The influences of preview horizon, control horizon and cost function are investigated. For the case of long preview and long control horizons, it is found that the predictive and LQ approaches give identical controllers. The results in this article provide a basis for identifying human steering behaviour from measured data.