Development of neural-network control of steer-by-wire system for intelligent vehicles

Abstract

An artificial neural network controller is developed for steer-by-wire of an intelligent vehicle. Prior to development of the control systems, the non-linear model of the steer-by-wire steering systems was developed. This was shown to be necessary to appreciate and understand the modelling difficulties and complexities inherent in modelling such steering systems. These difficulties and complexities were also shown to underpin the benefits and efficacy of neural-network control techniques for steer-by-wire system. By considering the non-linearity of steer-by-wire systems, a neural-network model reference adaptive control technique was used and this was integrated into the control of the steer-by-wire system test rig. The results of an extensive simulation and experimental study show that the non-linear dynamics of the steer-by-wire system have been well compensated by the adaptive compensator integrated as part of the controller.