Design and experimental implementation of observer-based adaptive neural network steering control for automated vehicles

In this paper, an observer-based adaptive neural network controller is developed for the Steer-by-Wire (SbW) system of automated vehicles with uncertain nonlinearity and unmeasured state. An observer is introduced to estimate the angular velocity of the front wheels, so the hardware cost and the complexity of mechanical structure and electronic circuits are reduced. Then, an observer-based adaptive neural network controller is proposed for the SbW system to achieve excellent steering precision. A radial basis function (RBF) neural network is used to model the uncertain nonlinearity, which mainly includes self-aligning torque and unknown friction torque with strong nonlinearity. The adaptive law of the RBF neural network designed by fuzzy basis functions rather than by its filtering is derived by Lyapunov stability theory and the Strictly Positive Real (SPR) condition. The tracking error and the observation error can be guaranteed to converge asymptotically to zero. Simulation and experimental results for two paths highlight the effectiveness of the proposed control algorithm.