Simultaneous Robust State and Sensor Fault Estimation of Autonomous Vehicle via Synthesized Design of Dynamic and Learning Observers
Siyou Tao, Jicheng Chen, Bin Zhou, Hui Zhang
Abstract
In this article, a scheme for estimating system state and sensor fault based on dynamic observer and neural network is proposed for autonomous vehicles. For simultaneous state and sensor fault estimation, the sensor fault is augmented as model state. Since the longitudinal velocity is a time-varying parameter which brings challenges to observer design, the vehicle model is converted to a Takagi-Sugneo (T-S) fuzzy form. Moreover, the modeling uncertainty caused by linearization, unmodeled dynamics, and exogenous disturbance degrades the estimation performance. To cope with the state-estimation problem with modeling uncertainty, a novel scheme is established. This scheme incorporates a neural network to accurately estimate modeling uncertainty, along with a robust dynamic observer to estimate the state of the augmented model with compensation of modeling uncertainty. Finally, simulation and experiment are carried out to verify the effectiveness of the proposed scheme. From the comparison results, we find that the proposed scheme has great improvement on estimation accuracy and convergence rate.