Neural Network-Aided Linear Parameter-Varying Modeling and Finite-Time Control for Autonomous Driving

Abstract

Irresponsible driving behavior remains one of the primary causes of road accidents. Given the limited effectiveness of repressive measures, this study proposes an intelligent-vehicle-based solution aimed at enhancing driving safety. The proposed approach focuses on the real-time estimation of unmeasured vehicle states and the torque applied by the driver, which are critical for improving shared control between human and automated systems. To achieve this, a linear parameter-varying model is developed to represent the vehicle’s nonlinear dynamics. The varying parameters of the linear parameter-varying model are estimated using an artificial neural network, ensuring both robustness and adaptability to dynamic driving conditions. Simultaneously, a finite-time sliding mode controller is designed to ensure rapid convergence of tracking error, enhancing system responsiveness. The integrated estimation and control framework is validated through MATLAB and Simulink simulations. Results demonstrate the effectiveness of the proposed method in accurately estimating key states and managing control under a human–machine cooperative driving scenario. This approach contributes to the advancement of intelligent vehicle systems by promoting safer and more reliable interaction between drivers and automated controllers.