Adaptive Model Predictive Control for Intelligent Vehicle Trajectory Tracking Considering Road Curvature

• Published in: International journal of automotive technology Vol. 25; no. 5; pp. 1051 - 1064
• Main Authors: Gao, Yin, Wang, Xudong, Huang, Jianlong, Yuan, Lingcong
• Format: Journal Article
• Language: English
• Published: Seoul The Korean Society of Automotive Engineers 01.10.2024; Springer Nature B.V; 한국자동차공학회
• Subjects: Adaptive control; Algorithms; Automotive Engineering; Back propagation networks; Chassis; Control algorithms; Control stability; Control systems design; Control theory; Controllers; Curvature; Electrical and Electronics; Engineering; Feedforward control; Intelligent vehicles; Lateral stability; Neural networks; Particle swarm optimization; Predictive control; Real time; Road conditions; Sampling; Steering; Stiffness; Tires; Tracking control; Traffic speed; Trajectory control; Vehicle Dynamics and Control; 자동차공학; Variable predictive horizon; Adaptive Model Predictive Control (AMPC); Trajectory tracking; Back Propagation (BP) neural network; Autonomous vehicles; Particle Swarm Optimization (PSO)
• ISSN: 1229-9138; 1976-3832
• DOI: 10.1007/s12239-024-00086-8

A parametric Adaptive Model Predictive Controller (AMPC) based on Particle Swarm Optimization-Back Propagation (PSO-BP) neural network has been developed in this paper, the primary focus is on improving the trajectory tracking performance of autonomous vehicles under varying road conditions. The PSO-BP neural network is employed for real-time adjustment of the controller's predictive horizon and sampling time. A vehicle dynamics model is established and an improved tracking control algorithm involving road curvature feedforward is proposed. In the design of AMPC, the real-time update of tire lateral stiffness is achieved through the adoption of the Recursive Least Squares (RLS) method, ensuring the precision of trajectory tracking for the vehicle under varying operating conditions. The simulation platform, which combines Carsim and Simulink, was employed for validating the proposed approach. The findings reveal that the proposed controller can promptly adjust the predictive horizon and sampling time according to the vehicle's state. Through the employed estimation strategy, real-time adjustments of tire lateral stiffness are achieved, allowing for dynamic alterations of vehicle speed and front wheel angle in response to road curvature. As a result, this approach significantly enhances control accuracy and lateral steering stability, especially in large curvature conditions.