Yi, Jian (Huzhou University), Wu, Xialai (Huzhou University), Lin, Ling (Huzhou University), Qin, Jiabin (Huzhou University), Chen, Junghui (Chung-Yuan Christian Univ)

Data-Driven Model Predictive Control for Vehicle Path-Tracking Using a Recurrent Neural Network

Scheduled for presentation during the Invited session "Estimation and Prediction" (FrB1), Friday, November 1, 2024, 11:50−12:10, Room T1

7th IFAC Conference on Engine and Powertrain Control, Simulation and Modeling, Oct 30 - Nov 1, 2024, Dalian, China

This information is tentative and subject to change. Compiled on January 2, 2025

Abstract

Path tracking control is crucial for autonomous vehicle driving. Traditional nonlinear model predictive control (NMPC) for path tracking demands extensive computation, making real-time implementation challenging. This paper introduces a Fast Model Predictive Control (FMPC) method for vehicle path tracking, utilizing a recurrent neural network (RNN) with symmetric saturating linear transfer functions (SSL-RNN) in the hidden layer. The proposed approach leverages the SSL-RNN model to accurately capture vehicle dynamics. Consequently, the optimal control problem in MPC is reformulated as a mixed integer linear programming problem, facilitating swift solutions. Simulation experiments validate the proposed method's efficacy. Compared to traditional vehicle mechanism model-based MPC and RNN-based NMPC, our FMPC demonstrates superior accuracy in path tracking and significantly enhances controller solution efficiency.