Vehicle Path Tracking Control Using Pure Pursuit With MPC-Based Look-Ahead Distance Optimization
Seungtaek Kim, Jonghyup Lee, Kyoungseok Han, Seibum B. Choi
Abstract
Optimizing look-ahead distance for the pure pursuit has been an important issue as it determines the relation between the smooth path tracking performance and tracking error. This article presents a novel model-based look-ahead distance optimization method to ensure both smooth path tracking performance and acceptable tracking error for pure pursuit. The conventional pure pursuit method was modified to give the look-ahead distance as the control input, and the optimization problem was formulated as model predictive control to optimize the look-ahead distance. The proposed method optimized the look-ahead distance by reducing it only when the vehicle was about to deviate excessively from the path and maximizing it elsewhere to improve cutting corner problems and ensure smooth path tracking performance. Simulation and experimental results of a test vehicle performing on various roads, including handling course, double lane change, and 90-degree turn, showed the ability of the proposed method to optimize look-ahead distance, gathering the advantages of the pure pursuit method with changing look-ahead distance.