Development of a Human-Like Model Predictive Path Tracking Control Algorithm for Autonomous Vehicles with Self-Tuning of Control Period

This study proposes a model predictive path tracking control algorithm for autonomous vehicles based on self-tuning of control period. A model predictive control (MPC) algorithm was designed to compute optimal steering angle using bicycle model-based error dynamics. The path tracking errors such as lateral preview error and yaw angle error were computed using waypoints and current position of vehicle. In order to construct human-like controller, self-tuning algorithm of control period of model predictive control algorithm was designed using control errors and weighting factor. The control period function was newly designed to represent human control characteristic and it was used to determine activation time of the MPC. The algorithm was constructed in CarMaker and Matlab/Simulink environment and performance evaluation was conducted under curved path tracking scenario. The evaluation results showed that the proposed control algorithm can represent human control characteristics in terms of control errors.