Overtaking path planning and tracking control of autonomous vehicles using optimal control method

Abstract

The overtaking path planning and tracking control algorithm that improves driving efficiency by passing slower preceding vehicle is essential for popularization and advancement of autonomous vehicles. The overtaking maneuver brings high risk due to relative velocity and complex interactions among vehicles. Furthermore, the complexity of technology increases for better ride quality of passengers. This thesis presents a hierarchical overtaking system using optimal methods combined with Model Predictive Control (MPC) based path planner and the Linear Quadratic Gaussian (LQG) based path tracking controller. The proposed overtaking algorithm considers interaction among vehicles and guarantees safety, efficiency, and ride comfort. Using the hierarchical structure, the algorithm utilizes the advantages of each method to plan the optimal overtaking path and ensure the stability and real-time performance of the path tracking. The cost function and constraints of the MPC are formulated in order to design each states systematically based on the conditional state machine. Furthermore, the stochastic MPC with chance constraint considers the uncertainty of prediction, thereby planning the safe optimal overtaking path with high efficiency and ride comfort. In addition, the Augmented LQG controller which considers the curvature disturbance is proposed to improve the path tracking performance. In this thesis, the performance of the proposed overtaking path planning algorithm is verified through statistical evaluation in traffic simulations for various scenarios. Finally, the controller is implemented in an autonomous vehicle for driving experiments in real-time.