Research on Highway Lane Change Control Strategy Considering Driving Styles

Abstract

To address the insufficient adaptability of existing automatic lane-change control technologies to driving styles, a highway lane-change control strategy with driving styles is proposed. Based on the highD natural driving dataset, 19 characteristic parameters are extracted, including lateral speed, acceleration, and headway distance, etc. using the principal component analysis (PCA). Driving styles are classified into three styles: cautious, normal, and aggressive, using the K-means clustering algorithm. A lane change trajectory planning model, based on a fifth-degree polynomial, is developed to balance safety and efficiency by dynamically adjusting the weight coefficients of lane change time and acceleration. Subsequently, LTV MPC tracking controller is designed which adapts to the dynamic needs of different driving styles by adjusting cost function weights and input constraints. Simulation results from CarSim and MATLAB/Simulink show that the cautious controller exhibits the smallest lateral tracking error and optimal yaw rate fluctuation range under various speed conditions. The aggressive controller has the fastest response speed, and the moderate controller provides the best overall performance. Compared to the traditional fixed parameter models, the proposed method significantly enhances lane change safety, efficiency, and personalized adaptation by the parameterized modeling of the driving style.