Left-turn decision-making model with the consideration of surrounding vehicles information at complex and dynamic urban intersections is critical for the development of autonomous vehicles. In this work, we propose an approach of developing the algorithms for left-turn tactical-level decision-making to ensure autonomous vehicles to cross intersections safely. Firstly, the trajectories data were collected at intersections by subgrade sensors and a retrofit autonomous vehicle. Then, the GPR model was trained to predict the trajectories of oncoming straight vehicles. After that, a conflict resolution method for left-turn behavior of autonomous vehicles at urban intersections and a driving action selection method considering multiple factors, such as safety, efficiency, comfort and altruism, were proposed to makes appropriate driving actions. Finally, a simulation and verification platform for algorithm was built based on the software of matlab/Simulink&Prescan. And the reliability and effectiveness of the tactical-level decision-making model was verified by both simulations and real road experiments. The results from this study can provide a theoretical basis for the in-depth study of left-turn driving decision-making under complex and uncertain intersection environment.

CICTP