Lane-changing decision and planning under humanized games for intelligent vehicles

Although autonomous driving vehicles have seen rapid development, their decision-making and planning technologies still fall short of meeting human-centric requirements. To address the issue of personalized driving within autonomous vehicles, a human-like autonomous driving framework is proposed. This framework divides the autonomous driving process into task decision-making and path planning. Firstly, during the modeling phase, a combined model of the driver and vehicle is integrated to form an integrated model for designing decision-making algorithms. Subsequently, a decision cost function is constructed during the decision phase to reflect driving safety, passenger comfort, and travel efficiency. Adjustments are made for different driving styles, employing non-cooperative game theory methods to resolve decision cost issues. Finally, an improved algorithm based on quintic polynomials is designed for trajectory planning to meet vehicle safety, comfort, and rapid operation requirements. Test results indicate that different driving styles lead to varying decision outcomes, yet the algorithms employed consistently provide reasonable decisions for autonomous driving vehicles, thereby enhancing road efficiency and safety.