领域(数学)
计算机科学
驾驶模拟
视野
模拟
心理学
数学
神经科学
纯数学
作者
Zhaojie Wang,Guangquan Lu,Haitian Tan
标识
DOI:10.1109/tits.2024.3465442
摘要
At uncontrolled intersections, drivers face a more complex traffic environment. The interactions between vehicles forces drivers to make complex decisions to avoid vehicle collisions. In most existing studies on modeling driving behavior at uncontrolled intersections, multi-vehicle interaction scenarios are often categorized as moving-across scenario and merging scenario. Separate driving behavior models are constructed for each scenario. Few studies have been conducted to model drivers' driving behaviors at uncontrolled intersections uniformly. In this study, based on risk homeostasis theory (RHT) and preview-follower theory (PFT), a risk-field based driving behavior model is constructed to simulate multi-vehicle interactions at uncontrolled intersections. Taking into account the driver's visual field characteristics, this model uses risk field theory to quantify the risk vehicle endured during preview time and find the trajectory that conform to the driver's behavior while satisfying the risk threshold constraint. An exploratory algorithm was contained in this model to solve intersection deadlock scenarios. By adjusting the acceptable risk level, vehicles can restart the traffic flow by finding a suitable acceleration from feasible action space. The model is able to uniformly describe the driving behavior of drivers within uncontrolled intersections, including merging behavior and moving-across behavior. The simulation results show that the model can accurately reproduce the priority and trajectory of vehicles crossing the intersection and resolve multi-vehicle conflicts within a reasonable time. The application of our model to improve the traffic efficiency in mixed traffic environments with automated and manual vehicles is also discussed. The model can be used for traffic simulation at uncontrolled intersections and to provide test validation for automated driving systems.
科研通智能强力驱动
Strongly Powered by AbleSci AI