Abstract: In view of the frequent collision accidents at unsignalized intersections, a dynamic risk detection and early warning strategy for networked vehicle collision at unsignalized intersections based on vehicle-vehicle communication is studied and proposed in the connected artificial driving environment. Firstly, a vehicle trajectory model based on polynomial fitting is designed by constructing a typical unsignalized intersection scene. Secondly, using time-to-collision and time exposed time-to-collision as risk detection indicators, a circular-bicircular vehicle model is constructed to realize vehicle collision risk detection. Thirdly, considering the heterogeneity of drivers and their interaction behavior, a two-level collision warning strategy based on game theory and genetic algorithm is proposed. Finally, a simulation environment is built based on SUMO to test and analyze the effectiveness and applicability of the early warning strategy. The results show that the proposed strategy can accurately identify all collision events and trigger early warning, and the success rate of early warning reaches 100%. Under different proportions of drivers in test conditions, the proposed early warning strategy can significantly reduce the collision rate and average collision kinetic energy.

Key words: unsignalized intersection, connected vehicle, dynamic risk detection, graded warning, driver heterogeneity, game theory

摘要： 针对无信控交叉口碰撞事故多发的问题，面向网联人工驾驶环境，研究提出了一种基于车车通信的无信控交叉口网联车辆碰撞动态风险检测与预警策略。在构建典型无信控交叉口场景的基础上，设计了一种基于多项式拟合的车辆轨迹模型；融合利用碰撞到达时间和风险暴露时间作为风险检测指标，构建了一种圆形-双圆车辆模型实现车辆碰撞风险检测；综合考虑了驾驶人异质性及其交互行为，提出一种基于博弈论和遗传算法的两级碰撞预警策略；基于SUMO搭建仿真环境对预警策略的有效性及适用性进行测试分析。结果表明，所提策略能够准确识别出所有的碰撞事件并触发预警，预警成功率达到100%；在不同比例驾驶人组成测试工况下，所提预警策略均能显著降低碰撞率和平均碰撞动能。

关键词: 无信控交叉口, 网联车辆, 动态风险检测, 分级预警, 驾驶人异质性, 博弈论