| 引用本文: | 杨逍遥,梁国华,陈亦新,杨潇洒,王宝杰.考虑右转车干扰的信号交叉口直行车辆轨迹预测[J].哈尔滨工业大学学报,2024,56(7):74.DOI:10.11918/202310015 |
| YANG Xiaoyao,LIANG Guohua,CHEN Yixin,YANG Xiaosa,WANG Baojie.Trajectory prediction of straight vehicles at signalized intersections considering interference from right-turning vehicles[J].Journal of Harbin Institute of Technology,2024,56(7):74.DOI:10.11918/202310015 |
|
| 摘要: |
| 相交道路右转车辆干扰下的直行车辆轨迹扰动现象成为城市道路信号交叉口交通运行的安全隐患,为提高直行车驾驶人对驶入右转车辆的应对能力并做出正确决策,对于直行车辆扰动轨迹的可靠预测至关重要。研究通过将不同状态下的直行车辆轨迹分布特征与相交道路右转车运行信息相关联,在车辆受扰轨迹识别的基础上将碰撞时间（TTC）指标作为输入层加入模型,构建了3层高斯混合-输入输出隐马尔可夫模型（GMM-IOHMM）,提出了一种考虑信号交叉口被交道路右转车对本面直行车作用程度的直行车辆受扰轨迹预测方法。实验结果表明:改进模型相比于传统的隐马尔可夫模型（HMM）在模型训练时能够更好地拟合实际的轨迹数据,且相比于传统的时间序列模型,GMM-IOHMM模型的拟合效果取得较大提升；可将TTC≤4.5 s且偏航角大于2.35°作为直行车辆轨迹扰动的判别标准。轨迹预测结果能够更加准确地判断直行车辆与周围车辆发生冲突的可能性,可作为受扰直行车及同行其他车辆辅助驾驶系统设计的重要依据。 |
| 关键词: 车辆轨迹预测 右转车干扰 受扰轨迹识别 GMM-IOHMM模型 累计频率曲线法 |
| DOI：10.11918/202310015 |
| 分类号:U491.2 |
| 文献标识码:A |
| 基金项目:国家自然科学基金(52172338)；陕西省科技计划项目(2024GX-YBXM-131) |
|
| Trajectory prediction of straight vehicles at signalized intersections considering interference from right-turning vehicles |
|
YANG Xiaoyao1,LIANG Guohua1,CHEN Yixin1,YANG Xiaosa2,WANG Baojie1
|
|
(1.School of Transportation Engineering, Chang′an University, Xi′an 710061, China; 2.Xi′an Outer Ring Branch of Shaanxi Communications Holding Group Co., Ltd., Xi′an 710100, China)
|
| Abstract: |
| The disturbance of straight vehicle trajectories under the interference of right turning vehicles on intersecting roads has become a safety hazard for traffic operation at signalized intersections on urban roads. To improve the ability of direct driving drivers to respond to right-turning vehicles and make correct decisions, it is crucial to reliably predict the disturbance trajectory of direct driving vehicles. This article associated the trajectory distribution characteristics of straight vehicles in different states with the motion information of right-turning vehicles on crossed approach. On the basis of identifying vehicle disturbance trajectories, the time to collision (TTC) was added as the input layer to the model, and a three-layer Gaussian Mixture Module-Input and Output Hidden Markov Model (GMM-IOHMM) was constructed. A method for predicting the disturbance trajectory of through vehicles was proposed, which considers the degree to which right-turning vehicles on the crossed road have an impact on the direct traffic on this surface at signalized intersections. The experimental results showed that the improved model can better fit actual trajectory data during model training compared with traditional HMM, and the fitting effect of GMM-IOHMM has been significantly improved compared with traditional time series models. And TTC is less than or equal to 4.5 s and "yaw angle is greater than 2.35 degrees" can be used as a criterion to determine whether a straight ahead vehicle is disturbed. The trajectory prediction results can more accurately determine the possibility of conflicts between direct vehicles and surrounding vehicles, and can serve as an important basis for the design of assisted driving systems for disturbed direct vehicles and other vehicles traveling together. |
| Key words: vehicle trajectory prediction right turn interference disturbed trajectory identification GMM-IOHMM model cumulative frequency curve method |
