基于行车风险场的高速公路交织区车辆轨迹预测方法  

Methods for predicting vehicle trajectories in motorway weaving zones based on driving risk fields

作　　者：秦雅琴[1] 董帅 谢济铭 陈亮[1] 刘拥华[1] 郭淼 QIN Yaqin;DONG Shuai;XIE Jiming;CHEN Liang;LIU Yonghua;GUO Miao(Faculty of Transportation Engineering,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]昆明理工大学交通工程学院,昆明650500

出　　处：《汽车安全与节能学报》2024年第6期952-961,共10页Journal of Automotive Safety and Energy

基　　金：国家自然科学基金资助项目(72261021)。

摘　　要：为提高交织区车辆轨迹预测精度,该文提出了一种融合行车风险场和车辆换道意图的车辆轨迹预测方法。分析交织区驾驶人驾驶需求变化,利用行车风险场模型统一表示车辆行驶时的交互风险;采用隐Markov模型识别车辆换道意图;通过深度置信网络在线学习机(DBN_OSELM)模型对输入特征进行多维度扩展和融合,提高交织区轨迹预测的准确率;最后,基于CitySim数据集对所提方法进行评估。结果表明:模型能以较高的准确率预测高速公路交织区的车辆轨迹,交织区驾驶人3类驾驶需求(汇入、保持、驶出)的车辆轨迹预测均方根误差(RMSE)分别为0.6835、0.2574、0.6315,平均位移误差(ADE)分别为0.46、0.21、0.48 m。该研究成果有助于提高复杂场景下的车辆轨迹预测精度,改善交织区的交通安全。A vehicle trajectory prediction method fusing the traveling risk field and vehicle lane-changing intention was proposed to improve the accuracy of vehicle trajectory prediction in the interweaving area.Firstly,the driving demand changes of drivers in the interweaving zone were analyzed,and the driving risk field model was used to uniformly represent the interaction risk when vehicles were driving;secondly,the Hidden Markov Model was used to identify the vehicle lane-changing intention;in addition,the input features were extended and fused in multiple dimensions by the Deep Belief Networks Online Learning Machine(DBN_OSELM)model,to improve the accuracy of the trajectory prediction in the interweaving zone.Finally,the proposed method was evaluated based on the CitySim dataset.The results show that the model can predict vehicle trajectories in the interweaving zone of highways with high accuracy,and the root mean square error(RMSE)of vehicle trajectory prediction for the three types(merging in the confluence area,maintaining the interweaving area,and driving out the diverging area)of driving needs of drivers in the interweaving zone are 0.6835,0.2574,and 0.6315,respectively,and the average displacement error(ADE)is 0.46,0.21,and 0.48 m,respectively.The research results are helpful to improve the accuracy of vehicle trajectory prediction in complex scenarios and improve traffic safety in the intertwined area.

关 键 词：智能交通 驾驶需求 行车风险场 换道意图 深度置信网络在线学习机(DBN_OSELM)模型 轨迹预测 

分 类 号：U492.84[交通运输工程—交通运输规划与管理]