基于行程时间预测的城市车辆出行路径推荐方法  

Recommendation Method of Urban Vehicle Travel Route Considering Travel Time Prediction

作　　者：李晓玉 邢雪[1] LI Xiao-yu;XING Xue(School of Information and Control Engineering,Jilin Institute of Chemical Technology,Jilin,Jilin 132022,China)

机构地区：[1]吉林化工学院信息与控制工程学院,吉林吉林132022

出　　处：《黑龙江交通科技》2023年第12期153-156,161,共5页Communications Science and Technology Heilongjiang

基　　金：吉林省科技发展计划资助项目(20210101416JC)。

摘　　要：路径推荐在提高出行者出行效率、减少出行成本和平衡路网运行状态、减少时间和空间上的拥堵方面具有重要意义。通过构建动态图卷积神经网络(DGCN)对路网各路段进行多步行程时间预测,根据出行需求并考虑避让拥堵区和均衡路网因素进行车辆出行路径推荐。其中,通过GN算法对结合路网结构和流量特征构建加权网络模型进行路网分区,并结合平均行程速度和预测所得行程时间,识别拥堵区域交通状态。实验结果表明:在实例数据集上使用残差拉普拉斯矩阵DGCN Res模型进行行程时间预测,比基于注意力的时空图卷积网络模型(ASTGCN)的MAPE提高4%、RMSE提高9.2%。在此基础上进行拥堵区避让的路径选择推荐,可有效降低路网拥堵均衡指数,因此从路网全局层面考虑拥堵避让机制可减缓路网整体拥堵程度。Route recommendation is of great significance in improving travel efficiency,reducing travel costs,balancing road network operation status,and reducing time and space congestion.A Dynamic Graph Convolutional Neural Network(DGCN)is constructed to perform multi-time-step travel time prediction for each road section of the road network,and vehicle travel route recommendation is performed according to travel demand and considering factors such as avoiding congestion areas and balanced road network.The road network partitioning is carried out by building a weighted network model combined with road network structure and traffic characteristics through the GN algorithm,and combined with the average travel speed and the predicted travel time,identify the traffic status in the congested area.Experimental results show that:using the Residual Laplacian Matrix DGCN Res model for travel time prediction on the instance dataset achieves a 4%improvement in MAPE and a 9.2%improvement in RMSE over the Attention-based Spatiotemporal Graph Convolutional Network model(ASTGCN).Based on the prediction results,the route selection recommendation for the avoidance of the congestion area can effectively reduce the congestion equilibrium index of the road network.Therefore,considering the congestion avoidance mechanism from the global level of the road network can reduce the overall congestion degree of the road network.

关 键 词：智能交通 路径推荐 动态图卷积网络 行程时间预测 拥堵避让 

分 类 号：U491.1[交通运输工程—交通运输规划与管理]