考虑交通事件影响的高速公路短时行程时间预测

本研究目的是基于历史行程数据并考虑交通事件,构建可提高高速公路短时行程时间预测精度的方法。基于深层机器学习理论,设计加权均方根相似(Weighted-RMSS)模型,利用经纬度将行程分段,考虑高速公路车辆时空流动性的时间传递,计算当前行程时间和历史案例行程时间的相似性,提高了行程时间预测精确度。在此基础上,结合交通事件数据建立交通事件影响矩阵,建立LGBM模型(Light Gradient Boosting Machine)用于短时行程时间预测,并利用广州高速公路平沙至机场南路段实测数据进行验证。研究结果表明,开发2个模型效果均优于传统KNN模型,且考虑了交通事件影响的LGBM模型的预测精度高于Weighted-RMSS模型,达到95.68%,比较不同未来预测时间得出预测5 min效果最佳,精度可达96.18%。本研究在短时行程时间预测上有显著的优越性,有助于为驾驶人提供准确的出行时间,有利于高速公路的交通管理。

Influence range of emergency under special events based on CTM

To determine the dynamic influence range of emergencies under special events, the spacial and temporal characteristics of the traffic flow are studied by simulation based on the cell transmission model （CTM）. Based on the traffic management measures used under special events, a semi-dynamic assignment algorithm is proposed, which is combined with an algorithm for logit multi-path traffic assignment and the CTM. In a simple calculation network, the spacial and temporal characteristics of traffic flows which vary with different traffic management schemes are studied, and a method to obtain the influence range of emergency is proposed by computing the jam time of the intersections. By contrasting the average delay of each vehicle, the dissipation effect is studied under two different traffic management schemes. The example shows that the spatial and temporal variety of the traffic flow can be easily simulated and the influence range of emergency can be confirmed by the method based on the CTM. The proposed method provides a new idea for decision-making on traffic management under emergency under special events.