基于反向传播神经网络模型的公交进站换道决策行为分析

公交车准备进站时将发生强制性换道,容易造成城市道路通行能力降低,诱发交通拥堵;还会对道路安全构成威胁。首先立足于现状调查,分析了进站决策点;并对公交车在车站上游的进换点分布情况进行了描述。其次提出了交通量、公交车数和离站路程为影响公交车换道的最主要因素;并将这3个因素作为输入变量,建立了以上游各区段进换点数量为输出的BP神经网络模型。最终,利用权积法对所建模型中的各输入变量进行敏感性分析。结果表明:交通量、离站路程与进换点数成负相关;而公交车数与之成正相关;其中,敏感系数最大的为离站路程,达0.220;3个影响因素在增加20%的扰动后,公交车数的敏感系数增速大于另外两者。

The Decision Making of Bus Entering Lane-Changing Behavior Analyzed by Back-Propagation Neural Network Model2.20

When bus enters bus stop, it has a mandatory lane-changing behavior which may cause capacity decrease of the upstream road. Firstly, based on the traffic survey, the lane changing location has been analyzed, and the bus entering lane-changing point distribution has been provided. This paper also figures out three main factors that impact bus entering lane changing, the traffic volume, the number of buses, and the bus location. Then the three factors are used as the input variable to model the back-propagation neural network (BPNN) of which the output is the entering lane-changing point in upstream road. Lastly, the weight product method is used to analyze the sensitivity of those three factors. The result shows that, traffic volume and bus location are positively correlated with the lane-changing point. However there is a negative relationship between the bus number and the point. Furthermore, bus location has the largest sensitivity coefficient with the value being 0.22. And when 20% perturbation is added to those three factors, the sensitivity of bus number grows higher than other two factors.