面向MFD的异质性城市交通路网分区方法

Heterogeneous Urban Traffic Network Partition for Macroscopic Fundamental Diagram

本文针对路网密度分布的空间异质性会导致宏观基本图(Macroscopic Fundamental Diagram,MFD)高度离散的问题,提出了一种将异质性城市路网划分为同质子路网的方法。利用张量分解算法提取路段交通状态在时间维度上的短期日变化和长期逐日变化特征,以此计算路段间相似度。以路段间相似度为边权,针对含权的城市道路交通网络改进传统的Fast-Newman快速划分算法,来保证划分后每个子路网内的路段交通状态相似且在空间上紧密分布。基于某市一个月的自动车牌识别数据对该方法进行实证分析,结果表明改进算法划分效果优于K-means算法及传统Fast-Newman快速划分算法,划分后每个子路网的MFD函数关系都有较好的拟合效果,子路网之间的交通流特征参数差异明显。

Aiming at the problem that the spatial heterogeneity of the network density distribution will lead to the high discreteness of macroscopic fundamental diagram(MFD), this paper proposes a method to divide the heterogeneous urban traffic network into homogeneous sub-networks. The tensor decomposition algorithm is used to extract both the short-term and long-term characteristics of traffic state in the time dimension, so as to calculate the similarity between links. Taking the similarity between links as the edge weight, the traditional Fast-Newman algorithm is improved for the weighted urban traffic network to ensure that the traffic status of links in each sub-network is similar and the links are closely distributed in space. An empirical analysis of this method is based on one-month automatic license plate recognition data in a certain city. Results show that the improved algorithm performs better than K-means algorithm and traditional Fast-Newman algorithm. The MFD function relationship of each sub-network after partition has a good fitting effect, and the difference of characteristic parameters of traffic flow between sub-networks is obvious.