面向边界控制的路网小区划分

在城市路网中，将较为拥挤的子路网从整个路网提取出来，通过边界控制调节拥堵区域的出入流，缓解拥挤区域的拥挤程度，是解决城市交通大范围拥堵的一个有效途径。作为边界控制的对象小区，具有紧凑型的宏观基本图（Macroscopic Fundamental Diagram，MFD）是必备前提。首先利用归一分割（Normal Cut，Ncut）法将路网分割成一定数量的子路网，然后建立了在动态合并中寻找具有最优拟合度的MFD子区的数学模型。然后对模型进行分析，提出求解模型的贪婪算法。以美国爱达荷州博伊西市中心区的路网为研究对象，分析对比了该路网分别在Ji的方法和所提方法划分下的区别，发现所提方法在寻找最优MFD拟合度小区的同时也可以优化子区的密度均匀度，对路网进行合理划分。最后通过对比分析定时控制和自适应控制下小区划分结果和子路网MFD的特征发现：①最优MFD拟合度子区不仅继承了原路网的MFD基本特征，而且与原路网相比有更高的道路使用率；②非最优MFD拟合度子区的MFD特征对该子区的交通态势有一定表征；③自适应控制对路网的优化效果可在原路网和最优MFD拟合度子区的MFD特征上得到表现。

Spatial Partitioning of Traffic Networks for Boundary Flow Control

It is an effective way for alleviating traffic congestion to use boundary flow control strategy to the extracted sub-network with the higher congestion degree in traffic network. Especially, existing compact Macroscopic Fundamental Diagram (MFD) in objective sub-network is an essential prerequisite for boundary flow controlling. Based on the initialization areas partitioned by Normal Cut algorithm, the math model to find an optimal fitting degree sub-network in dynamic merging process is built in this paper. Then, after analysis of the model, the greedy algorithm to solve this model is proposed. Finally, the method is applied to the network of downtown of Idaho in Boise. After analyzing the application and comparing the method in this paper and Ji's method, we find the method have a reasonable result, in addition, the network can be divided more evenly when the optimal solution is achieved. Hence, analyzing and comparing both the results of sub-network partitioning and characteristics of MFD between fixed-time control and adaptive control, we find that ① MFD characteristic of the sub-network with optimal goodness of fit inherits the one of origin network, and the road resource utilization in the optimal sub-network is higher than that in the origin network; ② Based on MFD characteristics of sub-networks with non-optimal goodness of fit MFD, the traffic situation of these sub-networks can be analyzed; ③ The optimization effect of using adaptive control in the network can be observed in MFD of both origin network and the sub-region with optimal goodness of fit MFD.