论文部分内容阅读
以道路网络的路段流量为基础进行OD分布矩阵估计.与以往文献不同的是本文应用了多层前馈神经网络的方法.由于路段流量与相关的OD矩阵分布之间存在连续性关系,这为神经网络模型的逼近特性提供了可行性.本文的方法适用于OD分布矩阵的实时校正.在已知OD分布矩阵的前提下,对两种情境———试验网络和实际Naples农村道路网进行仿真分析.主成分分析法的应用减少了变量个数并有利于改进输入数据.估计误差相对较低,与分析方法相反的是处理的时间几乎是实时的,因此这种方法可用于动态交通管理.本文的神经网络方法在误差和计算时间方面优于传统商业软件得到的OD估计结果.
OD distribution matrix estimation is based on the link traffic in the road network.Different from the previous literature, the method of multi-layer feedforward neural network is applied in this paper.Because of the continuity relationship between link flow and the related OD matrix distribution, The approximation property of neural network model is feasible.The method in this paper is suitable for real-time correction of OD distribution matrix.On the premise of known OD distribution matrix, two scenarios --- experimental network and actual Naples rural road network are simulated Analysis. The use of principal component analysis reduces the number of variables and facilitates the improvement of the input data. The estimation error is relatively low, in contrast to the analysis method, the processing time is almost real-time, so this method can be used for dynamic traffic management. The neural network method in this paper is superior to OD estimation results obtained by traditional commercial software in terms of error and computation time.