基于填充函数方法的OD矩阵估计

对现有关于求解OD矩阵估计的最小二乘模型所采用的逐次迭代算法的不足进行了分析，并引进了一种全局最优化算法即填充函数方法来找寻该模型的全局最优解。数值试验表明：所提出的填充函数算法有能力找到问题的全局最优解，且与初始值的选取无关，也有潜力解决较复杂网络的OD矩阵估计。通过数值结果发现，模型的权值选取对数值结果有明显影响。为此，引进了一种确定权值的评价指标RMSE，它能反映估计量与真实值之间的接近程度。利用该指标，可以选取较合适的权值。

OD Matrix Estimation Based on Filled Function Method

Origin-destination （OD） traffic demand matrix is one of the necessary and basic data for transportation modeling and plan- ning purposes. This paper points out some existing shortcomings of successive iterative method for estimating OD trip matrix from the literature [3-5]. To overcome these shortcomings, the filled function method is firstly introduced. Then numerical experiments show that the proposed filled function methos has the potential to find the global solution for the generalized least square model of OD matrix estimation under consideration and do not depend on the choice the initial point. Furthermore, we find that the choice of the weight parameter of the OD matrix estimate model has effect on the numerical results. Finally, an evaluation index as the statis- tical measure of estimators called the root mean squared error is also presented to show how to choose the weighting parameter appropriately.