一种动态OD矩阵估计算法的理论及应用

首先比较几种动态OD矩阵的估计方法（如广义最小二乘法、最大似然法，最小信息量法和极大熵法），优选出极大熵法，用组合理论推导出该方法的数学模型，进一步用拉格朗日乘子法求出模型的隐式解，由于实际解需迭代求出，文中介绍了一种迭代算法，并证明了它的正确性，最后，把所建立的模型和介绍的算法用于仿真案例和江苏省常熟市的交通诱导系统中，效果令人满意。

Theory and applications of dynamic OD matrix estimation algorithm

Maximum entropy method is adopted to obtain dynamic OD matrix, for it is the best one among the methods such as generalized least square method, maximum likelihood method and information minimizing method. Its mathematical model for estimating dynamic OD matrix is derived by using combinatorial theory. The implicit solutions of the model are derived by use of Lagrange multiplier. The actual solutions can only be obtained from iterative algorithm. The correctness of the iterative algorithm is strictly proved. Finally, the model and the algorithm are used to a simulation example and the traffic guidance system of Changshu city in Jiangsu Province in P. R. China. The results obtained are satisfactory