Parameter estimation new algorithms in static simulation models for urban traffic organization and optimization

From two aspects of better approximation and easier programming, this paper studied the parameter estimation algorithms for simulation models in urban traffic organization and optimization. It presented the maximum-likelihood method, normal number and B-splines approximation and also made a comparison of their characteristics and using conditions. The maximum-likelihood method, based on the mathematical statistics and experimental data, is fit for the condition of lower accuracy. The normal number approximation uses Euclidean norm (1-norm, 2-norm and infinitesimal norm) as its error measure, it has clear geometrical meaning, it is suitable for the condition when the rule of data curve is clear. When the data distribution rule is obscure, B-splines approximation is useful. An example shows that these three methods are feasible, the average absolute error can be ordered as B-splines approximation, normal number approximation and maximum-likelihood method by descending sort, and their average absolute errors are 0.0028, 0.0169, 0.0171 individually. In the actual engineering, these three methods should be used separately to choose the one with smaller error. 3 tabs, 1 fig, 12 refs.