摘要
Since it is difficult to fit measured parameters using the conventional traffic model, a new traffic density and average speed model is introduced in this paper. To determine traffic model structures accurately, a model identification method for uncertain nonlinear system is developed. To simplify uncertain nonlinear problem, this paper presents a new robust criterion to identify the multi-section traffic model structure of freeway efficiently. In the new model identification criterion, numerically efficient U-D factofization is used to avoid computing the determinant values of two complex matrices. By estimating the values of U-D factor of data matrix, both the upper and lower bounds of system uncertainties are described. Thus a model structure identification algorithm is proposed. Comparisons between identification outputs and simulation outputs of traffic states show that the traffic states can be accurately predicted by means of the new traffic models and the structure identification criterion.
Since it is difficult to fit measured parameters using the conventional traffic model, a new traffic density and average speed model is introduced in this paper. To determine traffic model structures accurately, a model identification method for uncertain nonlinear system is developed. To simplify uncertain nonlinear problem, this paper presents a new robust criterion to identify the multi-section traffic model structure of freeway efficiently. In the new model identification criterion, numerically efficient U-D factofization is used to avoid computing the determinant values of two complex matrices. By estimating the values of U-D factor of data matrix, both the upper and lower bounds of system uncertainties are described. Thus a model structure identification algorithm is proposed. Comparisons between identification outputs and simulation outputs of traffic states show that the traffic states can be accurately predicted by means of the new traffic models and the structure identification criterion.
基金
The work was supported by Chinese Science Foundation (No .60134010) .
