Novel Algorithm for Calibrating Speed-Density Model Parameters in Mesoscopic Traffic Simulator

摘要

This article is on the basis of the classical speed-density model which combines with the complex road traffic flow, to raise a way that using the weighted regression to calibrate speed-density model parameters in mesoscopic traffic simulator. After processing detector data, the densities are taken as the variables; locally weighted regression is used to build a nonparameter relationship for the number of the traffic flow, to calibrate the vehicle speed. The test with a huge amount of factual data shows that the methods proposed in this paper outperforms the common optimal algorithm: simplex method in the vehicles speed estimation precision, and can accurately descript the dynamic change regularity of the road traffic flow.