Simulation and Comparison of Two Fusion Methods for Macroscopic Fundamental Diagram Estimation

Accurate estimation of macroscopic fundamental diagram (MFD) is the precondition of MFD’s application. At present, there are two traditional estimation methods of road network’s MFD, such as the loop detector data (LDD) estimation method and the floating car data (FCD) estimation method, but there are limitations in both traditional estimation methods. In order to improve the accuracy of road network MFD estimation, a few scholars have studied the fusion method of road network MFD estimation, but there are still some shortcomings on the whole. However, based on the research of adaptive weighted averaging (AWA) fusion method for MFD estimation of road network, I propose to use the MFD’s two parameters of road network obtained by LDD estimation method and FCD estimation method, and establish a back-propagation neural network data fusion model for MFD parameters of road network (BPNN estimation fusion method), and then the micro-traffic simulation model of connected-vehicle network based on Vissim software is established by taking the intersection group of the core road network in Tianhe District of Guangzhou as the simulation experimental area, finally, compared and analyzed two MFD estimation fusion methods of road network, in order to determine the best MFD estimation fusion method of road network. The results show that the mean absolute percent error (MAPE) of the parameters of road network’s MFD and the average absolute values of difference values of the state ratio of road network’s MFD are both the smallest after BPNN estimation fusion, which is the closest to the standard MFD of road network. It can be seen that the result of BPNN estimation fusion method is better than that of AWA estimation fusion method, which can improve the accuracy of road network MFD estimation effectively.