Data fusion algorithm for the 3D-MFD estimation

Abstract

The concept of the Macroscopic Fundamental Diagram (MFD) has been recognized as a powerful tool to develop network-wide control strategies. Recently, it has been extended to the threedimensional MFD (3D-MFD), used to investigate traffic dynamics of multimodal urban cities where different transport modes compete for, and share road infrastructure. Due to the limited amount of available data used to develop the MFD or 3D-MFD, different estimation methods have been proposed. In most cases, the data comes from either loop detectors or GPS-equipped mobile probe vehicles. Recent research has shown the value of fusing those two data sources for improving the accuracy of an estimated MFD, but requires a priori information about the probe penetration rate (PPR). Considering that this information is not very often available or is very difficult to infer, implementation of such a fusion method has been constrained so far only to simulation data. In this study, however, we propose a methodology to estimate the 3D-MFD that does not require the PPR as an input. To that end, we have developed a fusion algorithm that combines information from probe vehicles and automatic vehicle location devices of public transport to estimate the average speed of cars and further a 3D-MFD in a mixed bi-modal urban network. The findings show that the proposed algorithm can significantly reduce the estimation error when compared to an estimation method that uses only one data source.