Estimation of fundamental diagrams in large-scale traffic networks with scarce sensor measurements

The macroscopic fundamental diagram (MFD) relates space-mean flow density and the speed of an entire network. We present a method for the estimation of a “normalized” MFD with the goal to compute specific Fundamental Diagram in places where loop sensors data is no available. The methodology allows using some data from different points in the city and possibly combining several kinds of information. To this aim, we tackle at least three major concerns: the data dispersion, the sparsity of the data, and the role of the link (with data) within the network. To preserve the information we decided to treat it as two-dimensional signals (images), so we based our estimation algorithm on image analysis, preserving data veracity until the last steps (instead of first matching curves that induce a first approximation). Then we use image classification and filtering tools for merging of main features and scaling. Finally, just the Floating Car Data (FCD) is used to map back the general form to the specific road where sensors are missing. We obtained a representation of the street by means of its likelihood with other links within the same network.