Titolo della tesi: A methodology for OD matrix estimation using the revealed paths of Floating Car Data on large scale networks
The increasing availability of historical Floating Car Data represents a relevant chance to improve the accuracy of model–based traffic forecasting systems. In particular, a more precise estimation of Origin–Destination matrices is a critical issue for the successful application of Traffic Assignment models.
In this research, a methodology for obtaining demand matrices is developed starting from a data set of vehicle trajectories, without any other prior information, except for the topology of the road network and a sample of link volumes. Unlike traditional correction approaches, here no assignment model is pre-required, thus avoiding the burden of calibrating route choice and traffic congestion before travel demand. This implies a clearer sequence of actions from the modeler, which breaks circularities that may slow down the planning process.
Three steps are defined. First, a Data Driven method is applied to determine: the observed O-D matrix; the node departure shares from origins and the attraction shares to destinations (which replace connectors); the splitting rates (or turning fractions) for each destination; the speed profiles (for dynamic assignment). Second, an assignment matrix is obtained through a specific network loading procedure. Third, the O-D matrix estimation based on traffic counts is formulated as a scaling problem of the observed FCD demand.
Four optimization problems with different design variables are proposed: a uniform scale factor; production and attraction factors for each origin and destination, respectively; iterative production and attraction; the whole O-D matrix.
The methodology was successfully tested on the networks of Turin and Rome. The results highlight the concrete opportunity to apply a data driven methodology that, independently from the reliability of the assignment model used in the simulation, minimizes manual and specialized effort to build and calibrate the transportation demand model.