Estimation of O/D Matrices from Traffic Counts: Theoretical Analysis and Simulation Experiments

We introduce a new binary integer linear programming model to formulate a generalised version of the sensor location problem, that is, to locate a minimal-cost set of counting points in order to derive the complete traffic flow vector on non-simmetric directed transport networks. We define a pair of heuristic algorithms that determine tight lower and upper bounds on the number of traffic sensors. Experimental results on a set of randomly generated test problems show that, even on medium-large size networks, estimated O/D matrices are closer to the true ones when obtained by the newly proposed sensor location strategies than by the common method, or, conversely, such strategies are more efficient (i.e., they require a lower measurement cost) than the common method to attain a given reliability level for the estimates. Preliminary simulation results on the real-world case of Salerno (an Italian medium-large town) are also presented.