Sensor Location Model to Optimize Origin–Destination Estimation with a Bayesian Statistical Procedure

For transportation planning and operations decisions, static origin–destination (O-D) matrices, which specify the number of trips between each O-D pair, are an essential input. One approach for estimating O-D matrices is to use data from traditional counting sensors on links with assumptions or models concerning how drivers choose routes on the network. The inverse problem is to locate a given number of counting sensors to obtain good O-D matrix estimates. A new model for locating sensors to minimize the uncertainties in route flow estimates is presented in this paper. A route choice set from each origin to each destination is assumed to be known, and prior estimates of flows on the routes in this set are given. The reliabilities for each O-D route prior estimate and the link flow measurements are assumed to be given as well. The sensors in this problem scenario are not necessarily perfect, and measurement errors may occur; however, the reliabilities of the sensors are assumed to be known. Extensive computational experiments and comparisons with existing sensor location models indicate that the proposed model consistently gives more reliable estimates of O-D flows.