Improved matrix estimation under congested conditions

The estimation of origin-destination matrices from UK traffic counts provides an attractive alternative to costly interview survey techniques. A robust and flexible analytical approach is described for the process of estimating trip matrices which extends that of the widely-used ME2. A variety of methods are available to accommodate uncertainty in each of the prior estimate of the trip matrix and the traffic counts, including entropy and likelihood: a broad view of the process of matrix estimation is used together with these concepts to formulate a joint objective function for a mathematical optimisation formulation. An objective criterion is presented for the selection of importance weights for each of the data used. Mathematical and numerical analyses have been used to determine several important properties of the estimation process and the way in which the estimates vary according to the choice of the form of the objective function. The results of calculations for an example network are given which illustrate many of the properties of the present approach. -from Authors