Estimating the O-D matrix from observed link volumes

Abstract Until now the traffic volume data have been used almost exclusively for the study of circulation control and road construction. We present a methodology to use these data for transportation planning where traditionally the data needed came from an O-D survey. We demonstrate that the traffic volume data are informative enough to be useful for this purpose. Most specifically we want to study the estimate of the O-D matrix obtained using observed link volumes for the class of proportional assignment methods. These methods are essentially the ones which do not take into account explicitly or implicitly any constraints of capacities. We show that for this class of method the estimation of the O-D matrix is equivalent to a linear regression problem. We discuss the problem of estimability of the parameters and relate this question with the sample of traffic volume needed. The discussion is illustrated by an example.