1. Introduction

The knowledge of the traffic flow from each origin to each destination is a very important element in much transportation engineering problems. These flows are usually organized in the form of a matrix which is called Origin/Destination matrix (O/D matrix): in particular each element of this matrix represents the traffic flow from a specific origin to a specific destination referred to a time unit. Different approaches have been proposed in the scientific literature in order to solve the nontrivial problem regarding the estimation and prediction of the O/D matrices. The research on this topic may be divided into several categories according, for instance, to the time influence (static or dynamic estimation), to the kind and dimension of the network (small and close networks or general networks), and so on. If the O/D matrix refers only to a specific time period (static estimation) the estimation could be performed from survey data or basing on traffic counts from that period (for a description of the main approaches to the static estimation of the O/D matrix from traffic counts, see [1]). The dynamic estimation of the O/D matrix leads to time-dependent matrices which are very important within dynamic traffic management systems (these systems include, for example, driver information systems through Variable Message Signs (VMS) and real time traffic control systems). According to [2], the research results on this last topic may be divided into two main categories: the estimation procedures applied on closed networks or on general networks. In the first case, all in-flows and out-flows are known (from traffic counts), while, for general networks, the exact knowledge of all entry and exit flows is practically impossible. Many contribution have been presented in the scientific literature on dynamic O/D matrix estimation both for closed networks [2], [3], [4] and for more general ones [5], [6], [7], [8]. The typical difficulties related to the knowledge of time-dependent O/D matrices are: the O/D flows should be estimated starting from traffic counts; the estimation of O/D flows generally requires a dynamic traffic assignment model (at present, a well established dynamic traffic assignment model is still missing and the proposed ones are too complex [2]).