An origin-destination (OD) matrix describes the travel demand between all pairs of origins and destinations in a traffic network. OD-matrices are essential input for most types traffic models, both for long-term planning, where the actions are meant to be permanent, and for short-term planning, where the operational situation is managed. This paper is a literature overview of models and methods for obtaining reliable OD-matrices.
We present the Four-stage model which is commonly used for sequentially computing the number of trips starting and terminating at the centroid nodes, the distribution to travel demand between origins and destinations, the split onto different travel modes and the choice-based assignment of routes and links in the network. Further, we identify the information contained in traffic counts as an important source for adjusting the computed OD-matrix. The generic OD-matrix estimation problem from link flow observations is formulated and we discuss different solution techniques and how they relate to the assumptions used for the assignment of travel demand to routes and links.
Time-dependence is a fundamental difference between the model types. A time-independent (static) model describes a steady state, which represents and average situation. If the interaction between sequential time periods is to be analyzed, however, a dynamic model is required. The time dimension is a complicating factor, since not only the route choice, but also the flow propagation must be taken care of in the model. Especially those models where the OD-matrix is to be estimated in real-time, must be simplified as to the computational effort.