Efficient OD Estimation

Origin and Destination (OD) estimation is the process of estimating where trips begin and end in a traffic network. The estimation is based on limited household surveys supplemented by indirect measures such as traffic counts.

The origin-destination matrix is the important input in many transportation analysis problems. It describes the demand for the transportation network in terms of vehicle flows from all possible origins to all possible destinations (OD pairs) in the network. This kind of information is used for strategic planning and analysis (like new road building, environmental analysis, etc.) as well as for operational planning purposes (traffic management systems). The direct information on the OD flows can be gathered though household sample surveys, though it is very expensive, and often not very efficient.

Other sources of information, so called indirect measures, are used to obtain more exact estimates on the real OD flows. Commonly, the information used to estimate OD flows is collected through sensors. They can record number of passing vehicles, their speeds or travel time between two sensor locations. Most of the research is based on only flow measurements for OD estimation. While for uncongested networks it can be enough for accurate OD estimation to use just traffic counts, in the congested case it is preferable to use other available information on the traffic conditions.

More and more often simulation is used as a part of the OD estimation process. Simulation software is used to map OD flows to the network. While it helps in creating some realistic assignment of flows to the network, it complicates the estimation process since simulation includes stochastic elements. Moreover, each simulation run can be computationally expensive, and use of algorithms which require many simulation runs can be unfeasible it terms of time.

The objective of the research is to develop methods, that are practical, computationally efficient so that they can be used for large networks and for a variety of applications (both off-line and real time), and general so that they can be utilized with a number of existing simulation and other tools. Furthermore, the developed methods will be flexible in the use of general traffic data (ranging from counts and spot speeds to travel times). The research will examine methods which have the potential to work even under congested and oversaturated conditions.

Period: 2010-01-01 to 2016-12-31

Keywords: Origin, destination, traffic, estimation, simulation, modelling Source of funding: Swedish Road Administration.