This paper will present a four-step framework for computing static origin-destination matrices for linear arterials with intersections. The method allows the user to control the resulting dynamic behavior by adjusting several tunable parameters which relate directly to the turning flows. These parameters include intersection weighting coefficients, which dictate the likelihood that a vehicle traveling along the arterial will turn at each intersection. Six additional parameters are required for each four-branch intersection, giving a total of seven. Three-branch intersections are defined solely by their intersection weight. All of the parameters have physical interpretations and can be computed from measurable quantities. They can be assigned constant or time-varying values. The main advantage of the procedure is that the parameters involved have a direct and explicit influence on the turning flows, and therefore on important dynamic quantities such as queue lengths and travel times. This is in contrast with approaches that employ tunable quantities such as the target matrix, whose relation to the resulting OD matrix is more complex. A test on a 1-mile stretch with 11 intersections is provided. In the example, parameters are tuned with a manual procedure involving a microscopic simulator. The microscopic model allows the user to tune the estimator based on dynamic quantities, but is not a necessary component of the OD estimation procedure. Results confirm that the method works well for the given linear network, and that is sufficiently flexible to accommodate different forms of supplemental traffic data, such as observed side street flows and queue lengths.
Abstract