SAMPLING STRATEGIES FOR 2-FLUID MODEL PARAMETER-ESTIMATION IN URBAN NETWORKS

The basic postulate of the two-fluid theory of town traffic relates th e average speed of moving vehicles to the fraction of moving vehicles in a street network; both representing averages over all vehicles in t he network. Data collection to provide estimates of the model paramete rs to-date has consisted of sampling the network with a test car, repl icating (as closely as possible) the trip histories of randomly select ed vehicles in the network. Because the two-fluid model is nonlinear, it cannot be simultaneously applied at the individual vehicle level an d the network level, as is shown in this article. However, due to the practical difficulties in collecting data for all the vehicles in the network, several potential sampling strategies are identified and are evaluated with computer simulation. The simulation experiments suggest that aggregating the trip histories of 10 to 20 test vehicles over 10 to 15 minutes yields parameter estimates very close to the true value .