Road test of a freeway model

A few years ago Newell reworked classical traffic wave theory into "A simplified theory of kinematic waves in highway traffic" (Newell, 1993a and Newell, 1993b). The simplifications - use of cumulative count curves instead of flows for most of the calculations and a triangular flow-density relation to describe traffic flows - were sufficient to allow him to apply the theory (Newell, 1993c) to complex situations with multiple bottlenecks and multi-destination flows. This paper tests Newell's model by comparing its predictions with conditions observed at three freeway test sites. The test data from San Francisco Bay Area freeways is old, but extraordinarily detailed, so provides the necessary input and observed densities for comparison. In the tests, the model did a very good job of predicting the growth and decay pattern of large queues and the effect of traffic entering and leaving the roadway within a congested area, but had difficulties dealing with light or sporadic congestion. However, the predicted travel times were quite accurate even for lightly congested roadways. The estimation of roadway capacities needed as input proved to be a major problem. The duration of queuing - both in the model and the real world - is very sensitive to the maximum rate at which vehicles can enter bottlenecks, and neither standard estimation tools nor the data set provided estimates of sufficient precision. This would seem to be a problem for any freeway model, but for many purposes there is no need for the level of accuracy sought here, and for others better data would be available. Overall, the results were very encouraging: the model requires very little calculation time and delivers excellent results for the severely congested freeways that are of the most practical interest.