A study was conducted by airborne instruments laboratory to evaluate deterministic models of traffic flow. The models considered represented solutions of the generalized car- following equation written as expressions relating average speed and average density. One of these solutions was found to be a best fit to two distinct sets of data--1224 velocity -density per-lane values averaged from all three lanes of the john lodge freeway. This best fit was determined by use of a nonlinear regression process applied directly to the data. The technique for selecting this best model included use of the sum of residuals as well as a statistical test of hypothesis based on rank-ordering. Values of the standard parameters free speed and maximum flow compared closely with previous estimates for the same data, but the exponents in the car-following equation were higher than in any previous model. Continuous multiregime models were constructed by combining models fitted to the low and high vehicle-density regions. A linearly weighted transition region, between 45 and 75 vpm, was defined by examination of a high-order polynomial curve fit. These multiregime models were better fits than the single-regime cases, but were analytically more complicated. A multilane experiment indicated that the best single-regime model class fitted the right-lane data of a two-lane road as well as it fitted the two-lane data itself. Additional lane-by-lane data, covering the full sensitivity range, is indicated as a necessary requirement to further support this conclusion.
Abstract