Derivation of Origin–Destination Distributions from Traffic Counts

Abstract

Origin–destination (O-D) traffic flows are an essential element in several freeway and network traffic models. In general, it is impossible to identify a unique O-D matrix from volume counts alone. Several methods for deriving O-D data exist, and they vary in complexity, cost, and analysis time. O-D matrices were generated with five methods that were applied to the same set of volume data. The resulting O-D matrices were used as the only variable input component in traffic simulation with INTEGRATION. Traffic simulation output was used to assess the implications of the O-D distribution. The O-D data produced by the deterministic method were clearly different from those of the other four methods. The proportionate and QueensOD (with default seed) output were somewhat dissimilar. All other combinations were similar, and the proportionate and FREQ O-D output were statistically identical. The simulation output with the proportionate-method O-D data met the 10/75 speed criterion at all five of the cross-section locations [it was within 10 mph (16 km/h) of field speed for 75% of the time periods]. Pearson's correlation of speeds at five locations indicated that all O-D methods except the deterministic one have a strong correlation with each other. A similar outcome was obtained with t-tests of the total travel time output.