Computational Issues in Increasing Spatial Precision of Traffic Assignments

The practical feasibility is demonstrated of a traffic assignment method, dubbed area-spread assignment, that incorporates information on the spatial relationships between nodes, links, and zone boundaries. Tests of area-spread assignment were performed on networks from Fredericton, New Brunswick, and Racine, Wisconsin. To eliminate the effect of errors from other model steps in the tests, comparisons were made with a detailed network in Fredericton that contained all streets in the city and extremely small zones. The tests indicated that there is a substantial difference between area-spread assignment and a traditional assignment using centroid connectors. Area-spread assignment did a better job of replicating volumes on the detailed network, although computation times were much longer.