Evaluation of traffic assignment models through simulation

Abstract

Assignment methodologies attempt to determine the traffic flow over each network arc based on its characteristics and the total flow over the entire area. There are several methodologies—some fast and others that are more complex and require more time to complete the calculation. In this study, we evaluated different assignment methodologies using a computer simulation and tested the results in a specific case study. The results showed that the “all-or-nothing” methods and the incremental assignment methods generally yield results with an unacceptable level of error unless the traffic is divided into four or more equal parts. The method of successive averages (MSA) was valid starting from a relatively low number of iterations, while the user equilibrium methodologies (approximated using the Frank and Wolfe algorithm) were valid starting from an even lower number of iterations. These results may be useful to researchers in the field of computer simulation and planners who apply these methodologies in similar cases.