A Microsimulation Comparison for Lane-Merging Driver Behaviors

Abstract

While driving behaviors have been widely studied with various microscopic simulations, comparatively little attention is paid to how different microscopic simulation environments measure the same human driving behaviors. To this end, we compared aggregate human driving behaviors for two microscopic simulators, SUMO and PTV Vissim, in zipper and early merge scenarios. We used standard embedded driver and traffic models for both microscopic simulators. Results showed the two simulators were generally equivalent in computational resource demand but produced delay estimates that were not consistent with one another, or internally. Another finding is that SUMO produced more variability than PTV Vissim, which may be an advantage when representing human behavior uncertainty that is highly variable. Ultimately, the choice of simulation environment depends on the research question, monetary costs, and human costs of development.