Parameter sampling strategies in traffic microsimulation

Abstract

The paper investigates the impact of different sampling strategies of car-following and lane-changing model parameters on traffic simulation results. The investigation considered seven possible sampling strategies including sampling parameters from independent normal distributions, which is customarily in commercial simulation software. Study results revealed that model performances in case of sampling from normal pdfs are extremely poor. In turn, results proved that parameter correlation, as well as the parameter distribution model, entail a big impact on model performances and should be properly take into account in the microsimulation practice.