Aggregate calibration of microscopic traffic simulation models

Abstract

The problem of calibration of microscopic simulation models with aggregate data has received significant attention in recent years. But day-to-day variability in inputs such as travel demand has not been considered. In this thesis, a general formulation has been proposed for the problem in the presence of multiple days of data. The formulation considers the day-to-day variability in all the inputs to the simulation model. It has then been formulated using Generalized least squares (GLS) approach. The solution methodology for this problem has been proposed and the feasibility of this methodology has been shown with the help of two case studies. One of them is with an experimental network and the other is with network from Southampton, UK. The results indicate that estimation of day-to-day OD flows is feasible. They also reinforce the importance of having good apriori information on the OD flows and locating the sensors so as to obtain maximum information.