An optimal downsampling procedure for microscopic simulation modeling of transportation networks: a proof-of-concept study

Abstract

This paper investigates and demonstrates the concept of scalability of microscopic traffic simulation systems as a means of reducing the associated computational requirements and maximizing their potential support for real-time traffic control and management functions. The primary goal of this research is to examine the feasibility of transforming the original simulation environment into a downsampled simulation environment, where fewer representative entities are simulated. This must be achieved, however, while retaining maximum fidelity to microscopic simulation properties and preserving most of the macroscopic characteristics. The methodology is presented as an optimization problem whose objective is to minimize the errors resulting from the transformation process and to seek optimal values of the behavioral parameters in the downsampled environment. In this proof-of-concept stage, experimental analysis was conducted on a homogeneous freeway segment using one of the well-known, and arguably sufficiently calibrated, car-following models developed by General Motors Laboratories (GM3). The results were promising and showed that optimal relationships between the behavioral parameters in both environments can be established to minimize the information loss associated with the transformation process.Key words: microscopic simulation, scalability, computational efficiency, downsampling, car-following models.