Fast Low Fidelity Microsimulation of Vehicle Traffic on Supercomputers

TL;DR

This paper presents a simple, high-speed microsimulation model for vehicle traffic that runs efficiently on supercomputers, enabling real-time analysis of large-scale traffic systems with detailed vehicle resolution.

Contribution

The paper introduces a low-fidelity, computationally efficient traffic simulation model optimized for supercomputers, capable of simulating over 4 million kilometers in real time.

Findings

Achieved real-time simulation of over 4 million km of traffic

Model provides realistic traffic behavior with simple rules

Suitable for high-resolution network routing applications

Abstract

A set of very simple rules for driving behavior used to simulate roadway traffic gives realistic results. Because of its simplicity, it is easy to implement the model on supercomputers (vectorizing and parallel), where we have achieved real time limits of more than 4~million~kilometers (or more than 53~million vehicle sec/sec). The model can be used for applications where both high simulation speed and individual vehicle resolution are needed. We use the model for extended statistical analysis to gain insight into traffic phenomena near capacity, and we discuss that this model is a good candidate for network routing applications. (Submitted to Transportation Research Board Meeting, Jan. 1994, Washington D.C.)