A distributed, scalable, and synchronized framework for large-scale microscopic traffic simulation

Abstract

Microscopic traffic simulation is effective for analyzing transportation networks. However, performing microscopic simulations for large-scale networks remains a huge computing problem. In this paper, we present ParamGrid, a scalable and synchronized framework that distributes the simulation across a cluster of ordinary-performance, low-cost personal computers (PCs) connected by local area network (LAN). Several key challenges are solved: dividing a large traffic network, routing vehicles traveling across tiles, transferring vehicles across tiles, and globally synchronizing the simulation time to achieve a complete solution to automate scalable concurrent simulations. These solutions are automated in ParamGrid. Our experimental results, using a real-world traffic network for the Long Beach Area in Southern California, show our new framework provides near linear speed-up with number of computers.