A pure distributed framework for large-scale microscopic traffic simulation

Abstract

Microscopic traffic simulation is effective for analyzing transportation networks, and distributed simulation modeling is the optimal approach for the simulation. In this paper, a pure distributed modeling framework (called DMTSS) for large-scale traffic simulation is proposed. DMTSS consists of low-cost PCs networked by high-speed local area network, the heavy computational task is divided into multiple light subtasks according to transportation subnets, and subtasks are distributed to these PCs. A distributed database is used to hold necessary data, and a distributed synchronization mechanism is employed to improve the efficiency of synchronization. To ensure the system stably running for long time, the failure recovery scheme is presented to recover the system from error state. Implemented details such as communication and are also discussed. Analyses indicate that the proposed framework is scalable and reliable and effective