GEMSim: A GPU-accelerated multi-modal mobility simulator for large-scale scenarios
A new publication on the Journal of Simulation Modelling Practice and Theory by Aleksandr Saprykin, Ndaona Chokani and Prof. Reza Abhari on the development of GEMSIM, an integrated solution to provide advanced traffic simulation and forecasting tools to planners and decision makers.
In order to accelerate large-scale agent-based mobility scenarios, and to provide advanced traffic simulation and forecasting tools for planners and decision makers, a GPU-accelerated (graphics processing unit) simulation platform, GEMSim, has been developed. GEMSim incorporates a mesoscopic, multi-modal, queue-based mobility simulator implemented on a GPU, including a detailed public transit model. Until now, no integrated solutions to run large-scale multi-modal mobility scenarios on a GPU have been presented. The simulator exploits activity-based demand modelling to consider human patterns of behaviour and to make scenarios more realistic in a sense of emerging transportation modes and mobility services. A large-scale scenario for Switzerland with detailed road infrastructure, public transit schedule and up to 5.2 million of agents with individual daily plans has been used to validate and to assess performance of GEMSim and to compare it with one of the most advanced existing agent-based simulators for transportation (MATSim). A factor of 68 speedup was achieved along with reduction in memory usage as a factor of 5, making it possible to simulate a full day of the whole Switzerland in less than 5 minutes even on a decent workstation.
You can find the article online on external page Elsevier