Download PDFOpen PDF in browserMulti-Agent Simulation of Epidemics' Distribution on Modern SupercomputersEasyChair Preprint 484611 pages•Date: January 2, 2021AbstractThe possibility of using modern supercomputers in solving the resource-intensive tasks of multi-agent modeling of the spread of mass epidemics based on the theory of growth of percolation clusters is considered in this article. The determination of quarantine zones during the spread of epidemics is based on the multi-agent percolation model. It includes the formation of a lattice of interaction between representatives of the population, modeling the spread of the disease, collecting information about the population, implementing a parallel algorithm for multiple labeling of percolation clusters with a linking technique for labels, visualizing the results. The article describes a variant of the algorithm for multiple labeling of Hoshen-Kopelman percolation clusters, improved for use on a multiprocessor system, and the current prototype of its implementation developed at the JSCC RAS - Branch of SRISA. The input to this algorithm is data in a format independent of the application. Therefore, it can be used in any field as a tool for differentiating large lattice clusters. The article provides estimates of the execution time of the multiple labeling algorithm for Hochen-Kopelman percolation clusters for various input parameters on the four main high-performance computing systems installed in the JSCC RAS - Branch of SRISA. Keyphrases: High performance computing systems, multi-agent simulation, percolation’s cluster epidemics' distribution, theory of percolation
|