DOI: 10.17587/prin.10.167-177

Development of Software Framework for Large-Scale Agent-Based Modeling of Complex Social Systems

V. L. Makarov¹, makarov@cemi.rssi.ru, A. R. Bakhtizin ¹, albert.bakhtizin@gmail.com, G. L. Beklaryan ¹, lbeklaryan@gmail.com, A. S. Akopov ^1,2, akopovas@umail.ru
¹CEMI RAS,
²National Research University Higher School of Economics, Moscow, Russian Federation

Corresponding author: Akopov Andranik S., Professor, Chief Researcher, CEMI RAS, Professor, National Research University Higher School of Economics, Moscow, 101000, Russian Federation, E-mail: akopovas@umail.ru

In this paper, a novel approach to designing software framework that is intended for large-scale simulation modelling of complex social systems is presented. Such systems require applying supercomputing and parallel computations technologies, as well as phenomenological descriptions of agent behavior and their interactions. An agent model of human crowd behavior in an airport has been developed, taking into account the influence of different factors, for example, the number of entries and exits, the number of check-in counters, the number passport control offices, the time of waiting for baggage and other characteristics. It allows defining the best values of important resource parameters of an airport that provide a dissipation of crowd clusters. A framework based on using parallel computations MPI (Message Passing Interface), programming language C++ and the boost library is designed. The feature of such approach is the comfortable procedure of developing, configuring and debugging models. At the same time, all main advantages of earlier known agent modelling systems (e.g. AnyLogic, NetLogo, Repast HPC, FLAME, etc.) are kept and provided with the possibility of using supercomputing technologies. Some simulation experiments are conducted and they demonstrate the possibility of dissipating human crowd clusters in the airport when improving some resource characteristics, e.g. increasing of check-in counters, adding new entry, etc. In the result, the efficiency of developed software framework for the implementation of large agent models of complex social systems is shown.

Keywords: agent-based modeling, social systems, parallel computations, crowd model, supercomputing technologies, С++, MPI, software framework for simulation modelling

pp. 167–177

For citation:
Makarov V. L., Bakhtizin A. R., Beklaryan G. L., Akopov A. S. Development of Software Framework for Large-Scale Agent-Based Modeling of Complex Social Systems, Programmnaya Ingeneria, 2019, vol. 10, no. 4, pp. 167—177.