Game Model of Ontological Project Support

Abstract

Context. In today’s information society with advanced telecommunications through mobile devices and computer networks, it is important to form a variety of virtual organizations and communities. Such virtual associations of people by professional or other interests are designed to quickly solve various tasks: to perform project tasks, create startups to attract investors, network marketing, distance learning, solving complex problems in science, economics and public administration , construction of various Internet services, discussion of political and social processes, etc.
 Objective of the study is to develop an adaptive Markov recurrent method based on the stochastic approximation of the modified condition of complementary non-rigidity, valid at Nash equilibrium points for solving the problem of game coverage of projects.
 Method. In this work the multiagent game model for formation of virtual teams of executors of projects on the basis of libraries of subject ontologies is developed. The competencies and abilities of agents required to carry out projects are specified by sets of ontologies. Intelligent agents randomly, simultaneously and independently choose one of the projects at discrete times. Agents who have chosen the same project determine the current composition of the team of its executors. For agents’ teams, a current penalty is calculated for insufficient coverage of competencies by the combined capabilities of agents. This penalty is used to adaptively recalculate mixed player strategies. The probabilities of selecting those teams whose current composition has led to a reduction in the fine for non-coverage of ontologies are increasing. During the repetitive stochastic game, agents will form vectors of mixed strategies that will minimize average penalties for non-coverage of projects.
 Results. For solve the problem of game coverage of projects, an adaptive Markov recurrent method based on the stochastic approximation of the modified condition of complementary non-rigidity, valid at Nash equilibrium points, was developed.
 Conclusions. Computer simulation confirmed the possibility of using the stochastic game model to form teams of project executors with the necessary ontological support in conditions of uncertainty. The convergence of the game method is ensured by compliance with the fundamental conditions and limitations of stochastic optimization. The reliability of experimental studies is confirmed by the repeatability of the results obtained for different sequences of random variables.

Game problem of assigning staff to project implementation

“…Other recurrent methods can be obtained from the complementary slackness condition (Kravets et al, 2020(Kravets et al, , 2021Neogy et al, 2018Neogy et al, , 2020, which is performed for Nash equilibrium points in completely mixed strategies:…”

“…. The theoretical order of the root-mean-square convergence rate of method ( 13) is equal ) , 1 min( Kravets et al, 2020Kravets et al, , 2021.…”

“…The game approach to covering the necessary ontologies of projects with existing ontologies of agents is considered in the work of (Burov et al, 2019;Kravets et al, 2019Kravets et al, , 2021.…”

“…1) Dynamic programming (De Lima & Carmo, 2018); 2) Greedy algorithm (Gupta & Singh, 2020); 3) Genetic algorithm (Dey et al, 2019); 4) Artificial neural networks (ANNs) (Philipsen & Stok, 1991); 5) Swarm intelligence algorithms (SIAs) or the emergent collective behaviour of groups of animals based on Particle Swarm Optimization (PSO), for example, Artificial Bee Colony, Bat Swarm Optimization, Salp swarm algorithm (SSA) or Salp Swarm Optimization (SSO) algorithm (Meraihi et al, 2019); 6) Ant colony optimization (ACO) algorithms (Dowsland & Thompson, 2008); 7) Multiagent Graph Colouring method (Blum & Rosenschein, 2008); 8) Game theory (Kravets et al, 2019(Kravets et al, , 2021Panagopoulou & Spirakis, 2008). As researchers (Frieze & Karoński, 2015;Raigorodskii, 2016;Raigorodskii, 2017;Raigorodskii & Karas 2022;Zhukovskii & Raigorodskii, 2015) emphasize, the colouring problem is much more complicated for time random graphs…”