мультиагентні системи

The development of effective monitoring software agents, essential components of modern multi-agent systems (MAS), increasingly relies on sophisticated model structures such as multi-layer machine learning ensembles. However, the growing complexity of these architectures presents significant challenges in ensuring the reliability, auditability, and, most critically, the reproducibility of experimental results. Addressing this challenge, this paper proposes a declarative approach centered on a newly developed domain-specific language (DSL).

The game model of multi-agent systems of self-organizing in the conditions of uncertainty is developed. The formulation of a stochastic game problem is carried out, criteria of self-organizing of strategies of players are defined, a recurrent method, algorithm and software of learning of multi-agent system to simulate the synchronised rhythmic luminescence of a colony of fireflies are developed.