Organization of Processes for Solving Functional Problems in Multi-Agent Systems
A. V. Gulay, В. М. Зайцев · Science & Technique · 2025
When developing technologies for managing production and technical systems, there is a tendency to use the principles of multi-agency, which reflect the development of the system concept of distributed processing of matter, energy and information. To perform work and solve target functional tasks in multi-agent systems, it is assumed that parallel functioning system objects are involved, which to a certain extent are endowed with subjective rights and play the role of executive agents. By introducing components of intelligent control of the components of the designed system, the processes of their individual and group functioning are implemented. With this in mind, the article analyzes the main schemes for organizing processes for solving functional tasks and suggests indicators of the efficiency of multi-agent technologies use. In particular, a variant of solving unrelated system tasks using the same type of executive agents is considered. The possibility of using the same type of agents to solve multivariate problems with subsequent consolidated results is shown. A variant of preliminary decomposition of tasks into separate logically completed phases and further obtaining an aggregated result is also given. An approach to solving functional tasks based on a parallel-sequential conveyor implementation of their individual stages and subsequent selection of an adequate result is presented. As indicators of the effectiveness of the application of multi-agent technologies to solving system problems, characteristics based on the assessment of time and probability of obtaining correct results are proposed and analyzed. Thus, the group of indicators characterizing the solution process includes the total time for completing tasks by a group of agents, as well as the compression coefficient in the form of the ratio of the average time of a single-agent sequential solution of a set of tasks to the standard time of their multi-agent execution. The process under consideration is also characterized by the calculated probability of providing the necessary set of solutions in a given standard time and the probability of obtaining a set of correct results with multiple parallel solutions of the problem using a group of agents.