UDC 004.021
ANALYSIS OF DECISION-MAKING METHODS IN DISTRIBUTED IOT SYSTEMS BASED ON MULTI-AGENT APPROACHES
M. M. Blagirev, post-graduate student, assistant of the department of instrumental and applied software of the Institute of Information Technologies MIREA –Russian Technological University., Moscow, Russia; orcid.org/0009-0008-2853-3411, e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
A. O. Kostyrenkov, post-graduate student, assistant of the department of instrumental and applied software of the Institute of Information Technologies MIREA –Russian Technological University., Moscow, Russia; orcid.org/0009-0007-0294-694X, e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
The article analyzes the existing methods of decision making in distributed IoT-systems based on multi agent approaches. Particular attention is paid to the architectures of multi-agent systems, data processing and prediction algorithms, deep learning models used in the systems under consideration. The aim of the paper is to identify advantages, disadvantages and areas of applicability of each approach, which will allow forming a holistic view of current state and development prospects of decision-making methods in distributed IoT. The paper discusses the architectures of SPDPs that incorporate modern methods of learning (MARL, VDN) and high-level control using large language models (LLM). Neural network models (GRU, LSTM) to work with temporal data and ensemble methods (Random Forest, XGBoost, CatBoost) for the tasks of deter mining accurate forecasting are used. The analysis of neural network models performance in tabular com parison form is given to reveal quantitative and qualitative indicators in comparison with model prediction with real values, as well as comparison of models by mean absolute error and coefficient of determination.
Key words: : IoT, multi-agent systems, decision-making, neural network algorithms, ensemble methods, LLM, MARL, forecasting, distributed systems.