OVERVIEW OF INTELLIGENT AGENTS STRUCTURES AND AN INNOVATIVE APPROACH TO THE INTEGRATION OF RECURRENT NEURAL NETWORKS INTO BDI ARCHITECTURE

Abstract

Intelligent agents have become a central theme in modern artificial intelligence research. They are an essen-tial component of contemporary information technologies, used to automate complex tasks such as data processing, decision-making, and adaptive learning. The development of intelligent agents is a complex, multifaceted process that includes the selection of architecture, the development of decision-making algorithms, and integration with the environment. Over the past decades, many new technologies and approaches have emerged that significantly /improve the efficiency of agents in real-world applications. Conceptual foundations of intelligent agents, various approaches to their design, and key technologies that enable the creation of autonomous and adaptive systems are explored. The authors examine the general algorithm of the intelligent agent's functioning, its main stages, and architecture. An overview of the concept of intelligent agents is presented, including their basic properties, models of interaction with the environment, and adaptive mechanisms. The article provides an overview of the current state of intelligent agent design, highlighting significant achievements and key challenges in the field. The focus is on the BDI (Belief-Desire-Intention) architecture, which is one of the most prevalent and effective approaches to creating intelligent agents. The beliefs, desires, and intentions of agents are modeled to ensure more adaptive and flexible behavior in dynamic environments. The article presents a novel method for incorporating recurrent neural networks (RNNs) into the BDI architecture to improve agents' ability to comprehend sequential data and evaluate the context of past events. RNNs provide efficient memory and analysis of temporal connections, which is critical for decision-making in uncertain and changing environments. Integrating RNNs enables agents to update their beliefs in real time while also improving the processes of developing desires and intentions, making them more accurate and adap- tive. The main findings of the research demonstrate that the implementation of RNNs in the BDI architecture significantly enhances the efficiency and productivity of agents, allowing them to quickly adapt their strategies to new conditions. This opens up new opportunities for applying such agents in various fields, including cybersecurity, energy management, healthcare, and many others. Overall, this article makes a significant contribution to the development of the theory and practice of intelligent agent design, offering an innovative approach to integrating recurrent neural networks into the BDI architecture, which ensures greater adaptability and flexibility in dynamic environments.