ARCHITECTURE AND METHOD FOR ADAPTIVE PERSONALIZATION IN A DISTRIBUTED E-LEARNING SYSTEM BASED ON AGENT TECHNOLOGIES

Abstract

The paper develops an architecture of a distributed personalized e-learning system based on agent technologies and proposes a method for adaptive personalization of the learning process that supports dynamic formation of individual learning trajectories with regard to the current state of the student. The relevance of the study is determined by the need to improve the effectiveness of e-learning in the context of distance and blended learning, the growth of digital educational content, and the demand for intelligent learner support in real time. Unlike conventional LMS platforms that mainly rely on test results, activity logs, and statistics of content access, the proposed approach takes into account the student’s behavioral and psychophysiological parameters as indicators of cognitive state, concentration level, fatigue, and emotional engagement. The system architecture includes a student state monitoring module, an agent-based decision-making environment, data synchronization tools, a unified database, and mechanisms for integration with learning management systems. Within the proposed method, the tutor agent applies the Q-learning algorithm to select adaptive pedagogical actions, while the student’s state parameters are integrated into the reward function to adjust learning content, presentation pace, and task complexity. Particular attention is paid to the coordination of interactions between system components, synchronization of data between the monitoring module and the agent’s cognitive model through a unified database, and privacy protection based on the Local Inference architecture. To validate the feasibility and effectiveness of the proposed solutions, simulation modeling was carried out in the NetLogo environment. The experimental results confirmed the effectiveness of the proposed architecture and method: the speed of personalization of learning trajectories increased by 38 %, academic performance improved by 28 %, and students’ cognitive discomfort decreased. The obtained results demonstrate the feasibility of using distributed agent-based systems for the development of intelligent e-learning platforms.