Article: A multi-agent reinforcement learning framework for educational resource management optimisation Journal: International Journal of Reasoning-based Intelligent Systems (IJRIS) 2026 Vol.18 No.7 pp.32 - 43 Abstract: This paper addresses the challenges of dynamic adaptation and multi-objective optimisation in educational resource management by proposing a novel multi-agent reinforcement learning framework. The framework utilises centralised teacher agents for global coordination and decentralised student agents for personalised recommendations. It incorporates an integrated optimisation mechanism to balance learning effectiveness, fairness, and efficiency, enhanced by a curriculum learning strategy that progressively trains agents from basic to complex tasks. Comprehensive experiments on the Junyi academy and assessments public datasets demonstrate that this approach improves knowledge mastery by 18% and enhances resource allocation fairness by 22% compared to traditional baseline methods, effectively narrowing the achievement gap among learners with different initial capabilities. The study provides an innovative and effective solution for next-generation intelligent education systems. The framework is designed to be applicable across various educational stages, including K-12 and higher education, to ensure broad relevance. Inderscience Publishers - linking academia, business and industry through research

Title: A multi-agent reinforcement learning framework for educational resource management optimisation

Authors: Yuanyuan Feng

Addresses: School of Marxism, Xinxiang Vocational and Technical College, Xinxiang, 453000, China

Abstract: This paper addresses the challenges of dynamic adaptation and multi-objective optimisation in educational resource management by proposing a novel multi-agent reinforcement learning framework. The framework utilises centralised teacher agents for global coordination and decentralised student agents for personalised recommendations. It incorporates an integrated optimisation mechanism to balance learning effectiveness, fairness, and efficiency, enhanced by a curriculum learning strategy that progressively trains agents from basic to complex tasks. Comprehensive experiments on the Junyi academy and assessments public datasets demonstrate that this approach improves knowledge mastery by 18% and enhances resource allocation fairness by 22% compared to traditional baseline methods, effectively narrowing the achievement gap among learners with different initial capabilities. The study provides an innovative and effective solution for next-generation intelligent education systems. The framework is designed to be applicable across various educational stages, including K-12 and higher education, to ensure broad relevance.

Keywords: multi-agent reinforcement learning; MARL; educational resource management; personalised learning; multi-objective optimisation; fairness.

DOI: 10.1504/IJRIS.2026.151422

International Journal of Reasoning-based Intelligent Systems, 2026 Vol.18 No.7, pp.32 - 43

Received: 01 Nov 2025
Accepted: 24 Nov 2025
Published online: 28 Jan 2026 *

Title: A multi-agent reinforcement learning framework for educational resource management optimisation

Keep up-to-date