A Comparative Study of Multi-Agent Reinforcement Learning Approaches for Resource Optimization in Wireless Networks

Integrated into wireless networks, Multi-Agent Reinforcement Learning (MARL) has shown considerable potential in addressing resource optimization problems including power control, spectrum allocation, and effective communication. When MARL is applied in dynamic wireless environments, including limited resource blocks, energy constraints, communication errors, and convergence delays, there are further challenges as well. By means of a comprehensive review of MARL approaches applied in wireless networks, this paper classifies current methods based on network characteristics, learning processes, and optimization objectives. We review important developments, alternative strategies, and flag up remaining problems like energy-efficient communication, scalability, and real-time convergence. One of the suggested future possibilities is the use of modern wireless technologies including Terahertz communication and Intelligent Reflecting Surfaces (IRS) to enhance MARL performance. This survey offers a basis for academics trying to design ideal MARL solutions for next-generation wireless networks.