Single and Multi-Agent Deep Reinforcement Learning for AI-Enabled Wireless Networks: A Tutorial

Deep Reinforcement Learning (DRL) has recently witnessed significant advances that have led to multiple successes in solving sequential decision-making problems various domains, particularly wireless communications. The next generation of networks is expected provide scalable, low-latency, ultra-reliable services empowered by the application data-driven Artificial Intelligence (AI). key enabling technologies future networks, such as intelligent meta-surfaces, aerial and AI at edge, involve more than one agent which motivates importance multi-agent learning techniques. Furthermore, cooperation central establishing self-organizing, self-sustaining, decentralized networks. In this context, tutorial focuses on role DRL with an emphasis deep Multi-Agent (MARL) for AI-enabled first part paper will present a clear overview mathematical frameworks single-agent RL MARL. main idea work motivate beyond model-free perspective was extensively adopted recent years. Thus, we selective description algorithms Model-Based (MBRL) cooperative MARL highlight their potential applications Finally, state-of-the-art fields Mobile Edge Computing (MEC), Unmanned Aerial Vehicles (UAV) cell-free massive MIMO, identify promising research directions. We expect stimulate endeavors build scalable systems based