The evolution towards next-generation of mobile networks demands for autonomous network management, emphasizing data-driven solutions based on Artificial Intelligence (AI), in particular machine learning. To attain such goals, this paper proposes a hybrid end-to-end learning approach that integrates imitation learning, deep reinforcement learning, simulation, domain adaptation, multi-agent cooperation, explainable AI, and generative AI. The work outlines a comprehensive vision for online agent learning about optimum network management policies while ensuring safety, interpretability, and adaptability in highly complex and dynamic use cases at the network periphery.

J. Moura was funded by FCT/MECI through national funds and when applicable co-funded EU funds under UID/50008: Instituto de Telecomunicações. The work of P. Santana was partially supported by FCT under ISTAR projects UIDB/04466/2020 and UIDP/04466/2020.

Programmable Networks,Reinforcement & Imitation Learning,End-to-End Learning,Network Automation,Transfer Learning,Safe RL,Generative AI,Explainable AI

• Ciências da Computação e da Informação - Ciências Naturais
• Engenharia Eletrotécnica, Eletrónica e Informática - Engenharia e Tecnologia