Toward AI-Driven Control of Open, Programmable, and Virtualized Wireless Systems

The Open Radio Access Network (RAN) paradigm is poised to transform the telecom ecosystem by promoting the evolution toward virtualized and disaggregated RANs controlled over multiple time scales through intelligent controllers to provide bespoke and adaptive services. Understanding O-RAN, its architecture, interfaces, and tools become of paramount importance for researchers in wireless networking.

This lecture will introduce the O-RAN architecture and discuss research challenges and early results on developing end-to-end AI-based control for Open, Programmable, and Virtualized Wireless Systems. It will also cover O-RAN AI/ML workflows, including a detailed overview of the procedures that regulate the AI/ML workflow in the O-RAN architecture, from data collection to the actual deployment and execution of network intelligence in real-time. It will also present and discuss OpenRAN Gym, a toolbox to generate datasets and train AI algorithms in the Open RAN. Last, it will discuss extensive applications of AI to control problems in ORAN for network slicing, spectrum sharing, end-to-end management, and orchestration, among others.