AI/ML for Radio Access Networks: Enabling Self-Evolving and Intelligent Connectivity
AI/ML for Radio Access Networks: Enabling Self-Evolving and Intelligent Connectivity
The transformation of radio access networks (RANs) in the age of artificial intelligence (AI)reflects a broad shift shaped by domain knowledge from both wireless communications and computer science. As mobile networks evolve toward the sixth generation of mobile systems(6G), AI is becoming a native, in-network capability, embedded directly into the architecture,managing its own model lifecycle, and operating with full awareness of computational load,processing constraints, and energy budgets. This chapter examines how growing servicediversity, rising traffic volumes, and increasing network complexity have exposed thelimitations of traditional design and optimization approaches, creating a need for predictive,adaptive, and self-evolving network behavior. First, the evolution of mobile networks isreviewed to explain why 6G must be designed with AI as an intrinsic component and howAI/machine learning (ML) transitions from an external aid to a fully native capability in next-generation systems. Building on this foundation, the chapter highlights three representativeresearch areas-prediction, optimization, and network management-by examining selectedstudies that are frequently emphasized in academia and industry. Next, the chapter discusseshow standardization bodies address AI/ML capabilities across specific use cases throughcommon architectures, unified data structures, lifecycle management frameworks, and well-defined procedures for training, inference, and deployment. This structured perspective linksearly-stage research to standardized, deployable solutions. Finally, the chapter summarizesthe scientific and operational challenges on the path to AI-centric 6G, including generalizationin non-stationary environments, privacy-preserving learning, multi-vendor interoperability,and ultra-low-latency execution. By integrating insights from research, architecture, andstandardization, the chapter contributes to the book’s broader vision of self-evolving,intelligence-driven connectivity.