Advances in edge-intelligent controllers for open radio access networks

Open radio access networks (O-RANs) expand 5 G interfaces and protocols to facilitate software-hardware separation, enhancing flexibility and interoperability. This paper focuses on the intelligent controller, a key component that manages networking models and radio resources. Unlike traditional RAN controllers that rely on static rules, O-RAN intelligent controllers enable data-driven, real-time decision-making at the network edge, making them essential for low-latency optimization and third-party innovation. We first introduce O-RAN architecture and key technical challenges. We then systematically survey intelligent controller design from three core aspects: (a) model training, (b) model inference, and (c) model deployment. Finally, we analyze future research trends and open challenges. This review provides insights to guide standardization and accelerate AI-native O-RAN deployments.