Nokia and China Mobile today said a trial of artificial intelligence (AI) on the operator’s live 5G radio access network (RAN) successfully forecast bandwidth and detected traffic anomalies with accuracy in excess of 90%.

The Finnish vendor said its RAN intelligent controller (RIC) served as a gateway for a series of China Mobile technologies designed to create a more efficient network. During the trial, Nokia’s RIC paired with its AirFrame Open Edge server platform to use AI to forecast traffic bandwidth in Shanghai and detect network anomalies in China Mobile’s network in Taiyuan.

The increased use of AI and machine learning (ML) in network infrastructure is widely expected to lower costs by automating functions that typically require human interaction. “Many communication service providers are already on a journey to become augmented service providers where AI augments human decision making for prediction, analysis, and new revenues,” Don Alusha, senior analyst at ABI Research, wrote this week in a report.

Operators are also investing in AI and ML to capture growth with the rapid development of new services, but internal overhauls of operations are equally important to achieve those goals, he said. 

“Communication service providers must realize that the new world in cellular must start with a foundation on software and API-led connectivity,” Alusha explained. “This may well mean that, in addition to bolting on software and intelligent capabilities, communication service providers need to learn how to build them as cloud edges, open radio access networks (RAN), and 5G core proliferates in the ecosystem.”

Nokia claims operators can realize cost savings of at least 70% by removing humans from the process of forecasting network performance requirements and detecting issues before they occur. The vendor said its anomaly detection software provides a distributed and split architecture, which makes it suitable for non-real-time and near-real-time RIC deployments.

AI-based user equipment accurately predicted traffic requirements in a live test on China Mobile’s 5G network in Shanghai with accuracy above 90%, according to Nokia. The vendor’s 5G AirScale base station, in real-time coordination with the RIC, estimated user equipment radio quality and related throughput for 100 milliseconds.

Nokia said it also worked with China Mobile subsidiary Shanxi Mobile and the China Mobile Research Institute (CMRI) to trial network anomaly detection technology on the operator’s 4G LTE and 5G network across more than 10,000 cells. 

“RIC plays a key role in enabling AI/ML capability in the RAN, which is of great significance to realize the concept of the i-wireless-intelligent and simplicity 5G network,” Huang Yuhong, deputy director at CMRI, said in a statement referencing the branding for China Mobile’s AI-based network technologies.