Fujitsu, the Japanese multinational information technology equipment and services company, has announced the development of an application that leverages AI technology to enhance mobile network communication quality while achieving energy savings and optimizing network operations. This development is part of the Research and Development Project of the Enhanced Infrastructures for Post-5G Information and Communication Systems (hereafter NEDO-led project) conducted by the New Energy and Industrial Technology Development Organization (NEDO).   

This application comprises three key technologies: 

• An AI-powered technology that can estimate smartphone users’ quality of experience (QoE) on their networks in real-time. 
• A technology that detects early signs of increased communication traffic and proactively prevents network quality degradation. 
• A technology that detects network quality degradation and redefines the area covered by peripheral base stations to maintain optimal quality. 

According to the official press release, Fujitsu verified the effectiveness of these technologies in August 2024 using real commercial data from mobile network operators under conditions closely resembling actual operating environments. 

This application is claimed to ensure a seamless and reliable connection for mobile network users, not only during normal operations but also during emergencies and periods of increased network traffic. It will ultimately enhance user convenience, satisfaction, and safety in critical situations. The company opined that for mobile network operators, the application will reduce operational costs and save power through optimized operations. 

Background 

As digital transformation (DX) accelerates globally, adopting 5G mobile networks is rapidly expanding as a critical infrastructure component. Mobile network operators are expected to further enhance their capabilities, including ultra-low latency and simultaneous connectivity while ensuring network quality at user and application levels. Furthermore, the RAN (Radio Access Network) domain is undergoing open and virtualized transformation based on the O-RAN concept, leading to anticipated reductions in total cost of ownership (TCO). 

The statement said the three technologies in this application operate on the RIC deployed within the O-RAN-compliant SMO, contributing to RAN’s intelligent automation and self-sufficiency.  

Development achievements 

Real-time QoE estimation and quality assurance using AI 

According to the company, this technology estimates QoE in real time and automatically switches users to other base station network areas when QoE degradation is detected. It enables the creation of AI models that can easily estimate QoE for individual applications by selecting feature values from statistical data (KPIs) calculated from high-speed packet analysis for 100 Gbps RAN traffic. This approach allows for flexible adaptation to various applications. 

“By accurately understanding the QoE of each user and assigning the necessary resources, this technology ensures user convenience and satisfaction while simultaneously suppressing excessive resource allocation. This results in a 19% increase in the number of users that can be accommodated per base station”, the statement read. 

Proactive base station activation/deactivation for quality maintenance and energy savings 

Fujitsu has reportedly developed a technology that utilizes AI to anticipate increased communication traffic and proactively activate previously dormant base stations to prevent degradation in user communication quality. 

Previously, energy savings were achieved by monitoring traffic in each area in real time and putting unnecessary base stations into sleep mode. “This technology goes a step further by detecting unusual increases in pedestrian traffic, such as those associated with local events, and predicting subsequent traffic increases at the grid level. This predictive technology has been proven to successfully activate base stations in advance without impacting user quality 99.8% of the time during the verification period,” the company said.  

This enables pinpoint activation and deactivation of base stations based on traffic conditions, achieving a balance between maintaining QoE and saving energy when combined with the energy-saving application announced by Fujitsu in December 2023. 

Detection of service quality degradation and area redesign for service quality maintenance 

Traditional anomaly detection technologies for single cells struggled to differentiate between simple load reductions and actual anomalies. Fujitsu claims that its new technology addresses this by comparing traffic trends across surrounding cells using AI, achieving a fault detection accuracy rate of over 92%. According to the company, this technology supports both supervised learning with limited fault data and unsupervised learning. By understanding the service impact, including cell overlap, it is possible to judge which areas should be recovered first. 

“When this anomaly detection technology identifies areas undergoing significant service impact, it utilizes a radio propagation prediction model that considers path loss in the real field and the direction and load conditions of surrounding cells to calculate the optimal tilt angle for surrounding cells. This minimizes the impact of faulty cells on service quality. As a result, recovery time from anomalies such as equipment failure, which previously took 24 hours, has been reduced to less than one hour, minimizing the impact on users”, the company added.  

Fujitsu