How Ericsson and KDDI are Pushing AI-RAN into Live Networks

The race towards autonomous telco networks has taken another step forward.

Ericsson and KDDI have completed a large-scale live network trial of an AI-driven RAN across thousands of 4G and 5G cells in Japan.

The trial saw Ericsson’s Uplink Interference Optimizer (UIO) rApp deployed on the Ericsson Intelligent Automation Platform (EIAP) across approximately 1,500 5G cells and 1,300 4G cells during Q1 of this fiscal year.

The companies say the deployment demonstrated measurable uplink performance improvements under live traffic conditions while also validating the operational readiness of AI-driven network automation.

Automation for network efficiency

For telco operators facing rising traffic demands from AI workloads, video applications and latency-sensitive services, uplink performance is becoming ever more important alongside traditional download speeds.

The results from the field trial point to how operators may use AI-based automation to improve network efficiency without major infrastructure overhauls.

The companies reported average throughput gains of 9.6% in 4G and 3.1% in 5G, alongside a 27% improvement in 5G Signal-to-Interference-plus-Noise Ratio (SINR).

Ericsson also said the application improved uplink modulation and spectral efficiency across both radio technologies.

Jean-Christophe Laneri, Head of Cognitive Network Solutions at Ericsson, says: “The successful field trial with KDDI confirmed the readiness of the UIO rApp, running on the EIAP, for a key role in achieving Autonomous Networks Level 4 (AN L4) live operation.

“This solution introduces advanced, robust automated optimisation within mobile networks, greatly improving subscriber experience.

The UIO rApp is now generally available for the global market, with additional AI-driven rApps to follow as part of Ericsson’s expanding rApp portfolio.”

AI-RAN moves from theory to deployment

The trial marks several milestones for Ericsson and KDDI.

According to the companies, it was the first global deployment of the AI-driven UIO rApp in a live commercial network environment and the first time EIAP had been used live in Japan.

It also represents a notable development in the wider push towards Autonomous Networks Level 4 capabilities, where networks can increasingly optimise and manage themselves with limited human intervention.

Ericsson said the UIO rApp achieved a Level 3.86 autonomy rating for uplink optimisation based on an adapted TM Forum ANLET methodology.

Rather than attempting to optimise every cell equally, the system selectively targeted cells carrying the majority of network traffic.

That method allowed the companies to improve uplink performance while maintaining network stability across the wider RAN environment.

The focus on uplink optimisation reflects changing traffic patterns across mobile networks. Applications tied to AI inferencing and edge computing are creating growing demand for low-latency upstream connectivity.

Hiroaki Ando, President and Representative Director at Ericsson Japan, says: “Highly reliable, low-latency uplink performance is a critical foundation for Physical AI use cases, which Japan has identified as a priority area in its semiconductor and digital industry strategy.

“Ericsson remains committed to continuing our close collaboration with KDDI and to contributing to the advancement of Japanese society and industry.”

Open RAN ecosystem gains momentum

Alongside the performance gains, the trial also tested interoperability between Ericsson and third-party software applications within the EIAP environment.

KDDI deployed the Japanese-developed FYRA Suite rApp alongside Ericsson’s own optimisation application, validating support for both Ericsson and non-Ericsson rApps through the R1 interface standard.

The companies said this demonstrated the platform’s compliance with open and interoperable O-RAN principles.

Ericsson said the trial highlighted how open rApp platforms and developer ecosystems could allow local software developers to build applications that scale internationally.

The Uplink Interference Optimizer rApp also reached General Availability status during the trial period, signalling readiness for broader commercial deployment as operators continue exploring AI-led automation across the RAN.