R&D project to build AI-native radio networks

A new €1.4 million research and development project, dubbed 6G-ROCKET, is seeking to accelerate AI and machine-learning solutions for future 6G radio access networks.

The University of Oulu’s 6G Flagship leads the project team, which also includes Nokia, NVIDIA, DNA, Keysight Technologies, MediaTek, and international partners Northeastern University and The University of Tokyo. Building on a GPU-accelerated AI-RAN (artificial intelligence-radio access network) testbed connected to the University of Oulu’s 5G/6G Test Network, the project aims to advance global progress in AI-RAN by combining 6G Flagship’s wireless research with its partners’ industry experience.

“6G-ROCKET is about building the foundations for AI-native mobile networks where communication, sensing and intelligence work together seamlessly,” said Dileepa Marasinghe, AI-RAN Research Manager at 6G Flagship. “Our goal is to move beyond simulations and demonstrate practical AI-RAN technologies in GPU-accelerated testbeds.”

“By integrating intelligence directly into the radio access network with AI-RAN, we are turning the base station from a signal processor into a high-performance computing node that can also interpret changes in the physical environment,” added Soma Velayutham, Vice President of Telecoms at NVIDIA. “It allows the network to see, reason and react in milliseconds, which is essential for robotics.

“The 6G-ROCKET project … is advancing 6G research for physical intelligence, helping developers build, test and optimise their next breakthrough on NVIDIA AI-RAN platforms.”

From reactive to proactive

Most mobile networks today are reactive, relying on manual adjustments and separate AI tools. 6G-ROCKET aims to change this by creating AI-native RAN technologies that proactively anticipate needs. The project covers AI-native RAN designs, real-time digital twins for network optimisation and control, GPU-accelerated testing platforms, multi-sensor perception for better awareness, and autonomous network management based on intent.

The project puts AI directly into the radio access network, making intelligence a native part of the RAN architecture rather than an add-on capability. This approach builds proactive capabilities into the network from the start — which makes the network flexible and scalable — rather than simply adding a perception system to existing networks.

Partners across the full AI-RAN stack

Each partner in the consortium brings their complementary expertise, beginning with 6G Flagship’s wireless research and the University of Oulu’s 5G/6G test network infrastructure. Nokia, in turn, offers expertise in mobile network systems and AI-native RAN evolution. NVIDIA provides AI-RAN infrastructure and AI Aerial open-source software libraries and tools. Keysight contributes expertise in digital twin technologies, including simulation, testing and validation capabilities. MediaTek provides chipsets and user-equipment expertise. DNA offers operator insight and real-world network experience. Northeastern University and The University of Tokyo add international research collaboration to align the work with global AI-RAN and 6G efforts. The project also supports Finland’s broader 6G strategy by strengthening connections between universities and industry and by helping develop the standards, technologies and skills the country will need.

“As an operator, we want to help ensure that future network technologies are developed around real customer needs and can be tested in practical environments from an early stage,” said Sari Leppänen, CTO at DNA.

“For DNA, the [6G-ROCKET] project is an opportunity to influence how AI-native networks develop, while gaining practical insight into solutions that can improve network efficiency, reliability and future services.”

“Validating AI-driven networks requires precise, repeatable measurements under realistic conditions,” added Michael Dieudonne, 6G R&D Manager at Keysight Technologies. “Keysight is pleased to contribute its simulation, measurement and channel validation expertise to 6G-ROCKET, helping the consortium advance from concept to demonstrated performance.”

“Device-side intelligence is essential to closing the loop between network and terminal,” said Chih-Ming Hung Associate Vice President of Technology at MediaTek. “MediaTek is glad to bring its chipset and user-equipment expertise to 6G-ROCKET, helping ensure these AI-RAN innovations work end to end.”

What will the next two years deliver

Over the next two years, 6G-ROCKET will create technology blueprints and prototype demos to prepare for smart, sensing-enabled AI-RAN and future 6G networks. An AI-RAN Perception Lab will also be established at the University of Oulu, which will support ongoing collaboration between universities and industry and help Finland remain a leader in 6G and AI-native wireless research.

By bringing together advanced AI computing, wireless systems and real-world testing, 6G-ROCKET is moving towards the next generation of smart, adaptable 6G networks. The project aims to strengthen Finland’s 6G plans, build local expertise in AI-native RAN and wireless innovation, encourage new international partnerships, and grow the country’s workforce and skills in AI and wireless communications.

Image credit: iStock.com/Crovik Media