One of the headline-grabbing announcements from Nokia’s launch of its AI-RAN portfolio was the company’s claim that its GPU-accelerated architecture would enable advanced computations that could, in certain circumstances, double spectral efficiency by 2028.
That’s a big contrast to the sorts of claims that Ericsson has been making for its intelligent RAN apps of late, which sit more in the 10-15% bracket, although of course Ericsson is talking about available technology and Nokia is talking about a two year time frame applicable to certain situations.
Nokia said that key to this uplift in efficiency will be the deployment of distributed applications (dApps) which will be enabled by the company’s adoption of support for a new O-RAN proposed interface called E3. A white paper from O-RAN Alliance’s next Generation Research Group (nGRG) said that E3 overcomes two key limitations in the current O-RAN xApp and rApp architecture.
“xApps and rApps are primarily designed to interact with data and operations associated with the RAN control-plane, thus not considering user-plane data, such as I/Q samples and packets for inference and optimisation (e.g., critical to sensing operations; and (does not enable control loops at timescales below the 10 ms enabled by xApps.
Being unable to process user-plane data from RAN nodes and perform inference and control below the 10 ms timescale limits the application of O-RAN technologies at the lower levels of the protocol stack, as well as the introduction of novel use cases and applications (e.g., RF fingerprinting, one- shot beam steering, anomaly and attack detection, spectrum sensing and incumbent detection, joint sensing and communications, to name a few) which will be at the center
of 6G.
So the E3 interface provides a mechanism for dApps to interact directly with O‑DU and O‑CU nodes, enabling low‑latency, closed‑loop control for functions such as dynamic spectrum allocation, positioning, and sensing‑assisted mobility or interference management.
This opens up a lot of opportunities for really bringing new innovation also into the base station, not just locking it
dApp ecosystem
Speaking to The Mobile Network, Nokia’s Head of AI‑RAN and Cloud RAN, Aji Ed, said that Nokia’s performance improvement claims are based on the AI-RAN platform’s ability to support compute-intensive AI-based algorithm features such as multi-user MIMO pairing, non-linear channel estimation, carrier aggregation using such estimation, and deep transmitters and receivers, with “many more algorithms in the pipeline”.
While not every site will gain from such algorithms and technology, fully loaded, highly congested sites with TDD Massive MIMO could benefit the most, Ed said.
In Ed’s words, the platform is “completely O-RAN compliant” in its support for fronthaul splits and the O1 and O2 interfaces, but the AI-RAN launch goes further still.
“What we are also launching now is distributed apps (dApps) which are very lower layer, connecting with the base station over an interface called E3. We are now in the phase of partnering or studying together with our partners. It has to be controlled, it has to be governed well, and we will only select the use cases or partners which really add value. But the good thing is this opens up a lot of opportunities for really bringing new innovation also into base station, not just locking it.”
For Ed, some areas that ”make sense” for dApps include geolocation accuracy and robotics automation, or ISAC (integrated sensing and communications) use cases.
“There are plenty of opportunities, and the good thing about the whole platform, what we are building today, is that there is already a developer ecosystem which is running around [Nvidia]s] CUDA. So we could leverage those developer ecosystems as well for building the new use cases around the dApps.”
Cohere’s attention
One company that will certainly be interested in Nokia’s support for dApps is Cohere Technologies. The company has been working to align with Universal Spectrum Multiplier software and, more recently, its Pulsone ISAC technology with OEM radio basebands. Although it has cycled through deployment models in operator trials such as the near real time RIC, and being deployed on a server adjacent to the RAN baseband, being able to deploy directly as a dApp would certainly make sense, and it’s surely no coincidence that the company today released a statement that it would be developing support for the E3 interface, describing the interface as a key development in Open RAN systems. Cohere and Nokia are both also part of the US-government and Nvidia-backed OCUDU programme, which seeks to build a platform that can deploy advanced RAN capabilities within an Open RAN environment.
In fact, Nokia’s claim that it would be able to gain 2x performance via dApp deployment aligns very neatly with Cohere’s claims for its USM tech. So was Nokia referring to Cohere and is Cohere aligning with the Nokia AI-RAN dApp platform? Cohere said it was “unable to comment”.
A competitive advantage?
Where the near real time RIC has failed to materialise on many operator roadmaps, Nokia has leap-frogged into support for dApps, opening up its lower layer PHY, although likely in a very managed and cautious manner, to innovation from other players.
This, says Ed, will give the company an advantage that other providers cannot yet bring.
“We have been really working with Nvidia for quite some time now, and from an industry standpoint that’s why we said this is the industry’s first AI RAN platform. This [Nvidia’s ARC-Pro upon which Nokia’s AI-RAN baseband is built] is an open platform, obviously everybody can jump in. But as an industry-first AI RAN platform, we are setting the direction for the industry, and we believe this is the game changer, because the development of the AI-native RAN ecosystem is what is important in the years to come.”
Ed believes being first-to-market will open up opportunities for Nokia and its partners that will not be easy for others to replicate.
“We work with many partners – not just Nvidia – as well because that is how the industry has to co-create and how we can really take innovation forward. That’s why we believe this can’t be easily replicated by our competition unless they change and move in this direction.”
