Qualcomm and Nokia Bell Labs confirmed how multiple-vendor AI fashions can work collectively in an interoperable means in wi-fi networks.
Carl Nuzman, Bell Labs Fellow at Nokia Bell Labs and Rachel Wang, principal engineer at Qualcomm, stated in a weblog put up that they demonstrated the flexibleness of sequential studying, which may facilitate community
decoder-first or gadget encoder-first coaching.
They stated Qualcomm and Nokia Bell Labs are persevering with to work collectively to reveal the worth of interoperable, multi-vendor AI in wi-fi networks. At Cell World Congress 2024, they first demonstrated over-the-air interoperability of AI-enhanced channel state suggestions encoder and decoder fashions.
These had been working in reference cell units with Qualcomm’s 5G modem-RF system and a Nokia prototype base station, respectively. These interoperable fashions had been developed by the 2 firms utilizing a brand new approach known as sequential studying. Now they’re again at MWC 2025 with extra.
Channel state suggestions helps the community determine one of the best ways to ship information to your gadget. As wi-fi situations change, so does the optimum route a transmission takes from the community to the gadget. Qualcomm and Nokia had been capable of make the community smarter and extra environment friendly by producing exact beams with AI.
With sequential studying, a number of firms can co-design interoperable AI fashions while not having to share proprietary particulars of their implementations. As an alternative, a coaching dataset of mannequin enter/output pairs is shared from one firm to the opposite.
Constructing on this proof-of-concept, the businesses have since continued working collectively to reveal the worth, flexibility and scalability of interoperable AI for channel state suggestions.
Wi-fi AI robustness in several bodily environments
As AI applied sciences are deployed in real-world networks, it is very important be sure that fashions work robustly in numerous environments. Coaching datasets ought to be sufficiently numerous for AI fashions to study successfully; nevertheless, it’s unrealistic for them to cowl all attainable eventualities.
Thus, it’s crucial for AI fashions to generalize their coaching to deal with new conditions. Within the collaboration, the corporations studied three very totally different cell websites: an out of doors suburban location and two totally different indoor environments.
Within the first state of affairs, they in contrast the efficiency of a standard AI mannequin educated with numerous datasets with hyper-local fashions which are educated at particular places. They discovered the widespread AI mannequin can work in several environments with comparable efficiency as hyper native fashions.
The businesses later tailored the widespread mannequin to incorporate information from Indoor Web site 2 (the Tailored Widespread mannequin). Then they measured the person information throughput at 4 totally different places inside Indoor Web site 2. The widespread mannequin got here inside 1% of the efficiency of the Tailored Widespread mannequin in all circumstances, displaying the robustness of the overall widespread mannequin to new eventualities.
AI-enhanced channel state suggestions permits the community to transmit in a extra exact beam sample, bettering the acquired sign energy, lowering interference, and in the end offering larger information throughput. We measured this enchancment by logging information throughputs skilled with AI-based suggestions and grid-of-beam-based suggestions (3GPP Kind I) because the cell person moved between varied places within the cell.
Use of the AI suggestions yielded larger throughput, with per-location throughput beneficial properties starting from 15% to 95%. The throughput beneficial properties that will likely be noticed in business techniques underneath AI-enhanced CSF will rely on many elements. Nevertheless, the outcomes of this proof-of-concept, along with quite a few simulation research, recommend that the throughput with AI enhancements will likely be persistently larger than the that achieved with legacy approaches.
Sequential studying may be carried out in two methods, both gadget encoder-first or community decoder first, which has totally different implications for deployment and standardization. To help 3GPP’s rising curiosity within the decoder-first method, this yr we changed our authentic encoder-first demonstrations with decoder-first mannequin coaching.
With the encoder-first method demonstrated in MWC 2024, Qualcomm designed an encoder mannequin, generated a coaching dataset of enter/output pairs, after which shared the dataset with Nokia, which subsequently designed an interoperable decoder.
This yr, with the decoder-first method, Nokia designed a decoder mannequin and generated and shared a coaching dataset of decoder enter/output pairs for Qualcomm Applied sciences to make use of in designing an interoperable encoder. We discovered that fashions designed by each modalities carried out equally nicely, inside a number of proportion factors.
Backside line
The prototype that Qualcomm Applied sciences and Nokia Bell Labs have collectively demonstrated represents a key step in transferring AI-enhanced communication from idea to actuality. The outcomes present that the person expertise may be considerably improved, in a strong means, through a number of studying modalities. As we study to design interoperable, multi-vendor AI techniques, we are able to begin to notice enhanced capability, improved reliability, and diminished vitality consumption.
