France's Orange has revealed details of an artificial intelligence (AI) research project with IBM and Nokia that could help it to predict demand patterns in future 5G networks and avoid the "overprovisioning" of network resources.
The CogNet project falls under the auspices of Europe's 5G Infrastructure Public Private Partnership (5G-PPP), a joint venture between the European Commission and companies in the region. Its participants also include Spain's Telefónica, cloud services provider Interoute and research institutions like Germany's Fraunhofer.
Researchers and data scientists from those organizations have together been exploring the application of AI, and specifically machine learning, to the management of networks based on 5G, a next-generation mobile technology that is moving rapidly through the standardization process.
The first 5G services are unlikely to appear in Europe until 2019 or 2020, but service providers are already concerned about the potential difficulty of managing 5G networks manually. Huawei Technologies Co. Ltd. and Ericsson AB (Nasdaq: ERIC), the world's two largest suppliers of mobile network equipment, have argued that 5G will bring a huge increase in the number of configurable network parameters. Without transferring responsibility to automation and AI systems, operators may struggle to cope.
The CogNet project kicked off way back in July 2015 but has kept a low profile ever since and only just announced details of its work. While there is no suggestion that Orange (NYSE: FTE) or Telefónica is poised to use AI technologies developed with IBM Corp. (NYSE: IBM) and Nokia Corp. (NYSE: NOK), the details of the research collaboration may unsettle Ericsson and Huawei, both of which see AI in networks as a major growth opportunity. (See Robot Wars: Telecom's Looming AI Tussle.)
With its Watson technology, IT giant IBM is regarded as one of the world's AI pioneers and could mount a strong challenge to network vendors in the 5G market. Until now, its commercial dealings with telcos have focused mainly on supplying Watson for customer-service chatbots and other digital assistants. But an IBM spokesperson says Watson can also be used for network management, and IBM's involvement with CogNet clearly hints at a 5G role in the future. (See Chatbot Takes Charge: Vodafone's Customer Services Overhaul and Orange Plans Bank Raid With AI, Digital Weapons.)
Presenting details of CogNet's findings at last week's AI Net conference in Paris, Orange's Imen Grida Ben Yahia said the project had looked partly at using AI to predict demand patterns and anomalies in dense urban areas. Researchers used real network data from Manhattan to see if they could make predictions about the occurrence of a large-scale event that might "drastically affect traffic demand."
In another CogNet study, they tried using AI to overcome the problem of the "noisy neighbor," whereby two or more virtual network functions on the same cloud infrastructure cause interference by unnecessarily hogging resources.
AI could also be used to manage service level agreements, said Yahia, who works as a senior research and development engineer at Orange Labs.
"We could map this to low-level metrics we are collecting to predict if a given SLO [service level objective] is going to be violated," she said at the AI Net conference.
Yahia told attendees that allowing data scientists to mingle with network professionals during the CogNet project had proved vital when it came to addressing challenges that might crop up in the 5G world.
In the meantime, both Ericsson and Huawei have been making plenty of noise about their own AI activities this year.
Ericsson showed off various "machine learning" products at this year's Mobile World Congress, including a "self-learning" tool that shifts network resources between basestations to suit traffic levels and user needs. (See Humans Beware: Ericsson Readies Machines to Run the Network.)
That service, which sounds similar to the applications that CogNet has been exploring, is already in trials with Vodafone Spain.
Huawei, meanwhile, is due to launch an AI-enabled compute platform called Atlas that will carry out data processing and analysis in central cloud facilities. It is also developing a Wireless Intelligence system to automate some processes in 5G radio networks. (See Huawei Commits Up to $20B for Annual R&D, Fleshes Out AI Pitch.)
The Chinese vendor currently dwarfs both Ericsson and Nokia on R&D investments. Last year, it spent $13.8 billion on R&D, with Ericsson managing $4.5 billion and Nokia stumping up about $5.2 billion. (See Huawei Dwarfs Ericsson, Nokia on R&D Spend in 2017.)
— Iain Morris, News Editor, Light Reading