Nokia is looking to take advantage of the latest in machine learning to more efficiently manage the complex nature of new technologies associated with next-gen mobile networks. That's according to a recent talk from made by one of the company's executives, Rajeev Agrawal during Taiwan's annual Computex trade show. Although the company isn't quite ready to publish the specifics of what that entails, Agrawal says the company is researching no fewer than three methods to approach the issues. Those include the use of machine learning to address more obvious issues such as the ability to scheduled beamforming over massive MIMO networks. However, the company is also looking into using AI for both locating and transmitting to devices positioned indoors and automating the configuration of uplink and downlink channels. Each of the solutions, used in conjunction with one another, should make reaching pinnacle of 5G networking a much less daunting task.
The first part of the three-part solution addresses the difficulties that arise in network management when dealing with networks configured to support multiple-input and multiple-output (MIMO) technologies. MIMO has been under experimentation, testing, and even deployment, in some cases, for quite some time but it doesn't necessarily get any easier and isn't being used to its full potential as a result. Effectively, signals are sent in parallel through a larger array of antenna, allowing for a more dense beam and more connectivity. In combination with beamforming, sending targeted beams to users instead of wide bursts, Massive MIMO can be very efficient and deliver huge improvements to bandwidth. Nokia wants to make the process of joining the technologies more efficient by training neural networks to discover and adapt to use the best scheduling possible in real time.
Tied in with that, the company also wants to utilize small cells equipped with A.I. to locate devices that are separated from the base station by walls, such as in a building. By locating the position of a device to within a claimed range of 9cm to 13cm, Nokia's networking equipment can better plot out the above-mentioned beams. The solution the company is researching involves the use of LTE eNB radiofrequency data to plot positions and the use of machine learning to teach A.I. to correctly identify a device's position. Finally, Agrawal believes that machine learning could be used to predict the equipment characteristics of a given user device and configure the beamed uplink and downlike appropriate for the best results based on that estimation. That wouldn't necessarily be too difficult to accomplish since there aren't an unlimited number of possibilities with regard to networking equipment found in smartphones or other connected hardware. Primarily, each of the solutions put forward by the exec would be implemented on the software side. So it's not impossible that they might eventually be utilized on the incoming 5G platform, which is set to arrive beginning in 2019.