5G performance – better signal with AI
I ended my recent blog about 5G network performance with ‘you ain’t seen nothing yet’ and since then we have achieved yet another 5G data rate world record - 4.7 Gbps on Nokia AirScale Base Stations deployed in major U.S. carriers’ commercial networks. This clearly shows the short-term potential of 5G on millimeter waves and provides for long term direction - remember that in 4G the peak data rates of commercial smartphones grew 20 times over the course of seven subsequent 3GPP releases.
On the other hand, the higher the frequency bands used, the smaller the coverage area per radio cell, so maximizing the network coverage area of each base station is more important than ever. In a city with a given grid of macro base stations, this means maximizing the cell edge performance and ensuring an excellent user experience, wherever the subscriber happens to be.
When it comes to maximizing the network capacity of each base station site in densely populated urban areas, there is a clear trend towards adaptive Active Antennas rather than conventional passive antennas. Here I’d like to share what we at Nokia have been doing, together with Vodafone, to optimize the coverage and cell edge performance of Active Antenna radio sites. The goal is to ensure the network can deliver excellent user experience with as few antenna sites as possible.
To understand the coverage challenge, let’s dive into a typical urban deployment scenario: high rise office buildings, next to a mall or restaurant areas in lower rise buildings, maybe adjacent to a park and a number of residential buildings. User density and traffic density will change hour-by-hour and day-by-day, for example, peaking in the high-rise buildings during office hours, moving to the street and the restaurants during the evening and to the park at weekends.
Within a conventional radio cell, this flow of traffic is dealt with by optimizing for the best compromise of antenna settings. This shows up one of the key advantages of modern 5G sites: the possibility to shape the channel transmissions allows to dynamically focus on areas within a cell, where most traffic is. “We realized with the 5G Active Antennas we could adapt the coverage of every cell to the real position of our customers. Then our main target for this project was to use Artificial Intelligence for every cell to shape the signals to maximize the number of users and the traffic’”, said Francisco Martin, head of Radio Product, Vodafone Group.
Getting ahead of the traffic
But how do we predict where the traffic will be? The answer is Artificial Intelligence applied to network data. In the Vodafone project, we analyzed the distribution of network traffic. We then developed a machine learning model and trained it using that data. Using the model, we were able to adapt the beam pattern of our AirScale Active Adaptive Antennas to the predicted geographic traffic distribution, for example, to the higher floors of an office building or to a residential building. The optimized beam pattern forms the basis for efficient signal shaping and ensures an excellent typical user experience.
Based on field measurements, we could achieve an average performance improvement of 78%. We could also overcome penetration losses and make the network available within individual buildings, where previously the smartphone could not find the network.
Another plus was that we no longer wasted spectrum resources, as we could turn the conventional homogenous cell patterns with their unavoidable interference into beam patterns. These beams are continually focused on those areas within the cell that need mobile broadband capacity.
Using machine learning allows our mobile broadband networks to adapt themselves to changes in geographic traffic distributions, getting more from existing network infrastructure and frequency spectrum.
An initial 78% improvement without any hardware modification is more than promising. As I concluded in my previous blog – You ain’t seen nothing yet.
Share your thoughts on this topic by joining the Twitter discussion with @nokianetworks or @nokia using #5G #smartcities