NIST Researchers Use A.I. to Boost 5G Efficiency

Billy Hurley, Digital Editorial Manager

5G already promises super-fast, real-time data connections between people and their technologies.

And researchers at the National Institute of Standards and Technology (NIST) are working on a way to make the cellular network standard even faster.

Using a machine learning formula, NIST engineer Jason Coder and his team have demonstrated a highly efficient way for 5G and other wireless networks to select and share communications frequencies.

Described at the 2020 IEEE Vehicular Technology Conference virtual online conference this month, the formula could be programmed into software on transmitters in many types of real-world networks.

The mathematical model is meant to operate in a network environment that contains a variety of frequencies, including bands with License Assisted Access (LAA) and unlicensed bands like Wi-Fi. Think a hospital, airport, or building with multiple wireless access points and cellphone operations.

Through the sharing of unlicensed frequency ranges, the formula allows wireless systems like 5G to boost their efficiency.

“This work explores the use of machine learning in making decisions about which frequency channel to transmit on,” said NIST engineer Jason Coder . “This could potentially make communications in the unlicensed bands much more efficient.”

The NIST study focused on an indoor scenario in which Wi-Fi competes with cellular systems for specific frequencies, or subchannels. The machine-learning method enables transmitters to rapidly select the best subchannels for successful and simultaneous operation of Wi-Fi and LAA networks in unlicensed bands.

{youtube} https://www.youtube.com/watch?v=2j9BGVKbzS4 {/youtube}

"Our research group’s mission is to look at sharing the radio spectrum efficiently in any frequency band – licensed or unlicensed," Coder told Tech Briefs in the Q&A below.

Without communicating to each other, the transmitters each learn to maximize the total network data rate. The “Q-learning” technique maps environmental conditions — such as the types of networks and numbers of transmitters and channels present — onto actions that maximize a value, known as Q, that returns the best reward.

The algorithm runs through the options and determines which channel provides the best outcome; each transmitter then learns to select the channel that yields the best data rate under specific environmental conditions. If a transmitter selects an unoccupied channel, then the probability of a successful transmission rises. When interference is minimized, the signal is strengthened.

In the NIST computer simulations, the allocation method assigns channels to transmitters by searching all possible combinations for the best total network data rate. The NIST study found that an exhaustive effort to identify the best network latency would require about 45,600 trials, whereas the NIST formula could select a similar solution by trying only 10 channels.

Coder and the NIST team now plan to model the method in larger-scale outdoor scenarios and conduct physical experiments to demonstrate the effect.

In a short Q&A below, Jason Coder tells Tech Briefs what's possible when 5G gets even faster and more efficient by sharing the spectrum.

Tech Briefs: If communication in the unlicensed bands becomes more efficient, what does that efficiency enable? What could, say, a manufacturer do with this kind of efficiency?

Jason Coder: From the research we presented in the paper, I think the biggest change would be an improvement in network latency. Use of the proposed A.I. method has the potential to reduce the amount of time spent searching for a channel to transmit on. This enables the transmitter to select a channel quickly and send its data.

One of the big thrusts for 5G has been to reduce network latency. Lower network latency enables applications like high-speed control systems, collaborative robotics, and augmented reality. This method could be a step in that direction for unlicensed communications.

In general, the use of A.I. has the potential to make licensed and unlicensed bands more efficient. Either by sharing spectrum more efficiently, by optimizing network configurations, or many other possible improvements. As a community, we’re starting to demonstrate the value of A.I. in communications, but it may not be until “6G” that we fully harness its potential.

Tech Briefs: In what kinds of environments and applications will this kind of frequency selection be most valuable?

Jason Coder: I see two areas where this type of frequency selection can make an impact. First, in crowded spectrum environments. In today’s connected world, this typically means anywhere you have a dense gathering of people. Hospitals, airports, convention centers, and residential complexes can all see crowding in the unlicensed bands. Using the proposed method could help relieve the congestion by increasing spectrum efficiency and reducing the time it takes devices to locate the open channel.

Second, channel selection can be very important when the path between your device and the base station (or receiving device) is poor. This can occur when the receiving device is far away, obstructed by a building, or when the environment is dynamically changing. The good news is that signal propagation characteristics are frequency dependent. For example, lower frequency signals tend to propagate better through buildings compared to higher frequency channels. The trick then becomes finding the channel that works best. The proposed method could be adapted to make that selection process much more efficient.

Tech Briefs: What inspired this work?

Jason Coder: Our research group’s mission is to look at sharing the radio spectrum efficiently in any frequency band – licensed or unlicensed. We’re a neutral entity in the communications space; non-regulatory, and non-enforcement. We seek to develop tools that industry can use to better achieve their wireless communication goals as well as tools they can use to measure the performance of their systems.

Specifically, this work was inspired by some of our other work in LTE Licensed Assisted Access (LAA). We’ve looked extensively at coexistence between LTE-LAA and Wi-Fi, and this method is an offshoot of that work. We realized that if we change the way systems select channels, we may be able to improve their ability to coexist with each other.

Tech Briefs: And what are you working on next?

Jason Coder: A big push for us has been to develop measurement methods for examining the coexistence between two wireless systems. Ideally, we’d like to develop these methods proactively; before systems are deployed and problems are uncovered at the last minute. The millimeter wave unlicensed bands (e.g., 60 GHz) present an opportunity for us to develop coexistence measurement methods before use of that band really takes off in the coming years. Coexistence is a bit different at these frequencies, and we need to build on the foundations we’ve created at lower frequencies to develop robust methods at 60 GHz (and beyond).

Tech Briefs: What is most exciting to you about the role of 5G in the future?

Jason Coder: To me, the most exciting parts about 5G are the applications we don’t know about yet because they haven’t been created. There are so many different possibilities for 5G applications, it’s difficult to predict how it will change our lives. Applications like remote surgery via 5G links, augmented reality, and autonomous vehicles have the potential to change our lives for the better. But people are still innovating and developing new applications. I think the biggest surprises may still be in store.

What do you think about the possibilities of 5G? Share your questions and comments below.