Mathematical formula could help 5G networks share communication frequencies efficiently

Researchers at the National Institute of Standards and Technology (NIST) have developed a mathematical formula that computer simulations suggest could help 5G and other wireless networks select and share communications frequencies about 5,000 times more efficiently than trial and error.

NIST engineer Jason Coder performs mathematical calculations for a machine learning formula that could help 5G and other wireless networks efficiently select and share communication frequencies.

The novel formula is a form of machine learning that selects a range of wireless frequencies, called a channel, based on prior experience in a specific network environment. The formula can be programmed into software on transmitters in many types of real-world networks.

The formula is a way to help meet the growing demand for wireless systems, including 5G, by sharing unlicensed frequency ranges, also known as spectrum bands. For example, Wi-Fi uses unlicensed spectrum, which is spectrum that has not been allocated to specific users by a regulatory agency.

Quickly select excellent sub-channels

The study focused on situations where Wi-Fi competes with cellular systems for specific frequencies, or subchannels. What makes this situation challenging is that these cellular systems are increasing their data rates by using a method called licensed assisted access (LAA), which combines unlicensed and licensed bands.

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

This formula enables transmitters to quickly select the best subchannels to successfully and simultaneously operate Wi-Fi and LAA networks in unlicensed bands. Each transmitter learns to maximize the total network data rate without communicating with each other. The scheme quickly achieves overall performance close to that based on an exhaustive trial-and-error channel search.

The research differs from previous machine learning studies on communications in that it takes into account multiple network "layers," the physical devices, and the channel access rules between base stations and receivers.

The formula is a Q-learning technique, meaning it maps environmental conditions (such as the type of network and the number of transmitters and channels present) to actions that maximize a value (called Q) that returns the best reward.

By interacting with the environment and trying different actions, the algorithm learns which channel provides the best results. Each transmitter learns to choose the channel that produces the best data rate under specific environmental conditions.

Increase data rates

If both networks select channels appropriately, the efficiency of the combined overall network environment will increase. This approach improves data rates in two ways. Specifically, if the transmitter selects an unoccupied channel, the likelihood of a successful transmission increases, resulting in a higher data rate. Also, if the transmitter selects a channel that minimizes interference, the signal is stronger, resulting in a higher received data rate.

In computer simulations, the best-of-breed method assigns channels to transmitters by searching through all possible combinations to find one that maximizes the total network data rate. This formula produces results close to the best-of-breed results, but with a much simpler process.

The study found that the exhaustive work required to find the best solution required about 45,600 trials, while the formula could select a similar solution by trying 10 channels, with just 0.02% of the effort.

The study targeted indoor scenarios, such as buildings with multiple Wi-Fi access points and cell phone use in unlicensed bands, but the researchers now plan to model the approach in larger outdoor scenarios and conduct physical experiments to demonstrate its effectiveness.