Sometimes people combine things you wouldn’t really think would go well together. Chocolate and peanut butter come to mind, a wonderful mix that doesn’t seem like it should work as well as it does.
Well, technology also frequently mixes and matches things to see what works best. A lot of people probably didn’t think something as modern as artificial intelligence would go well with radio frequencies, but they actually work great together.
Artificial intelligence (AI) is being used to grow radio frequencies (RF) and bring this retro technology to the modern sphere. With RF, we can expand our signals further than before, working faster than with the tools we currently have.
To be clear, we aren’t trying to move backward or get rid of the technology we have now. We’re just finding interesting ways to mix radio into our modern applications and coming up with incredible results. Here’s what the future holds for the meshing of AI and RF.
AI detects offshore RF signals
The National Institute of Standards and Technology (NIST) and the Federal Communications Commission (FCC) have worked together to create a way for the United States Navy to share a popular 150MHz section of RF for communicating. By using AI, NIST can detect when offshore radars are working so multiple channels don’t try to use the same band at once. The AI will tell commercial users when they have to pause in using the 3.5GHz band.
The FCC created the Citizens Broadband Radio Service (CBRS) in 2015. It allows commercial vendors and providers to use the 3.5GHz band when it’s not needed for operations from the United States Navy. Companies like AT&T and Google jumped at the chance because the band was expected to grow their markets and provide better coverage and data rates. Without the AI and a little legislature, companies and citizens alike wouldn’t be able to use this very popular band.
The Defense Advanced Research Projects Agency (DARPA) noticed a problem with too much happening all at once. There are a lot of RF signals put into the same spectrums, making things difficult to discern and hear. DARPA created the Radio Frequency Machine Learning Systems program (RFMLS) to fix this problem.
RFMLS can pick out and discern one signal from the next, identifying more important ones and those that don’t follow the rules. With the internet of things (IoT) gathering data from all sorts of sectors, including radio, RFMLS helps collect important, relevant information for this exact purpose. DARPA is hoping to continue the implementation of AI to create better spectrum sharing in the future, much like what NIST has worked on.
Enhancing system performance
As a company that works to bring new engineering technologies to the United States Department of Defense, Alion Science & Technology works to lower costs and improve efficiency. It has also found the same problem as DARPA, with the RF spectrum becoming congested with useless information that drowns out important data. Instead of picking out certain signals, Alion has found a way to boost the ones it needs.
Alion designed a custom AI network to find any signal in raw in-phase/quadrature data with signal-to-noise (SNR) values from +10dB to -10dB. These SNR values are on different sides of the extreme, with -10dB being afflicted with a lot of noise to the point of being nearly discernable. However, Alion’s found that certain networks can understand different aspects of SNR data. AI is used to pair the two up to effectively boost the signal.
RFI and causes
Radio frequency interference (RFI) is what happens when another electronic signal impedes upon the original one. These days, because most things have moved past radio technology, we often call this phenomenon electromagnetic interference, or EMI. The two terms, RFI and EMI, are used synonymously — but it’s important to know where exactly the differences lie, mostly with the product being affected.
The main cause for RFI is usually a congested space of other electronic devices using the same spectrum, like when a radio buzzes because your cellphone is getting a call. The problem could also be due to the design of the product, creating low absorption loss. Both of these are pretty prominent problems, but the core issue is often a design flaw.
A technique called RF shielding is often used to minimize RFI. The shield protects important circuit functions and prevents emissions, lowering the chance for RFI to occur. One could use conductive foam, a metal box or metallic film. Either way, RF shielding should be done as early into the design as possible.
You can also use a filter to protect your device from RFI and prevent it from being emitted. Most filters apply a three-pole approach to target a range from 150kHz to 30kHz. These are available at both the commercial and consumer levels for anyone to use. However, this is not a DIY project like RF shielding.
Machine learning and RF
A lot of AI implementation comes down to necessity, as evidenced by machine learning and RF. The military, companies and average people all use radio these days, which can congest the systems on a ridiculous level. Radio communication isn’t obsolete yet.
While a lot of people have become used to satellites, radio is still a viable and quick way to get information across large areas. Unlike many satellites, radio will also always remain dependable — no matter what the situation or how far into the future we look.
Getting AI involved only helps RF evolve into modern-day usage. Now congestion doesn’t have to be an issue. We can pick out certain frequencies, boost even the lowest of signals or tell others to get off when something more important needs to go through.
With AI, radio can rise in popularity again and become the prominent figure it once was. Satellite communication won’t be going anywhere, but perhaps the radio can take us further than any of us previously thought possible.