As the U.S. prepares for future conflicts where victory will hinge on using data to accelerate decision making, integrating 5G and 6G technology into the military’s war-fighting capabilities is critical in maintaining a competitive advantage. Yet, sharing the limited spectrum dedicated to 5G with other commercial and non-defense industries is an omnipresent challenge. Efficient use of the limited resource of the radio spectrum requires maximum utilization of spectrum resources in the time, geographical and frequency domains.
The good news is the White House recently unveiled its National Spectrum Strategy to drive innovation and give U.S. industrial competitiveness a boost. As part of this strategy, additional spectrum bands have been identified for in-depth study to determine if they can potentially be repurposed. In addition, the NTIA just announced its second round of grants from the Public Wireless Supply Chain Innovation Fund to support the development of open and interoperable wireless networks.
While these are important steps in the right direction, there is simply too much demand even with the expansion of the spectrum pipeline — from consumers, the private sector and the military alike. Looking at smartphone usage alone, the monthly average mobile data usage per smartphone in North America is anticipated to reach 58 GB in 2028, as access to unlimited data plans and 5G network coverage improves. This overcrowding of wireless communication channels leads to high risks of interference for critical Department of Defense (DOD) missions.
While spectrum sharing has been heralded as a potential solution, handling the possibility of interference continues to be challenging, given U.S. operators tend to operate at higher power levels compared to other parts of the world. However, there are increasingly positive signs that applying artificial intelligence (AI) can enable safer spectrum sharing — finally at scale.
Historical success with AI for 5G
5G technology did not have AI natively incorporated into its standards. As a result, AI must be manually added and prioritized based on where there is both value in doing so, and the technical ability to implement it. Spectrum optimization, where AI techniques are applied to interference detection to enable spectrum sharing, is one exciting area for AI that has shown immense promise. As spectrum is a limited resource, any technology that can improve the efficiency of how it is used warrants immediate attention.
Recent projects have involved prototyping the application of AI to radio signal processing while successfully accelerating access to the analytics required for smarter decision making. Working off large unprocessed radio frequency datasets, prototypes leveraging AI are proving that they could triage regions of interest and drive discovery of conflicting commercial or military 5G signals, while providing early warning of adversary jamming effects.
AI has also been useful in quickly suggesting interference-mitigation measures, such as alerting to the need to vacate the channel due to a higher priority incumbent signal. The ability to use AI technology to accurately identify and characterize interference in terms of power, frequency, bandwidth and timing will unlock access to large swaths of valuable midband spectrum — if it can be implemented economically at scale.
Challenges to widespread AI implementation
Widescale implementation of AI to enable spectrum sharing is where things get complicated. AI does a good job in detecting spectrum interference types, but it is limited by what it can be trained in advance to recognize. As an industry, we must make sure it is holistically trained. Here are some approaches to consider when exploring wide-scale implementation:
- Test with legacy systems: Legacy technologies are still heavily in use by incumbents and their operational details must be part of any spectrum sharing solution. Older tech impacts the efficacy of AI, so it is critical to be realistic and meet stakeholders where they are on the digital continuum. AI solutions for spectrum sharing must also be tested in mixed-signal environments that simulate the effect of real-world conditions, such as land clutter, atmospheric obstruction and the effects of buildings and terrain that result in signal fading and multipath effects which can distort actual received signals.
- Consider multiple stakeholders. A strict focus on technical approaches is almost certainly doomed to fail if consideration is not also paid to the competing demands of multiple disparate stakeholders and legacy technologies still in use. Research and development efforts toward spectrum sharing must be coordinated with stakeholder policy to ensure that consensus is built towards an acceptable solution. Joint industry and government working groups, such as the National Spectrum Consortium (NSC)’s Partnering to Advance Trusted and Holistic Spectrum Solutions (PATHSS) Task Group, are a compelling channel to advance information sharing. They help ensure that all sides have input to both the technical solution and the policy recommendations.
- Leverage hybrid approaches. In cases that require high sensitivity or low power consumption, AI techniques are not the best option, as they do not yet equal the accuracy and precision of existing techniques in those situations. Hybrid approaches are the solution. A combined approach that allows AI to do what it does best and traditional methods to do what they do best is the best path forward.
- Optimize use of spectrum real estate: It is still a reality that some spectrum bands are used only a portion of the time and remain unused otherwise. The 3.1-3.45 GHz portion of the midband, for example, is currently in the spotlight as a leading candidate for sharing between future 5G networks and existing DOD systems. Itinerant use of this band, along with deterministic solutions for radar detection and mitigation, make this band a strong candidate for AI-based spectrum coexistence.
The U.S. government is moving in the right direction with the new National Spectrum Strategy and NTIA grants, and AI will be an important piece of the puzzle moving forward. AI-based systems can detect the presence of a military system, and then can temporarily instruct a nearby 5G base stations to reduce their power until the operation ceases. Now, the time has come to scale this technology. While 5G didn’t have AI baked in, 6G networks will be the first generation of wireless networks developed to natively take advantage of AI from the beginning. 6G technologies will continue to push the envelope of spectrum bands — and AI will be a key part of the adaptive technology that unlocks spectrum efficiency.
