Operators, spectrum leads, and coalition partners weigh in on edge AI, survivability, and spectrum coordination
At the Defense Communications Forum, a series of military operators, spectrum policy leads, and coalition partners came together to answer one question: how do you turn edge AI and private 5G from slideware into something a forward-deployed unit can actually use? Moderated by retired Colonel Michael Black of AFCEA International, the session zeroed in less on what’s theoretically possible and more on what’s deployable.
“The network is no longer just an enabler, it’s the battle space,” Black said, pointing to a demand signal that looks the same whether you’re talking homeland security or a contested expeditionary environment — resilient, low-latency, secure connectivity, under any condition. That’s a tall order, and the panelists discussed where the technology is ahead of the process, and where the process is ahead of the technology.
Architecting survivable 5G networks
Venkatesh Ramaswamy, chief technologist for NextG at MITRE, laid out what a defense-grade reference architecture actually needs to look like. The shape is broadly similar to a commercial 5G network, but with some important differences. At the top sits a mission layer that defines intent and priority — who gets what. That feeds an orchestration and security layer, which translates centralized control into distributed execution. Below that is the edge, which Ramaswamy described as the heart of the system: where the data lives, where the compute happens, where data sovereignty is preserved, and where latency is kept in check. Underneath it all is the connectivity layer, or the 5G transport everyone is interested in.
What makes the connectivity layer different in a defense context comes down to survivability, programmability, and adaptability. “Nothing is planned. Everything is rapidly changing, and the network better be able to change in a rapid, dynamic fashion,” Ramaswamy said. The architecture has to assume conditions will shift faster than any pre-planned config can keep up with.
He was also blunt about a tendency to over-engineer. The industry, he argued, often builds for what’s possible rather than what’s necessary — designing a Swiss army knife when a knife will do. The cost is complexity, and complexity breaks. His litmus test? Is the network quickly deployable, can a soldier operate it without much training, and does it produce a measurable improvement to the mission? If the answer to all three is yes, it probably isn’t over-engineered.
Spectrum superiority and coordination challenges
Salvador D’Itri, chairman of the National Spectrum Consortium, framed spectrum superiority as a coordinated maneuver across air, land, and sea — and noted that 5G has to operate at that same speed of coordination if it wants a seat at the table. The gap, in his view, isn’t really about regulators. It’s about the military’s own coordination machinery. The 1494 spectrum certification process, in particular, is showing its age against the pace of modern radio hardware.
D’Itri’s ideal looks something like this: a 5G system gets issued a token, and whether it’s operating in DOD-controlled territory, the UK, CBRS in the U.S., or somewhere in between, that token tells a frequency manager the system is authorized and good to go. We’re not there yet. He also pushed back on the idea that 5G will swallow everything else, noting that the military “rarely walks into any battle with one communication system. So we keep talking about this like it’s the system 5g it’s going to take over its wide band. Everything else is antiquated. That’s, that’s, that’s just never going to happen, not in my lifetime.”
NATO Band 79 came up repeatedly as a candidate sandbox for allied coordination — high enough in the sub-6 range to be useful, and common enough across NATO to be worth investing in, but with real range limitations in difficult terrain. The deployment model is also shifting. CBRS has already proven mil-on-commercial sharing at scale, and work at sites like Marine Corps Logistics Base Albany is now demonstrating mil-on-mil 5G sharing. OUSD R&E’s Advanced Spectrum Coexistence effort, led by Josh Weaver, is pushing into harder bands like 3.1–3.45 GHz, where incumbents are faster and less predictable than CBRS-style sharing can handle.
The remaining bottleneck, D’Itri argued, is the back end. “We’re not still trying to figure out how to allocate frequencies to a Ferrari” was how he put the goal — modernize the spectrum management tooling so that when 5G or 6G shows up, the military can actually execute. Ramaswamy added that not every problem has to be solved at the physical layer. Future G has funded work at MITRE on coexistence approaches that let existing radios keep using their designated spectrum while still operating as part of a unified network. Same problem, different layer.
Navigating the maritime edge and coalition operations
Florin Constantinoiu of the Romanian Navy Headquarters offered a coalition and maritime perspective, with the standard disclaimer that he was speaking from his own experience. Maritime is its own animal. There’s no dense tower infrastructure, signals scatter off waves, salinity corrodes equipment, and the platforms themselves are moving through international waters. Coverage beyond 50 to 70 kilometers from shore generally requires a hybrid model combining satellites and floating base stations.
He pointed to NATO’s REPMUS exercises in Portugal, which his Centre of Excellence has participated in for the last five years, as a useful working example. Starting with REPMUS 2023, a mobile private 5G network was used to test uncrewed maritime systems and stretch their operational range — from roughly 10 kilometers in 2023 to almost 30 kilometers in 2025. The 5G bubble delivered the expected benefits, including high throughput, low latency, device identification and traffic control — and support for high device density. It also came with the expected costs — deployment complexity, spectrum access, and vulnerability to jamming, as with any wireless system. Last year’s exercise tried to bridge two bubbles from different providers, with mixed results that he hopes will be sorted in the next iteration.
Interoperability is the other recurring issue. Most uncrewed systems are commercial-off-the-shelf with proprietary data formats, which makes coalition operations harder than they should be. Constantinoiu was, on balance, optimistic — NATO’s CMRE has developed protocols that have been tested in REPMUS, and industry is starting to align with NATO requirements (formal standards are still being written). But he kept coming back to the same point: “in the end, I think it’s all about people, because we need to have trust within ourselves and to understand that we need we need each other.”
Operational tempo and assured command and control
Lieutenant Colonel Benjamin Pimentel of the U.S. Marine Corps, speaking on Assured C2 and Project Dynamis, gave the operator’s read on when edge plus private 5G actually moves the needle. The pairing of a high-throughput network with multi-access edge compute, or MEC, lets a forward unit ingest sensor and logistics data, run an AI/ML decision aid against it, and act — without having to reach back to cloud resources that may not be available. “That high speed network is really facilitating speed of operational tempo at the edge, because now you can make decisions faster,” he said.
The flip side is just as important. When over-the-horizon connectivity is good, a small forward element can use a high-speed last-tactical-mile network to reach back for expertise it doesn’t carry organically. Maintenance, medical, whatever the situation calls for. He also cited a Future G project supporting an austere airfield construction effort, where an edge 5G bubble let operators coordinate using mobile devices and map-based apps instead of hauling additional radios.
On the deployment side, Pimentel made a practical argument for leveraging existing RAN sharing agreements with host nation mobile operators where possible. That keeps operations inside an established regulatory and licensing framework and sidesteps the slower 1494 process. “I think aligning operational requirements behind economic incentives is generally the fastest way to get to that capability that you want,” he said.
Assured command and control in contested, mobile, or disconnected conditions is where mission autonomy and PACE plans intersect. Pimentel walked through the autonomy stack — platform autonomy, collaborative autonomy across platforms and domains, and mission autonomy where an operator can hand a platform an objective and let it work. Combine that with platform-level PACE plans, where a single UAV or unmanned surface vessel might carry MANET radios, cellular, and LEO SATCOM, and you get continuity of C2 across a much wider band of battlefield conditions.
Table of Contents
