The DoD's Most Expensive Capability Gap

The two missing layers that prevent real robotic teaming

The Department of Defense just announced a $100 million prize to solve a problem most people don't even know exists. In January 2026, the Defense Innovation Unit, working with the Defense Autonomous Warfare Group and the Navy, launched the Autonomous Vehicle Orchestrator challenge: an urgent call for technology that can translate a commander's intent into coordinated action across entire fleets of robotic systems. The prize money signals desperation, not generosity. After decades of building platforms, control stations, and onboard AI, the Pentagon has realized it's missing the layers that actually matter for warfighting at scale.

‍

The autonomy stack that exists today has three solid layers

Start at the bottom. The Platform Layer is now mature. The physical robots work. MQ-9 Reapers fly persistent ISR missions. Switchblade loitering munitions can engage targets autonomously. Robotic Combat Vehicles navigate terrain. Unmanned Surface Vessels patrol littorals. The Replicator initiative has contracts with over 35 companies, 75% of them non-traditional defense contractors. The hardware problem is largely solved.

Above that sits the Control Layer: the ground control stations, data links, and human interfaces that let operators manage individual platforms. NATO STANAG 4586 standardized UAS control architectures. The UCS framework provided common middleware. Commercial software like QGroundControl enables mission planning and telemetry monitoring. This layer handles the mechanics of piloting one robot at a time.

Then comes the Autonomy Layer: the onboard intelligence that lets platforms operate without constant human input. Modern systems can follow waypoints, detect obstacles, track targets, return home when communications drop, and increasingly make decisions without human approval for each maneuver. DARPA's Assured Autonomy program and dozens of commercial efforts have pushed this layer forward rapidly.

These three layers function. What they don't do is enable the force structure the Pentagon needs: hundreds or thousands of coordinated autonomous systems operating as coherent combat formations across air, land, and sea.

‍

Two critical layers are missing entirely

Between the Autonomy Layer (what individual platforms can do) and actual combat employment sits a gap the Defense Department has only recently named. The DIU Orchestrator challenge calls for a technology layer that translates "commander's intent from voice, text, and haptic input into machine execution." This is the Orchestration Layer: the missing software that turns plain-language commands like "establish a picket line five kilometers out, prioritize fast-movers, maintain emissions control" into coordinated tasking across dozens of heterogeneous platforms.

Lt. Gen. Frank Donovan, who leads the Defense Autonomous Warfare Group, articulated the requirement precisely:

"We want orchestrator technologies that allow humans to work the way they already command, through plain language that expresses desired effects, constraints, timing, and priorities, not by clicking through menus or programming behaviors."

The Orchestration Layer must interpret natural language, understand military doctrine implicitly, decompose intent into platform-specific tasks, allocate resources across the fleet, monitor execution, and adapt when conditions change, all while operating at the edge without reliable cloud connectivity. It cannot modify the underlying platform autonomy stacks. It must work with whatever robots the DoD already owns.

Above the Orchestration Layer sits another gap: the Domain Bridge Layer. Current autonomous systems evolved in service-specific silos. A 2006 GAO report found that "the services' UAS development programs have been service-specific and insufficiently attentive to joint needs." That assessment remains accurate nearly two decades later. Air systems speak different protocols than ground systems. Maritime autonomy operates on completely different timescales and communication constraints than air domain systems. Each domain developed its own architecture. STANAG 4586 for air, JAUS for ground, UMAA for maritime. These architectures were never designed to interoperate.

The Domain Bridge Layer would enable what actual combat requires: a drone providing overwatch for a ground robot convoy, USVs coordinating with loitering munitions for littoral defense, air-ground-sea task forces executing joint missions. Project Convergence demonstrated pieces of this, USVs autonomously offloading UGVs, quadcopters coordinating with robotic combat vehicles, but these remain experimental one-offs, not operational capabilities.

‍

Why the operator bottleneck makes this the DoD's most expensive capability gap

The missing layers aren't just technical shortfalls. They're blocking the entire theory of victory for great power competition. Consider the numbers. Lt. Gen. Tony Bauernfeind, Air Force Special Operations Command commander, put it bluntly:

"It takes over 150 personnel to maintain a single MQ-9 orbit. That doesn't seem too unmanned to me."

The Marine Corps has only 38 of the 68 MQ-9 operators it needs, a 44% shortfall. Scale that to the thousands of systems Replicator envisions deploying, and the math simply doesn't work.

The current model requires one highly trained operator per platform, sometimes more. A pilot and sensor operator for the MQ-9. Dedicated personnel for ground control stations. Intelligence analysts to interpret feeds. Maintenance crews per aircraft. The "unmanned" system demands enormous human capital. As one analysis noted: "A battalion can't practically run 'one operator per vehicle' if swarming, distributed unmanned fleets are the new battlefield reality."

Worse, traditional interfaces don't scale. Current ground control stations use graphical interfaces: dropdown menus, map displays, point-and-click interactions. These work for one platform. For twenty platforms, they become cognitively overwhelming. For two hundred, they're useless. One industry CEO stated it plainly:

"Graphical user interfaces break down when you have lots and lots of objects and you can't build enough dropdown menus and buttons to communicate all the constraints you want to an intelligent system."

And consider the warfighter's reality. A pilot can't take their eyes off the cockpit to operate a touchscreen. Dismounted infantry can't carry laptops through contested terrain. Sailors on small boats have no room for elaborate control stations. Voice becomes the only viable input, but voice that actually works, not memorized keywords that fail under stress. The DIU challenge explicitly requires systems that function under "intermittent connectivity" and "edge environments without access to the cloud."

‍

What this reveals about the real path to robotic warfare

The Orchestrator challenge tells us the Defense Department finally understands what it's missing. The prize structure is telling: $100 million in awards, ten days from selection to first sprint, six months to demonstration. This is emergency acquisition tempo for a capability gap that's been quietly growing for years.

But the Orchestration Layer alone won't solve the problem. Without Domain Bridges, commanders get better control over air swarms, or ground swarms, or maritime swarms, but not joint task forces. They get faster single-domain operations, not multi-domain maneuver. The real requirement is both layers working together: natural language intent translated into coordinated action across any combination of platforms, in any domain configuration, under contested communications.

The services have spent twenty years building increasingly capable individual robots while neglecting the software that makes them fight as formations. They've optimized the autonomy of the parts while ignoring the architecture of the whole. That's now recognized as a strategic failure.

The next phase of defense autonomy won't be won by the company that builds the best drone, the fastest USV, or the most capable UGV. It will be won by whoever solves the coordination problem: by building the missing layers that transform a collection of robots into an actual fighting force. The Pentagon's hundred-million-dollar prize is an admission that this capability doesn't exist today and that they need outside help to create it.

The platforms are ready. The control interfaces exist. The onboard AI works. What's missing is everything that turns machines into teams.

‍