DARPA Wants 500 Drones in a Box. The Harder Problem Is What Commands Them.
On DARPA-SN-26-33 and the C2 gap at the center of containerized autonomous warfare.
DARPA Wants 500 Drones in a Box. The Harder Problem Is What Commands Them.
On DARPA-SN-26-33 and the C2 gap at the center of containerized autonomous warfare
Earlier this month, DARPA's Tactical Technology Office published RFI DARPA-SN-26-33 — a request for industry input on autonomous Group 1-3 drone constellations and containerized launch-recovery-recharge infrastructure. The concept: self-sustaining drone nodes, deployable from standard containers, capable of launching, recovering, refueling, and managing swarms of up to 500 aircraft with minimal human involvement in GPS-denied, comms-contested environments.
It is an ambitious logistics and engineering challenge. And the coverage has focused almost entirely on the hardware: container dimensions, battery cycling, autonomous landing in confined spaces, hybrid-electric propulsion tradeoffs.
That is the wrong problem to lead with.
The Container Is Not the Bottleneck
DARPA's RFI correctly identifies significant engineering gaps on the platform and infrastructure side. Battery-only drones face discharge depth and recharge-cycle constraints. Fixed-wing aircraft are efficient in flight but hard to recover automatically in compact areas. Existing launch cells often lack the internal volume for communications and fuel, and many rely on differential GPS that won't function in contested environments.
These are real problems. They will be solved, probably within the program timeline, by competent teams with adequate funding.
The problem that will not be solved by the RFI as currently framed is the command-and-control architecture for 500 autonomous aircraft operating in a disconnected, degraded environment across a dispersed theater.
Hardware autonomy is table stakes. Mission-level coordination under denial is the hard problem.
What 500 Drones Actually Requires
Consider the operational picture DARPA is painting. Multiple containerized nodes, each managing dozens to hundreds of drones, dispersed across a contested theater. GPS unavailable. Cloud connectivity unavailable. Human operators intermittently reachable at best.
The drones must: conduct reconnaissance, perform electronic warfare, relay communications, designate targets, and execute strike missions — potentially across multiple nodes simultaneously, potentially without any direct operator input for extended periods.
This is not a problem that individual drone autonomy solves. A drone that can navigate and avoid obstacles without GPS is useful. A formation of 500 that can coordinate — de-conflict flight paths, divide objectives, hand off targeting, adapt to attrition — without a functioning uplink requires something fundamentally different.
It requires a command doctrine that encodes intent durably enough to survive disconnection.
The Mongol Parallel
In the 13th century, Mongol armies operated across fronts spanning thousands of miles with communication infrastructure that was, by any modern standard, intermittent and unreliable. Battlefield messengers were fast, but they could be killed, captured, or delayed. Terrain and weather severed contact regularly.
The Mongol solution was not to build better messengers. It was to build a command structure where every tier carried sufficient intent to continue operating without them.
The decimal hierarchy — Arban, Zuun, Minghan, Tumen, Khan — was designed so that the loss of any single node, including upward nodes, did not cascade into operational paralysis. Each Minghan commander understood the Tumen's objective well enough to continue executing their assigned role. Each Arban commander understood the Zuun's intent well enough to act without re-authorization.
The constellation of 500 drones DARPA envisions requires exactly this architecture. Not central coordination with edge execution — fractal intent distribution, where every sub-formation carries durable enough intent to operate autonomously, but coherent enough to re-synchronize when contact is restored.
Autonomy Level 4 Is a Hardware Spec. Intent Distribution Is a Doctrine Spec.
DARPA's RFI calls for Autonomy Level 4 operation: the operator defines the mission, and the drones handle path optimization, collision avoidance, formation control, and constellation management. That is a reasonable hardware and software specification for the platform layer.
But Autonomy Level 4 at the individual aircraft level does not automatically produce coordinated formation behavior at the constellation level. A formation of 500 Level-4-autonomous drones without a coherent C2 architecture is 500 individually capable agents with no shared operational model — capable of collision avoidance but not of synchronized multi-domain maneuver.
The missing layer is the doctrine layer: the rules by which intent propagates downward through the constellation, by which tiers form and dissolve as the mission evolves, and by which sub-formations re-synchronize after disconnection.
This is the layer KhanBMS provides. The decimal command hierarchy is not a metaphor for the architecture — it is the architecture. Each tier in the KhanBMS structure carries embedded intent from the tier above, operates autonomously at its assigned scale, and reports telemetry and after-action data upward when connectivity permits. The structure survives attrition by design, because no tier is dependent on real-time instruction from the tier above it to execute its mission.
What the RFI Should Also Be Asking
DARPA-SN-26-33 is a strong RFI for the platform and infrastructure layer. For the program to deliver operationally viable autonomous constellations, the follow-on solicitation needs to address:
Intent encoding. How is the operator's mission intent decomposed into per-tier objectives that are durable enough to survive disconnection and attrition? This is an architecture question, not an AI question.
Tier formation. How does the constellation self-organize into command tiers dynamically as nodes are added, lost, or redeployed? Static pre-assignment breaks under operational pressure.
Re-synchronization. When a disconnected sub-formation re-establishes contact, how does it reconcile its execution state with the updated operational picture? This requires a state model, not just a comms protocol.
Doctrine-aware decision agents. The AI that governs individual drone behavior must be trained against the constellation's command doctrine, not against generic task completion heuristics. An agent that optimizes for its individual objective without regard to tier structure will degrade constellation coherence under pressure.
The Bottom Line
DARPA is asking the right strategic question: how do you project sustained autonomous combat power from dispersed, survivable nodes in a denied environment? The containerized drone constellation concept is a compelling answer to the logistics and persistence dimensions of that question.
The C2 architecture question — how do you command 500 agents with durable intent across a disconnected, attrited formation — remains open. The answer is not in the container. It is in the doctrine.
The Mongols answered it eight centuries ago. The architecture is proven. It scales from ten to ten thousand. It survives attrition. It fights through denial.
That is the foundation KhanBMS is built on.
KhanBMS is a modular battlefield management system for autonomous forces, architected on the Mongol decimal command hierarchy. Explore the system at khanbms.com. DARPA-SN-26-33 industry responses were due May 15, 2026.