7.10.4 — Autonomous Military Systems — maturity: live

AI-Directed Counter-UAS

Using satellite-cued, AI-driven sensor fusion to detect, classify and cue effectors against hostile unmanned aerial systems at operationally relevant timescales.

Auto-drafted reference page — content reviewed for shape, individual references not yet link-checked.

Small UAS are the defining tactical threat of the current decade: cheap, proliferating and invisible to legacy air-defence radars optimised for fast jets. A nation relying on rented commercial space data to feed its counter-UAS kill chain surrenders both the detection latency advantage and the ability to tune AI classifiers against its specific threat set — the adversary's exact drone signatures, flight profiles and electronic emission patterns. Without sovereign collection, the algorithm that decides what to shoot is trained on someone else's threat library.

A purpose-built LEO constellation combines wide-area RF survey payloads to detect UAS control-link emissions, multispectral EO for daytime optical cueing and onboard edge-inference to compress detection-to-tip latency to under two minutes from tasking. The space segment does not engage; it performs persistent wide-area surveillance and feeds a national AI fusion engine that correlates space, airborne and ground radar tracks into a common operational picture, then passes prioritised cue packages to ground-based effectors — jammers, directed energy weapons or interceptor drones. The AI classifiers run on sovereign compute, trained continuously on operationally collected data that never leaves national infrastructure.

The operational payoff is scalability: a 16-satellite walker provides near-continuous revisit over a nation's full sovereign territory and maritime approaches, something no ground radar network achieves economically for low-observable targets at low altitude. When drone swarms are used as a saturation tactic — as seen in Ukrainian, Red Sea and Nagorno-Karabakh engagements — the satellite layer is the only sensor that can simultaneously track the launch origin, transit corridor and terminal approach without geographic gaps. Commanders get prosecution authority with confidence; political leadership retains escalation control because the decision chain is entirely national.

What matters

Detection latency under 90 seconds from UAS emission to ground cue is operationally meaningful; rented data pipelines cannot guarantee it.
AI classifiers trained on a sovereign threat library will out-perform generic commercial models against a nation's specific adversary drone inventory.
Satellite RF survey is the only sensor that covers the full kill chain — launch site, transit, terminal approach — without ground radar line-of-sight constraints.
Export controls on counter-UAS AI (US EAR, EU dual-use regulation) mean a purchased service can be legally suspended at the worst possible moment.

Sovereignty score: 9/10 — The algorithm that decides what to shoot at must be owned, trained and operated by the nation doing the shooting — renting that capability from a foreign provider is an operational and constitutional impossibility.

Rules of engagement and targeting AI cannot legally or ethically be delegated to a foreign commercial operator; sovereign control of the kill-chain decision layer is a non-negotiable legal requirement under international humanitarian law.
US Export Administration Regulations and EU dual-use controls classify counter-UAS sensor fusion systems as munitions-adjacent; a commercial provider can be compelled to suspend service during precisely the crisis in which the capability is most needed.
Threat classifier performance is directly proportional to access to operationally collected signature data; a foreign vendor's model, trained on a different adversary set, degrades exactly when the national threat is most specialised and contested.
A sovereign satellite layer provides escalation control — the national command authority can surge revisit, redefine engagement zones and deny data to allies without negotiating with a commercial SLA or a foreign government.

Reference architecture

Payload: Dual-payload microsatellite: (1) RF survey receiver, 400 MHz to 6 GHz covering primary UAS control-link and video-downlink bands, 800m geolocation CEP at 500km altitude; (2) multispectral EO imager, 3m GSD, 20km swath, MWIR channel for hot exhaust plume detection
Bus class: ESPA-class microsat, 150kg dry, 450W total power, 300W payload allocation, dual-polarisation patch antenna array for RF survey, gimballed EO telescope on nadir face
Orbit: Sun-synchronous LEO at 480-520km, 16-satellite Walker Delta constellation, 55-degree inclination variant available for higher-latitude coverage, 70-minute average revisit over any point in national territory
Ground segment: 2-station national ground network (X-band downlink at 150 Mbps, S-band TT&C); airgapped mission-planning terminal at national air-defence headquarters; no commercial ground station use for operationally sensitive passes
Software & data
Latency
Cost band
Lead time

References

NATO Counter-UAS: A Comprehensive Approach — NATO formally recognises UAS as a top-tier asymmetric threat and calls for layered detection, identification and defeat capabilities. The document stresses interoperability and national ownership of critical sensor nodes.
HawkEye 360 RF Geolocation Technology Overview — HawkEye 360 demonstrates that clustered LEO smallsats can geolocate RF emitters, including drone control links, to sub-kilometre accuracy. Their commercial architecture is the direct technical antecedent for a sovereign RF survey layer.
CISA Cybersecurity and Infrastructure Security: Counter-UAS Authorities and Roles — CISA catalogues the legal and operational frameworks for domestic counter-UAS, emphasising that detection data and AI decision logic must remain under national authority to comply with civil liberties constraints and rules of engagement.
RAND Corporation — Countering Small Unmanned Aerial Systems — RAND finds that no single sensor mode is sufficient for reliable UAS defeat and that fusion of RF, EO and radar cues reduces false-positive rates by an order of magnitude. Satellite-derived wide-area cueing is identified as a critical gap in current national architectures.
ESA Space for Defence and Security Programme — ESA's defence programme explicitly funds sovereign dual-use space capabilities including surveillance and AI-enabled cueing, providing a European route to non-US-controlled smallsat buses and payload supply chains.