REQUIEM

Mission

REQUIEM detects, classifies and attacks emitters. When it meets a signal it has never seen, it writes a new technique in flight.

It infers a model of the receiver on the other end, synthesizes candidate techniques against that model, evaluates them on its own compute before radiating, applies the best one, and scores it by measuring what the adversary actually does. Effective techniques promote into the local library and propagate across REVENANT MESH in forty seconds.

Then it hash-chains every step of that derivation into REVENANT LEDGER. That is the product. Cognitive electronic warfare has existed as a research topic for fifteen years and has not fielded, because a technique generated by a learned system in flight cannot be released by an authority that cannot reconstruct how it was made. We made the reconstruction the deliverable.

240ms Detect to technique, library match

11s Synthesis of a technique never seen before

40s Fleet propagation over REVENANT MESH

41kW Ground node effective radiated power

22min Ground node emplacement, 2 people

Specification

Airborne pod and ground node

-A goes forward in a WRAITH mission bay. -G sits on the flank and listens. They share one technique library and one ledger.

REQUIEM-A — airborne pod
Weight	1,180 lb
Dimensions	14.1 ft × 16 in diameter
Carriage	WRAITH mission bay, single station (module EW-1), or any 30 in lug pylon on a Group 5
Transmit coverage	0.4 – 18 GHz continuous
Receive coverage	0.1 – 40 GHz
Effective radiated power	4.2 kW peak, 6 – 18 GHz
Apertures	2 × 512-element GaN AESA, ±70° electronic steering each
Simultaneous transmit beams	8
Tracked emitters	24
Concurrent techniques	6
Prime power	14 kW at 270 VDC
Heat rejection	9 kW liquid, to a ram-air / fuel exchanger
Latency, library match	240 ms detect → classify → apply
Latency, synthesized technique	11 s
Unit cost at rate	$6.4M

REQUIEM-G — ground node
Form factor	20 ft ISO half-height flatrack
Weight	6,900 lb
Emplacement	2 people, 22 min from arrival to radiating
Transmit coverage	1.2 – 18 GHz
Receive coverage	0.03 – 40 GHz
Effective radiated power	41 kW peak
Apertures	2,048-element GaN AESA, 4 faces on a 12 ft mast, 360° azimuth
Prime power	62 kW — one 60 kW tactical generator or site grid
Battery endurance	4 h at 30% transmit duty
Direction finding	1.1° RMS bearing
Networked geolocation	40 m at 30 km with 3 nodes on a cooperative MESH baseline
Secondary roles	REVENANT MESH bearer; GPS spoof and denial detection, with geolocation of the source
Unit cost at rate	$3.1M

The cognitive loop

Seven steps, all of them signed

Six of these steps exist in every cognitive EW research programme. The seventh is why this one can be fielded.

Observe. The receive chain captures a signal that matches nothing in the onboard library — a new pulse-repetition pattern, a new agility scheme, a new waveform.
Infer. REQUIEM builds a model of the receiver most likely to be on the other end from the emission's structure: bandwidth, coherent processing interval, likely sidelobe behavior, likely constant-false-alarm-rate design.
Synthesize. It generates candidate techniques against that inferred model and evaluates them in a digital twin on the pod's own compute, before radiating anything.
Apply. The best candidate is transmitted.
Measure. The adversary's behavioral response is the score. Did the emitter change PRF, change frequency, drop the track, go silent?
Promote or discard. An effective technique enters the local library and propagates to every other REQUIEM node over REVENANT MESH within forty seconds.
Sign. Every one of steps one through six — the raw observation, the inferred receiver model, each candidate, the twin evaluation, the transmitted waveform, the measured response, the promotion decision — is hash-chained into REVENANT LEDGER with timestamps and node identity.

Concept of operations

Listen first, radiate last

Every watt of transmit is a decision. Most of the work happens before one is spent.

Four phases

Survey

REQUIEM-G nodes emplace on the flanks of the operating area and listen. Three networked nodes geolocate emitters to 40 m at 30 km over a cooperative MESH baseline without radiating a watt.
Prepare

The observed order of battle is compared against the technique library. Emitters with no matching technique are flagged and, where the collected signal is rich enough, pre-synthesized on the ground — where compute and time are cheap and nobody is being shot at.
Attack

REQUIEM-A goes forward in a WRAITH mission bay and generates in flight against anything the ground survey did not see. Techniques discovered airborne propagate back to the ground nodes in forty seconds.
Release

After the sortie the LEDGER export goes to the release authority as a signed derivation. Techniques validated in combat become releasable library entries for the whole force, instead of an anecdote in somebody's debrief.

Engineering rationale

Why it is built this way

Three decisions, and the price we agreed to pay for each one.

Evaluate in a twin before radiating

A candidate technique is tested against the inferred receiver model on the pod's own compute before a single watt goes out. That avoids broadcasting a technique which reveals capability and achieves nothing. The tradeoff: it costs 11 s from first observation to first transmission, against 240 ms for a library match. In a pop-up SAM engagement 11 s is a long time, and we would rather lose the eleven seconds than teach the adversary what we can synthesize.

Behavioral scoring, not signal scoring

The only honest measure of an electronic attack is what the adversary does about it. REQUIEM scores on emitter behavior — PRF change, frequency change, track drop, silence — rather than on a modeled jam-to-signal ratio. The tradeoff: this is a permanent limitation, not a defect. Against a receiver that does not visibly change behavior under attack there is no scoring signal, and REQUIEM falls back to library techniques.

The ground node earns its transport slot

A 41 kW jammer that is only allowed to transmit on some days is a bad deal on a strategic airlifter. REQUIEM-G is also a MESH bearer and a GPS-spoofing detector and geolocator, so it does useful work every day. The tradeoff: those secondary roles cost about 700 lb of receiver, processing and antenna that a pure jammer would not carry, and they are the reason the node is a 20 ft flatrack rather than a trailer.

The design decision

The deliverable is releasability

Generating a technique in flight is the part of cognitive electronic warfare that has been demonstrated for fifteen years. Getting one authorized is the part that has not. Our differentiator is not detection sensitivity and it is not effective radiated power.

A REQUIEM technique arrives at the release authority with a signed, replayable derivation attached — the observation that prompted it, the receiver model that was inferred, the candidates that were rejected, and the effect that was measured, all hash-chained into REVENANT LEDGER. Synthesis of a technique never seen before takes 11 seconds; propagation across the fleet over REVENANT MESH takes 40. Without the derivation, a novel in-flight technique cannot be certified, cannot be shared across the force, and in practice will not be authorized — so the record is the product and the waveform is the by-product. Secondary: REQUIEM-G doubles as a mesh bearer and a GPS-spoofing geolocator, so the node earns its keep on the days nobody clears it to transmit.

Carried by Wraith

Program candor

Open engineering risks

One of these is a permanent property of the approach, not a bug we intend to fix. One of them we have no solution for at all.

Four of them

Two modes, chosen by the emitter. Against an adaptive receiver the cognitive loop scores an observable response and generates against it. Against one that does not change behaviour there is nothing to score — and nothing to generate, because a fixed emitter is already solved: the library answers it in 240 ms rather than 11 s, and the technique is deterministic and releasable. The library is the correct mode for that case, not a degraded one. What is genuinely constrained is a novel emitter that also refuses to react; that one gets library coverage until a human analyst adds it.
Blue-force fratricide. A 41 kW ground node will jam our own datalinks unless the MESH schedule and the EW schedule share a time reference. They do. If that reference is lost, the node automatically drops to a conservative mode 6 dB down with a reduced band plan — which costs effect in order to protect our own communications.
Thermal. At 45 °C ambient the airborne pod is duty-cycle limited to 60% in the high band. The fix is a larger exchanger costing about 40 lb, and 40 lb of the WRAITH bay budget is a real price we have not yet agreed to pay.
Spectrum clearance for training. Realistic training requires transmitting on frequencies peacetime spectrum management will not always clear, which limits how much operator proficiency can be built before a conflict. This is an industry-wide problem and we do not have a solution for it.

All air systems

Wraith Carries REQUIEM-A internally as bay module EW-1.

Revenant LEDGER makes the technique releasable. MESH spreads it in 40 s.

Poltergeist The passive-RF nose that services what REQUIEM geolocates.

Electromagnetic spectrum operations

Bring us an emitter we cannot classify

Technique libraries, release packages and blue-force deconfliction are discussed in a cleared facility. Start here and we will arrange it.

Request a briefing

Every figure on this page is a design target derived from first-principles analysis, not a measured result. No REQUIEM payload has been built and no technique has been generated in flight. Where a number here is later measured and comes out worse, we will publish the measured number. No export-controlled information is published on this site.

Key figures

Survey

Prepare

Attack

Release

Evaluate in a twin before radiating

Behavioral scoring, not signal scoring

The ground node earns its transport slot

Bring us an emitter we cannot classify