STYX

Mission

Banshee

Put a real magazine and a real sensor on a hull with nobody aboard, and send it into the water a crewed combatant cannot afford to enter.

The reason to take the crew off is not autonomy for its own sake. It is that the crew is what makes a warship expensive to risk. A frigate is not held back because the hull is precious; it is held back because two hundred people are aboard. Take them off and the calculation changes: STYX can hold a chokepoint inside an adversary's weapons envelope, because the worst case is a hull on the seabed and an insurance line, not a casualty list. Attrition becomes an option rather than a catastrophe.

There is also a systems reason this ship exists that has nothing to do with weapons. In the maritime domain the surface vessel is the gateway — the only node that can hold a satellite link, an aircraft datalink and an acoustic link to something submerged at the same moment. Without a surface vessel, an undersea picture stays undersea. STYX is the boat that carries the sea domain's data across the surface, and while it is doing that, it is also a combatant.

10t Modular payload bay — effectors, sensors or stores. Design target.

0 People aboard — no accommodation, no helm, nothing to evacuate

60d Certified unattended endurance. Everything beyond it is tankage.

$11M Unit cost at rate, design target. A crewed Mk VI patrol boat is about $15M.

Parameters

STYX ASC-1 — principal characteristics (design targets)
Class	Autonomous surface combatant — uncrewed, armed
Length overall	61.0 ft / 18.60 m
Beam	17.1 ft / 5.20 m — welded steel semi-planing monohull, deep-V forward
Draft	3.9 ft / 1.20 m hull; 5.6 ft with the transducer trunk lowered
Lightship	18.5 t — welded steel hull, aluminium deckhouse
Full load	32.0 t — fuel 6.2 t, payload 10.0 t, stores 0.8 t
Modular payload	10.0 t bay amidships — effectors, sensors or stores, struck down pierside from a flat-rack
Vertical launch	8 compact cells in the payload bay — strike and point air defence
Gun	1 × 30 mm remote weapon station, stabilised, forward
Undersea	2 × lightweight torpedo tubes, one per side
Sprint speed	40.0 kt on 2 × 1,000 kW commercial marine diesels, waterjets
Loiter speed	8–12 kt diesel-electric, mains declutched and cold
Silent approach	5 kt for 60 min on a 260 kWh LFP pack
Fuel-limited range	2,000 nm at 25 kt
Certified unattended endurance	60 days — set by machinery run certification, not by tankage
Sea state, transit	SS4 at any speed; SS5 at 12 kt
Sea state, survival	SS7 / 9 m significant wave height, unattended autonomous heave-to
Weapon release	Gated by a LEDGER-signed authorisation carried aboard, scoped to an engagement envelope. No network required to fire; no network can create permission.
Survivability	Two-compartment flooding, machinery duplicated on separate buses, fixed suppression with no manual fallback
Sensors	360° panoramic EO/IR (8 × 5 MP visible + 4 × 640 LWIR, 1.1 mrad), solid-state X-band nav radar 0.8 m, ESM, AIS, two hull hydrophones
Comms	REVENANT MESH: line-of-sight optical 40 Mbps at 14 km with zero RF emission; HF and UHF fallback; acoustic modem to submerged nodes; REVENANT LINK to Link 16, TTNT and IBCS
Autonomy	REVENANT CORE plus the declarative COLREGS engine
Crew	0 — no accommodation, no galley, no head, no helm station
Delivery transit	Autonomous under a filed voyage plan; no delivery crew, no chase boat
Boarding provision	Recovery harness and a towing eye. The boat is not designed to be gone aboard at sea.
Unit cost target	$11M at rate

Mission deck capacity
BANSHEE-M / -R	8 — six in the magazine, two in deck cells
SHADE-M / -R	24 in two 12-round tube launchers
POLTERGEIST-150	4 canisters in deck cells
TIDEWALL nodes	12-node laying spread, over-stern chute
Launch	1 × 2.4 t uncrewed surface tender and davit — laying, recovery and inspection
Autonomous weapon station	1 × 30 mm, bow cell — cued by the boat, released under LEDGER-signed authorization
Reconfiguration	Under 4 h pierside on the rail grid

Fighting a ship with nobody aboard

01 · The problem

Every weapon on a crewed warship ultimately terminates in a person who can be asked whether they are sure. Remove the crew and that person is somewhere else, on the other end of a link that an adversary is actively trying to sever. The hard problem in an uncrewed combatant is not steering it. It is the release of a weapon.

STYX does not solve this by keeping a human in the loop, because a link that can be jammed is not a control measure — it is a single point of failure that happens to be reassuring.

02 · What we do instead

Weapon release is gated by a LEDGER-signed authorisation carried aboard before the vessel sails, scoped to an engagement envelope: which target classes, inside which geometry, for how long. The node verifies the signature against a key it already holds. No network is required to fire, and no replayed message can create permission that was not granted.

Outside that envelope STYX does not engage, and it does not ask. It records what it saw, why it declined, and continues. The authorisation, the identity behind it, and every release decision are written into the same hash-chained record the rest of the system uses — verifiable offline, afterwards, by someone who does not trust us.

03 · Damage control without people

A crewed hull survives damage because people fight it. STYX has to survive it structurally: watertight subdivision sized so any two adjacent compartments can flood without loss, machinery duplicated port and starboard on separate buses, and a fixed suppression system with no manual fallback because there is no manual.

The honest limit: a hull with nobody aboard cannot shore a bulkhead, patch a hole or restart a flooded switchboard. Past a certain damage state STYX is lost where a crewed vessel would be saved. That is the trade the price buys.

COLREGS is a legal problem, not a path-planning problem

Collision avoidance is not path planning. It is a legal problem with a path-planning component, and treating it as the former is why unmanned surface vessels are still shadowed by a chase boat with a licensed mariner aboard.

Rules 13 to 17 — overtaking, head-on, crossing, give-way action, stand-on action — are geometry, and geometry compiles. Rules 2, 5, 6, 7, 8 and 19 are judgement, and judgement is what gets litigated. The worst of them is Rule 17, which requires the stand-on vessel to act when it becomes apparent that the give-way vessel is not taking appropriate action. That is a machine deciding a human mariner is failing, and then overriding him. There is no way to write that rule without writing an opinion.

So we sign the opinion. The COLREGS engine is a declarative rule set inside REVENANT CORE, and every avoidance manoeuvre is LEDGER-signed together with the contact picture that justified it. After an incident the investigator receives a signed, replayable record: which contacts were held, at what CPA and TCPA, which rule the engine believed governed, the moment it concluded the give-way vessel was not acting, what it did, and what the alternatives it rejected would have cost. No other USV can produce that record, and it is the only way a flag state, an insurer or an admiralty court can treat an unmanned vessel as anything other than an unattributable hazard.

Autonomous navigation — behaviour and acceptance
Rule 5 — look-out	Panoramic EO/IR detects an unlit 6 m hull at 2.9 km and a floating container at 900 m. Radar holds a 5 m² target at 8 nm in SS3. And two hull hydrophones — Rule 5 says "by sight and hearing", and we hold ourselves to both halves of it. STYX hears a fog signal and an approaching screw.
Rule 3(g) / Rule 18	While operating BANSHEE the vessel is restricted in her ability to manoeuvre, and displays the lights and day-shapes that say so. That is not decoration — under Rule 18 it obliges power-driven vessels to keep clear while the deck is hot.
Rule 8(b) — manoeuvre policy	Large and early: a 30° course change at 6 nm, not a 5° nudge at 2 nm. An alteration must be "readily apparent to another vessel observing visually or by radar." A mathematically optimal micro-correction is legally indefensible and illegible to the mariner it is aimed at. We give up efficiency to be readable.
AIS and running dark	AIS receives always and transmits by default. Going dark is a commanded state entered only on a LEDGER-signed authorization naming the human who ordered it and the window it is valid for.
CPA envelope	0.5 nm open ocean · 200 m in a traffic separation scheme · 60 m in a harbour approach at 6 kt or less
Acceptance gate	Zero COLREGS violations across 14,000 nm of witnessed trials in Class A and B traffic, adjudicated by licensed masters reviewing the signed record

CONOPS

Phase 01 · Sortie

Any pier, any well deck, under its own power

No launch equipment, no crane, no mother ship. The launch cells are struck down and reloaded pierside from a standard flat-rack, which means the boat that sailed with a strike loadout comes back configured for air defence without going near a shipyard.

Phase 02 · Transit

10 kt on the loiter drive, mains cold

AIS transmitting, lights and shapes correct, COLREGS engine holding a half-mile CPA envelope. 2,000 nm of fuel available; 60 days certified. The main engines are not started for a transit — running a 720 kW diesel at 26 horsepower destroys it, which is the entire reason the loiter drive exists.

Phase 03 · Barrier establish

Forty-mile spacing, one picture

Optical mesh to neighbours comes up at 40 Mbps with zero RF emission. Emissions go receive-only. An acoustic handshake reaches any LEVIATHAN or TIDEWALL field beneath, and the barrier becomes the surface end of the undersea network.

Phase 04 · Persistent watch

Radar, EO/IR, ESM, passive acoustic — fused and signed

Every contact goes into one barrier-wide picture, hash-chained at the edge. SHADE is launched for close-look identification of contacts the radar cannot resolve, at four pounds a copy.

Phase 05 · Engage

On a signed token, and any hull can shoot on any hull's track

BANSHEE against air and cruise threats, POLTERGEIST against surface and land targets. Engagement authority is a LEDGER capability token that is issued once and consumed once — there is no ambiguity about who authorized a shot and no mechanism by which two hulls service the same target twice. Unexpended BANSHEEs come home to the compensated deck and regenerate a sortie in nine minutes.

Phase 06 · Sustain

Any vessel with a five-tonne crane

Deck cells are replenished at sea from whatever is available, or the boat returns to any pier. There is no specialised tender in this concept, because a specialised tender is a coordinate and a single point of failure.

Phase 07 · Withdraw or scuttle

It is an $11M object

If the boat cannot be recovered, its data has already gone out over the mesh and the hull is scuttled on a signed authorization. Nothing about the campaign plan depends on getting any particular hull back.

Why it is built this way

The unit price is the requirement, and everything else is downstream of it. A steel monohull is heavier and slower than the aluminium hull a naval architect would prefer, and we chose it because steel is cheap, every commercial yard can weld it, and the whole argument collapses if these cost what a patrol boat costs. The engines are commercial marine diesels off a catalogue, not marinised gas turbines, for the same reason.

Everything else follows. The hybrid drivetrain exists because a 60-day boat cannot run its sprint engines at 4% load. Two widely separated shaft lines exist so a single-engine casualty is a limp-home rather than a tow. And the COLREGS engine is declarative and signed because an unmanned vessel that cannot explain itself after a collision will never be allowed out of sight of a chase boat.

The argument · mass

The cheapest thing on a warship is the steel. The expensive part is the crew — and the crew is what makes it too valuable to risk

A combatant you cannot afford to lose is a combatant you cannot afford to use. That is the actual constraint in a contested littoral: not magazine depth, not top speed, but whether a commander will send the hull inside the envelope at all. STYX is built so the answer is yes. No accommodation, no galley, no damage-control party, no helm — and therefore no reason to hold it back.

What that buys is numbers. One exquisite hull covers one place. Twelve STYX cover twelve, and an adversary who wants the chokepoint has to service every one of them with a weapon that costs more than the target. The arithmetic only works if the unit price stays low enough that we mean it, which is why the design is deliberately unremarkable: a steel monohull, commercial diesels, and no capability that exists to impress anyone.

What it gives up, stated plainly. STYX carries what it carries. When the cells are empty it has to withdraw and reload alongside — there is no regenerating magazine and no aircraft coming back to it. It cannot repair itself under way. And it is worse in heavy weather than a larger crewed hull, because seakeeping is bought with displacement we deliberately did not buy.

Where the program actually stands: STYX is in design. No hull has been laid, nothing has been fabricated, and nothing has been to sea. Every figure on this page is a design target or a calculation — not trials data — and we present it as such.

FORGE-2 and the outfitting quay

What is hard, stated plainly

Open engineering risks — STYX
Unattended machinery reliability	This is the whole program. A crewed boat has an engineer who notices a weeping seal at 0300. Sixty unattended days means the boat diagnoses itself — 340 sensed points and a vibration-signature model per rotating machine. First-year MTBCF target is 1,400 h. The mission needs 2,600 h. We are not there and we know exactly how far.
The compensated platform	2.1 t of hexapod exposed to green water, salt and thousands of cycles per deployment. The failure mode is graceful — it locks level and reverts to a fixed deck — but recovery then drops to SS2 and BANSHEE's reusability argument collapses with it.
Not self-righting	Stated above rather than engineered around.
Fouling	Sixty days at 6 kt in warm water costs about 22 % of speed by day 60 in the Gulf. Foul-release coating helps. A hull-crawler would fix it. We have not built the crawler.
Regulatory	Flag-state and classification treatment of a 48 t unmanned vessel is unsettled. We build to ABS autonomous-vessel guidance and DNV's ARV rules and fully expect both to change underneath us mid-program.
Weapon release policy	Whether an unmanned surface vessel may fire in peacetime waters is a policy question we do not get to answer. The LEDGER token architecture is built to make it answerable. That is not the same as answered.

Related systems

All sea systems

Banshee Eight rounds aboard. The nine-minute turnaround is measured on this deck, with nobody standing on it.

Shade Twenty-four in two tube launchers. Recovery into a soft catch cone is trivial at 4.1 lb.

Leviathan The thing on the other end of the acoustic link, 400 m down.

Government & Primes

Request a STYX briefing

Mission-deck interface control documents, the COLREGS engine's rule set and trial adjudication plan, and barrier-concept modelling are available to US government customers and allied ministries of defence under an executed agreement.

Request a briefing programs@fantom.aero

Every figure on this page is a design target derived from first-principles analysis, not a trial result. STYX is in detail design; sea trials are planned for 2029. No export-controlled information is published on this site.