How Japan's Coast Guard Got an Autonomous Boat | Reading the Procurement Record

Hello, this is Ryuta Hamamoto from TIMEWELL.

On 24 August 2026, the Nikkei ran a short piece: "Yanmar delivers unmanned vessel to Japan Coast Guard, first domestic automation from departure through navigation"¹. The opening lines read: "Yanmar Holdings has developed technology, a first in Japan, that automates a vessel from departure through offshore navigation, and has begun demonstrating it on a Japan Coast Guard unmanned boat. It is intended for gas detection work at maritime accidents, improving both efficiency and safety."

Brief as it was, the piece caught me. Unmanned boats are not new. So what exactly qualifies as a first? And which Coast Guard vessel is this?

Following the thread led to a single small boat the Coast Guard has been building for three years. Its name is Albatross, and at 3.8 metres and 0.3 tons it is roughly the size of a car. Better still, the Coast Guard's published procurement records let me trace the whole thing down to contract values and the number of bidders. This article walks through what the boat is for, how it was procured, and what technology holds it together. No maritime background required.

What is hard about "departure through navigation"

Start with the claim itself.

Boats that run without a crew are not new. Remotely piloted craft have existed for decades, and survey vessels that follow a preset course are in commercial service. What is being called a domestic first is joining the whole cycle together: leaving the quay, running offshore, and coming back alongside, with nobody touching the controls.

The difficulty is that each segment demands wildly different precision. Offshore navigation is actually the easy part. There is little to hit, and drifting off course by a few metres does not immediately cause an accident. Docking is another matter. Thirty centimetres of error puts the hull into the quay wall, and wind and current change by the second. Picture parallel parking a car while the ground itself keeps sliding.

Boats also have no brakes. Stopping means reversing thrust, and there is a lag before it bites. A human helmsman reads that lag from experience and works the rudder and throttle around it. Getting a machine to do the same requires control that predicts how the hull will behave rather than merely reacting.

"Departure through navigation" sounds like a modest phrase, but it means the hardest segment and the easiest one were connected without a person in between. If a human takes over for docking, half the value evaporates. The entire point is keeping people away from the scene, and that fails if someone has to be aboard for departure and arrival anyway.

What Albatross is actually for

Which Coast Guard boat is this? The answer sits in a Coast Guard notice dated 6 April 2026².

The hazardous material detection unmanned vessel Albatross, in development since fiscal 2023. It is 3.8 metres long, 1.9 metres wide, displaces 0.3 tons and makes over 5 knots. The name has a nice origin: it comes from the albatross used as the motif on the insignia of the Coast Guard's hazardous-materials response unit.

The purpose makes the boat make sense. When hazardous or toxic substances spill at sea, Coast Guard officers have until now gone to the scene in chemical protective suits and breathing apparatus, carrying gas detectors by hand. A vessel has capsized, flammable or toxic gas is escaping, and people go stand next to it. You have to approach before you know what is leaking, in order to find out what is leaking. That circularity is the cruel part of the job.

The unmanned boat takes that first step instead. Send a small craft carrying detection equipment ahead, establish remotely what is escaping and in what concentration, then decide how people move. Whether to suit up, whether to approach at all, becomes a decision made from data.

Testing by the hazardous-materials response unit began in fiscal 2026 in the waters off the Yokohama Maritime Disaster Prevention Base. Three items are listed: speed from land transport to launch, accuracy of the detection equipment, and function of the remote control system. On 30 June 2026 the boat was reportedly used in a live exercise simulating spills of substances including LNG and hydrogen³.

That first test item deserves attention. Speed from land transport to launch. Accidents do not wait where you keep your equipment, so the boat goes by truck to somewhere near the scene, and the clock that matters is how many minutes until it is in the water and moving. The automated departure the Nikkei reported bears directly on that number. Every step that no longer needs a person aboard is time recovered.

What the procurement record says

Here is the part I enjoyed most. The Coast Guard's own disclosures let you reconstruct how this boat came to exist.

Start with fiscal 2023. A tender titled "design work for hazardous material detection unmanned vessel plus two items, manufacture" was opened on 31 August 2023⁴. Yanmar Power Technology won it. Contract value ¥110 million against a ceiling price of ¥119 million, with delivery due 22 March 2024.

The result sheet includes a column listing the companies that bid. There was one name in it. Yanmar Power Technology bid ¥100 million excluding tax in the first round and took the contract.

Then fiscal 2025. A contract titled "development and manufacture of an unmanned control system for hazardous material detection" was signed on 25 June 2025⁵. Same counterparty: Yanmar Power Technology (Chayamachi, Kita-ku, Osaka; corporate number 7120001071567). It was classified as government procurement, meaning an open tender large enough to fall under the WTO Government Procurement Agreement. Against a ceiling price of ¥200.6 million the contract came in at ¥200.42 million, a ratio of 99.91 percent.

¥110 million for design, ¥200.42 million for development and manufacture. Around ¥310 million all in. Set against a patrol vessel that costs over ¥10 billion to build, this is a very small line item in national equipment spending.

A 99.91 percent ratio means the contract landed just under the ceiling. Numbers like that make some readers suspicious. I would not read this one as evidence of anything murky. As the single bid in 2023 shows, there is barely a field of competitors here. The number of Japanese firms that build their own marine engines, mould their own hulls, and hold the automatic docking control stack in-house can be counted on one hand. When one company can meet the specification, prices sit near the ceiling. Competition did not fail; the pool of possible competitors is simply tiny.

As an aside, the same fiscal 2025 disclosure shows Yanmar Power Technology also supplying the Coast Guard with a 560 kVA diesel generator set and 1,417 kW diesel engines. This is a company that has long put engines into patrol vessels. Not a newcomer that appeared from nowhere, but an incumbent machinery supplier that became the vehicle for automation.

One practical note: from fiscal 2025, Coast Guard tender notices moved to the government procurement portal (GEPS). Individual bid results are consequently harder to find on the Coast Guard's own site, and I only located this contract on page 22 of the annual disclosure summary. For anyone tracking procurement transparency, the work got slightly heavier.

The technology underneath

Now the engineering, in plain terms.

At the centre sits RTK-GNSS. GNSS is the general term for satellite positioning, the technology in your car's navigation system. Car navigation is accurate to a few metres, which is nowhere near good enough to put a boat against a quay. RTK, real-time kinematic, layers on top: a ground reference station at a precisely known position continuously measures the error in the satellite signal and broadcasts a correction, which the vessel applies to sharpen its own fix dramatically. In its 2019 release, Yanmar describes using RTK-GNSS that combines satellite positioning with correction data from a relay unit of its own design⁶.

Accurate positioning alone will not dock a boat. You need a path. The same release describes generating a docking trajectory that combines a target position with a target heading toward the designated berth. Not merely steering to a coordinate, but deciding what orientation to arrive in. A boat has to come alongside parallel to the quay, so attitude control is baked into the path itself.

The hull uses Yanmar Shipbuilding's GRP moulding, glass-reinforced plastic, a material suited to small low-cost production and consistent with a 3.8-metre boat. The control software runs on a platform built around middleware. Middleware is the common layer between hardware such as sensors and motors and the control programs above; with it in place, swapping a component does not force a rewrite of everything above. It is standard practice in robotics.

The lineage matters too. The 2019 robotic boat came out of SIP's "Next-generation Technology for Ocean Resources Exploration", the Zipangu-in-the-Ocean project, and served as a surface relay platform for JAMSTEC's seabed resource surveys, linking underwater vehicles to the surface. In other words, technology that spent seven years being hardened in civilian ocean survey work moved across into Coast Guard disaster response. A research prototype did not simply get adopted into government service.

I should note what I could not verify. Docking accuracy in centimetres, endurance in hours: numbers for both circulate online, but I could not trace them to a primary source, so they are not in this article.

Why it worked, and where it goes

So why did this succeed?

Three reasons, I think. First, the technology matured in civilian use. Automatic docking was originally developed for fisheries and leisure boating, industries squeezed by labour shortages. It was not built from scratch for a government requirement, and because something working already existed, the project could start at ¥110 million for design.

Second, the use case was narrow. Hazardous material detection: dangerous for humans, and genuinely substitutable. Trying to build an all-purpose unmanned boat inflates the specification until the project collapses under it. Narrowing the job to taking the first step for a suited-up officer is what made a 3.8-metre boat sufficient.

Third, the state showed up as a buyer first. When no market exists, a company has no path to recovering development costs. Here the Coast Guard commissioned design, then two years later commissioned development and manufacture. From the company's side, the exit for the investment was secured from the beginning.

That third point is now being written into policy. Japan's Growth Strategy, approved by Cabinet on 21 July 2026, puts marine unmanned vehicles at the head of its ocean sector, with ¥1.2 trillion in public and private investment and ¥9.4 trillion in economic ripple effect through fiscal 2040⁷. The stated goal is a 30 percent share of the world market. Among the measures: "work toward establishing a system to promote market formation and expansion through anchor tenancy". Anchor tenancy means the government acting as the first major customer to bring a market into being. What happened with Albatross has now been given a name and an institutional home. I went through the strategy in full in a separate article on the 17 sectors and 62 technologies.

The same strategy contains a sentence worth pausing on. On marine unmanned vehicles: "the importance of unmanned assets in the security field, and their importance as dual-use technology, are both increasing." Dual-use means usable for civilian and military purposes alike.

For component makers, that is not somebody else's problem. A docking controller or a precision positioning module is, on its own, a civilian part destined for a fishing boat. Built into an unmanned platform, depending on the end use, it can touch export control. Without visibility into where your part travels after it leaves through an integrator, you can find yourself inside a transaction that required classification you never performed. If you want a quick read on how your products are treated, our three-minute export compliance check is a reasonable starting point, and TRAFEED supports classification against METI criteria and counterparty screening, though final classification always rests with your own export control officer. On the security side of unmanned systems, the Aerospace Self-Defense Force reorganisation covers adjacent ground.

In summary

• Albatross is a hazardous material detection unmanned vessel: 3.8 metres, 0.3 tons, over 5 knots, in development since fiscal 2023 and under test since fiscal 2026 off the Yokohama Maritime Disaster Prevention Base
• The name comes from the albatross on the insignia of the Coast Guard's hazardous-materials response unit
• It replaces work previously done by officers in chemical protective suits carrying gas detectors to the scene
• Two contracts produced it: ¥110 million for design in August 2023 and ¥200.42 million for development and manufacture in June 2025, both to Yanmar Power Technology, with a single bidder in the 2023 tender
• The core technology is RTK-GNSS positioning plus a docking trajectory combining target position and heading, descended from the 2019 robotic boat and JAMSTEC ocean survey work
• Marine unmanned vehicles are one of the 62 priority technologies in the Growth Strategy, named explicitly as dual-use

What stayed with me was how small ¥310 million is. National unmanned capability calls to mind enormous defence budgets, and what actually changed this particular job was a project that took civilian technology, narrowed the mission, and put it on a car-sized boat. Unglamorous. But if it means nobody has to stand beside leaking toxic gas in a protective suit, that is a real result.

If your technology is heading toward government demand or an unmanned systems supply chain and you want to start with export control, talk to us directly.

---

Footnotes

1. Nikkei, "Yanmar delivers unmanned vessel to Japan Coast Guard, first domestic automation from departure through navigation" (24 August 2026) https://www.nikkei.com/article/DGXZQOUF084VS0Y6A700C2000000/ — the body is behind a paywall, so this article draws only on the published opening lines ↩

2. Japan Coast Guard, "Verification begins toward operation of the hazardous material detection unmanned vessel" (6 April 2026) https://www.kaiho.mlit.go.jp/info/kouhou/post-1299.html ↩

3. Nippon TV News NNN, report on the unmanned vessel used in a hazardous substance response exercise (30 June 2026) ↩

4. Japan Coast Guard, bid result for "design work for hazardous material detection unmanned vessel plus two items, manufacture" https://www.kaiho.mlit.go.jp/ope/nyusatsu/r5/20230831tokutyou1168kekka.pdf ↩

5. Japan Coast Guard, disclosure of competitive tender information for goods and services, fiscal 2025, page 22 https://www.kaiho.mlit.go.jp/ope/nyusatsu/r8/ippankyousou202600409.pdf ↩

6. Yanmar Holdings, "Development of core technology for an auto-navigation robotic boat and an automatic docking system" (17 January 2019) https://prtimes.jp/main/html/rd/p/000000025.000034384.html ↩

7. Cabinet Secretariat, "Japan Growth Strategy" and the Public-Private Investment Roadmap (21 July 2026) https://www.cas.go.jp/jp/seisaku/nipponseichosenryaku/pdf/jgs2026.pdf ↩