Autonomous Ships | Hacker News Reader

Autonomous Ships

bloomberg.com

37 pointsprotomyth69 comments

I spent two years working on autonomous ships at NASA JPL, primary on developing deep learning algorithms for ship detection and classification to obey maritime traffic rules (COLREGS), avoid obstacles, help with decision making, etc.

Regarding piracy, autonomy would cut down on costs a lot. Based on my understanding, besides cutting down on life-support costs and having extra room for more cargo, the ships would also be able to eliminate the cost of the armed guards that many have kept on their ships since the rise of piracy. That's one of the main reasons why we don't hear about piracy anymore [0]. With no controls on the ship, pirates wouldn't be able to take control.

Note also that Rolls Royce is looking into drone ships (rather than full autonomy): https://www.rolls-royce.com/media/press-releases/yr-2016/pr-...

[0] http://www.marineinsight.com/marine-piracy-marine/18-anti-pi...

What would stop heists on the high seas? For example, I could imagine someone breaking into a container with "valuable" goods (remember a lot of these pirates come from impoverished backgrounds, so being able to flip cargo still net a nice profit).

Possible, but unloading cargo would be more involved. Also, without access to electronic manifests, I am not sure there's any easy way to know what's in a container, not to mention many of them would be stacked in such a way as to inaccessible. All of which is a lot harder than ransoming a crew. I suppose it might be possible for pirates just to threaten to sink the ship.

One would simply disable propulsion and communications, and then haul the ship elsewhere for classification and sale of its cargo.

With no humans on board, then there is no incentive not to use deadly force for disabling critical subsystems.

equivalently, there's no real incentive to not use deadly force to clean the ships of pirates as a deterrence and take the write off on any damages.

I suspect with drone ships and the lack of crew liability, there will arise contract security companies with pre-positioned ex-seal teams in hotspots to deal with piracy.

With access to electronic manifests, you could program the logistical system to deliver the goods to the destination of your choosing. Which might be easier than piracy.

Another thing are systems like the one done by ATG - planning better routes for ships with multi-agent models. For example the paper Vaněk, Ondřej, Ondřej Hrstka, and Michal Pěchouček. "Improving group transit schemes to minimize negative effects of maritime piracy." IEEE Transactions on Intelligent Transportation Systems 15.3 (2014): 1101-1112.

> With no controls on the ship, pirates wouldn't be able to take control.

It doesn't matter if there aren't any switches or touch screens. Once someone has access to the engine room and steering gear compartment, all bets are off. They can physically adjust valves to get the ship going where they want it to.

Even if they don't have any tools to gain entrance into the ship, they just have to block the air-intakes and the engines will shut down. If it's battery-powered, they just have to let the ship run over some appropriately-sized lines which will get caught in the propellor and seize the shaft.

Or just get onboard and start smashing stuff until the boat stops. It would be pretty hard to prevent a grapple hook from working (one could imagine automatic water cannons though). Best idea I've got is flexible rods sticking out from the perimeter of the ship to block the hook. The hook gets caught in the rods and when you pull it slides off the end.

It's possible to make doors that are extremely hard to cut through... but they can always just cut through the wall instead.

Drone ships sound interesting. The shipping company can definitely afford having a person remotely check in on millions of dollars of cargo every ~15 minutes or so, and so full autonomy seems like the 20 in 80/20 (maybe more like 95/5).

With no controls on the ship, pirates wouldn't be able to take control.

20 lbs of C4 in the control and /or engine room might just do it. So they don't take control (ship can be $urrendered later in pristine condition), they do worse.

>With no controls on the ship, pirates wouldn't be able to take control.

So they'll just figure out how to disable the ship. And then they have a big drifting fortress to squat on/ransom/move cargo off of.

Taking hostages is the only reason why pirates can (sometimes) ransom the ship instead of getting cleaned off by PMCs or even foreign militaries patrolling the waters.

A ship without hostages means that a pirate has a good chance of not getting home ever again - the patrols aren't allowed (rightly) to stop a "fisherman" boat in these waters before it's done anything, but once they are on board and making a ransom note, they'd be lucky if they survive to be arrested.

Former Marine Engineer here, I know there's at least one other HN users that's an ME. You might be able to automate navigation, but never maintenance. Something small that doesn't get fixed underway could cause a lot more damage than whatever crew costs would have been saved.

Things break and leak all the time, even on new ships. Automatic valves get stuck and things get clogged. I'm an air traffic controller now, and my break is over, but I'll try to expand more later.

Edit: I have a few minutes. A lot of people are talking about pirates only boarding ships to hold the crews hostage, and not to steal cargo. How about this: a pirate demands a crypto currency payment within X number of hours, or they sink the ship. It's very easy to sink a ship once you're inside.

> Things break and leak all the time, even on new ships.

> [...] a pirate demands a crypto currency payment within X number of hours, or they sink the ship

Based on your response and toomuchtodo's [1] it appears we're trying to automate from the wrong end - we should examine automation of defence/security and maintenance before stripping a ship of its crew.

[1] https://news.ycombinator.com/item?id=14363456

I will agree with you that steam or turbine power plants require continuous maintenance. I've got experience, too. :-)

But, read the article. The ship they're talking about is a short-haul transport with battery-electric propulsion. Perfect scenario to pioneer this concept. If it works, I'm betting that it would be possible to create a propulsion plant that will survive longer hauls without a crew.

You could fly a helicopter out to any ship that isn't crossing the pacific or atlantic, but that would be $100,000-$300,000 per incident and it would very quickly not be worth it. There isn't much in-between for reducing crew since a couple people alone on a boat together will probably be homicidal in short order.

I'm curious, what breaks on a ship? I have no experience with anything over a few dozen feet. I imagine hydraulic systems could break and lead to Consequences, but I can't imagine there are many big systems that can be fixed at sea, even with a crane onboard. What kind of things get fixed at sea?

You might as well ask what doesn't break on a ship. Salt water corrosion, thermal stress, vibration, residue accumulation, and hull flexing due to large swells are really hard on equipment. Here's a good case study on the type of repairs that a merchant ship crew might have to perform in the middle of the ocean.

http://www.brighthubengineering.com/marine-engines-machinery...

Ocean going merchant ships spend much of their time outside of practical helicopter range. And even when close to shore, winching a repair technician down to a disabled autonomous vessel that can't even keep its bow into the weather would be a dangerous affair.

By one consultant's estimate, moreover, carrying sailors accounts for 44 percent of a ship's costs.

I find this hard to believe even if autonomous ships can be done without a bridge. The Emma Maersk has 15,000 TEU and a crew of 13 http://www.emma-maersk.com/specification/

"Crew costs of $3,299 a day account for about 44 percent of total operating expenses for a large container ship, according to Moore Stephens LLP, an industry accountant and consultant."[1]

The Maersk triple E's have a crew of 13 and consume ~136 tonnes/day of bunker fuel[2] which costs currently ~$315/tonne[3], for a daily fuel cost of $42,840. So right off the bat that number is very suspect. Even for other ships like the CSCL Globe with a crew of 31, that's a 12x difference.

I don't understand how that estimate can possibly be true. Ship captains and engineers make ~80k annual and crewmen make under 40k. For a crew of 20-30, $3,299 would be basically bang on for a daily crew cost but there's no fucking way fuel and depreciation only cost $4,200 a day. Even a 4000 TEU ship (tiny) at 17 knots (snail) burns $15,000+ a day.[4]

[1] https://www.bloomberg.com/news/articles/2014-02-25/rolls-roy...

[2] http://www.scdigest.com/ontarget/13-09-12-1.php?cid=7401

[3] http://www.tsacarriers.org/calc_bunker.html

[4] https://people.hofstra.edu/geotrans/eng/ch8en/conc8en/fuel_c...

While I agree with your analysis, one counter-argument I can think of: Is the ship always using that much fuel?

I know a lot of ships spend days waiting to enter the panama canal, or days in port. Crew is certainly paid for those days as well. What percentage of the time is the ship fully underway and using fuel?

That 44% figure might be more reasonable if you're only burning fuel 1/3 of the time.

I did think of that, and also that you may need other shifts on retainer, and the normal food/insurance/fees/overhead of employing people across borders. Still though, for the triple E the expected fuel cost is 30x the expected crew cost.

The stretches to explain that gap would be ludicrous eg 3 shifts, each costing the company 3x more than their actual salary, and the engine being turned off 70% of the time. Even still that requires you to disregard depreciation, which for the triple E is at least $17,000 a day.

I suspect that the answer more likely lies in the direction of "there's a lot of ships smaller than the Emma Maersk".

Canals, port clearances (bridges, channel depth, width), dockside facilities, etc.., all put limits on ship sizes. There is a whole slew of "max" and other classifications: Panamax, Suezmax, Chinamax, Seawaymax, etc. "Handysize" is still my favourite of the lot.

https://en.m.wikipedia.org/wiki/Cargo_ship#Size_categories

For small and local transit, think the US Great Lakes, the Baltic Sea, English Channel, Mediterranean, South Pacific islands (many small populations and ports), and coastal transit along Africa, South America, and Asia, outside the major ports, there's still call for many much smaller ships.

As with other areas, automating smaller instances may be where the payoffs come as the relative labour costs are greater relative to total operating costs. Part of that comes from simple logistics:

* You need multiple shifts. * You need a helmsman, and an engineer, for each shift. * You need a cook, and other housekeeping.

That's already a crew of about 9 - 12.

Regardless of whether your DWT is 20k tonnes or 400k tonnes, your staffing actually doesn't change all that much, and may well increase as smaller ships frequently have (and I presume staff) their own offloading equipment.

What about food? What about living supplies? Also, dock living services? And other dock related expenses for crew? Don't forget, all this has to be insured on sea and on land.

A crew that has to be paid, fed, insured, provided habitable accommodations on the ship including sleep, cooking, and personal hygiene, and provided benefits. I'm sure that's not cheap (though perhaps not 44%, either).

Rough calculations: 1 truck is 2 TEU/1 driver; 1 ship is 15K TEU/15 crew members. Time shifts and speed differences aside, ships are several hundred times more efficient than trucks in terms of TEU per unit human labor.

I could believe human labor being 44% of the cost of operating a ship, but the cost of operating a ship is also next to nothing. It's the last mile that engulfs these costs. Economically speaking it may not make a lot of sense to make ships autonomous.

But making ships autonomous is probably a lot easier than cars or airplanes, and could probably considered low-hanging fruit in the tech world. You don't need fancy computer vision algorithms to deal with a ship; GPS alone will probably do. Ships don't get within several hundred meters or even kilometers of each other.

Maybe security these days? But they can get ex-special forces soldiers from third world countries for $15k a year. Not sure how it works with visas, being armed while visiting so many ports, insurance and all.

As far as my knowledge goes about the shipping industry: ships are already 90% automated (since they simply go straight forward for days/weeks). There is not much to do for a captain/officer whilst in the middle of an ocean.

AFAIK most of the humans on board of a vessel are there to perform maintenance (most of which can be done during transit).

Yes, we can automate the last 10% to make the ship sail itself, but we'd still need humans on board to perform maintenance.

Maybe fully autonomous ships should have maintenance performing robots then?

That is a way harder problem than automated sailing.

Also a way harder problem than just designing in longer service intervals.

Honest question. Have you ever been in the engineering spaces in a ship/offshore rig/etc.? Do you have any idea of the diverse type of work a ship's engineer and his various reports do?

I don't mean to pick on you too much but the idea that you can have little Hueys, Deweys, and Louies tooling around a container ship doing whatever maintenance is required is ludicrous.

And the other suggestions about having crews that you transport from ship to ship isn't quite as unrealistic but is needlessly complicated in the way that software often is. Among other things--have the people suggesting this sort of thing ever been in the North Atlantic in bad weather?

That or have a small service vessel that sails with a school of autonomous ships handling maintenance.

I don't get it. What will they do when an autonomous ship loses power or steering in the middle of the ocean outside of helicopter range? It can't just pull over to the side of the road and stop.

Ships require constant maintenance to not fall apart. That maintenance can't be automated with anything like current technology. If the crew doesn't do maintenance en route then it will have to be done at the dock, thus taking the ship out of service and costing more money.

Why's it need to pull over? It's in the middle of the ocean. Unless you're a spaceship it's hard to get any more out of the way.

Presumably you send out a repair crew by boat.

No that's not how it works. Disabled ships don't just sit there. The wind and currents eventually push them into rocks. Disabled ships also present a hazard to other vessels. Repair boats don't move fast enough to get there in time, especially in rough weather.

How does it work currently? So much redundancy that there's never an unfixable failure that would stop the ship?

Unfixable failures are very rare in critical systems. Usually the crew is able to catch incipient failures and take preventative action, or else repair the failure at least well enough to get close to a port. Ships engineers are good at jury rigging temporary solutions.

There are a few incidents every year where large merchant vessels lose power and drift onto rocks, or break apart and sink in a storm.

There are about ten major shipwrecks annually -- call it one per month:

http://www.outsideonline.com/1922111/monsterwellen

I'd initially turned up that fact whilst considering the implications of nuclear propulsion for merchant shipping. I'm considering it not particularly attractive.

https://www.reddit.com/r/dredmorbius/comments/3c52ll/shippin...

"The U.S. Coast Guard has estimated that human error accounts for up to 96 percent of all marine casualties."

But if we somehow manage to account for 90% of all eventualities when programming, how are we going to be safer?

Prior HN coverage: https://news.ycombinator.com/item?id=14315036

I'm interested in taking a shot creating a SaaS business related to the freight forwarders / containers shipping logistic or similar.

Does anybody here works on that sector and is willing to share some of the troubles they perceive on the industry that can be solved with software / machine learning?

Look into flexport.com they already do this.

it seems that Flexport is a freight forwarder on its own, I was thinking of something more in the line of solving problems logistics operators have without forcing them to change providers

isn't Flexport doing that? or am i mistaken?

I would love to work on this kind of software.

It's exactly the type of things I find interesting, it's a shame I didn't make different career choices when I was younger but it is what it is I guess.

What makes you think you can't work on this type of software?

I'm 37, I have few academic qualifications so to transition would require 4-5 years more education and then I'd be a 42-43 year old competing against guys coming straight out of university with a background in desktop and web application development.

Practicalities really.

As guess this would qualify as great news for sea-faring pirates.

Pirates are after hostages, not cargo, generally. It's very hard to crack the cargo anyways - it's usually stored in a solid steel 20-to-40 foot shipping container with solid steel lashings bolting everything together. You'd need to be able to cut through a lot of steel, but then moving the actual cargo itself would require, well, a large vessel. Selling it would be a very involved process. All of this puts sea piracy of drone vessels well outside of the scope of a group of armed men in a motorboat.

Potentially you could also establish measures that make it hard for a human to get aboard the ship, such as no walkable spaces, or tall fences, now that you no longer need to accommodate human crew.

It's very hard to crack the cargo anyways - it's usually stored in a solid steel 20-to-40 foot shipping container with solid steel lashings bolting everything together.

The other plan would be go on board and reprogram the ship to go elsewhere, the nearest deserted coast say, maybe carry the ship's beacons a ways away in a different direction before dumping them in the sea.

I'm sure if you crash a whole container ship on the coast of Somalia, your confederates could manage to get money out of the contents.

Edit: Also I recall stories of pirates who hacked container companies to actually find where the valuable stuff was. And containers are steel but they aren't designed for any real security.

Ships are trackable by satellite. Photos, I mean; they show up on photos. Totally passive.

So if you have successfully hijacked a Pacific container ship full of automobiles bound for Los Angeles... where are you going to dock it? Who does the offloading? And where do you sell 5500 sedans at once that makes all this risk worthwhile?

I think that planting remotely controlled explosives and extorting payment against sinking the cargo is the way to go.

It is suprisingly easy to lose track of things in the ocean. Even big things. Ships. Aircraft.

If the transponders are headed to X whilst the ship is headed to Y, it may take you a while before you realise that the ship isn't also headed to X. At which point you now have the problem of determinging which of those fuzzy, cloud-obscured smudges is your ship, and which is someone else's.

Radar can pick up ships, but it also detects, say, rogue shipping containers, of which there are a suprisingly large number.

Existing tracking systems rely on AIS transponders:

http://www.bbc.com/news/science-environment-28372461

Yes, Google can track ships at sea. But it's also relying on AIS, rather than imagery:

http://breakingdefense.com/2012/05/google-satellites-can-tra...

LA-bound cars might be hard. A shipment bound for the Philippines, or India, or elswhere along the Indian or South China seas, quite possibly much easier.

Sure. As long as you don't care about actually living to spend anything. It's all insured. I'm sure you'll make a memorable example for the next guys with a similar idea.

Isn't that possible now, by spoofing GPS? Would the crew have any idea until they suddenly saw shore?

And no crew that you could threaten to obtain the keys to move freely.

Also, an armed rescuing force would have no hostages to worry about.

No one to threaten to make the ship stop either. There's a fair chance a would-be pirate would end up in Hamburg.

And if they do stop the ship, no point sending in an armed rescue force. Just park a destroyer outside and wait for the pirates to run out of food.

For that matter, you tell the ship to report to the destroyer.

Hacking shipping route info, then sending pirate boarding parties would be pretty cyberpunk (and too easy for these drone ships to actually catch on).

Or maybe not. There's also nobody to make hostage or force changing a course.

A fully weaponized autonomous ship would be pretty badass.

https://www.theverge.com/2016/4/8/11391840/us-navy-autonomou...

Until commandeered electronically within your harbour.

But what if we just reduced the crew to a few security personnel?

Don't want to suggest automated security systems...too terminator for me at this point.

Most pirates aren't actually after cargo I'd assume. They're just taking hostages and ransoming them.

Couldn't they ransom the cargo?

"Pay us bitcoins or we'll detonate the IED we planted in your ship and sink it"

Time to add robotic turrets to the sides of the ship...

Lets call the sailors mechanics and the ships frail tech in harsh environment. Suddenly not a good idea.

Reply on news.ycombinator.com