How hard can it be to design a self-driving boat for inland shipping? (2020) | Hacker News Reader

How hard can it be to design a self-driving boat for inland shipping? (2020)

thomas.toye.io

37 pointsdalben32 comments

Railways outcompeted canals for shipping, and then road trucks outcompeted railways.

Controlling the boat near the dock is probably more difficult than cars or air planes since swells/waves/winds can move the boat relative to the dock. The robot needs to read the water and wind to protect the boat from collisions.

As taxis, docking/taking on passengers is tricky since it needs to handle situations when people cling onto rails and risk getting stuck between the boat and the dock.

There are already gps-based systems that can do this on large ships but they are designed for open-water and can keep the boat surprisingly stable even in high winds as long as the ship has thrusters.

You are right though, being by the dock is harder although at this point, you are likely to have some kind of personnel to help tether the boat to the quay.

All depends on how far you go with your maneuvering capabilities, which is simply related to costs. Dynamic positioning [0] has existed for a long while and has now moved down even to small dual engined boats [1], on the other hand it requires much more tech and sideways and rotational thrust capabilities than what regular cargo vessels have.

A single prop with a bow thruster, or especially without, can't move freely in the 2D space which makes it much more complicated because you have to move towards the dock with the correct angle depending on wind, current and what not. Then time your maneuvers while taking the environment and for example prop walk [2] and other effects into consideration.

[0]: https://en.wikipedia.org/wiki/Dynamic_positioning

[1]: https://www.youtube.com/watch?v=491RkaOYfr4 Volvo Penta Assisted Docking

[2]: https://en.wikipedia.org/wiki/Propeller_walk

But the space is not 2D unless it's a small calm canal. Waves/swells/wakes can make the difference in vertical position of the bow relative to the stern change by meters (or whatever depending on length) but it can be significant relative to human size. In some situations it might be impossible to take on/let off passengers, so the boat would need to plan its motions taking this into account while otherwise not obstructing or causing danger.

The goal then for the robot is not to position itself in 2D space with meter-level precision, the goal is to behave reliably and predictably in the harbour under chaotic circumstances, with sub-meter precision. That requires predicting movements and drift, taking wind into account. Since this is almost impossible, skippers learn to read the water/wind to understand what is appropriate for safety distances, speed, angles, etc.

A road vehicle could just refuse to move if the environment becomes unrecognizable, but a boat keeps on moving, so the robot can not just "give up".

It's true that thrusters increase the manoeuvrability.

In the spaces where waves and wakes make a difference the tolerances are larger. The one exception is for example smaller commuter ferries without a ramp, then you might have to time it as you say.

Wind and current doesn't really matter as long as you have dynamic position which abstracts it away as long as you keep within your power budget. The real hard part here is moving through a shifting environment, for example the current often reverses along banks. Especially in a river based harbor environment where you've created piers and what not disrupting the flow.

In a previous life I worked as a skipper in just those conditions, passenger ferries in the 25-40m sizes with a couple of hundred passengers in river harbour environment. Two props and bowthruster and you have to keep all the considerations you mention in account. Especially since you can't angle the stern without angling the bow (duh) when you're applying reverse thrust to dock. And with some speed forward you move the center of rotation forward making the bow thruster less effective.

It was actually easier with quite a strong wind since then that would overcome the current easily and you would lie on the wind margin side of things. If you come too high just make the docking take a bit longer so you would blow down and land perfectly. In almost calm conditions you would instead have to guess which would win that time, wind or current.

That said, the ones with two pods, one at each end you just balance things out and go straight to the dock.

> Controlling the boat near the dock is probably more difficult than cars or air planes since swells/waves/winds can move the boat relative to the dock.

For a layman, it sounds as the same kind of difficulty as controlling a drone in the air. It's not perception and identification of millions of different objects, neither it is predicting other driver's behaviour on the road — it's a physical process. Computers seem to be quite effective at adapting to that.

For those interested in this sort of stuff have a look at YouTube channel rctestflight - very cool diy projects. Less ML and more solar powered waypoints but still found it insightful on the unique challenge water provides

This is an interesting article but it needs a big disclaimer.

They are designing a self-driving boat without using sonars not radars due to their budget. It's fun but of little relevance to the real world. Most of their problems would be solved or at least heavily alleviated with access to the relevant sensors.

Still, that's quite impressive for a college project built on a shoe string.

From a practical sense though, wouldn’t the staffing costs pale in significance to the other costs of running a large vessel?

It depends. For last-mile delivery and public transport with small vessels, that may not be the case.

The city of Ghent was eyeing autonomous small crafts as "water taxis", and a big shipping provider wanted to do last-mile(s) delivery for shops, restaurants, and small parcel pickup points along a river. For the latter case, the vessel would still be manned, but instead of having to navigate, that person could sort and prepare parcels for delivery.

If we’re talking practical, you’d always need people on board to deal with mechanical breakdown or emergencies. And you’d need all the support systems and supplies need for them. After factoring that in, how large is the bridge crew on a container ship?

Also, pirates would become a problem because taking over a ship without a crew would easier and way less risk.

The argument for easier piracy assumes that a self-driving ship would have a standard "people accessible" format. There's no reason for packaging of the vessel to remain the same.

Consider patrol drones. They look nothing like something an individual can get into.

This was already addressed. Most of it would still have to be human accessible for maintenance/emergency purposes. If you're going to have to potentially service the thing underway, there is no point redoing the packaging.

Besides which, an automated ship is about a trio of tugs, a computer replacement, and a buncha fuel away from anywhere.

In terms of maintenance you can put the access below the cargo which significantly increases what it takes to hijack a ship. However for hijacking’s the simplest approach is to simply scuttle the ship remotely. If they get nothing then it’s rather hard to get people to hijack the next ship.

> scuttle the ship

You mean deliberately sink a $10 million ship, causing a massive environmental disaster. Not to mention being a hazard to other shipping if the water is shallow enough or the ship breaks apart.

Do you blow up a grocery store because someone attempts to steal a jar of pickles?

Ships sink all the time, unless it’s an oil tanker they really aren’t an environmental disaster. As to cost, if someone’s hijacking a ship you already lost the ship now it’s just an insurance question.

The question for an insurance company isn’t just about this ship, but the cost of paying out future hijackings. On top of this there is the rather more interesting option of refloating the ship.

...so you just hand out an instant logistical DDoS button to any sufficiently motivated group? What also makes you think insurance is going to pay out for you sinking your own ship. Moral hazard much?

Every rocket launch a range safety officer (RSO) keeps track of the trajectory and is ready to hit the self destruct button if needed. Insurance pays out if you hit it as that’s been agreed to ahead of time, the same would apply here or you wouldn’t sink the ship.

As to DDoS, sinking isn’t any more of a risk than someone taking over control of your self driving boat and aiming it at the coastline. From a pure safety standpoint you need some method of remotely disabling the ship.

A rocket launch is generally done much less frequently, premiums to insure would be through the roof due to the smaller risk pool, and don't involve cargo lots measured in the thousands of TEU.

The value of a single rockets cargo can easily be worth more than a container ships cargo. The average TEU at Port of Los Angeles is worth ~43k USD and the words largest container ship is only 24k TEU, that’s up to 1 Billion in 2021 USD and plenty of satellites have been worth more than that.

The ships can be worth quite a bit when new, but their also depreciating assets with only a few percent of that in scrap value. Only using the least valuable ships in areas that can be pirates seems like an obvious optimization.

You have the absolute oddest set of risk management practices I've tried to wrap my head around, but to each their own I guess. I don't see the sense in trusting a global supply chain that's built on JIT principles to a paradigm where your first reaction is "welp, scuttle it". Those TEU's may not be the most glorious things, but they are the inputs that keep modern civilization going. To do otherwise... Well... It just rings foolhardy to me. Might look good on the ledgers, but doesn't make good sense or shake out in the real world.

Piracy is an extremely rare event, and keeping it that way is important for ensuring the global supply chain works.

At the same time actively preventing piracy is expensive. Having a significant armed security force on a boat is effective, but paying for them when 99.999% of the time when their not needed isn’t cost effective. Making it clear that attacking a ship is pointless on the other hand is a great way to avoid getting attacked in the first place, but you need to be willing to carry through.

>Having a significant armed security force on a boat is effective, but paying for them when 99.999% of the time when their not needed isn’t cost effective.

Who cares? You seem to be assuming that security force does nothing else but sit around in case of piracy. In reality, we call these people "mariners", they are trained specialists at maintaining the hardware with which they ply their trade, they know the customs and rules to follow in diverse ports of call, and they are there to provide boots on the ground and hands on tools if something goes sideways.

There is a huge difference between bodies on the boat, and no body on the boat. You cross that line, and leave no bodies on the boat to look after things, and you now have a floating puzzle full of other people's stuff, owner's of which are not necessarily going to be happy that you decided to drop their load to the bottom of the ocean to "prove a point" where you can't even guarantee that the entire point wasn't the pirates making sure some shipment didn't show up in the first place.

Hell, if it wasn't you, the shipper, making the scuttle decision, and was actually the local Navy on justification of "no negotiation", fine. Though that really just punts the issue to international waters.

No one will blame you or foster ill-will for doing everything you could and failing anyway. People will have hard feelings if you come up and say, we blew up all your stuff because those damn dirty pirates. They paid you to get their stuff from A to B, you decided C was better.

All of this "let's go unmanned" just really seems to me to be a solution looking for a problem, and not being shy about creating a few more while we're at it.

You need to improve reliability and have a mobile crew you can dispatch on an emergency. Most mechanical engines do not need a maintenance crew on-site all the time, including the much larger and more complex electricity generators and the disaster-prone airplanes.

About pirates, no, they work will become much harder if you don't have anybody they can point a gun to and tell to stop the ship.

I don't think it is the costs necessarily but the availability of people with the right skills. Even with the money, you can't just magic up qualified mariners.

Also, a significant amount of money is spent providing 24 hour cover of navigation in the bridge so I guess you could save something by automating a lot of it.

You are probably right that in the extreme, it would be better to have some humans to make sure your containers make it to their destination!

Economics of scale are the whole point of large vessels so yeah, there's much less money to be saved eliminating crew on ships vs semi truck drivers (or even shipyard equipment operators).

there is a dutch startup called roboat thats working on autonomous boats for transportation of humans. Eventually we will also have autonomous Cargo ships

As far as I'm aware, dot ocean does it too ( Belgium - https://www.dotocean.eu/ )

By coincidence, they also had an experiment in Amsterdam.

Note: Friends of me.

When I met them (years ago), they seemed to be more focused on open waters. In particular, I believe they were providing monitoring vessels in the Mediterranean during the refugee crisis.

There are multiple products as far as i know (it's not an easy market to get into)

- Their IOT big data platform ( Atlantis - https://www.dotocean.eu/products/atlantis/)

- Dredging canals cheaper

- Autonomous boats

- Underwater soil detection with a probe

I'm not really sure if there is a specific focus on open waters though, but i don't think so.

Looks awesome. Bookmarked the page

tell your friends I'm rooting for them thats an interesting project

hmmm..stuff still happens on inland waterways...bad stuff...sometimes the gales of november come early

Reply on news.ycombinator.com