Wave Gliders [1] have been making trips like that for years. They're powered by waves, rather than wind, and they're quite controllable. They can generally stay within 30m of the desired track. They use GPS and Iridium. It's a commercial product, and you can add small sensors.
Sounds like they set non trivial challenges. But in the right conditions, the Atlantic is a doddle. Start late November in the Canary Islands. Point your ship at an island in the Caribbean. Sail with the trade winds behind you for twenty five days. I did it two years ago. The only manoeuvre we needed was rounding the top of St Lucia on the way into harbour at the end.
I realize the ocean is a hostile place but I can’t imagine it’s hard compared to AI driving challenges. A message in a bottle could potentially do this journey.
This seems like a challenge in construction/durability more than in control and navigation. I thought there must be some kind of limitation to make it truly interesting (no gps, no batteries etc) but didn’t find any mentioned in the article.
I’m sure I must be underestimating the difficulty but unsure how I’m underestimating it.
Consider that AI driving challenges happens on roads, which we have engineered to be easy to traverse on. And the direction of travel is independent of the force used for propulsion. Sailing has more variables, and many are not independent.
For example, you frequently can’t just point your boat in the direction you want to go; that’s only going to work when you’re going downwind. You’ll often need to tack back and forth, rarely actually facing the direction you want to go. When and how frequently to tack will all have to be determined automatically. In a driverless car, that’s two independent variables: steering wheel orientation and gas paddle depression.
Thought of this way, cars really are a marvel of engineering and UI design. They reduce transportation down to only three controls: steering wheel orientation, gas pedal depresssion and brake pedal depression. The complexity of the car is hidden under the hood. Just about all other modes of transportation require much more input.
But, of course, the actual navigation of driverless cars (as in, how to steer the wheels and modulate the accelerator/brakes) is probably mostly trivial. The hard parts are detecting paths, junctions, signs, obstacles, etc. and behaving predictably so as to cooperate with other drivers.
I’m sure the actual navigation of sailing vessels is more complicated than cars, but I can’t imagine that experienced sailors and talented programmers and engineers would have much trouble working out an algorithm.
But clearly they do! That's what the entire article is about. Boats also have to deal with other boats - not as much, but it's definitely there, and the results of getting it wrong tend to be worse. alkonaut's question was, basically, what's hard about this compared to driverless cars? My claim is that you have roughly the same requirements (yes, including paths and obstacles) but way more variables.
That’s true but navigating in and around ports isn’t part of this competition it’s 100% open water sailing. The start lines are defined as lines of longitude with an autonomous run up of 40 nautical miles before hitting the start latitude and it’s roughly from a ways off the coast of Newfoundland to off the north coast of Spain both well away from land.
They demonstrate it is not _impossible_ to have ship to ship collisions, but considering how rare these events are they tell us more about the state of seaman skills in the navy than about navigational hazards.
The collisions weren't really "near" any port, and were due to really shitty ship handling by the person in charge, possibly due to how the navy treats its sailors.
Tacking is only necessary when traveling directly into the wind, thus programming it would be trivial with a digital weather vane to determine wind direction. If desired direction == wind direction, then vary direction by several degrees for a minute or two, then switch to the other direction in a minute or two.
Also, driverless cars may just have two controllable independent variables, but the environments they work in are incredibly complex, with hundreds if not thousands of objects to track at any one moment. Boating in the middle of the ocean would still be complex, but it wouldn't compare to autonomous driving on land.
Well don’t forget that the current/surf are just as important, it’s not just wind. Then there’s also the fact that you don’t really ever stop, unless you’re heaving to or setting anchor. These factors when added together sort of compound the difficulty in programming intelligent decision making, because you can’t avoid accidents by hitting the breaks as easily or just steering out of the way, everything has to work with this universe of waves and wind and (almost) never stopping.
That’s just under fair conditions. Under foul conditions, I can’t really imagine how you can have a program make sense of this stuff, since all the sensors are going to be all over the place, and you sort of need a human to make sense of it all. So under those conditions the it’d probably be easiest to just heave to, but then it’d probably get off course.
Also don’t forget that GPS is unreliable. The ship could probably do dead reckoning better than Columbus, but that’s something it absolutely must be taught in addition to everything.
I think a fair analogy would be to flying. You could make the argument that flying isn’t as complex as autonomous driving, but they’re just different. Sailing is about one degree of freedom more simple than flying.
I think the worst part is ropes and stuff getting tangled. Unless you run some sort of fixed wing sailboat, which has it’s own share of problems (how to reduce sail in strong winds or make a full stop?)
What I am interested is using CV and sensors inside ropes to optimise for speed or comfort, without user intervention. Takes quite a bit of training to learn how to read sail.
Well, of course designing an autonomous system to operate in an environment where malicious actors are actively spoofing and jamming your sensors with pathological inputs is a pretty huge challenge - admittedly one the military is probably very interested in - and is not the purpose of any of these competitions. It's hard enough when you trust your sensors, at the moment!
Sailing is slightly more complicated than that. Take a look at a polar diagram for a typical boat. They're fastest sailing at ~90 to ~120 to the wind, and generally you can't get closer to the wind than 30 degrees or so. Because of this the best velocity made good is often by alternating on two tacks rather than sailing directly toward your destination, even when sailing off the wind.
There's a lot to consider when sailing. None of it is some sort of magic that can't be automated, but there is quite a bit more to it than how you describe it in your first paragraph.
Ocean is powerful. Before being a surfer and following storm forecasts to chase good swells, I had no idea how much energy a storm on the open ocean could generate.
For who is interested on sailing, I recommend all the books written by Amyr Klink. He describes in a beautiful and poetic way his journeys across the seven seas. It’s very interesting when he describes the day he faced a 4 hundred meters iceberg (this is just 10% of it’s real size) that traveled oposite to the wind, on his direction. The iceberg follows the sea currents, it does’t care about the wind direction.
Thank you for sharing. There are also really good speches on youtube. I also follow him on Instagram, and it´sfunny to see a normal day of a such exotic human being.
He is on his sixties now, I wish to shake his hand in a near future.
I looked into joining a crew on the Atlantic Rally for Cruisers once and I remember being told that long-distance sailing is less about sailing and more about making repairs at sea. Controlling the boat seems like the easy (I know, "easy") part. What are they going to do when their robo-boat's rigging is damaged?
I'm sure they've accounted for this—they're obviously not dummies. I'm just curious.
[1] https://www.liquid-robotics.com/platform/how-it-works/