A fish that can steer its tank around the room | Hacker News Reader

A fish that can steer its tank around the room

youtube.com

69 pointsAtomicOrbital28 comments

nice expression of embodiment

This is undoubtedly some cool technology, but it seems traumatizing for the fish. I have no idea what degree of consciousness a goldfish has, but my guess is a) not enough to be aware it's controlling the movements, yet b) aware enough of the jostling and movement that it's highly panicked. The result is that it's flailing around frantically; its fight-or-flight mode is always engaged.

So I'm a little conflicted here. Love the tech; concerned about the ethics.

That's my thought as well. I can't help but think the poor fish is just trying to make the earthquake stop and in the process creating his own hell.

I was thinking something like a smooth toy railroad track would be less traumatizing.

The current design doesn't seem to let the fish just swim around without moving the car. There should be a slightly bigger area for the fish to swim and only make the car move when the fish is in a small area of the tank in the front or back.

It would be interesting to have the track go into different areas/rooms with different colors or lighting to see if the fish spends more time in certain areas. That might shed some light on whether the fish is aware that it can control where the train/car goes.

I think part of the problem is that the steering is not responsive enough, and that might be preventing it from realizing it can control direction.

It should also probably move slower to keep it from jostling the fish around.

I think the concept is pretty neat, but the implementation leaves a lot to be desired. For one thing the inertial effects on the water should be contained (a lid would help there) and secondly the response time is way too high. But given the mechanism, its possible to do some really interesting fish intelligence studies.

For example. set the cart to move very very slowly, so that basically the fish staying on the edge would move maybe 1 or 2 feet per minute. Second, use roller wheels so that the orientation of the tank stays the same regardless of direction, and only its position changes. Finally put different "scenes" around the room and see if the fish gets the tank to move over to the scene or not.

Then I would be LED lamps at different parts of the room, when one was the target color if the fish moved the tank to it, it would drop food in the water and the light would turn off. Then another of the lamps would turn the target color. Could you train the fish to move the tank to the correct light color?

Any idea how well a fish can see (and process/understand) outside its aquarium?

The eyes vary with the adaptions of each species, but overall fish have a strong vision sense, similar to us. They can focus, have day and night vision (rods and cones), most see in color and some can see into the UV spectrum. So, in the vision department, fish are capable.

I don't know how much a fish navigates the space based on vision alone (and I don't think it plays a big role, since visibility underwater is limited), but I know some from the Cichlidae family like the Oscar are able to recognize it's owner, which is quite a feat.

No, but after my experiments I would hope to be able to make some pretty definitive statements about that :-)

A lid seems a pretty primitive solution plus it wouldn't help against feedback on the fish's position itself. Furthermore it would obstruct the vision of the camera. If you're designing a craft if possible you want to avoid the movement of the craft changing the control inputs. If not, and I'm betting this thing already experienced that, it can get into a feedback loop, either creating very "shocky" movement, or unbounded acceleration.

Instead, prevent having waves in the first place. Just like a person can move a glass without generating (large) waves in the liquid. What you want to do is to avoid using the position of the fish directly as input to the motor voltage. Instead, use it as an input to the control circuitry.

Then calculate the force generated on the fish : voltage is linearly related to speed. And the differential of voltage gives you the force on the boat/fish. So let the fish move and this sets a "target", then every timeframe you bring the voltage, say, 0.01V closer to what the fish "wants" to set it to, limiting the force exerted on the bowl.

Btw: there will be a "resonant" value that will be relatively low in value, but will build massive waves. Avoid at all costs. In a project like this I might just assume I'm very unlikely to configure just the resonant value, but in a real project I'd make DAMN sure I knew what the resonant value(s) is/are and put in a warning.

You probably think that this will restrict the movement a lot but it won't. Find the right values and you will hardly see any waves at all.

Is the fish aware that it has control of the movement, or it just swimming in random terror?

There's no positive feedback, so I doubt the fish will ever develop control over it.

Put some caterpillar treads under that bad-boy and the tank can also be a tank! (Actually, that's where tanks got their name. The British WWI code words referred to them as "water tanks" to keep them out of the attention of spies.)

I read somewhere that Survival Research Labs did something similar with a giant legged robot and a hamster, with the added feature of a flame thrower that went off whenever the hamster would squeak.

That's interesting, I didn't know about the "water tanks" code word. A fun bit of related trivia: those early WW1 tanks were piloted by navy sailors and, at least officially, were called "landships".

Ok, let's crowdsource one big enough for a dolphin, a couple million should be enough.

Shouldn't the headline be: A fish named wander

This takes walking into a room and forgetting what you came for to a whole new level.

Klaus from American dad needs this so much! Not sure about regular fish though

I would like to see how the fish behaves in a bigger tank where there is a larger dead point in the middle where the cart doesn't move at all. If it still likes to explore around maybe it does like the experience.

A few years ago I saw a cockroach on a trackball that controlled a similar vehicle. You could chase around by visual stimuli so in turn you would control the cockroach. It was a bit perverse but very interesting at the same time.

More interesting would be a quad copter fish.

I tried once with a cockroach: http://rockolo.com/#aeroflux

How about quad copter cat?

http://www.ohgizmo.com/2012/06/05/the-web-is-full-of-wtf-tax...

Bad fish! https://www.youtube.com/watch?v=pvRQ4d76ZV0

Seems to take some inspiration from a similar device for a parrot: https://www.youtube.com/watch?v=R5N3-QtMdf8

I can see this kind of technology being developed towards a method for controlling "vehicles" while in a liquid capable of absorbing hits.