MIT's Soft Robotic Fish Explores Reefs in Fiji | Hacker News Reader

MIT's Soft Robotic Fish Explores Reefs in Fiji

spectrum.ieee.org

82 pointssohkamyung21 comments

This is completely flippant to say, since I've never built anything like this and don't understand the complexity behind it, but I'm surprised at how awkwardly it looks swimming. It's clear to my eyes that it's not a real fish.

I wonder if other fish will interpret it as injured and attempt to eat it :)

It's doubtful - it's swimming more powerfully than other fish, not less. It's why it looks so odd to humans - the tail strokes are far too heavy. Fish have a lot more subtlety in their own.

But the fact that it spends more energy, and it is less efficient, should be interpreted as a sign that something is wrong, isn't it? The gestures are powerful but the fish does not advance a lot.

Limping on one leg is a more extenuant exercise precisely because you need (and show) more power (and that is less sustainable). I don't think a lion would see it as a warning signal!

I know next to nothing about submarine dynamics, but I am an RC plane pilot, and it looks to me like this fish swims in a constant stall. If it was a plane, I'd be inclined to say that its center of gravity was too far back.

The video states that it can control its own bouyancy, yet it looks like it's constantly fighting to swim level instead of sinking. If it could achieve a neutral bouyancy then I bet it would glide along much more like a real fish.

I think the reason is that they tried to minimize the number of actuators (due to both dollar and energy costs), so more refined, undulating movements cannot be flexibly reproduced - instead, they are built-in in the hardware design.

Maybe it's intentional, but the tail looks much too heavy. It only looks remotely natural during transitions where it starts to dive. The motive tail is providing significant lift. As a later commenter (ibnjmn) mentions... it looks like the center of gravity is too far back.

Maybe they designed this so that the robo-fish could stop quickly, but it certainly doesn't look efficient for something that is nominally neutrally buoyant.

A short list of the other attempts on bio-inspired underwater robots:

* Fish https://seagrant.mit.edu/publications/MITSG_00-4J.pdf * Octopus https://phys.org/news/2015-04-octopus-inspired-robotthe-fast... * Snake https://eelume.com/ * Jellyfish http://www.emdl.mse.vt.edu/projects/alex.html * Manta Ray https://www.usnews.com/news/news/articles/2017-12-04/singapo...

There's lots of very interesting research in this area inspired by biology. Check out http://www.people.fas.harvard.edu/~glauder/FishRoboticsLaude...

It's a little amusing how the controller is a super nes controller

The paper on this provides a lot of interesting insight: https://groups.csail.mit.edu/drl/wiki/images/a/a5/Hydraulic_...

CSAIL has published some other papers[1] on the fish-bot which look interesting, too.

1: http://groups.csail.mit.edu/drl/wiki/index.php?title=Soft_Ro...

The video mentions that the controller uses ultrasonic signals to communicate. I wonder if those signals would disturb the other fish, negating the “neutral observer” aspect of the project.

I like that it has a "fish-eye" camera in the front

I hope larger fish don't get confused and attempt to eat these robots.

So I wonder what are the biggest predators that live near these reefs.. and how they are able to identify prey.

Reefs attract some fairly large predators. Barracuda, various sharks, grouper, moray eels, among others. All can get to well over a meter, or a couple meters in some cases.

Sharks prefer colder waters. It's more likely destroyed by trigger fish.

Perhaps the 'un-fishy' smell of the robot fish might put off predators.

They need to put as much stabilisation on that camera as they can.

Pretty cool otherwise.

That's pretty interesting

I want one.

Reply on news.ycombinator.com