3 ms·

Undersea drones are a remarkably (and fascinatingly!) complex field. Ignoring all the mechanical challenges (airtight, watertight machine that can operate unde

Undersea drones are a remarkably (and fascinatingly!) complex field. Ignoring all the mechanical challenges (airtight, watertight machine that can operate under crushing pressures and still has a way to separate delicate, saltwater-sensitive electronics and actuators from external-facing motive devices), the communications are also a challenge: radio does awful under water (VLF can go 100 feet, but now you're trying to control something inside the Titanic with radio signals that bounce off all the still-existent metal in wild ways; problem is a bit harder than "get full WiFi coverage of your house"). Sound would actually probably be a great tool, but sound also behaves strangely under water that deep. The most well-researched, well-tested solution is a long cable back to the control ship, but that constrains how deep you can get into the Titanic (one idea I've never seen floated: attach a fleet of robots to the cable, a main one for exploration and observation and multiple wire-managers who's full job is to anchor to something and manage the cable slack... But now you're anchoring to the wreck, so that's also not good). Fully-autonomous robots might actually be the best option here, and the only thing stopping that is the actual research and development to build a machine that can make independent decisions exploring an environment with unknown shapes, dangerous sharp edges, and possible currents, in full 3D. We've put more research into rovers on Mars than into this problem.

Teever 2y ago
I think you'll find this youtube channel very interesting: https://www.youtube.com/@CPSdrone https://www.youtube.com/@CPSdrone
bluGill 2y ago
> exploring an environment with unknown shapes, dangerous sharp edges, We know the original blueprints. We can program that into a robot, along with existing known differences. Then the robot needs to just detect anything that isn't like expected and photograph it and return. After a few rounds of that you would get most of the interesting areas explored. Currents are handled just by comparing expected return signals to what you really see. OF course worth it is questionable. The above fails in the general case only because generally in navigation we also have safety considerations. Your self driving car cannot run over a child no matter what. However there is much less to worry about in the Titanic. (don't confuse that with this is easy, only that if you don't worry about safety it is much easier)
WalterBright 2y ago
Since there are already submersibles and cameras that work at that depth, that cuts down the size of the problem space quite a bit.

Teever 2y ago

I think you'll find this youtube channel very interesting: https://www.youtube.com/@CPSdrone https://www.youtube.com/@CPSdrone

bluGill 2y ago

> exploring an environment with unknown shapes, dangerous sharp edges, We know the original blueprints. We can program that into a robot, along with existing known differences. Then the robot needs to just detect anything that isn't like expected and photograph it and return. After a few rounds of that you would get most of the interesting areas explored. Currents are handled just by comparing expected return signals to what you really see. OF course worth it is questionable. The above fails in the general case only because generally in navigation we also have safety considerations. Your self driving car cannot run over a child no matter what. However there is much less to worry about in the Titanic. (don't confuse that with this is easy, only that if you don't worry about safety it is much easier)

WalterBright 2y ago

Since there are already submersibles and cameras that work at that depth, that cuts down the size of the problem space quite a bit.