For sure. I'd love to see a more efficient way of doing it too. A common problem is that sometimes devices requires a larger voltage than the board delivers. Plus you don't want the options narrowed because of those limitations. Since relays are cheap as are the 595's, until there's a better option, that would be my immediate choice.
You could pulse the motor to manage speed, which is actually quite ingenious when you think about it. I don't think pulsing would be inefficient either.
An object of neutral buoyancy also wouldn't need to conform to any particular shape. You could use a ping pong ball, for example, attached to shaft attached to two rotary encoders to determine the objects tilt for staying flat in the water, and using thrusters for making corrections. The ROV could be the shaped like a ball, hypothetically.
Once I'm over my current back log of work, and other projects I'm working on, I'd definitely like to put all that to the test.