[1] http://www.apl.washington.edu/projects/seaglider/summary.htm...
The sea gliders are interesting (see Liquid Robotics) but they have limited cargo capacity. 15,000 lbs would be like 5,000 sea gliders, that is a big foot print.
The game will of course really change when the human operators step out. Most drug cartels seem unwilling to risk their load to robotic pilots but at some point that will change. In a submersible you could gain a tremendous advantage if you didn't care how 'deep' you went relative to the Coast Guard.
Once we get electro-stimulus drugs this will become moot of course but in the mean time it makes for some interesting engineering challenges.
The beauty of current AI is that the compartment can be set to notice divers nearby and "walk" to the other side of the hull. Since sonar can't pick up small attached objects, boom you got yourself the equivalent of a smuggling bed bug. The crew can be all detained while this happens because they are not in on it.
The only issue would be noise made if you have listener crews. That one would be tough to solve. That's why you attach yourself to large boats, not to small smuggle ships.
As for power, why not a turbine powered generator? The host ship won't miss the few tens of watts extra drag.
- construct a buoyancy-controllable drogue which would magnetically attach to a cargo ship, then hang behind and below it during transit
- load up the pseudoephedrine, or whatever, in (say) shanghai or during a stop somewhere near a supplier
- wait until the ship came near the US, then release the drogue offshore and pick it up with a boat
Seems almost foolproof, although I'd imagine marine capable technology is harder than my web developer's brain is envisaging, and you'd be betting lots of money that your little parasite didn't fall off.
I'm probably overthinking this, however, and you'd just pay someone onboard to keep some suitcases in their cabin, or something.