This could end up complicated.
Is it to stabilise attitude, or position, or to provide guidance (navigate to a point), or a combination of these functions. Under water, it is easy enough to get sensors to provide inputs for water speed and depth or height above bottom (well maybe not in a fish tank, it is too small). Think more of swimming pool.
Acceleration can be derived from speed. Direction can usually be obtained with a magnetic compass, and this can give 3D vectors using speed, depth and direction. In fact a compass with three axes might give tilt too, as the magnetic field lines are at a fixed angle to the horizon in a given location. Attitude can be determined with tilt sensors in X and Y. Currents and waves might need consideration as these add noise to such readings. For example waves are superimposed on a pressure sensor reading used for depth, but attenuated according to depth. You will need to understand trigonometry and physics and calculus to some extent to make these sensors work. Microprocessors are limiting sometimes.
For navigation the problem is getting a position reference - there are few or no visual cues. There tends to be only acoustic beacons or dead reckoning. Radio navigation and positioning can only be considered at the surface.
That is no doubt why you are considering inertial sensors. The inertial sensors have various purposes. Speed and direction sensors really provide heading (the direction the object is pointing) and speed through the water, and you want speed and direction over the bottom for navigation or stabilisation of position (considering drift due to currents).
A gyro gives an inertial reference for attitude, heading. A multi axis accelerometer can be used to determine acceleration in X, Y, Z and so integrated to get velocity vectors. Using a known starting point, the position can be determined by further integration. Any error is also integrated, so the error gets bigger over time. This is dead reckoning. Generally a new "fix" is used to reset the position periodically. The accelerometer and gyro may be used just to provide a faster response for a closed loop control system. This is because the speed and position sensors can take longer to update.
Generally the small low cost gyros and accelerometers used in model planes and helicopters are not accurate enough for navigation. They are used for stabilisation, short term stuff of a few seconds. As above, they supplement speed and direction of other sensors for better closed loop control.
The computer has to be low powered for battery operation underwater, so more likely some kind of microprocessor. You will need to know how to use this in the hardware and software sense. Sensor outputs are connected to its inputs, using signal processing to match the sensor to the computer inputs. Once the desired control outputs are determined by the program, they are used to operate an output port. For a motor, this involves a driver circuit capable of interfacing the computer to a motor. In this case, I would think you need "thrusters" (meaning motors with propellers) in three axes. This might be more than 3 thrusters. The computer can vary the motor speed appropriately, using the software you create.
Unfortunately, a significant power (in watts) is needed to stabilise someone under water against currents amounting to a meter or two per second. This means big batteries. Another issue is that all this stuff needs to be made waterproof, sealed in containers, and can be quite bulky and expensive. This involves skills in a number of fields including mechanical design and fabrication.
Phidgets are a family of input and output (i/o) devices that use the USB port for connection to a PC. Not much good for underwater, as a PC tends to be impractical, and microprocessors don't use that interface generally. More likely they use serial ports (RS232), with analogue and digital inputs and outputs of various kinds that interface more directly with sensors.
You can look up robotics and autonomous aerial vehicles (UAV) and autonomous underwater vehicles (AUV) for more info. You will find plenty of useful info, but start with wikipedia.
It seems this is just too full on for getting started. Too many issues making an AUV, and too many things to get familiar with. Add to this the additional complications of making it work with a human...
Try a model vehicle with wheels that moves slowly. No problems with leaks then, and you can see what is going on, and only 2 dimensions, X and Y.