Whenever I tell anyone, technical or otherwise, about our group's goal to create an autonomous underwater vehicle that will explores the waters around NYC, one line of thought always emerges: Who is controlling the robot? How does it know where it is?
To the first question, no one controls it-- it's autonomous. Controlling an underwater robot is actually rather difficult. You need either a tether (typically an ethernet line) or an extremely expensive military-grade equipment, which has its own difficulties. Radio waves do not travel through water very well. Very Low Frequency (VLF, 3-30kHz) radio waves might travel 20 meters, and Extremely Low Frequency (ELF, 3-300Hz) radio waves might travel several hundred meters. Keep in mind, typical WiFi operates at either 2.4GHz or 5GHz frequency.
To the second question, there actually exist some adequate solutions. Unfortunately, most of them are too expensive for us (hello, sponsors!). The best and most expensive is a Doppler Velocity Log (DVL). The device contains several, angled sonar devices that send out signals, and then measure the time of flight and angle of return. From this, the DVL estimates its velocity, direction of travel, and/or position. If you maintain an acceptable depth relative to the ground, and if you have ~$25k to burn, this solution works well enough. Another expensive solution is a high precision Fiber Optic Gyro (FOG) Inertial Measurement Unit (IMU). An IMU typically contains a gyro, an accelerometer, and a magnetometer, and it is used primarily to measure orientation. With highly accurate IMUs, you may attempt to integrate over the measured acceleration in order to obtain distance, but that double integration only exponentiates the errors. Then the FOG + compass can tell you along which axis that distance was traveled. Some high-grade FOG IMUs have such a low amount of error that they can provide these measurements accurately enough, but they cost upwards of $8k.
That leaves us at the solution which we decided to pursue-- high frequency acoustic triangulation. Black box acoustic locator beacons, which operate at 37.5 kHz can be detected up to 3 miles away in good conditions. We plan to hang such an acoustic beacon, subsurface, from a floating device with a GPS attached to it. On the AUV, we'll attach four hydrophones (essentially, specialized microphones) at odd corners. All hydrophones will listen for the acoustic beacon's signal, measure the time at which each hydrophone receives the signal, and then determine the location of the acoustic beacon. With that location, the AUV always knows its position, relative to the acoustic beacon. Then, later, the GPS + acoustic beacon + timestamp data may be used to position the AUV globally. This solution will cost approximately $5k.
There it is. Positioning ain't easy, and positioning ain't cheap. It is, by far, the most expensive part of this project. However, the ability of a robot to position itself frees it to explore untethered, bounded only by power.