It's a lot easier to send a bloodhound to track a criminal, or your kid to pick up groceries, than it is to get a deep-sea robot to find something on the seafloor. The dog will pick up the scent and know to turn around if the bad guy doubles back. The kid can use his eyes and brain to wend his way to the vegetable aisle and recognize a red pepper. But deep-sea robots are typically dispatched on programmed missions, swimming up and back along designated track lines, like a mower cutting a lawn, and blindly collecting data. Only after the vehicle resurfaces can scientists download the data to see what the robots may have found.
These autonomous underwater vehicles, or AUVS, move autonomously in that they swim unmanned and unconnected to a ship via a tether. But they have lacked that essential capability to think autonomously-to interpret cues they gather from the environment, make decisions on the fly, and change course as necessary to accomplish missions more effectively and efficiently.
Scientists are beginning to experiment with ways to give AUVs more decision-making capabilities, and on a research cruise in September 2009 off Santa Barbara, Calif., scientists from Woods Hole Oceanographic Institution (WHOI) demonstrated that they could teach a new robot some old tricks.
They equipped an AUV called Sentry with a high-tech "nose"-an underwater mass spectrometer developed by WHOI scientist Rich Camilli that can detect and identify minute quantities of chemicals in seawater. Then they gave Sentry some brains, in the form of software to analyze the chemicals it was sniffing-in this case, naturally occurring oil, gas, and other hydrocarbons leaking from the seafloor in areas known as "cold seeps." Finally, they gave Sentry the ability to change its pre-programmed course and operating mode, so it could home in on targets and begin sampling right away, instead of waiting to return on a subsequent mission.
In engineers' parlance, it's called "dynamic re-tasking to execute non-deterministic missions." In other words, Sentry took its first baby steps on the road toward making its own decisions without input from people.
Read the complete article in Oceanus magazine here.