Untethered in the Deep
Autonomous underwater vehicles advance–not to mention stop, turn, and hover.
The two-meter-long robotic submarine Odyssey IV could easily be mistaken for a half-finished undergrad project. Although its electronic innards and motors are in place, its body is a work in progress–a stainless-steel frame with a few chunks of foam tacked on for flotation. Some pool floats holding a temporary power cable are attached to the top. But once the vehicle is lowered into the water, it acquires a sudden grace, coasting forward, stopping to hover in place, and executing perfect turns. For now it’s merely cruising the 25-yard Alumni Pool in Building 57, but soon, the finished version could be exploring some of Earth’s most remote underwater destinations.
Odyssey IV is on its way to becoming the newest addition to the family of autonomous underwater vehicles, or AUVs, that have been conceived, designed, and built through the MIT Sea Grant College Program since 1988. Thanks to great advances in mechanical design and programming, these independent and–by submarine standards–inexpensive swimming robots have explored the depths of the Antarctic ocean, searched for ancient shipwrecks off the coast of Greece (see “Excavating the Deep,” May 2003), scoured underwater minefields, and even inspected the hulls of ships. And MIT researchers are convinced that these recent successes are just the beginning,
In the mid-1980s, Chryssostomos Chryssostomidis, now director of the program, grew frustrated with the technology used for deep-sea exploration. The top robotic submarines of the day were remote-operated vehicles, or ROVs, controlled via cables by engineers on large surface ships. The size of the surface ships, and of their crews, made the subs expensive to deploy, and the cables, which also provided the subs with power, limited their range and maneuverability. “The tether was a nuisance,” says Chryssostomidis, who looks and sounds like a Greek fishing captain. “So we asked ourselves, ‘Could we make it tetherless?’ ”
This wasn’t a simple problem. Getting rid of the tether meant that the sub needed to be independent. Radio waves don’t travel well through water, so there was no way for scientists to wirelessly pilot the vehicles from the surface. The robots would have to steer themselves.
Inspired by his computer science colleagues’ continual efforts to push their field in new directions, Chryssostomidis decided to try to do the same with subs. He challenged his students to build, for less than $100,000, a robotic submarine that could dive to 6,000 meters (ROVs with tethers couldn’t go deeper than 1,000 meters). He also wanted it to be light enough for two undergrads of average strength to carry it down to the Charles River for testing.
In 1992, led by research engineer James Bellingham, PhD ‘88, who now runs the AUV program at the Monterey Bay Aquarium Research Institute, the students produced Odyssey I, a robot that met Chryssostomidis’s main requirements. Well, almost. It could dive twice as deep as its remotely operated brethren, and it weighed just one-sixth as much. But it cost closer to $140,000, and a typical pair of physics majors probably wouldn’t have been able to lug it more than a few feet. “It needed two very strong undergraduates to carry it,” Chryssostomidis jokes. (In fact, students used a cart to get it to the river.)
The U.S. Office of Naval Research, which had initially invested $50,000 in the AUV Lab, was impressed enough by Odyssey to boost its backing to as much as $5 million annually. By the late 1990s, MIT had produced a fleet of 20 autonomous underwater vehicles. To meet the navy’s growing demand and serve an expanding client base, the AUV Lab spun off a company called Bluefin Robotics, which has become one of the leaders in the field. Today, oil companies use sonar-equipped Bluefin vehicles to survey the ocean floor, and the navy employs them to search for mines in potentially dangerous coastal waters.
The advantage of these AUVs is that they can cover large areas without constant monitoring–and they don’t require a large surface vessel to launch them, or trained operators on board to pilot them. They are programmed to head out on their own. An AUV engaged in surveying might be instructed to hit a series of waypoints, or underwater targets, in a given area. As it coasts, it relies on different instruments to estimate its position; it might use a GPS-based intelligent buoy-tracking system, for example, pinging acoustic beacons on the surface. Onboard navigation software uses the position estimates to determine whether the vehicle has remained on course and adjusts its heading if necessary. A scientist aboard a small vessel in the vicinity needs nothing more than a laptop to track data collected by the AUV.
As useful as AUVs have already proved in the real world, MIT engineers like Franz Hover, who runs the Odyssey IV project, are working hard at making them even smarter and more capable. MIT’s first three Odyssey-class vehicles are like sharks, in that they always have to be moving forward. That’s fine when you’re trying to survey the ocean floor, but it’s not so great when you want to stop and take a closer look.
Odyssey IV can really move–tests suggest it should be able to cruise underwater at about 2.5 meters per second–but just as important, it can stop. Four thrusters–two on either side, plus one each mounted on the bow and stern–enable the robot to turn in all directions, and to stay put when necessary.
The ability to hover could prove important for the marine-archaeology expeditions Chryssostomidis loves, since the vehicle could stay in place and study interesting objects instead of simply grabbing a sonar reading as it cruised past. It could also benefit energy companies that are currently assessing the feasibility of drilling wells far out in the Gulf of Mexico, beneath thousands of feet of water. Maintaining such wells will be tricky. Stop-and-go AUVs like Odyssey IV could cruise over a large area, perform close-up inspections, and monitor the wells for damage. In fact, from the AUV’s perspective such monitoring would be easier than exploring the vast ocean floor, because the wells themselves would provide reference points for the vessel’s high-powered Doppler velocity log. The DVL would bounce sound waves off the wellheads and pumps, just as it does with the walls of the Alumni Pool today. Then it would recapture those waves, calculate their Doppler shift, and derive the AUV’s velocity to estimate its position.
Although the researchers aren’t ready to toss Odyssey IV into the Gulf of Mexico or the Adriatic just yet, AUV Lab graduate student Jim Morash says the tests in the Alumni Pool have been encouraging. “It’s very smooth and very steady in the water,” he says.
Not only are AUVs starting to look as comfortable in the water as fish, but in the future they might start to move like them, too. Researchers from MIT’s Bioinstrumentation Laboratory are working with scientists at Drexel, Harvard, and George Washington universities to develop a mechanical fin for aquatic maneuverability. The group, led by Professor Ian Hunter and James Tangorra, PhD ‘03, who is now an assistant professor at Drexel, studied the pectoral (or forward) fins of the bluegill sunfish, which can maintain control in fast-moving streams and turbulent waters. Creating an artificial copy would be impractical; Tangorra notes that the bone structures inside it are controlled by 59 muscles. And the group’s prototype fins are not as rugged as the thrusters on Odyssey IV. Still, they’re lifelike and elegant. “They look like pieces of paper flowing back and forth in the water,” says Tangorra.
To enable the mechanical fin to move like the real thing, Hunter, Tangorra, and the rest of their group worked with Timothy Swager, head of MIT’s chemistry department, and his lab to develop new actuators based on conducting polymers that alter their shape depending on the electrical current. These actuators, or artificial muscles, could eventually be used in other machines, too.
As the bodies of these vehicles evolve, so do their brains. At the Center for Ocean Engineering, visiting scientist Michael Benjamin has successfully demonstrated that the autonomous navigation system he developed for a group of MIT’s autonomous kayaks, or SCOUTs (surface craft for oceanographic and undersea testing), also works with undersea vehicles (see “Autonomous Kayaks,” January/February 2007). This system generally hews to the same waypoint-following strategy that most AUVs employ. But now Benjamin is upgrading it so that it can capitalize on new acoustic communications technology that will help the AUVs talk to each other, and to buoys or boats on the surface–reducing much of the guesswork involved in underwater navigation.
But better communication is only part of the story. Benjamin also attributes the success of his system to lowered costs, which mean that he and a colleague can drag a pair of kayaks down to the Charles whenever they please. The result, he says, is that he can be more adventurous when writing the code–he’s not limited to one test a year, during which everything has to go absolutely right. Lower costs also mean that more groups can experiment with AUVs. Add open-source software and open collaboration to the mix, he says, and you get exponential progress.
Such collaboration seems like a natural step for AUV research, which already draws on fields from mechanical engineering to AI. And in Chryssostomidis’s view, the more collaborators, the better. “To be frank,” he says, “there are enough challenges that we could put the whole of MIT to work on it.”