Saturday, April 09, 2005


Technology in Extreme Places: Underwater


By Davis D. Janowski
April 07, 2005

The research vessel Aegaeo lumbered along the Greek coast to a point 85 nautical miles southwest of Athens: the island of Kythira. Here a team from MIT's Autonomous Underwater Vehicle (AUV) Lab launched the yellow, torpedo-shaped Xanthos robot on yet another mission to demonstrate its capability as an undersea archaeological survey tool. Getting to Greece had been no simple task. MIT shipped over 2,700 pounds of robotics, batteries, and support equipment via air cargo from Boston to the port of Piraeus, loaded it aboard a ship, and kept a crew, researchers, engineers, and archaeologists ready to go on a tight timetable.

Running the untethered Xanthos along an ancient trade route 8 to 12 feet above the bottom of the sea (280 feet deep at this point), they discovered and photographed the anchor and other probable remains of a 17th century shipwreck. Sidescan sonar images collected by Xanthos, coupled with gigabytes of digital still images, assured the team and their hosts from the Hellenic Center for Marine Research that they'd discovered something human divers could not have reached. That means scientists will have a better set of tools for discovering and surveying the thousands of shipwrecks lying undisturbed and full of information—not just in Greek waters, but all over the world.

Deep Sea
People often forget that even as astronomers photograph planets in other star systems and as NASA scientists continue their off-road exploits on Mars, large swaths of our own planet remain unexplored. Most terra incognita is located beneath the oceans, at an average depth of 2.5 miles. And that's not even much: The deepest known seafloor on the planet is a place called The Challenger Deep, almost 7 miles down, at the bottom of the Pacific Ocean's Marianas Trench. These are places the fragile human body can never hope to reach—not without reinforced titanium, that is.

That's where AUVs come in. They are beginning to help us explore environments in which humans simply can't survive. While SCUBA gear and other breathing apparatus have been a boon to human exploration in the upper reaches of the ocean, they can't keep us alive beyond a few hundred feet. Making the trip to deeper reaches requires a submersible, like the venerable Alvin of Titanic fame, but only a handful of these vessels exist, and they have very long waiting lists. Such craft must also have a mother ship and crew, and so the expenses mount.

AUVs, on the other hand, are relatively inexpensive to build and operate, small enough to launch easily from shore, robust and self-sufficient enough to go about their business completely unattended, and able to return home on their own. No ship is required, no tether for power or control, no large support crew, and no humans endangered by the crushing pressure of the deep sea.

There are many practical applications for such vehicles, of course, including oil exploration, military work, seabed mapping, underwater archaeology, and deep-sea exploration, to name just a few. There are also many universities and institutions working on these systems, both in the US and abroad. Two of the three most respected development centers are found in Massachusetts, one at MIT and the other at the Woods Hole Oceanographic Institution (

Beginning in the mid-1980s, the AUV Lab at MIT received grants from the Office of Naval Research and the National Science Foundation to design and build AUVs. Though given specific design requirements (mine detection, in the case of the Navy) the folks at MIT hoped to produce a platform for the scientific community as well. One of the Navy's restrictions gave them a good starting point: The vehicle had to fit into a torpedo tube.

"In its purest sense there's this idea of the lab being an ethereal place where you come up with these ideas and then build them," explains Rob Damus, Chief Software Engineer with the AUV Lab. In reality, innovation is more often defined by the demands of the customer. "This really limits your ability to be sitting on a mountaintop coming up with the Red October drive," he points out. That single design constraint left the professors and students to determine the remaining hardware and software, get the craft to work, and fit into a relatively small, odd shape.

Though new craft are always on the drawing boards, MIT's Lab currently operates two AUVs: Xanthos, of the Odyssey IId Class (originally rolled out in 1995 but since upgraded several times) and a newer model called Caribou, of the later Odyssey III Class. The Caribou is now being produced by a company spun off from the Lab, called Bluefin Robotics Corporation ( ).

It's impressive because of its relatively low cost—around $400,000 per vessel (base price)—and small size, at about a fifth the size of comparable AUVs.

Using these vehicles, MIT's team has traveled far afield, to New Zealand, the Haro Strait off the coast of British Columbia, the Labrador Sea near Greenland, Elba, and twice to Greece. During their most recent trip in June 2004 Xanthos spent over 10 hours on the seafloor, surveying half a square kilometer of ocean bottom, and taking 10GB of 1.3-megapixel images.

Let's Talk Hardware
In order to keep costs down, virtually all onboard electronics in Odyssey-Class vessels are commercial, off-the-shelf parts. The main vehicle computer is a PC/104 motherboard with a 166-MHz Pentium-MMX processor. The team's current CPU of choice is the Kontron MOPSlcd6, though it uses more power than they would like (7-8W). "It's proven reliable, and it's fast enough to compile things in a semi-reasonable amount of time," explains Jim Morash, an electrical engineer with the Lab.

You might also recognize wireless cards from Cisco and a wireless bridge from Netgear, albeit somewhat modified, for use in short range communications while the AUV is surfaced. At the other end are modifications including use of a Hyperlink Technologies amplifier attached to an access point (on land the FCC might frown on this, but offshore it's not interfering with anyone). In fact, 30 to 40 percent of all the components and subsystems are off-the-shelf; the Lab uses glass instrument housings from Benthos, for example, which are rated to 22,000 feet.

But even with the off-the-shelf parts, the team had to do a lot of modification to get the components to work together. And some parts simply had to be developed from scratch. "Propulsion was totally 100 percent in-house," explains Damus. Thrusting and turning are functions that must be translated into power, and that meant getting a custom manufacturing job and a housing to mount it in. In the end, that meant two in-house mechanical engineers to design it, work with the manufacturer, and then test it.

MOOS, Sounds Like Moose
What runs all this hardware? None other than the Mission Oriented Operating Suite, or MOOS. It is the fourth AUV OS written by Paul Newman, who has since left MIT and become a lecturer with the Oxford University Robotics Research Group. Newman designed MOOS to be a lightweight piece of software for coordinating and executing many different AUV operations. It manages all on-board sensors, navigates, controls the actuators, and monitors the vehicle's safety. MOOS can currently run on Linux, Solaris, and Windows 2000, in addition to other less suitable OSs; the MIT group runs it on the Woody version of Debian.

The suite adheres to a star topology; each application within a MOOS community (a MOOSApp) connects to a single MOOS database. Three central libraries make up the code base of this object-oriented system. At heart, it's a client/server system—with MOOS acting as both client and server.

At any given time there are up to 15 processes running on the AUV. Three are responsible for navigation and driving, with the highest priority among them being a self-preservation process; among other things, it keeps track of the bottom and avoids it.

As the name suggests, when an AUV is submerged it operates autonomously using dead reckoning data fed to the main computer by a CrossBow AHRS (altitude, heading, and reference system, composed of a three-axis accelerometer and magnetometer). The AHRS is housed in the control sphere. In addition, the vehicle has a Datasonics PSA-916 sonar altimeter and Paroscientific Digiquartz pressure sensor. Any communications between the vehicle and staff go through a FreeWave spread spectrum radio modem housed in the aft sphere. While on the surface, data from a GPS is added to that of the other sensors and all of this is supplied to the main vehicle computer.

The Trouble with Running MOOS on GOATS
For a research institution, working with the military means more than funding. It also means acronyms …lots of acronyms. In 2002, for example, those of us at MIT's AUV Lab found ourselves preparing for a JRE between US and NATO for MCM in the VSW LSZ using SAS during the GOATs cruise. Whew.

What's all that mean? Basically, we'd been asked to drive Autonomous Underwater Vehicles (AUVs) that would showcase the latest work of our colleagues in the Acoustics group of MIT's Department of Ocean Engineering. This would take place during a joint research endeavor (JRE) between the US and NATO on the subject of mine countermeasures (MCM) in the very shallow water (VSW) littoral surf zone (LSZ) off the coast of Italy's Cinque-Terra.

(Who said robotics re-search is all bad?) We would be using a synthetic aperture sonar (SAS) at Generic Oceanographic Array Technologies (known as GOATs—a mostly at-sea symposium to demonstrate working hardware and software).

We planned on doing more than just driving, however. We intended to usher in a whole new era, an age of autonomous intervention through vehicle-to-vehicle interaction. MCM was a hot topic for the program managers at the Office of Naval Research (you guessed it, ONR), and they'd funded a lot of our work over the last four years. We needed to impress them. During GOATs we'd be running a fancy new Odyssey III-Class AUV affectionately dubbed Caribou. Its 1.1-meter payload area would be stuffed to the gills with the life's work of one of our Acoustics grad students: a state-of-the-art synthetic aperture sonar system.

Our planned demonstration called for Caribou and its SAS payload to prowl around searching for underwater mines (simulated, of course). When it found a target of interest, Caribou would transmit the coordinates to her companion AUV, Xanthos. Since Xanthos had an on-board camera, she would fly over the target and photograph it, providing visual evidence of the find.

Key to this, of course, was AUV-to-AUV communications. In the weeks leading up to our demonstration, I worked closely with Dr. Paul Newman, then completing his ingenius post-doctoral work in field robotics at the Lab. He'd fathered our new AUV operating system, known as the Mission Oriented Operating Suite, or MOOS. After tearing our hair out over scenarios that could threaten the vehicle's safety, and working through programming challenges from ThirdPartyTasks to SessionTimeOuts, we were ready.

On the last day of GOATS it was time to run our demo. Caribou began her survey and entered the minefield. Over the hydrophone array we could hear her contacting Xanthos with target coordinates to inspect. Triumph was at hand. On the tracking computer, Xanthos could be seen leaving her stationary "orbit" and heading toward…

"What's it doing?" Paul asked from the aft-bridge of R/V Alliance. "It looks like she's floating to the surface," I said, concerned. "Ops, Sonar. I no longer hear Xanthos' propeller," came the voice of Matt, our acoustics specialist. Sure enough, we soon saw the yellow hull of our normally reliable Xanthos bobbing to the surface. A post-mission analysis showed that our slaved-over ThirdPartyTask had indeed been invoked, but the simplest of obstacles had stopped us cold: The batteries aboard Xanthos had died.

The Future
The AUV world is constantly changing along with our understanding of the vast world beneath the waves. Some engineers are working on greatly improved hovering vehicles, flatfish- rather than torpedo-shaped. Such a platform will be more advantageous to deepwater archaeologists and marine biologists, who need to focus on a particular spot rather than cover ground. Future designs will also carry manipulator arms, collection baskets, and other gear.

Gliders are another promising AUV design for long-term study. These vehicles are propelled not by electrical motors, but by changes in their buoyancy achieved by pumping seawater in or out of their hulls. Such vessels can gather data on currents or temperature gradients for months at a time. Some researchers are exploring the possibility of AUV networks that collect data over a large area simultaneously. Scientists recently held the first such experiment in Monterey Bay.

Perhaps the most promising idea is that researchers will one day be able to sit in the comfort of a virtual reality lab and ride along with a vehicle. "Imagine being able to survey a deepwater shipwreck site in mere hours or days—a tiny fraction of the time and resources required today," said Damus. "The true gain, as with all robotics, is that using them you're not wasting a human's time…you're multiplying it."


Comments: Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?