Professor, former grad student taking different routes in autonomous research


The primary goals of faculty members in the School of Engineering and Applied Science (SEAS) are to conduct first-class research and groom students into future colleagues by providing them with a solid education and a love of engineering.

Naomi E. Leonard '85, a professor of mechanical and aerospace engineering, has accomplished both of these goals with aplomb.

Professor Leonard specializes in the study and design of autonomous underwater vehicles. She passed her zest for the topic onto her former graduate student Craig Woolsey *01. Professor Woolsey has joined the faculty at Virginia Polytechnic Institute and State University, more commonly known as Virginia Tech. There, he has begun his own research program that is related to the work he performed in Professor Leonard's lab.

Although Professor Woolsey and his mentor are proceeding along different paths, they are both studying control theory with the purpose of designing vehicles that will effectively respond to their environment autonomously.

School of gliders ready for testing in Monterey Bay

by Peter Page

Most people, if asked to think "sardine" with their eyes shut, will understandably first picture a rectangular can packed with many oily, little fish.

Living sardines are almost as chummy, spending their lives in close-knit schools. Like starlings of the sea, large sardine schools moving smartly in unison intimidate predators, and graze optimally, an adaptation that inspires Naomi E. Leonard's work.

A professor of mechanical and aerospace engineering and director of the Dynamical Control Systems Laboratory, Professor Leonard is developing a control theory for stringing a sensory necklace from a dozen handcrafted, diving robots that, gram for gram, are far simpler than a solitary sardine.

The gliders can't yet even communicate with each other while submerged; however, with frequent enough communication on the surface, it is planned that the gliders will emulate a sardine school's uncanny, synchronized response to environmental feedback.

"A school makes decisions and responds at a higher level of intelligence than the individuals," Professor Leonard said. "The fish school is our metaphor."

Schooling synergistically improves the capacity of ocean dwellers to respond effectively to a dynamic environment. Professor Leonard is confident that what works equally well for witless sardines and brainy dolphins and whales will make drones smart enough to descend a temperature gradient toward the turbulent thermal front of an upwelling event and return smarter from each dive.

She and her collaborators in the ONR-sponsored Autonomous Ocean Sampling Network II (AOSN II) project will set loose a school of gliders for six weeks this summer in Monterey Bay, Calif.

The gliders, working in two sets, will probe in and around the upwelling of cold, nutrient-rich water from an ocean canyon conveniently located within sight of land.

Every two hours the gliders will surface, orient themselves to each other via GPS, share data, and resume the dive.

"We want to collect specific, targeted data that will be best for understanding and forecasting, in this instance the upwelling event," Professor Leonard said. "In advance, we are writing these control theory algorithms that say 'Given what you are measuring and what you want to do, this is how you should respond.'"

Because so little has been measured about the Monterey Bay upwelling and so much uncertainty exists, the measurements Professor Leonard needs to hone feedback algorithms for the gliders won't be available until the gliders take them.

In the meantime, the necessarily adaptive algorithms to guide the gliders on their first dive are being tuned on the existing historical data, as well as on experience in the still pool at the Dynamical Control Systems Laboratory.

"We have lots of challenges," Professor Leonard said. "The sea is pretty dynamic, and we only know a little bit about what is going on down there. There are just so many parameters we don't know."

Upwelling is the result of persistent winds pushing aside warm surface waters, causing nutrient-laden cold water to rise from below. The exploratory voyage beneath Monterey Bay is equal parts an effort to gather physical and biological data on the upwelling from sea floor to surface, and applied, experimental control theory to network autonomous vehicles to gather the data.

The mission of the gliders, Professor Leonard said, is to find and sample features such as fronts and eddies associated with the upwelling, e.g., by descending temperature or salinity gradient fields. The interface of the upwelling and the ambient waters is turbulent, so much so that a single glider seeking colder water is unlikely to find the source. It could easily settle on a ledge or a trough (in a temperature field) and believe it had found the cold spot, but if the gliders work in a school, another will find colder water and signal the others so that they all continue toward the goal, Professor Leonard said.

The fish school is constantly playing "follow your neighbor," Professor Leonard said. There is no leader or hierarchy, yet the school processes all the information collected by the individual fish and reacts collectively with the quickness and focus of an individual.

"Each fish in the school is somehow contributing what information it gleans into a collective pool of information that the school reacts to collectively," Professor Leonard said. "A school seeking food succeeds better than a solitary fish that might be distracted by a small source of food and overlook a richer cache. The school has the advantage of detecting both sources and selecting to feed where there is more to eat. We are trying to emulate that."

The hand-built gliders have a Kitty Hawk simplicity that underscores the potential for someday deploying large schools of inexpensive and durable gliders. Professor Leonard foresees using control theory to design quick feedback mechanisms so the glider school can determine collectively where to range and how to deploy, which optimizes its opportunities to gather desired data.

In the upcoming experiment in Monterey Bay, Professor Leonard and her scientist shipmates will use an ocean forecast to perform daily planning for the glider school in addition to the re-planning automatically done every two hours. They will work on a three-day cycle such as this--tomorrow's plan is set, the plan for the day after that will be finalized tomorrow, there are various possibilities for the day after that that will be sifted as the data gathered by the gliders is considered and the model forecasts are updated.

At the heart of applying control theory, Professor Leonard said, is feedback and responding to feedback.

"We often don't understand a system perfectly, but we do have the ability to measure things," she said. "By measuring what can be measured about the system, we can base a decision on how to influence the inputs to achieve a desired outcome on both the modeled and measurable behavior of the system.

"The glider network problem is especially interesting because the gliders measure not only their own behavior but also the behavior of the ocean environment. In this context, we want our glider school to move in such a way, in response to what it is measuring, that it does a better job of measuring it."

Alumnus designs 'finless' spherical underwater vehicle

by Liz Crumbley

With spinning wheels, moving masses, and $675,000 in research grants, Craig Woolsey *01 of Virginia Tech aims to help improve the maneuverability, robustness, and reliability of underwater, air, and space vehicles.

Professor Woolsey, who joined Virginia Tech's aerospace and ocean engineering faculty in 2001 after receiving his Ph.D. in mechanical and aerospace engineering under the guidance of Naomi E. Leonard '85, is studying the design of advanced controls and control mechanisms for unmanned vehicles.

Suppose the U.S. Air Force's Predator Aerial Vehicle, in addition to taking off, flying within a limited range, and snapping photographs as ordered, could sense an anti-aircraft missile coming its way and take evasive action? Or suppose an unmanned submarine could be sent out to sea on its own, without being tethered to a ship, to track the boundaries of El Niño?

Muscles and brains

Such vehicles would have to use sophisticated control devices and advanced control algorithms--the muscles and brains of any unmanned vehicle--to perform complex maneuvers, Professor Woolsey said. His research will extend new methods of advanced control design to underwater vehicles by incorporating the important effects of lift, drag, and other fluid forces.

"Lift--the force that keeps an airplane in the air, for example--is an important consideration for air and ocean vehicles, and even some space vehicles," he said.

Building model

Professor Woolsey and his graduate students are building a spherical underwater vehicle with internal rotors (see illustration). The vehicle will be tested in a water tank being constructed in Virginia Tech's Randolph Hall, funded by the Research Division ASPIRES program.

"As a first step, we'll program the vehicle and have it perform maneuvers similar to those of an unmanned spacecraft," Professor Woolsey said. "The next step will be to add a streamlined hull and propeller and control how the vehicle swims."

Professor Woolsey also is exploring the use of moving masses for underwater vehicle control. One goal of his project is to find ways to perform successful maneuvers with most of the controls inside the vehicle, where they are protected from corrosion and biological fouling problems such as seaweed.

The devices and control strategies Professor Woolsey is developing also can be used for spacecraft, where contr ols have to be protected from intense forces and heat.

Another goal is to design controls that will enable the underwater vehicle to move at a low velocity, or even hover, without being thrown off-track by disturbances from waves or currents.

Professor Woolsey became interested in the field when he helped analyze foreign missile systems as a cooperative education student with the U.S. Central Intelligence Agency while an undergraduate at Georgia Tech.

The underwater vehicle consists of a glass spherical instrument housing containing three reaction wheels (not to scale here), a computer, an inertial measurement unit, and, to regulate depth, a ballast actuator with pressure sensor. It represents the first component in a modular experiment to study control of underwater vehicles using internal actuators.

This story first appeared in the summer 2002 issue of Virginia Tech Research Magazine and is reprinted here with permission.