Autonomous Systems and Controls Lab
Read more and see videos of the robots at the Autonomous Systems and Controls Laboratory website.
Whether they’re adding brains to existing systems or designing entirely new robots, researchers in the Autonomous Systems and Controls Laboratory (ASCL) don’t just simulate: their systems dive into real waters.
Applying its expertise to both surface and underwater vehicles, the team analyzes each situation and adds whatever is necessary to achieve the autonomy required. Sometimes this means developing fundamentally new algorithms and integrating them with an existing boat, sometimes creating a new robot with added physical capabilities, and frequently integrating both physical and software features of the projects.
Research that is so dependent on successful fieldwork presents special challenges. According to ECE professor Dan Stilwell, director of the ASCL, “there’s a big difference between working in the lab and working in the field, and that gap is huge. There are all sorts of new failure modes, and all sorts of risks. Our vehicle disappears on purpose, so if it doesn’t work right we won’t know. It just won’t come back. In our case, that means we’ve just wasted two years of work, so our tolerance for risk is very low.”
There are also special challenges posed by field work. “A team may spend long days in the hot sun or the bitter cold, depending on the season. In these circumstances, effective teamwork is a necessity.” But, as Stilwell says, “we do get to go out to the Gulf of Mexico and hang out in a boat.” His team has tested systems not just in the Gulf, but also in coastal and fresh water rivers and lakes of Virginia, the rivers of Louisiana, and the Chesapeake Bay.
In collaboration with the Naval Postgraduate School and Craig Woolsey of aerospace and ocean engineering (AOE), Stilwell and his team are adding intelligence to existing surface vehicles. They are developing a sensing and autonomy package that can make any boat autonomous, giving the Navy the option to let their boats transition from manned to unmanned at the flip of a switch. Stilwell explains the project as “the ability to make a boat smart, not the design of a new boat. The Navy doesn’t need another robot,” he continues, “the Navy needs to make its robots smart.”
A large part of this project is their development of new classes of guidance algorithms. “At some point the vehicle has to decide where to go and how to get there,” says Stilwell. Their new algorithms allow an autonomous vehicle to operate in extremely large and mostly unknown environments.
“We can’t run global algorithms in real-time, because the domain is just too big,” Stilwell explains. “The amount of time needed to run calculations is consistent with how large an area you’re planning over.” To deal with this, Stilwell continues, “our algorithms switch between thinking globally and thinking locally.” The global domain is a big picture of the larger area, while the local domain includes trajectories for where the vehicle needs to go immediately, at a very detailed level, and can be computed in real-time.
Incorporating info from the global domain, the local algorithms can make decisions rapidly. “The vehicles start with a map, but that map will have lots of errors,” according to Stilwell. As long as the local trajectories and data correspond relatively well to what is known of the global information, the vehicle will continue based on the local trajectories.
“We don’t have to process globally until we see that what we’re doing locally and what we would have done globally are different in a meaningful way,” says Stilwell. “We don’t have to think globally very often,” he continues, “not until the theory says it absolutely has to be done, and in practice that turns out to be seldom. The vast majority of the time we’re only looking ahead a minute or two. We’re able to effectively combine global and local in a smart way.”
Working with surface vehicles also presents challenges the team doesn’t face in their work with underwater or ground robots. “Dynamics count for a boat,” explains Stilwell, “A boat can slip around a corner. It can go sideways. We had to develop some new theory that deals with the fact that dynamics really matter.”
Most autonomous vehicles are closely monitored during operation, but Stilwell and Wayne Neu of AOE are leading a team that has built a different kind of autonomous underwater vehicle (AUV), in collaboration with the Naval Oceanographic Office (NAVO). This new AUV operates alone and can go long periods of time without communication. Even more unusual, the AUV is self-mooring: it can anchor itself in a precise location of the ocean without drifting.
“This AUV can travel a long distance to a specified location, anchor itself to remain in that location for an extended period of time, then return home when its mission is complete,” says Stilwell, who, with Wayne Neu, oversaw a successful deployment of the AUV in Panama City, FL last summer.
Six feet long with a diameter of seven inches, the new AUV weighs almost 100 pounds and is more than twice the size of the 475 AUV, which has been used for many applications since 2007.
The main weight of the AUV is its battery payload. Because of its independence, the AUV must tote all its power onboard. “The AUV’s efficiency was a big concern for the hydrodynamic design,” says Neu. “We designed the shape to reduce drag and the propulsion system is as efficient as possible.” Approximately half of the battery power is used in traveling to the destination and back; the AUV reserves the remaining power for operating sensors while anchored and collecting data.
For surface communications and navigation, the AUV is outfitted with a GPS receiver, an RF modem with a couple-mile reach, and a satellite communications system. The AUV has an acoustic modem for communication while submerged, but a ship needs to be nearby in order to talk to it.
To get a GPS reading, the AUV surfaces periodically while traveling to its assigned location. Upon reaching its destination, the AUV must rapidly anchor itself—before it has time to drift.
According to Neu, developing the mooring system was the greatest challenge of this endeavor. The solution is a vacuum-attached false nose, which becomes the anchor. “When we were brainstorming what shapes we needed to grab and hold to the ocean bottom, we realized the anchor doesn’t have to hold. Down deep in the ocean, currents are small,” notes Neu.
The team found that the anchor took too long to separate from the AUV if they merely released the pressure and allowed it to drop. So, a cartridge of compressed air is used to blow the anchor off.
The anchor release system is another crafty solution: when it’s time to come home, the AUV uses a galvanic release to cut the wire attaching it to the anchor. A galvanic release relies on an anode and cathode separated by seawater and won’t work in freshwater bodies. When a potential is applied to one side, the 3-4mm stainless steel wire rusts through in about 10 minutes, freeing the AUV to return home. The anchor is left at its location. “The anchor is a cheap part,” Neu explains, “It’s just a hunk of steel appropriately shaped.”
The control algorithms for this AUV were created via the same process developed for the 475 AUV, but the development timeframe was weeks instead of the many months it took before. “We have robust control architectures and a good development process,” says Stilwell. Developing the process took time initially, but now the process is part of his laboratory’s toolkit.
In addition to military applications, applications for AUVs with these capabilities include environmental monitoring, acoustic monitoring for whales and other sea life, or taking readings during extreme events such as hurricanes when it isn’t safe for humans. Since the AUV can remain at rest for many months, it can conduct long-term marine life population assessments.
“What’s cool about this kind of work,” Stilwell says, “is that it’s extremely cross disciplinary. We have hydrodynamics issues, we need level-headed mechanical design, we have cost constraints; we have issues with electronics components and need to get some custom designed—but not too much, and we need autonomy software that can operate robustly without human intervention throughout all phases of an extended mission”
“All these different disciplines come together in one vehicle and it turns into a real system. We take all this knowledge from aerospace and ocean engineering and electrical and computer engineering and actually see it fly.” Students on the project leave with a great portfolio, he says. “They have taken a project from initial concept, through design, fabrication, and field trials.
“It’s great engineering.”
The Navy is also interested in a drifter that can operate successfully in rivers, and Barron Associates in Charlottesville, Va. has asked the ASCL to create one.
A drifter is mostly passive, taking measurements as it drifts. The specific challenge for drifting in rivers, however, is that all the drifters will converge to the same streamline—getting measurements from only one of many possible streamlines. “They would like to measure what’s going on in the river,” says Stilwell, “but they can only get this one streamline.” So, working with Craig Woolsey and Wayne Neu, Stilwell’s team is building a drifter with an actuator that can sense if it is following the drifter ahead and change to another streamline.
For this project, they can’t use flaps or fins because they would get fouled in a river. So the system includes only one propeller, carefully screened from “river muck,” and a reaction wheel inside. The reaction wheel consists of a metal plate that spins to rotate the craft.
The drifter is mostly spherical, with nothing that could easily get caught by any debris in the river.
The challenge, according to Stilwell is “how to deploy several drifters and have them talk to each other and avoid each other’s streamlines, and do it in such a way that it minimizes energy: the propeller should spin rarely.”
“Half our program is making vehicles smart,” he continues, “the other half is making new types of vehicles.”