Do physical connections improve the control and performance of robot swarms?
A Lehigh University research team explores the question by building novel control algorithms and linked robots. Prototypes will be put to the test in the Defend the Republic Drone Competition Nov. 17.
Grant and Award AnnouncementIMAGE:
EIGHT TEAMS, INCLUDING LEHIGH AND SEVEN OTHER UNIVERSITIES, WILL COMPETE IN THE Defend the Republic Drone Competition NOV. 17 ON LEHIGH’S MOUNTAINTOP CAMPUS.
view moreCREDIT: LEHIGH UNIVERSITY
In nature, swarms can accomplish amazing things. Schools of fish can more efficiently find food and migrate. Flocks of birds can confuse predators. Bees, ants, and termites can work together to feed, defend and build their colonies.
Robotics researchers have long been trying to harness this ability to explore environments, capture objects and build structures using robot swarms.
“But the problem of how do we efficiently control all those robots is still unsolved,” says David Saldaña, an assistant professor of computer science and engineering in Lehigh’s P.C. Rossin College of Engineering and Applied Science.
Saldaña, who leads the SwarmsLab and is a member of the Autonomous and Intelligent Robotics (AIR) Lab at Lehigh, recently received a three-year, $755,000 grant from the Office of Naval Research to design aerial swarms and develop control methods that rely on physical inter-robot interaction.
Typically, he says, robot swarms are manipulated by controlling each robot individually. And to be successful, swarms must operate in well-defined spaces and environmental conditions—with negligible or constant wind, for example—and be tested repeatedly in that space.
“If I send the robots to a different environment where there might be obstacles, moving objects, and wind, it’s a different story,” says Saldaña. “We’re still very limited with what we can do.”
Saldaña and his team hypothesize that if the robots are physically connected—like ants that use their bodies to form a bridge for others in the colony to traverse—they may be easier to control and more robust to challenging conditions. The researchers will test three types of physical connectors: rigid bars, flat sheets and ropes. They will use each to perform a variety of tasks and measure how effective the connected robots are in grabbing and transporting objects.
What makes the team’s approach particularly novel, he says, is that they are developing new control algorithms while creating these robots from scratch.
“In our lab, we don’t buy robots, we make robots,” he says. “And since the robots we’re using don’t exist anywhere else, we have to push the boundary in control theory. As we’ve advanced the design, we’ve discovered some very interesting control problems that haven’t been explored in the literature. We’re constantly going back and forth between hardware and software, and that’s what makes this project so interesting.”
The team has already developed prototypes using each approach and will test them in November at the annual Defend the Republic Drone Competition. This year’s event is being held Nov. 17 on Lehigh’s Mountaintop Campus (in High Bay C2). It will feature eight university teams competing in a Quidditch-esque game (see: Harry Potter) in which autonomous robots vie to capture floating helium balloons and deliver them to the opponent’s goal: Lehigh University, George Mason University, University of Florida, Baylor University, West Virginia University, Indiana University, Virginia Tech and Drexel University The purpose of the competition is to drive research and innovation in vehicle design, multi-agent control, swarm behaviors, and communication.
“Lehigh has been to this competition twice, and this year, as hosts, we plan to have at least 30 robots entered,” he says. “It gives us the opportunity to evaluate them in a real setting doing repetitive work. And what’s cool about this event is that everyone is trying to achieve the same goal using different research approaches, and they’re all willing to share their solutions to different problems. We’ve learned a lot from our colleagues.”
Ultimately, the work Saldaña and his team are doing could be applied to tasks like drone delivery by providing more information on the types of connections necessary between robots to move packages of varying sizes. It could also help in search and rescue efforts when the swarm size necessary for the task isn’t clear. “If you don’t know how bad the situation is, you might not know if you need large or small robots, so if you can take many smaller robots and simply attach them depending on the situation on-site, that could solve the problem.”
Harnessing a robotic swarm and successfully manipulating its components to do essential tasks will be exceedingly difficult. But for Saldaña the challenge is not only endlessly interesting but also a great way to entice the next generation of computer scientists interested in both theory and practice.
“When I talk to students, I tell them, we do robot hardware, we do mathematics, we do physics, we do control theory, modeling, and rigorous experimentation. We’re constantly switching between software and hardware because that’s where robotics is—right in the middle.”
Research reported in this story is supported by the United States Navy/Office of Naval Research/Department of Defense, under award number N000142312535.
Related Links
No comments:
Post a Comment