New Data Could Lead to More Agile Autonomous Vehicles
As the sun sets in a remote stretch of South Texas, an engineer from Boston University’s Intelligent Mechatronics Lab (IML) launches a two-pound unmanned aerial vehicle (UAV) toward a dense cloud of Brazilian free-tailed bats gushing forth from a cave like water from a fire hose. Thousands of the bats flap harmlessly past the UAV, while three ground-based, high-speed, infrared cameras and an onboard, 3-D, high-definition camera capture their flight paths.
Directed by College of Engineering Research Engineer, Kenneth Sebesta, and supported by recent graduates, Dane Sarcone (ME ’11) and Ryan Hunter (ECE ’11), the IML conducted the experiment to determine how large numbers of bats can fly so close to one another and past unexpected obstacles without colliding—a capability they hope to translate into flight control systems that will significantly boost the agility of UAVs and other autonomous vehicles.
The IML team’s effort is part of a five-year, Office of Naval Research-funded project called AIRFOILS (Animal Inspired Flight with Outer and Inner Loop Strategies) led by biologists, computer scientists and engineers from Boston University, the University of Washington, the University of Maryland and the University of North Carolina at Chapel Hill. The objective of the $7.5 million project is to investigate how diverse airborne species sense and react to their environment in forests, caves and other cluttered spaces—and apply that new knowledge to the design of more agile unmanned vehicles for disaster recovery, environmental sensing, military surveillance and other applications.
Toward that end, Sebesta and his team aim to parse out the choices bats make that allow them to maintain their stable, tight formation without colliding.
“There are, in some caves, literally a million bats emerging within half an hour,” said Sebesta. “Humans could never do this; put a million pilots in a similar situation, and very quickly you’ll have 10 people left and a bunch of scrap metal on the ground.”
It took Sebesta two tries to produce a four-rotor, “quadcopter” UAV that could serve as an unexpected obstacle, collect data, and avoid harming the bats.
Concerned that specialty replacement parts in the Texas backcountry would be impossible to find, his design centered on components that could be found anywhere in the U.S. Fashioned from Home Depot aluminum towel racks, fiberglass kite rods, bamboo and netting, the frame for the first version of the UAV proved no match for the stiff winds and hot temperatures of South Texas in June; it now hangs on a wall at the IML. Shifting from aluminum to lighter-weight carbon fiber rods, Sebesta assembled the second version, “Batcopter 2.0,” in a single afternoon.
“The basic concept is simple: You want the four motors to point up, you want the UAV to be rigid so the frame won’t twist, and you want it to be light,” said Sebesta. “So I sat down and started creating.”
Using carbon-fiber arrow shafts donated by a hunting shop in Texas, he formed a tic-tac-toe-shaped frame; bound the shafts with glue and twine; attached four motors, a control unit and battery; softened the edges with blue packing foam; added netting around the craft to protect the bats; and used heavy-duty zip-ties, velcro and double-sided tape to hold everything together.
Remotely controlled using OpenPilot, an open-source autopilot platform for small UAVs, Batcopter 2.0 endured four, five-minute flights before overheating motors brought the experiment to an end. The flights yielded about 20 gigabytes of video showing the bats maneuvering past the UAV.
AIRFOILS computer scientists will subsequently convert the high-speed, high-resolution, infrared images into a three-dimensional computer model of bat trajectories. Based on patterns they see in this model, Sebesta and colleagues at the IML will attempt to develop flight control algorithms that approximate the bat flight—and ultimately test them on real vehicles.
“We’ll look at that data, try to understand what the bats are doing, and see if we can find a generalized theory we can apply that predicts the way the bats are going to behave,” said Sebesta.