For years, researchers seeking to broaden the flying capabilities of robots have endeavored to decode one of nature’s best-kept secrets: the flight of the hummingbird, which is a remarkable hybrid of avian flight and insect-like hovering.
The value of such a skill could have applications ranging from finding victims trapped in collapsed buildings to drone-based espionage, the latter of which was explored by military supplier AeroVironment in the early 2010s. More recently, a team from Stanford studied the flight capabilities of hummingbirds and bats — the only other vertebrates with the ability to hover in place — with the goal of bringing new insight to robotic flight.
Now, researchers at Purdue University have trained flying robots with machine-learning algorithms based on hummingbird techniques, with high-flying results:
- After learning from a simulation, the robot "knows" how to move around on its own like a hummingbird would, such as discerning when to perform an escape maneuver.
- Artificial intelligence, combined with flexible flapping wings, also allows the robot to “teach itself new tricks.”
- Though the robot can’t see (yet), it can sense by touching surfaces. Each touch alters an electrical current, which the researchers can track.
- The robot remains steady through turbulence, which the researchers demonstrated by testing the dynamically scaled wings in an oil tank.
In developing their robots, the researchers studied hummingbirds for multiple summers in Montana. Key maneuvers such as making rapid 180° turns were documented, then translated into computer algorithms that could be learned in a simulation environment. Further study on the physics of both insects and hummingbirds allowed the researchers to build even smaller robots without compromising the quality of flight.
The team has built an insect-sized robot weighing just one gram; the hummingbird model, measuring in at 12 grams, can lift more than twice its own weight. Both are comprised of 3D-printed bodies, carbon-fiber wings and laser-cut membranes. They require only two motors, though they currently need to be tethered to an energy source for flight. It is anticipated that a battery will eventually be added, along with sensing technology such as a camera or GPS.
The robots’ silent flight, another trait borrowed from hummingbirds, increases their potential for covert operations. Each wing can be controlled independently of the other, which is one secret to the highly agile maneuvers performed by flying animals in nature.
The researchers will present their work at this year’s IEEE International Conference on Robotics and Automation (ICRA) in Montreal, which runs May 20-24, 2019. Open-source simulations of the technology are available on GitHub.
