Hummingbird Robots Use AI to Go Where Drones Can't

Hummingbird Robots Use AI to Go Where Drones Can't

But their search-and-rescue skills are still being refined.

Researchers at Purdue University engineered artificially intelligent hummingbird robots, designed to easily maneuver through collapsed buildings and cluttered spaces to find trapped victims.

Trained on machine learning algorithms to behave like the delicate birds they’re modelled after, the devices “know” how to move on their own, and can teach themselves new tricks.

The hummingbots, for instance, can’t see (yet); they sense objects by touching surfaces. Each touch, according to a Purdue press release, alters an electrical current, which researchers can track.

“The robot can essentially create a map without seeing its surroundings,” Xinyan Deng, an associate professor of mechanical engineering at Purdue, said in a statement.

“This could be helpful in a situation when the robot might be searching for victims in a dark place,” she continued. “And it means one less sensor to add when we do give the robot the ability to see.”

The team will present their work at next week’s 2019 IEEE International Conference on Robotics and Automation in Montreal.

Drones simply can’t be made much smaller; they wouldn’t be able to generate enough lift to support their weight. Hummingbirds, however, don’t rely on conventional aerodynamics.

“The physics is simply different,” Deng explained. “The aerodynamics is inherently unsteady, with high angles of attack and high lift. This makes it possible for smaller, flying animals to exist, and also possible for us to scale down flapping-wing robots.”

Hummingbirds have long been the subject of robotics research: In 2011, AeroVironment (commissioned by DARPA), built a cyborg that was heavier and slower than the real thing, with helicopter-like flight control and limited maneuverability.

Deng & Co. took a hands-on approach to their studies, spending multiple summers in Montana examining hummingbirds. The team documented key maneuvers (like a rapid 180-degree turn), then translated them to computer algorithms for the robots to learn.

Like their IRL inspirations, Purdue’s hummingbots can fly silently and stay steady through turbulence, making them ideal for covert operations. Each device requires only two motors, and can control each wing independently.

“An actual hummingbird has multiple groups of muscles to do power and steering strokes,” Deng said. “But a robot should be as light as possible, so that you have maximum performance on minimal weight.”

Moving forward, researchers hope to add a battery and sensing technology (like a camera or GPS).

Simulations of the technology are available open-source on Github.

未经允许不得转载:国际 » Hummingbird Robots Use AI to Go Where Drones Can't