Small helicopters have learned to fly themselves through challenging aeronautic routines with an “apprenticeship” to an expert radio control pilot, researchers say. The clever robots were first steered through the maneuvers several times by the pilot while the helicopters’ computers recorded every movement; then the computers used an algorithm to determine the “ideal trajectory” that the pilot was aiming for on each loop or flip, and replicated those motions when they set off on their own into the wild blue yonder.
Researchers say the helicopters’ self-taught skills are particularly impressive due to difficulty of flying helicopters and their nature to always tend to an unstable state. “The helicopter doesn’t want to fly. It always wants to just tip over and crash,” said Garrett Oku, the pilot [TG Daily]. Because helicopters have to constantly adjust to changing wind currents, the inventors couldn’t simply program them to fly a set routine.
The artificial-intelligence helicopter, an off-the-shelf model other than its new brains, can do traveling flips, rolls, loops, stall-turns with pirouettes and more. It can even do the “tic toc,” in which the helicopter, while pointed straight up, hovers with a side-to-side motion as if it were the pendulum of an upside down clock [LiveScience]. A video shows that the helicopter could eventually complete the stunts with more precise and consistent results than the pilot could achieve.
Researchers say the technology could be used to scan a forest fire for dangerous hotspots, but the most obvious applications involve the military. But while self-teaching war bots might, in theory, seem like a great idea, in practice commanders prefer to keep close tabs on their robots just like they do their human soldiers. “The real problem with autonomy is … I think military guys want to have control over everything they have in their battlespace,” Army researcher Steve Petty said. “They don’t trust autonomy” [Wired News].
Image: Ben Tse