Robot Helicopter self-taught in flight
Assisted by a new artificial intelligence system, a robotic helicopter is able to teach itself to fly and perform stunts, by watching and simulating these maneuvers done by other helicopters.
This autonomous helicopter relies on observation and registration rather than on software. Its observation model is an expert in what the Stanford University team calls “apprenticeship learning”, a 4-foot model helicopter flown by radio-control pilot Garret Oku, known to fly amazing acrobatics on any helicopter.
The study conducted by the Stanford University team, made up from professor Andrew Ng and his research graduate students proves that computers can acquire the same skill. Therefore, the artificial intelligence helicopter can do rolls, loops, traveling flips, stall-turns with pirouettes – “ by far the most difficult acrobatic maneuvers flown by any computer-controlled helicopter”, according to the Stanford professor.
The machine is also remarkable through its underlying technology: aftermarket instrumentation such as accelerometers, gyroscopes and magnetometers, since controlling a helicopter is known to be difficult – “ it always wants to just tip over and crash” , says Oku.
He then went on a seven-fight winning streak before losing his last two bouts, both under the Strikeforce banner. Much like Day, he will be looking to this bout to help propel him back into the limelight.
QJ: “I don’t really even NEED Chuck.”
This is one of the best blogs I