Autonomous Helicopters Teach Themselves Aerobatics
A fleet of scale-model autonomous helicopters operated by Stanford computer scientists can learn to fly complex stunts by “watching” other helicopters perform the same maneuvers, the research team said this week. The project illustrates the capability of “apprenticeship learning,” in which robots learn by observing an expert, rather than by following pre-programmed software instructions. Using artificial intelligence, the autonomous helicopters are able to fly a complex routine while correcting for variables such as wind gusts. During a flight, instruments monitor the position, direction, orientation, velocity, acceleration and spin of the helicopter in several dimensions. A computer crunches the data, makes quick calculations, and beams new flight directions to the helicopter via radio 20 times per second — with no human input. The technology could be useful in “training” autonomous helicopters to search for land mines or wildfires, said Andrew Ng, director of the Stanford research team. “In order for us to trust helicopters in these sort of mission-critical applications, it’s important that we have very robust, very reliable helicopter controllers that can fly maybe as well as the best human pilots in the world can,” he said. Stanford’s autonomous helicopters have taken a large step in that direction, according to Ng.
A fleet of scale-model autonomous helicopters operated by Stanford computer scientists can learn to fly complex stunts by "watching" other helicopters perform the same maneuvers, the research team said this week. The project illustrates the capability of "apprenticeship learning," in which robots learn by observing an expert, rather than by following pre-programmed software instructions. Using artificial intelligence, the autonomous helicopters are able to fly a complex routine while correcting for variables such as wind gusts. During a flight, instruments monitor the position, direction, orientation, velocity, acceleration and spin of the helicopter in several dimensions. A computer crunches the data, makes quick calculations, and beams new flight directions to the helicopter via radio 20 times per second -- with no human input. The technology could be useful in "training" autonomous helicopters to search for land mines or wildfires, said Andrew Ng, director of the Stanford research team.
"In order for us to trust helicopters in these sort of mission-critical applications, it's important that we have very robust, very reliable helicopter controllers that can fly maybe as well as the best human pilots in the world can," he said. Stanford's autonomous helicopters have taken a large step in that direction, according to Ng.
The autonomous helicopters can perform traveling flips, rolls, loops with pirouettes, stall-turns with pirouettes, knife-edge flight, Immelmann turns, inverted tail slides, and the hurricane (described as a "fast backward funnel") as well as a maneuver called the "tick tock," 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.