Nov. 6, 2012 at 2:25 PM ET
Researchers have built a robot that uses a single camera and an algorithm to gain bird-like obstacle-avoidance smarts as it flies around the great outdoors.
That is, the robot is smart enough to avoid smacking into trees.
Robots have previously been programmed with the smarts to avoid objects while flying in human-built environments such as hallways and parking garages, but the great outdoors presents the challenge of more difficult route finding through larger distances.
To overcome it, Ashutosh Saxena and colleagues at Cornell University programmed the robot to turn an image from its video camera into a 3-D model of the environment, determine what features are obstacles and compute safe a path through them.
To do this quickly, the team is “running their algorithm on a neuromorphic hardware platform based on the collective firing of a network of artificial neurons,” IEEE explains. In other words, the hardware acts like a bird brain.
In 53 tests in obstacle-rich environments, the robot — an off-the-shelf card-table-size quadrotor — crashed only twice, both times due to winds.
In future versions of the robot, the team aims to improve its ability to account for the winds as well as avoid moving objects such as real birds. To test the latter, Saxena suggests people throw tennis balls at the flying robot.
Ultimately, a robot with such autonomous flying skills may be well suited for search and rescue missions beyond the reach of traditional radio communications, such as on the battlefield. That may explain why the project is funded by DARPA, the military’s futuristic research arm.
Check out the video below to see the quadrotor in action. More information is available from Cornell.
– via IEEE