There are a lot of stories about drone crashes out there. The unmanned aerial vehicles are, and crash without any help into and on . Now, new research out of MIT's Computer Science and Artificial Intelligence Lab (CSAIL) may make drones a lot better at not getting grounded by obstacles.
CSAIL Ph.D. student Andrew Barry developed a drone with an obstacle detection system and then set it loose in a tree-filled field where it automatically made adjustments to avoid smashing into trunks and limbs as it reached speeds of up to 30 miles per hour (about 48 kilometers per hour). Barry worked on the system as part of his thesis with MIT professor Russ Tedrake.
Barry built his self-preserving drone using standard components that cost about $1,700 (roughly £1106, AU$2373), including a camera on each wing providing the drone with stereoscopic vision. The drone also had two processors "no fancier than the ones you'd find on a cellphone," according to a report about the project from CSAIL.
What really makes the drone stand out from other more clumsy fliers though, is the algorithm Barry embedded into software that allows the drone to sense its surroundings in real time and steer away from obstacles. The superfast software extracts depth information from the camera feeds (which provide 120 frames per second) at a speed of 8.3 milliseconds per frame.
Helping the software keep up with the drone's 30 mph speed is the fact that it only takes readings every 10 meters (33 feet) out. This is different than other slower-moving drones that analyze readings from multiple shorter distances (like every 1 or 2 meters).
"You don't have to know about anything that's closer or further than that," Barry said. "As you fly, you push that 10-meter horizon forward, and, as long as your first 10 meters are clear, you can build a full map of the world around you."
Barry is now working to improve his algorithm to function at different depths, which would allow the drone to fly in denser settings like a forest or a crowded city.
"Our current approach results in occasional incorrect estimates known as 'drift,'" he said. "As hardware advances allow for more complex computation, we will be able to search at multiple depths and therefore check and correct our estimates. This lets us make our algorithms more aggressive, even in environments with larger numbers of obstacles."