It isn't easy maneuvering drones past obstacles, with many ending up shattered victims of trees.
Well, one MIT researcher sought to change all that in keeping drones from hitting trees. Andrew Barry, MIT computer science and artificial intelligence lab Ph.D. student, recently uploaded an experiment to YouTube, pitting his autonomous drone against a human-piloted drone in a forest.
The autonomous drone was hooked up with two quad-core CPUs and a pair of stereo cameras, working together in harmony to detect and avoid obstacles like trees—even at speeds of 30 mph in a forest. Its stereo-vision builds a real-time map of the area, helping it to detect and avoid trees and other objects that could spell its demise upon impact.
"Everyone is building drones these days, but nobody knows how to get them to stop running into things," Barry said in an MIT press release. "Sensors like Lidar are too heavy to put on small aircraft, and creating maps of the environment in advance isn't practical. If we want drones that can fly quickly and navigate in the real world, we need better, faster algorithms."
Barry's autonomous drone has algorithms that run 20 times quicker than existing software. Although his drone cost nearly $1,700 to build, the obstacle-avoidance technology will probably need to be a part of drones' makeup in the near future, leading up to the Federal Aviation Administration eventually regulating them. After all, drones crashing in places is one of the FAA's top concerns about the vehicles.
To see Barry's obstacle-avoiding autonomous drone in action, check out this video.
Via: FastCompany
