Flying a drone around obstacles is a difficult task for people as we have to rely mostly on our visual perception of the unmanned craft and if the number of obstructions in the path are more than what we can handle, the only answer is to use sensors and AI to our advantage. It is believed that eventually even the human element in drones will be decreased in the coming years. Flying past objects lying in the path of a robot is the first step towards making it completely autonomous.
Various organizations have been working on algorithms and AI systems for obstacle detection and evasion systems. MIT, the world’s premier engineering institution is also at the forefront of drone research and has recently come up with a new algorithm that dips, dives and evades through obstacles at a good speed of 30 mph. In a video released by the MIT researchers, we see how the drone can fly through the obstacles with ease. It also doesn’t even come close to crashing.
Andrew Barry is a PhD student at MIT and he uses a stereo camera to shoot video at 120 fps and it can map objects 33 feet from it and thus it can avoid them.
Here is the footage of the autonomous flight across the trees: