It does not do this through a self-aware artificial intelligence, but rather, the drone will fall back on a new obstacle-detection system which has been developed by Barry, since that is part of his thesis with MIT professor Russ Tedrake.
Barry shared, “Everyone is building drones these days, but nobody knows how to get them to stop running into things. If we want drones that can fly quickly and navigate in the real world, we need better, faster algorithms.” You know what they say, if you want something done, you’ve got to do it yourself, and so that is what Barry went ahead to do.
This stereo-vision algorithm is said to be 20 times faster compared to current software, letting the drone detect objects while building a full map of its surroundings in real time, and it operates at 120 frames per second, while boasting of depth information processing at a speed of 8.3 milliseconds per frame.