MIT Autonomous Drone Avoids Objects Using Obstacle-Detection System

Posted on November 6, 2015

MIT autonomous drone

Andrew Barry, a researcher at MIT's Computer Science and Artificial Intelligence Lab (CSAIL), developed an obstacle-detection system for drones as part of his thesis. The software enables drones to fly through a tree-filled field at upwards of 30 miles per hour without crashing into a tree.

Barry says in a statement, "Everyone is building drones these days, but nobody knows how to get them to stop running into things. 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."

The open-source software is available on GitHub. The stereo-vision algorithm runs 20 times faster than existing software. MIT says the software enables the drone to detect objects and build a full map of its surroundings in real-time. Barry was able to cut the algorithm speed when he realized the world a drone sees does not change much between frames. He made his algorithm compute using a smaller subset of measurements than traditional algorithms thus shortening the processing time. His algorithm uses distances of just 10 meters away.

Barry says, "You don't have to know about anything that's closer or further than that. 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."

The drone weighs just over a pound and has a 34-inch wingspan. It was built using off-the-shelf components costing about $1,700. Here is a video of the drone in action:



Photo: MIT