MIT Researcher Develops Drone that Avoids Obstacles

“Everyone is building drones these days, but nobody knows how to get them to stop running into things,”  MIT researcher Andrew Barry said in a press statement.  The statement announced the release of an open source technology that will allow drones flying at 30mph to avoid obstacles without operator influence.  Barry, an MIT PhD candidate, developed the system as part of his thesis with professor Russ Tedrakein MIT’s Computer Science and Artificial Intelligence Labs (CSAIL.)

In his thesis, Barry cites the ability of birds to fly quickly through crowded environments as his inspiration for the technology:

The stereo-vision algorithm that Barry has developed is 20 times faster than any existing software.  It enables the drone to detect obstacles and create a detailed map of its environment in real-time. The software operates at 120 frames per second, and can extract depth information at 8.3 milliseconds per frame.

Barry built a drone to demonstrate his software from off-the-shelf components for about $1,700.  The 34-inch drone weighs just over a pound, and carries two cameras (one on each wing) and two small processors of the kind used in cellphones.  The software eliminates the need for additional sensors on the drone to detect obstacles.

“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,” says Barry.

Barry’s open source software, poetically called the “Flying through forests project”  is available here at Github; and a video of the drone in action can be seen here.