MIT develops collision avoidance technology for small UAS

By Patrick C. Miller | November 12, 2015

Technology developed at MIT’s Computer Science and Artificial Intelligence Lab (CSAIL) enables small unmanned aerial systems (UAS) to detect and avoid obstacles while flying at speeds up to 30 miles per hour.

A video released by CSAIL shows a small, fixed-wing UAS flying through trees and making maneuvers to avoid them. The camera view onboard the UAS demonstrates how the aircraft automatically identifies obstacles along its path as green—no threat—or red—a collision danger. The drone automatically steers clear of identified threats.

“Everyone is building drones these days, but nobody knows how to get them to stop running into things,” said Andrew Barry, the CSAIL Ph.D. student who developed the system as part of his thesis with MIT professor Russ Tedrake.

“Sensors like LIDAR are too heavy to put on small aircraft, and creating maps of the environment in advance isn’t practical,” Barry explained. “If we want drones that can fly quickly and navigate in the real world, we need better, faster algorithms.”

According to CSAIL, Barry’s system runs 20 times faster than current software. The stereo-vision algorithm he developed enables the drone to detect objects and build a map of its surroundings in real-time. Operating at 120 frames per second, the software extracts depth information at a speed of 8.3 milliseconds per frame.

The drone used for the test weighs slightly over a pound and has a 34-inch wingspan. It was made from off-the-shelf components costing about $1,700. It’s equipped with a camera on each wing and two processors similar to those found on a cellphone. Barry’s software is open-source and available for online download.

Barry said previous approaches to sUAS collision avoidance were computationally intensive and made calculations too slowly, limiting speeds to five or six miles per hour. But he realized that at faster speeds, the surrounding environment didn’t change significantly between frames, which enabled his program to compute a small subset of measurements 10 meters (about 33 feet) away.

“You don’t have to know about anything that’s closer or further than that,” Barry said. “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.”

Barry hopes to further improve the algorithms to operate at more than one depth and work in environments such as a dense forest.

“Our current approach results in occasional incorrect estimates known as ‘drift,’” he said. “As hardware advances allow for more complex computation, we will be able to search at multiple depths and therefore check and correct our estimates. This lets us make our algorithms more aggressive, even in environments with larger numbers of obstacles.”

 

For more on the UAS Industry, follow us on Twitter @UASMagazine