Watch MIT’s new drone juke and jive through trees on its own

We may earn a commission from links on this page.

One of the biggest issues preventing drones delivering packages—other than many, many regulations—is figuring out how to make sure they don’t crash into things. But new research out of MIT may have figured out how to turn the dreams of Google, Amazon, and myriad other companies into reality.

Andrew Barry, a PhD student in MIT’s Computer Science and Artificial Intelligence (CSAIL) lab, has developed a system that allows a drone to detect objects up to 10 meters (32 feet) away about twenty times faster than any other existing software. According to a release from MIT, Barry’s drone carrying the new setup used only off-the-shelf parts—so no fancy laser-radar systems like those in Google’s self-driving cars—and only cost about $1,700 to put together. (That’s roughly the same price as a high-end consumer drone.)

In the video, Barry’s fixed-wing drone—which is shaped like a tiny plane—can be seen whipping round tree branches without any human input at up to 30 mph (48 kmh). Previous systems for self-flying drones have relied on cameras that are computing what’s in front of the drone at various depths. This takes up a lot of processing power, and means the drones can only fly at a few miles per hour. According to MIT, Barry realized that when traveling at faster speeds, things don’t change too frequently. The drone’s camera sees its surroundings at 120 frames per second, managing to perceive depth per frame in about 8 milliseconds, always looking about 10 meters ahead.

“You don’t have to know about anything that’s closer or further than that,” Barry told MIT. “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 also uses on-board odometry systems, and uses previous map it created to help it navigate. Barry is working on improving his detection algorithm further so that the drone could duck in and out of denser forests on its own. He’s also open-sourced the algorithm online so that anyone can try it out. Perhaps the companies trying to start up drone delivery services—or any drone that could crash into things outside—should take a look.