Drones are usually blind—even those carrying cameras rely on a human to navigate safely and spot obstacles. But the octocopter in the video below used a prototype of a new compact, military-style radar system to detect its surroundings and track other aircraft.

As the drone takes off, you can see the traces of barbed wire fences and trees in the radar image on the left. Later, the radar detects another drone nearby and tracks its progress. In other tests the system showed it could detect a small single-propeller plane and an ultralight aircraft.

The technology, from startup Echodyne, could address a major challenge holding up dreams like Amazon’s plan to deliver packages by drone. The U.S. Federal Aviation Administration says that systems that enable drones to detect and avoid other aircraft are needed before they can be allowed to operate out of the sight of supervising humans (see “Now You Can Finally Use Your Drone to Make Money”).

Currently, no solution exists. Companies such as Intel and startup Skydio have shown systems that use cameras to sense and steer around ground-level objects such as trees (see "Daredevil Drone Flies Through Trees Like an Ace"). But they don’t have the range needed to avoid aircraft when flying hundreds of feet off the ground.

Echodyne says the radar system it will release early in 2017 can fill that gap. The company has developed a way to shrink a component called a phase shifter needed for a radar to scan its beam around the world to identify and track objects (see “Metamaterial Radar May Improve Drone and Car Vision”). Electronically scanning radar is currently mostly limited to military uses, because existing phase shifters are bulky and complex, and can cost hundreds of thousands of dollars.

Eben Frankenberg, CEO and cofounder of Echodyne, says that multiple drone companies are already testing his technology, but declines to name them, citing nondisclosure agreements.

The company is also testing its radars on cars, with a view to helping efforts to build autonomous vehicles.

Prototype driverless cars rely heavily on lidar, which uses laser light to build up a picture of the world (see “Fist-Sized Laser Scanner to Make Autonomous Cars Less Ugly”). Electronically scanning radar can’t collect such a detailed image of the shape of objects. But Frankenberg says it has other benefits. For example, because radar uses radio waves and not light, rain and snow aren’t a problem, the range is greater, and it can instantly report back on the speed at which cars and other things are moving, he says.