Self-Driving Cars Should Look Down, Not Just Ahead

There’s a reason Waymo, Uber, and even grocery store giant, Kroger, are testing their shiny new autonomous vehicles in southwestern cities like Phoenix. Yes, Arizona’s regulations are friendly to them, but the year-round good weather is the major draw. As sophisticated as these machines are becoming, they still struggle when fog reduces visibility, or when snow covers lane markings.

But eventually, AVs will have to learn to navigate the wintry mix. That's where WaveSense, a Boston-based startup, sees an opportunity. The company, launching formally today, wants to take technology developed at MIT for the military, and use it to give self-driving cars an extra sense. It says equipping vehicles with a radar looking down, penetrating the ground, will give them a new way to map exactly where they are in the world without relying on visual clues or GPS.

Autonomous vehicles use a suite of sensors to monitor the world around them. Most of their creators have settled on a combination of lidar laser sensors, radar, cameras, short-range ultrasonic sensors, and GPS, combined with on-board maps. But the final component mix that will turn cars into fully autonomous vehicles is far from finished. (Waymo, for example, added microphones too, to hear sirens, and Tesla says it can do it all with cameras, no lidar.)

For a robo-car to be able to navigate a city, it needs to know where it is, and where it should be, within an inch or so. Before carmakers let their creations loose, they usually send a mapping vehicle around the environment they’re going to drive in, to build up a detailed lidar picture of the signs, the access roads, the bike lanes, and the awkward junctions. Then, when an autonomous vehicle is cruising, it can use that map as a reference, even when GPS satellite navigation signals are blocked by tunnels or tall buildings.