Bouncing Light Beams: How LiDAR Helps Self-Driving Cars See

Picture a spinning dome on top of a car, firing more than a million invisible laser pulses every single second in every direction. That's LiDAR, and it's how many self-driving cars 'see' the world well enough to spot a cyclist, a stop sign, or a ball rolling into the street — even in total darkness. Waymo's self-driving cars use a rooftop LiDAR unit that spins 360 degrees ten times per second, sending out over a million laser pulses and catching every single one that bounces back. That's how the car builds a 3D map of everything around it, updated ten times every second.

- What the letters in LiDAR actually stand for - How a bouncing laser beam turns into a distance measurement - How thousands of those measurements become a 3D picture - Why cars use LiDAR alongside cameras and radar instead of just one sensor

LiDAR stands for Light Detection and Ranging. It works by shooting out a short pulse of laser light and starting a stopwatch the instant it fires. When that pulse hits something — a tree, a curb, a person — some of the light bounces straight back to the sensor. The sensor stops the stopwatch the moment it catches the returning light. Light travels at about 300,000 kilometers per second, so even a round trip to an object 30 meters away and back takes only about 0.0000002 seconds (200 nanoseconds). A computer inside the sensor does the math — distance equals speed of light times time, divided by two because the light made a round trip — thousands of times per second, for every single pulse.

One laser pulse only tells you the distance to one point. A LiDAR unit fires over a million pulses per second in a spinning pattern, so within a fraction of a second it has measured distances to hundreds of thousands of points all around the car. Put all those dots together and you get what engineers call a point cloud — a 3D map made entirely of measured points, shaped exactly like the real world around the car, without needing a single photograph.

Self-driving cars almost never rely on just one sensor. Cameras see color and read text, like stop signs and lane paint, but they struggle in the dark or in bright glare. Radar bounces radio waves instead of light — it works great in fog and rain but gives a blurry sense of shape. LiDAR gives extremely precise distance and shape information in 3D, day or night, but it can get confused by thick fog or heavy snow, since those scatter the laser light. Engineers combine all three so the car's computer can cross-check what each sensor reports.

LiDAR isn't perfect. Heavy rain, thick fog, and falling snow scatter the laser pulses in every direction, so fewer of them make it back to the sensor cleanly — that can shrink how far the car can reliably 'see.' Early LiDAR units also cost tens of thousands of dollars each, which is part of why engineers keep working on cheaper solid-state versions with no spinning parts at all. And each spinning unit generates a firehose of data — over a million points every second — that the car's computer has to sort through in real time to find the pieces that actually matter, like a pedestrian stepping off a curb.