LiDAR

LiDAR works by shooting out light and timing how long it takes to bounce back, which tells you the distance. The name itself means "light-based radar."

Imagine walking into a pitch-dark room. LiDAR fires beams of light in every direction and plots each reflection as a point — three meters here, five meters there — building up the shape of the room point by point. This cloud of points is called a point cloud. The distance measurements are extremely accurate. What it can't do is read what's written on the wall. Since it only times how long light takes to return, color and text simply don't exist as information for LiDAR.

Cameras are the opposite. They can read text and see color, but figuring out how far away something is has to be computed from the image alone. That's why using a camera to judge distance requires much more sophisticated AI. This is the root of the long-running split in the self-driving and robotics world between the LiDAR camp and the camera camp.

Cost used to be another dividing line. There was a time when a single LiDAR unit cost more than a car, and that gave the camera camp a strong cost argument. Prices have dropped a lot since then, so that argument doesn't carry the weight it used to.

"Matic's cleaning robot sees the home with five cameras instead of LiDAR" — this means it skips that spinning sensor that sweeps the ceiling and relies on cameras alone. In other words, it ditched the measuring tape and decided to get by on eyesight instead.