What Lidar Is Actually Good At

A radar gun at a ballpark does one job superbly: it reports that a pitch was 97.3 miles per hour. It does not tell you whether it was a fastball or a cutter, whether the pitcher was tiring, or whether the catcher expected it. One measurement, high precision, silence on everything else.

Lidar is in a similar spot. It sends laser pulses and times the return, producing a cloud of points with directly measured distances. That range is typically accurate to centimeters, and it is measured rather than inferred. There is no extra guess sitting between the pulse and the number.

It does not tell you whether it was a fastball or a cutter, whether the pitcher was tiring, or whether the catcher expected it.

It also works when a scene is dark. A camera needs photons from headlights, streetlights, or the sun. Lidar brings its own illumination, so an unlit road at 2 a.m. is not automatically harder than noon. It can also help in some low-contrast cases that give cameras trouble, such as a gray vehicle against a gray wall.

Those advantages are why lidar became common on research vehicles. When a team is proving a concept, measured geometry removes a whole class of argument.

The live question is not whether lidar is good at ranging. It is. The live question is whether precise geometry is the binding constraint of driving, and what it costs to put that precision on millions of vehicles instead of a few hundred.