3D LiDAR and Point Clouds

A 3D LiDAR stacks many laser beams at different tilt angles and spins them together. Each beam traces a ring, and together they build a 3D point cloud of everything around the robot. This is the main sensor on most self-driving cars.

Number of beams Horizontal resolution Distance to the person Color points by height

Drag the 3D view to look from another angle.

Where the rings come from

A beam tilted down by angle $\phi$ from a sensor at height $h$ hits flat ground at distance

$$r = \frac{h}{\tan\phi}$$

so each beam draws one ring. Beams that point down steeply make tight rings near the robot. Beams that point nearly level make rings that spread far apart.

What to notice

Move the person-sized box away and count the points on it. Close up, dozens of points land on it. At 30 or 40 m only a handful do, from just one or two beams, which is why detecting pedestrians at long range is hard. More beams and finer horizontal resolution help, at a higher price and more data to process. A 64-beam sensor at 0.2° resolution makes over 100,000 points per turn.