Velocity sensing for moving targets

The problem

Lidar outputs the position of points. Velocity is inferred by comparing consecutive frames — and the further, sparser and more occluded the target, the less stable that inference. Yet velocity is what decides whether to brake: is that point a stationary guardrail, or a car that is slowing down?

How we approach it

FMCW emits a continuous beam whose frequency varies linearly. The beat frequency between the return and the local oscillator encodes range and Doppler shift at once — range and radial velocity come from the same measurement.

The boundary is worth stating: Doppler gives the component along the line of sight. A target crossing laterally has little radial signature, and that motion still needs frame-to-frame information. Radial velocity earns its keep in longitudinal dynamics — closing rate, braking ahead, telling stationary from crawling. That is exactly the quantity time-to-collision needs.

Coherent reception also brings structural interference immunity: direct sunlight, oncoming headlights and other vehicles" lidar never enter the mixing path.

Where it applies

Longitudinal decisions in ADAS and autonomous driving, drone and low-altitude obstacle avoidance, target separation in dense traffic.