robotics//sensor//radar

2025-06-13

A radar is a sensor that transmits radio waves and analyses their echoes to measure the range of objects and their radial velocity, and it is what lets a car keep its distance in fog and an air traffic system follow aircraft through clouds. The range comes from the round-trip time of the wave (or, in the frequency-modulated radars of cars, from the frequency difference between what is sent and what returns), and the velocity from the Doppler effect: an echo from a target moving towards the radar at speed \(v\) comes back shifted in frequency by

fD=2 v fc.f_D=\frac{2\,v\,f}{c}.fD=c2vf.

At the 77 GHz of automotive radar, 1 m/s of closing speed is a shift of about 513 Hz, so velocity is measured directly in each echo, where a camera or a lidar has to difference positions over time. Automotive radars work between 76 and 81 GHz.

It sees through what blinds light: fog, rain, dust, smoke and darkness barely affect it, which is why it is the sensor of adaptive cruise control and of many collision warnings.

Its weakness is angle. Its wavelength is millimetres, against the nanometres of light, so with antennas of a practical size its angular resolution is poor: two cars side by side far ahead can merge into one echo. Metal structures bounce the waves, producing ghost echoes of objects that are elsewhere, and a manhole cover or a bridge can look like an obstacle.

Its output is a list of detections with range, angle and velocity, with false alarms and misses, so it feeds a tracker (multi-target tracking) rather than a controller directly (perception). It is usually paired with a camera, and sometimes a lidar. Imaging radars that also resolve elevation are a niche advancing quickly.