Sensor Fusion: Why Self-Driving Cars Use Many Sensors
Every sensor has a weakness. Combining cameras, radar, LiDAR and motion sensors gives a picture better than any one alone.
The IMU reports tilt and spin rate, and the controller pushes back instantly.
A camera sees colour but struggles in glare and darkness. Radar sees speed and works in fog but has low detail. LiDAR gives shape but can fail in heavy rain.
Sensor fusion merges all these readings, weighting each by how much it can be trusted, often with a filter such as a Kalman filter, to get one steadier estimate.
The robot keeps working when one sensor is fooled, and gets a more accurate position and a more reliable view of obstacles.
Relying on one sensor means one glare, one fog or one failure can blind the whole machine.
More sensors add cost, calibration work and the problem of what to do when they disagree.
Why combine several sensors?