How is the Kalman filter used for object tracking in ADAS?

The Kalman filter is an algorithm that estimates object state (position, velocity) by combining predictions from a motion model with noisy sensor measurements. For ADAS tracking: Prediction step uses kinematic model (constant velocity, constant acceleration) to estimate next state; update step incorporates new sensor measurements, weighted by measurement uncertainty. The filter maintains state covariance representing estimation confidence. Extended Kalman Filter (EKF) handles non-linear models; Unscented Kalman Filter (UKF) provides better non-linear approximation. Multi-object tracking uses multiple Kalman filters with data association algorithms (Hungarian algorithm, JPDA) to match detections to tracked objects across frames.