Instantaneous lateral velocity estimation of a vehicle using Doppler radar

High-resolution image radars open new opportunities for estimating velocity and direction of movement of extended objects from a single observation. Since radar sensors only measure the radial velocity, a tracking system is normally used to determine the velocity vector of the object. A stable velocity is estimated after several frames at the earliest, resulting in a significant loss of time for reacting to certain situations such as cross-traffic. The following paper presents a robust and model-free approach to determine the velocity vector of an extended target. In contrast to the Kalman filter, it does not require data association in time and space. An instant (~ 50 ms) and bias free estimation of its velocity vector is possible. Our approach can handle noise and systematic variations (e.g., micro-Doppler of wheels) in the signal. It is optimized to deal with measurement errors of the radar sensor not only in the radial velocity, but in the azimuth position as well. The accuracy of this method is increased by the fusion of multiple radar sensors.