Pioneer Study on Near-Range Sensing with 4D MIMO-FMCW Automotive Radars

This paper explores the feasibility to detect and classify static or moving targets with weak radar cross sections i.e. pedestrians with a 4D MIMO-FMCW automotive radar. A straightforward classification using data integration over time for pattern recognition is not possible due to the low point cloud density. Hence a particle filter is employed, where the probability density function of the points are used for object classification instead. Measured results of an adult person walking and a child dummy on the ground classified with a particle filter has been shown to be promising. Their locations relative to the radar are tracked and the pedestrians' heights can also be also estimated.