Improved radar tracking using a multipath model: maximum likelihood compared with eigenvector analysis

The performance of the MUSIC algorithm and many other superresolution methods degrades severely with the highly correlated multipath signals encountered in radar low-angle tracking. The paper presents a new eigenvector-based method in which the search vector is replaced by a deterministic specular multipath model. The performance is then compared with that of maximum likelihood using the same model and the well known MUSIC algorithm using spatial smoothing. Simulations and experiments at X-band indicate that the use of a specular model combined with an array radar having frequency agility gives much more accurate tracking than the conventional approaches. The experiments were conducted at Sylt (North Sea), Germany, using corner reflectors mounted on poles inserted in the sea bed; sea conditions varied from sea-state one to five