Evaluation of Track Filters and Bias-Corrected PCRLB for Radar Elevation Multipath

In the presence of multipath, radar sensors produce biased measurements when the multipath is unresolved by the signal processor. The focus of this paper is to determine a lower bound on the accuracy of the estimated target when the raw radar measurements are contaminated by multipath. The uncertainty in the raw radar measurements usually sets an upper limit on the acceptable performance, and the Posterior CramérRao Lower Bound (PCRLB) sets the lower limit for the tracker posterior covariance matrix. These limits establish bounds for the uncertainty of the estimates produced by trackers. This research compares the performance of multiple tracking algorithms using simulations of a radar operating in elevation multipath conditions tracking maneuverable targets. It also combines the contributions of prior work by incorporating corrections for multipath biased measurements into a PCRLB algorithm. The results show that the PCRLBs (with & without bias correction) form a lower bound, but that the bias-corrected PCRLB represents a tighter limit. Kalman filters with either converted measurements or non-linear uncertainty representation provide the best position estimates; whereas particle filters give less uncertainty when multipath elevation biases are present. The PCRLB gives the lowest possible errors on the output of the tracking filter. This allows for the performance comparison between the lower bound and the output of various tracking algorithms. The bound can also be used for the evaluation of tracking filters for radars where closed source software is used.