A comparison of two Crame´r-Rao bounds for nonlinear filtering with Pd<1

The paper presents a comparative study of two recently reported Crame´r-Rao lower bounds (CRLBs) for nonlinear filtering, both applicable when the probability of detection is less than unity. The first bound is the information reduction factor CRLB; the second is the enumeration method CRLB. The enumeration method is accurate but computationally expensive. We prove in the paper that the information reduction factor bound is overoptimistic, being always less than the enumeration CRLB. The theory is illustrated by two target tracking applications: ballistic object tracking and bearings-only tracking. The simulations studies confirm the theory and reveal that the information reduction factor CRLB rapidly approaches the enumeration CRLB as the scan number increases.