Derivation of the CRLB in nonlinear filter and application to target tracking
Wei Wei · 2007
For the nonlinear filter with overlapped zero-mean Gaussian white noise, a theoretical error performance Carmer-Rao Low Bound (CRLB) was derived using the unscented transform method. A suboptimal algorithm-- the Range Parameterized Unscented Kalman Filter (RPUKF) was presented to deal with the bearings-only target tracking using the opto-electronic signal. The performance curves of the proposed algorithm and the Range Parameterized Extended Kalman Filter (RPEKF) were compared with the CRLB in the simulation. The results show their asymptotical agreement with the theoretical bound and the advantageous RPUKF performance.