Adaptive Cubature Kalman Filters And Their Applications In Radar Target Tracking

Cubature Kalman filter proposed recently is a good nonlinear filter which attracts much consideration. In standard CKF, the involving process and measurement noises are assumed to be the known independent Gaussian white noise sequences with zero means to meet the approximate linear minimum mean square error condition. However,the noises assumptions are often violated in the applications, which directly lead to performance degeneration even divergency of CKF. To overcome the limitations and enhance the robustness of CKF, aiming at different abnormal noises, several improved adaptive CKFs are proposed and applied in radar target tracking simulations tracking. The main works are as follows.1. Concerning negative impact on performance of CKF due to unknown system noises, an adaptive SCKF is proposed based on the standard SCKF and modified Sage-Husa noises estimator. Maneuvering target tracking simulations demonstrate that the proposed algorithm has the better accuracy and stability than the standard SCKF with unknown constant and time-varying process noises.2. To solve the estimation accuracy degradation of standard CKF with colored measurement noise, an improved CKF and its square root version are presented on the basis of a first-order Markov model, a Gaussian filter and a third-degree cubature rule. The target tracking simulation results verify the effectiveness of two proposed filters.3. An improved high degree cubature H-infinity filter based on the frame of nonlinear H-infinity filter and fifth-degree cubature rule is proposed for a class of nonlinear discrete-time systems with uncertain noises. Simulation results show its effectiveness.4. Two improved high degree CKFs (HCKFs) are proposed through fifth-degree cubature rule and two different decorrelated principles to solve the accuracy degeneracy of traditional HCKF with cross-correlation between process and measurement noises at the same time. Target tracking simulation results show the two proposed filter can overcome the limitations of CKF.5. The performances of three classical adaptive cubature Kalman filters (ACKFs) including ACKF based on maximum a posterior estimation (ACKF-MAP), strong tracking CKF(CKF-STF) and cubature H-infinity filter (CH∞F) are investigated through stability analyses and accuracy comparison simulations. The stability analyses show that stability of CH∞ depends on the value of scalar β, and stability of other two ACKFs are affected by the initial process noise. The simulation results demonstrate that all the three ACKFs can improve the performance of CKF and each ACKF has its own merits.