Investigation of Derivative-free Estimator by Mixing State Dependent Factorization Model with EKF Solver

Abstract

Nonlinear filtering techniques, like the Divided Difference Filter (DDF) and the θ-D filter, which explore the derivative free modeling approach to ameliorate the effect of high order term ignorance of the baseline Extended Kalman Filter (EKF), have recently gained a lot of attentions in the estimation and control community, especially in the context of robust and adaptation capabilities. This paper employs the State Dependent Factorization (SDF) approach to accomplish derivative free modeling in the dynamic process and measurement process while borrowing the EKF framework to achieve the performance of a nonlinear estimator. The proposed mixed SDF/EKF estimator is evaluated using a nonlinear target tracking problem. Its performance is evaluated against other mainstream nonlinear filters, the Unscented Kalman Filtering (UKF) and also the baseline EKF. Current evaluation of the Mixed SDF/EKF compared with EKF and UKF via Monte Carlos simulation for this particular target tracking problem has demonstrated the feasibility of the proposed approach toward nonlinear estimation problems.