D FILTER VS NONLINEAR FILTERING TECHNIQUES IN NONLINEAR TARGET TRACKING APPLICATIONS
Quang Lam, Bruce Anderson, Ming Xin · 2010
Nonlinear filtering and control techniques ( -D filter and controller) have recently gained a lot of attentions in the estimation and control community, especially in the context of robust and adaptation capabilities. This paper examines the -D filter performance for a nonlinear target tracking problem and evaluates its performance and implementation simplicity versus other nonlinear filtering techniques such as unscented Kalman Filtering (UKF) or Extended Kalman filtering (EKF) techniques. The -D filter is derived by constructing the dual of a new nonlinear regulator control technique, D approximation which involves approximate solutions to the Hamilton Jacobi Bellman (HJB) equation. The structure of this filter is similar to the State Dependent Riccati Equation Filter (SDREF). However, this method does not require an online computationally intensive solving of the algebraic Riccati equation at each sample time as compared with the SDREF. A complex highly nonlinear tracking problem is employed to evaluate this new class of nonlinear filter vs other nonlinear filtering contenders such as UKF and/or EKF. Current evaluation of the D filter compared with EKF and UKF via Monte Carlos simulation for this particular target tracking problem has illustrated a very attractive and encouraging result toward nonlinear estimation problems.