Nonlinear trajectory estimation using extended state-space recursive least squares.

Abstract

In a variety of filtering applications including target tracking, global positioning systems, and autonomous robots, the nonlinear nature of the system model makes estimation tasks challenging. This study presents an extended state-space recursive least squares (ESSRLS) filter for nonlinear trajectory estimation. The paper focuses on ESSRLS filter derivation for non-autonomous systems and investigates its performance against the extended Kalman filter and unscented Kalman filter for maneuvering aircraft trajectory estimation applications. The major accomplishment of the proposed approach is nonlinear filtering, independent of a priori information about noise statistics and the provision of the tuning parameter (forgetting factor). This makes the ESSRLS a more suited candidate for practical nonlinear filtering applications. The simulation results show that in the presence of model uncertainties, data outages (occlusion), and large initial condition deviations, the proposed method gives superior estimation performance as compared to the state-of-the-art.