Simultaneous estimation and modeling of nonlinear, non-Gaussian state-space systems

This paper presents a framework for simultaneous estimation and modeling of nonlinear, non-Gaussian state-space systems. In the proposed approach, uncertainty in motion model parameters is incorporated to avoid overconfidence in state prediction and better account for modeling inaccuracies. The additional original contribution of a model correction stage improves nonlinear model parameter estimates in order to enhance the accuracy of state estimation. The presented nonlinear/non-Gaussian Simultaneous Estimation And Modeling (SEAM) approach was compared with contemporary estimation techniques using a Monte-Carlo simulation study. This study showed that the proposed method successfully reduces estimation error relative to existing approaches even when substantial model parameter uncertainty and multi-modal sensor noise are present. The framework has potential use in a wide range of applications where state-space estimation is employed, including robotics, signal processing, and controls. Published by Elsevier Inc.