A novel nonlinear state estimation technique based on sequential importance sampling and parallel filter banks

The sequential importance sampling (SIS) method, also known as particle filtering (PF), has emerged as a promising filtering technique for nonlinear and non-Gaussian systems in which state estimation is the ultimate objective. In the framework of PFs, the accuracy of this estimation depends on the choice of the proposal distribution, which is the problem that is addressed in this paper. Here, we propose a novel particle filtering algorithm which is based on state partitioning and a bank of extended Kalman filters to render a more accurate proposal distribution and hence yielding a more precise estimation of the state. Moreover, because of the improved proposal distribution, the new filter can achieve a given level of performance using fewer samples than its conventional SIS counterparts. Our results show that this new approach yields significantly unproved estimates of the state