Multimodal Nonlinear Filtering Using Gauss-Hermite Quadrature
نویسندگان
چکیده
In filtering problems the (posterior) state distribution p(x) is recursively estimated given observations y and state dynamics. For nonlinear observation functions, the state distribution can becomemultimodal. Common Solutions are 1.Unimodal Gaussian filter using linearization of obs. function. 2. “Bank of filters” using multiple independent unimodal filters. We present a variational approach for recusively fitting a mixture of Gaussians (MoG) to state posteriors that minimizes the KL divergence between a MoG and the true posterior state distribution.
منابع مشابه
Sigma-Point Filtering and Smoothing Based Parameter Estimation in Nonlinear Dynamic Systems
We consider approximate maximum likelihood parameter estimation in nonlinear state-space models. We discuss both direct optimization of the likelihood and expectation– maximization (EM). For EM, we also give closed-form expressions for the maximization step in a class of models that are linear in parameters and have additive noise. To obtain approximations to the filtering and smoothing distrib...
متن کاملConditional Gauss-Hermite Filtering With Application to Volatility Estimation
The conditional Gauss–Hermite filter (CGHF) utilizes a decomposition of the filter density p(y1, y2) into the product of the conditional density p(y1|y2) with p(y2) where the state vector y is partitioned into (y1, y2). In contrast to the usual Gauss–Hermite filter (GHF) it is assumed that the product terms can be approximated by Gaussians. Due to the nonlinear dependence of φ(y1|y2) from y2, q...
متن کاملSparse Gauss–Hermite Quadrature Filter with Application to Spacecraft Attitude Estimation
A novel sparse Gauss–Hermite quadrature filter is proposed using a sparse-grid method for multidimensional numerical integration in the Bayesian estimation framework. The conventional Gauss–Hermite quadrature filter is computationally expensive for multidimensional problems, because the number of Gauss–Hermite quadrature points increases exponentially with the dimension. The number of sparse-gr...
متن کاملAn Improved Particle Filter with Applications in Ballistic Target Tracking
In this paper, we present an improved particle filter algorithm for ballistic target tracking, the quadrature Kalman particle filter (QKPF). The proposed algorithm uses quadrature Kalman filter (QKF) for generating the proposal distribution. The QKF is a recursive, nonlinear filtering algorithm developed in the Kalman filtering framework. It linearizes the nonlinear functions using statistical ...
متن کاملGaussian filtering and smoothing for continuous-discrete dynamic systems
This article is concerned with Bayesian optimal filtering and smoothing of non-linear continuous-discrete state space models, where the state dynamics are modeled with non-linear Itô-type stochastic differential equations, and measurements are obtained at discrete time instants from a non-linear measurement model with Gaussian noise. We first show how the recently developed sigma-point approxim...
متن کامل