Abstract:Practical Bayes filters often assume the state distribution of each time step to be Gaussian for computational tractability, resulting in the so-called Gaussian filters. When facing nonlinear systems, Gaussian filters such as extended Kalman filter (EKF) or unscented Kalman filter (UKF) typically rely on certain linearization techniques, which can introduce large estimation errors. To address this issue, this paper reconstructs the prediction and update steps of Gaussian filtering as solutions to two distinct optimization problems, whose optimal conditions are found to have analytical forms from Stein's lemma. It is observed that the stationary point for the prediction step requires calculating the first two moments of the prior distribution, which is equivalent to that step in existing moment-matching filters. In the update step, instead of linearizing the model to approximate the stationary points, we propose an iterative approach to directly minimize the update step's objective to avoid linearization errors. For the purpose of performing the steepest descent on the Gaussian manifold, we derive its natural gradient that leverages Fisher information matrix to adjust the gradient direction, accounting for the curvature of the parameter space. Combining this update step with moment matching in the prediction step, we introduce a new iterative filter for nonlinear systems called \textit{N}atural Gr\textit{a}dient Gaussia\textit{n} Appr\textit{o}ximation filter, or NANO filter for short. We prove that NANO filter locally converges to the optimal Gaussian approximation at each time step. Furthermore, the estimation error is proven exponentially bounded for nearly linear measurement equation and low noise levels through constructing a supermartingale-like property across consecutive time steps.

Recursive filtering with non-Gaussian noises

l1 norm‐based recursive estimation for non‐linear systems with non‐Gaussian noises

Recursive Performance Ranking of Kalman Filter with Mismatched Noise Covariances

Gaussian Sum Filtering Methods for Nonlinear Non-Gaussian Models

Adaptive Kalman Filtering For Recursive Both Additive Noise And Multiplicative Noise

Filtering with degenerate observation noise: A stochastic approximation approach

Nonlinear Bayesian Filtering with Natural Gradient Gaussian Approximation

Nonlinear/Non-Gaussian Bayesian Sequential Filtering

The modified extended Kalman filter based recursive estimation for Wiener nonlinear systems with process noise and measurement noise

A Novel Kalman Filter for Nonlinear Non-Gaussian Systems and Its Applications in Self-localization

Kalman Filter Auto-Tuning With Consistent and Robust Bayesian Optimization

A New Adaptive Kalman Filter with Inaccurate Noise Statistics

Nonlinear Kalman Filtering with Reparametrization Gradients

A new method for the nonlinear transformation of means and covariances in filters and estimators

Generalized Iterated Kalman Filter and Its Performance Evaluation

Nonlinear Kalman Filter Using Cramer Rao Bound

Kalman Filter Auto-tuning through Enforcing Chi-Squared Normalized Error Distributions with Bayesian Optimization

Kalman Filter and its Modern Extensions for the Continuous-time Nonlinear Filtering Problem

Polynomial Filtering Algorithm Applied to Navigation Data Processing under Quadratic Nonlinearities in System and Measurement Equations. Part 1. Description and Comparison with Kalman Type Algorithms

A Gaussian approximation filter for nonlinear systems with compound noises

Rational-quadratic kernel-based maximum correntropy Kalman filter for the non-Gaussian noises