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.

Filtering problem with nonlinear observations and drift terms equal to gradient vector field plus affine vector field

Solution of Filtering Problem with Nonlinear Observations

Linear Filtering with Nonlinear Observations

Finite-dimensional Filters with Nonlinear Drift. V: Solution to Kolmogorov Equation Arising from Linear Filtering with Non-Gaussian Initial Condition

Direct Method for Yau Filtering System with Nonlinear Observations

Finite-dimensional Filters with Nonlinear Drift, VI: Linear Structure of Ω.

Finite-dimensional Filters with Nonlinear Drift. III: Duncan-Mortensen-Zakai Equation with Arbitrary Initial Condition for the Linear Filtering System and the Benes Filtering System

Finite Dimensional Filters with Nonlinear Drift, XI

Real Time Algorithm for Nonlinear Filtering Problem.

Finite-Dimensional Filters with Nonlinear Drift. XII: Linear and Constant Structure of Wong-Matrix

Finite Dimensional Estimation Algebra for Time-Varying Filtering System and Optimal Transport Particle Filter: A Tangent Flow Point of View

Addendum to "Finite-Dimensional Filters with Nonlinear Drift": A Response to F. Daum'S Comments

Solution to Brockett's Problem on Finite-Dimensional Estimation Algebras of Maximal Rank in Nonlinear Filtering

Special Solutions to Some Kolmogorov Equations Arising from Cubic Sensor Problems.

New Algorithms in Real Time Solution of the Nonlinear Filtering Problem.

Finite-dimensional Filters with Nonlinear Drift X: Explicit Solution of DMZ Equation

Linear Filtering System With Arbitrary Initial Conditions

Recent Results On Nonlinear Filtering - New Class Of Finite-Dimensional Filters

A novel suboptimal method for solving polynomial filtering problems

Lie Algebraic Method in Nonlinear Filtering with State Space Dimension Up to 6

Nonlinear Bayesian Filtering with Natural Gradient Gaussian Approximation