Abstract:The main purpose of this paper is to address a fundamental open problem in linear filtering and estimation, namely, what is the steady-state or asymptotic behavior of the Kalman filter, or the Kalman gain, when the observed stationary stochastic process is not generated by a finite-dimensional stochastic system, or when it is generated by a stochastic system having higher-dimensional unmodeled dynamics. For example, some time ago Kalman pointed out that the usual positivity conditions assumed in the classical situation are not in fact necessary for the Kalman filter to converge. Using a ''fast filtering'' algorithm, which incorporates the statistics of the observation process as initial conditions for a dynamical system, this question is analyzed in terms of the phase portrait of a ''universal'' nonlinear dynamical system. This point of view has additional advantages as well, since it enables one to use the theory of dynamical systems to study the sensitivity of the Kalman filter to (small) changes in initial conditions; e.g., to changes in the statistics of the underlying process. This is especially important since these statistics are often either approximated or estimated. In this paper, for a scalar observation process, necessary and sufficient conditions for the Kalman filter to converge are derived using methods from stochastic systems and from nonlinear dynamics-especially the use of stable, unstable, and center manifolds. It is also shown that, in nonconvergent cases, there exist periodic points of every period p, p greater-than-or-equal-to 3 that are arbitrarily close to initial conditions having unbounded orbits, rigorously demonstrating that the Kalman filter can be ''sensitive to initial conditions.''

A New Algorithm for Optimal Filtering of Discrete-Time Stationary Processes

Optimal Filtering of Continuous-Time Stationary Processes by Means of the Backward Innovation Process

Some New Non-Riccati Algorithms for Continuous-Time Kalman-Bucy Filtering

Some Reduced-Order Non-Riccati Equations for Linear Least-Squares Estimation : the Stationary, Single-Output Case†

Iterative and Algebraic Algorithms for the Computation of the Steady State Kalman Filter Gain

On Fredholm integral equations, Toeplitz equations and Kalman-Bucy filtering

Optimal Linear Filtering for Discrete-Time Systems with Infinite-Dimensional Measurements

Linear Least-Squares Estimation of Discrete-Time Stationary Processes by Means of Backward Innovations

On integrals in multi-output discrete-time Kalman-Bucy filtering

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

Direct Method Without the Riccati Equation for a Linear Filtering System with Arbitrary Initial Conditions

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

Synthesis of quasi-optimal fast filters by the least square criterion

Constructing numerically stable Kalman filter-based algorithms for gradient-based adaptive filtering

On the Nonlinear Dynamics of Fast Filtering Algorithms

Anisotropy-based optimal filtering in linear discrete time invariant systems

Optimal nonlinear filtering using the finite-volume method

Data-driven Optimal Filtering for Linear Systems with Unknown Noise Covariances

An Alternate View on Optimal Filtering in an RKHS

Improved Discrete-Time Kalman Filtering within Singular Value Decomposition

Extended Kalman Filtering for Recursive Online Discrete-Time Inverse Optimal Control