The Kantorovich inequality for error analysis of the Kalman filter with unknown noise distributions

The state-space model with random noises is widely used to model dynamic systems. In many practical problems the noise terms will have unknown probability distributions, which is contrary to the usual assumption of Gaussian noises made in state-vector estimation; this assumption is usually made for reasons of mathematical tractability. A problem arises, however, in that inference based on the incorrect Gaussian assumption can lead to misleading or erroneous conclusions. This note shows how the Kantorovich inequality from probability theory has potential for characterizing the estimation error of a Kalman filter in such a non-Gaussian (distribution-free) setting.

[1]  J. Spall,et al.  Asymptotic distribution theory for the kalman filter state estimator , 1984 .

[2]  S. Gupta,et al.  A generalization of Anderson’s theorem on unimodal functions , 1976 .

[3]  B. Anderson,et al.  Optimal Filtering , 1979, IEEE Transactions on Systems, Man, and Cybernetics.

[4]  S. Gupta,et al.  ON ANDERSON'S PROBABILITY INEQUALITY , 1983 .

[5]  Guanrong Chen.  Approximate Kalman filtering , 1993 .

[6]  James Ting-Ho Lo,et al.  Finite-dimensional sensor orbits and optimal nonlinear filtering , 1972, IEEE Trans. Inf. Theory.

[7]  Ludwik Kurz,et al.  An adaptive robustizing approach to kalman filtering , 1983, Autom..

[8]  J. C. Spall,et al.  The distribution of nonstationary autoregressive processes under general noise conditions , 1993, Proceedings of 32nd IEEE Conference on Decision and Control.

[9]  Sylvia Schnatter.  Integration-based Kalman-filtering for a dynamic generalized linear trend model , 1992 .

[10]  Samuel Karlin,et al.  Studies in econometrics, time series, and multivariate statistics , 1983 .

[11]  J. Spall.  THE DISTRIBUTION OF NONSTATIONARY AUTOREGRESSIVE PROCESSES UNDER GENERAL NOISE CONDITIONS , 1993 .

[12]  Achim Clausing.  Kantorovich-Type Inequalities , 1982 .

[13]  H. Sorenson,et al.  Recursive bayesian estimation using gaussian sums , 1971 .

[14]  Joel M. Morris,et al.  The Kalman filter: A robust estimator for some classes of linear quadratic problems , 1976, IEEE Trans. Inf. Theory.

[15]  G. S. Watson,et al.  A method for discovering Kantorovich-type inequalities and a probabilistic interpretation , 1987 .

[16]  Y. Tong.  Some majorization inequalities in multivariate statistical analysis , 1988 .

[17]  W. Chan,et al.  Unimodality, convexity, and applications , 1989 .

[18]  Y. Tong.  Probability Inequalities in Multivariate Distributions , 1980 .

[19]  Morris L. Eaton,et al.  Concentration inequalities for Gauss-Markov estimators , 1988 .

[20]  Robert Kohn,et al.  Prediction mean squared error for state space models with estimated parameters , 1986 .

[21]  G. Kitagawa.  Non-Gaussian State—Space Modeling of Nonstationary Time Series , 1987 .

[22]  James C. Spall.  [That BLUP is a Good Thing: The Estimation of Random Effects]: Comment: The Kalman Filter and BLUP , 1991 .

[23]  C. Masreliez.  Approximate non-Gaussian filtering with linear state and observation relations , 1975 .

[24]  R. J. Kelly.  Reducing geometric dilution of precision using ridge regression , 1990 .

[25]  Han-Fu Chen,et al.  Lp‐stability of estimation errors of kalman filter for tracking time‐varying parameters , 1991 .

[26]  D. Varberg.  Convex Functions , 1973 .

[27]  M. E. Welch,et al.  Bayesian analysis of time series and dynamic models , 1990 .