Heavy-tails in Kalman filtering with packet losses

Abstract In this paper, we study the existence of a steady-state distribution and its tail behaviour for the estimation error arising from Kalman filtering for unstable linear dynamical systems. Although a large body of literature has studied the problem of Kalman filtering with packet losses in terms of analysis of the second moment, no study has addressed the actual distribution of the estimation error. First we show that if the system is strictly unstable and packet loss probability is strictly less than unity, then the steady-state distribution (if it exists) must be heavy tail, i.e. its absolute moments beyond a certain order do not exist. Then, by drawing results from Renewal Theory, we further provide sufficient conditions for the existence of such stationary distribution. Moreover, we show that under additional technical assumptions and in the scalar scenario, the steady-state distribution of the Kalman prediction error has an asymptotic power-law tail, i.e. P [ | e | > s ] ∼ s − α , as s → ∞, where α can be explicitly computed. We further explore how to optimally select the sampling period assuming an exponential decay of packet loss probability with respect to the sampling period. In order to minimize the expected value of the second moment or the confidence bounds, we illustrate that in general a larger sampling period will need to be chosen in the latter case as a result of the heavy tail behaviour.

[1]  Soummya Kar,et al.  Kalman Filtering With Intermittent Observations: Weak Convergence to a Stationary Distribution , 2009, IEEE Transactions on Automatic Control.

[2]  A. Brandt The stochastic equation Yn+1=AnYn+Bn with stationary coefficients , 1986 .

[3]  Lihua Xie,et al.  Mean square stability for Kalman filtering with Markovian packet losses , 2011, Autom..

[4]  A. Sahai,et al.  Intermittent Kalman filtering: Eigenvalue cycles and nonuniform sampling , 2011, Proceedings of the 2011 American Control Conference.

[5]  A. Grincevičius,et al.  One limit distribution for a random walk on the line , 1975 .

[6]  João Pedro Hespanha,et al.  A Survey of Recent Results in Networked Control Systems , 2007, Proceedings of the IEEE.

[7]  Andrea Censi,et al.  Kalman Filtering With Intermittent Observations: Convergence for Semi-Markov Chains and an Intrinsic Performance Measure , 2011, IEEE Transactions on Automatic Control.

[8]  Bruno Sinopoli,et al.  Kalman filtering with intermittent observations , 2004, IEEE Transactions on Automatic Control.

[9]  Bruno Sinopoli,et al.  A characterization of the critical value for Kalman filtering with intermittent observations , 2008, 2008 47th IEEE Conference on Decision and Control.

[10]  H. Kesten Random difference equations and Renewal theory for products of random matrices , 1973 .

[11]  Benoîte de Saporta,et al.  Tail of the stationary solution of the stochastic equation Yn+1=anYn+bn with Markovian coefficients , 2004 .

[12]  C. Goldie IMPLICIT RENEWAL THEORY AND TAILS OF SOLUTIONS OF RANDOM EQUATIONS , 1991 .

[13]  Francesco Bullo,et al.  On Kalman Filtering for Detectable Systems With Intermittent Observations , 2009, IEEE Transactions on Automatic Control.

[14]  Michel Loève,et al.  Probability Theory I , 1977 .

[15]  Anant Sahai,et al.  The Necessity and Sufficiency of Anytime Capacity for Stabilization of a Linear System Over a Noisy Communication Link—Part I: Scalar Systems , 2006, IEEE Transactions on Information Theory.

[16]  R. Jungers,et al.  Lower and upper bounds for the largest Lyapunov exponent of matrices , 2013 .

[17]  Bruno Sinopoli,et al.  Kalman Filtering With Intermittent Observations: Tail Distribution and Critical Value , 2012, IEEE Transactions on Automatic Control.

[18]  Thomas Mikosch,et al.  Stochastic Models with Power-Law Tails: The Equation X = Ax + B , 2016 .

[19]  I Ya Gol'dsheid,et al.  Lyapunov indices of a product of random matrices , 1989 .

[20]  Pawel Hitczenko,et al.  PERPETUITIES WITH THIN TAILS REVISITED , 2009, 0912.1694.

[21]  Keyou You,et al.  Survey of Recent Progress in Networked Control Systems , 2013 .

[22]  R. Murray,et al.  Estimation with Information Loss: Asymptotic Analysis and Error Bounds , 2005, Proceedings of the 44th IEEE Conference on Decision and Control.

[23]  Ling Shi,et al.  Kalman Filtering Over a Packet-Dropping Network: A Probabilistic Perspective , 2010, IEEE Transactions on Automatic Control.

[24]  Ling Shi,et al.  An improved stability condition for Kalman filtering with bounded Markovian packet losses , 2015, Autom..

[25]  Minyue Fu,et al.  Kalman Filtering With Intermittent Observations: On the Boundedness of the Expected Error Covariance , 2014, IEEE Transactions on Automatic Control.

[26]  Charles M. Goldie,et al.  Perpetuities with thin tails , 1996, Advances in Applied Probability.

[27]  Karl Henrik Johansson,et al.  Wireless networked control system co-design , 2011, 2011 International Conference on Networking, Sensing and Control.

[28]  Luca Schenato,et al.  Optimal Estimation in Networked Control Systems Subject to Random Delay and Packet Drop , 2008, IEEE Transactions on Automatic Control.

[29]  Subhrakanti Dey,et al.  Stability of Kalman filtering with Markovian packet losses , 2007, Autom..

[30]  Subhrakanti Dey,et al.  Heavy-tails in Kalman filtering with packet losses: confidence bounds vs second moment stability , 2018, 2018 European Control Conference (ECC).

[31]  John G. Proakis,et al.  Digital Communications , 1983 .