Selection of noise parameters for Kalman filter

The Bayesian probabilistic approach is proposed to estimate the process noise and measurement noise parameters for a Kalman filter. With state vectors and covariance matrices estimated by the Kalman filter, the likehood of the measurements can be constructed as a function of the process noise and measurement noise parameters. By maximizing the likelihood function with respect to these noise parameters, the optimal values can be obtained. Furthermore, the Bayesian probabilistic approach allows the associated uncertainty to be quantified. Examples using a single-degree-of-freedom system and a ten-story building illustrate the proposed method. The effect on the performance of the Kalman filter due to the selection of the process noise and measurement noise parameters was demonstrated. The optimal values of the noise parameters were found to be close to the actual values in the sense that the actual parameters were in the region with significant probability density. Through these examples, the Bayesian approach was shown to have the capability to provide accurate estimates of the noise parameters of the Kalman filter, and hence for state estimation.

[1]  Nicholas P. Jones,et al.  SIMULTANEOUS ESTIMATION OF SYSTEM AND INPUT PARAMETERS FROM OUTPUT MEASUREMENTS , 2000 .

[2]  C. N. Ng,et al.  Recursive estimation of model parameters with sharp discontinuity in non-stationary air quality data , 2004, Environ. Model. Softw..

[3]  Roger Ghanem,et al.  Structural System Identification. II: Experimental Verification , 1995 .

[4]  Richard A. Brown,et al.  Introduction to random signals and applied kalman filtering (3rd ed , 2012 .

[5]  R. E. Kalman,et al.  A New Approach to Linear Filtering and Prediction Problems , 2002 .

[6]  A. Mikami,et al.  Simultaneous identification of time and space variant dynamic soil properties during the 1995 Hyogoken-Nanbu earthquake , 2005 .

[7]  Zhang Yigong,et al.  Nonlinear structural identification using extended kalman filter , 1994 .

[8]  Toshiro Matsumoto,et al.  Application of BEM with extended Kalman filter to parameter identification of an elastic plate under dynamic loading , 2004 .

[9]  Arthur Gelb,et al.  Applied Optimal Estimation , 1974 .

[10]  Masaru Hoshiya,et al.  IDENTIFICATION OF CONDITIONAL STOCHASTIC GAUSSIAN FIELD , 1996 .

[11]  J. Beck,et al.  Updating Models and Their Uncertainties. I: Bayesian Statistical Framework , 1998 .

[12]  R. Shumway,et al.  AN APPROACH TO TIME SERIES SMOOTHING AND FORECASTING USING THE EM ALGORITHM , 1982 .

[13]  Lambros S. Katafygiotis,et al.  Bayesian time–domain approach for modal updating using ambient data , 2001 .

[14]  H. Sorenson.  Least-squares estimation: from Gauss to Kalman , 1970, IEEE Spectrum.

[15]  M. Hoshiya,et al.  Structural Identification by Extended Kalman Filter , 1984 .

[16]  R. Ghanem,et al.  Structural-System Identification. I: Theory , 1995 .

[17]  S. C. Tam,et al.  Solving Harmonic Sea-Level Model with Kalman Filter: A Macau Case Study , 2002 .

[18]  Chung Bang Yun,et al.  Identification of Linear Structural Dynamic Systems , 1982 .

[19]  Masaru Hoshiya,et al.  ESTIMATION OF DYNAMIC PROPERTIES OF A MULTIPLE DEGREE OF FREEDOM LINEAR SYSTEM , 1984 .

[20]  R. E. Kalman,et al.  New Results in Linear Filtering and Prediction Theory , 1961 .