Abstract
In this paper a square root covariance form of the Kalman filter is presented which is in many ways the covariance analog of the Paige and Saunders information filter [SIAM J. Numer. Anal., 14 (1977), pp. 180-193] and possesses many of the advantages of their algorithm. For example, it is a true square root implementation; but the distinctive feature which it shares with the Paige and Saunders algorithm is a compact, convenient, and effective square root implementation of the interpolation smoother. The algorithm is based on the Duncan and Horne generalised least squares form of the filter equations [J. Amer. Statist. Assoc., 67 (1972), pp. 816-821], but uses these in a recursive manner.
Original language | English |
---|---|
Pages (from-to) | 1324-1331 |
Number of pages | 8 |
Journal | Siam journal on scientific and statistical computing |
Volume | 12 |
Issue number | 6 |
Publication status | Published - Nov 1991 |
Externally published | Yes |
Keywords
- KALMAN FILTER
- COVARIANCE FILTER
- INFORMATION FILTER
- SQUARE ROOT FILTER
- GENERALIZED LEAST SQUARES
- SMOOTHING STEP
- DISCRETE