Description
| Abstract: | In this paper, a computational aspect of real-time estimation algorithm implementation is considered, in which the estimation algorithm to be used is defined having the standard optimal Kalman filtering structure but with the actual inverse matrix within the Kalman gain being replaced by an expedient approximation at each time instant. In real-time applications, most Kalman filtering schemes are approximate to a degree as a consequence of computational errors in calculating matrix inversion due to numerical roundoff. Convergence properties and error estimates of such approximate mechanizations of the Kalman filtering algorithm are obtained in this paper to provide a theoretical basis for gauging the utility of using computational approximations of the Kalman gain matrix at each time instant viz use of exponentially convergent sequences for comparison. A new exponentially convergent scheme is also suggested for approximating the inverse matrix within the Kalman gain. Conditions are offered when on-line approximate matrix inversion can be eliminated as the scapegoat cause of Kalman filter divergence in real-time mechanization leaving only model mismatch as the likely culprit when divergence occurs. |
|---|
| Item Description: | "January 1992."
"This paper is dedicated to Professor Mingjun Chen for introducing this author to the field of Kalman Filtering twenty year ago when the author was a high school student"--Page [i]. |
|---|
| Physical Description: | 18 pages : illustrations ; 28 cm |
|---|
| Bibliography: | Includes bibliographical references (pages 17-18). |
|---|