MASSACHUSETTS UNIV AMHERST DEPT OF ELECTRICAL AND COMPUTER ENGINEERING
The Kalman filter has been used in many applications, however, practical implementation of the filter has required exact knowledge of the various system parameters input and measurement noise covariance so as to yield optimum performance. The paper develops a minimax technique for the direct synthesis of Kalman-like estimators when there are large uncertainties in the a priori statistics of the plant and measurement noises. Both continuous and discrete estimators are considered. General properties of the filters that satisfy the various minimax performance indices are discussed and a number of examples of both continuous and discrete applications are then presented to demonstrate the technique. Author
Pub. in Proceedings of AISS Guidance and Control Conference, Stanford, Calif., 14-16 Aug 72, n72-878 p1-9.