Abstract
This correspondence discusses some possible limitations on the application of a new algorithm for recursive linear estimation. An algorithm proposed by Kailath using Chandrasekhar equations to calculate the Kalman gain matrix has been shown to be computationally advantageous in some special cases. This correspondence considers the case of controllable and observable linear systems and shows that the conditions required for the algorithm to be advantageous form a closed set of Lebesgue measure zero.

This publication has 5 references indexed in Scilit: