MINIMISING THE RISK OF COLLATERAL DAMAGE WITH THE KALMAN-LEVY FILTER

This paper examines the use of Kalman filters and variants of the standard Kalman filter in guidance and navigation systems. In particular, the paper is concerned with the properties of a filter that has been developed recently for use with systems with non-Gaussian stable noise or where the noise converges slowly to the Gaussian limit. In such situations, minimisation of the estimated variance (which is the basis of the standard Kalman filter) is not compatible with the minimisation of the large errors that could be associated with collateral damage.