A method and computer program product are provided for estimating the states of a dynamic system based upon a combination of first and second feedback signals. The first feedback signal is based upon anticipated state changes and is typically represented by a state propagation model. The second feedback signal is based upon the difference between the anticipated and actual measurements of the sensors. This difference may be weighted based upon a predetermined criteria, such as the credibility of the actual measurements and/or the degree to which outlying sensor measurements are to be discounted. The weighted difference is converted into a corresponding change in the state estimates. At least one feedback signal may then be modified to define its relative contribution to the state estimates. The first and second feedback signals are thereafter combined to produce the state estimates, subject generally to limitations upon the permissible changes in the state estimates.