Parameter errors in dynamic and measurement models of dynamic systems can result in poor state estimates when using a traditional Kalman filter structure. In dealing with these parameter errors it is possible to: 1) Ignore them completely; 2) Add the parameters as additional states to be estimated; or 3) "Consider" the error in the state covariance matrix by introducing additional parameter covariance matrices. This paper analyzes the effect of using all three of these types of filters on a simple asteroid rendezvous scenario to determine the applicability of each. Two types of consider Kalman filters are explored, namely an Augmented Measurement Consider Kalman Filter and a Minimum Variance Consider Kalman Filter. This paper finds that a Minimum Variance Consider Kalman Filter can provide not only improved state estimates to a traditional Kalman filter, but also produces consistent results from a statistical perspective. Copyright 2010 by Drew P. Woodbury and John L. Junkins.
