00079 ²4addfcc1ad8f7adb124bf9726a6931

78

Hembree & Zimmer

FigurÄ™ 13. Weighting Function.

model is a performance comparison between various configurations of the DBAF and the truth fiiter. The performance criterion to be used for the comparison is the covariance of the fiiter estimate.

The approach used to do this is to propagate the covariances for each of the truth models and propagate the covariances for each of the various configurations of the adaptive filters and compare the results. Recall that for our model, the covariance of the State estimate for the truth fiiter is independent of the data and therefore converges to steady-state values. However, the covariances for the adaptive fiiter are not independent of the data siÅ„ce the estimates are a function of the residuals. In order to overcome this problem the expected value of the residuals, as developed in this previous section, is used instead. This then allows the expected covariances for the fiiter to be propagated without having to resort to Monte Carlo simulation.