For two-sensor systems with unknown cross-covariance, a covariance intersection(CI) fusion robust steady-state
Kalman filter is presented, which has robustness with respect to unknown cross-covariances. It is rigorously proved that its
the actual accuracy of the fitter is higher than that of each local filter, and is lower than that of the optimal fusion Kalman
filter with known cross-covariance. The geometric interpretation of the accuracy relations is given based on the covariance
ellipses. Further, the above results are extended to the multisensor case. A Monte-Carlo simulation example for a tracking
system shows that its actual accuracy is close to that of the optimal Kalman fuser with known cross-covariance.

multisensor information fusion;covariance intersection fusion;robust Kalman filter;covariance ellipse