Covariance intersection (CI) and bounded covariance inflation (BCINF) estimators can be used to handle unknown correlation data fusion problems encountered in aerospace guidance, navigation, and control. Understanding the uncertainty tradeoff inherent to CI/BCINF is important for explaining the seemingly counterintuitive performance of these filters. The impact that the number of states has on the uncertainty tradeoff can have filter design implications. Additionally, covariance normalization is necessary to conduct a meaningful uncertainty tradeoff as part of the fusion.
Bibliographical noteFunding Information:
The authors acknowledge the U.S. Department of Homeland Security (grant number 2008-ST-061-BS0002) and the National Science Foundation (National Robotics Initiative Large grant IIS- 1328722) for support of this work.