Consistency analysis for data fusion: Determining when the unknown correlation can be ignored

In this paper we examine the conditions in which data fusion can be performed by neglecting the unmodeled correlation between two information sources without compromising the consistency of the system. More specifically, we explore those situations in which one can disregard the correlation information and achieve a consistent estimate by simply adding the respective estimates' information matrices. This estimate will deliver considerably better performance than the widely employed Covariance Intersection (CI) algorithm in terms of estimation uncertainty.