Linear Fusion under Random Correlation of Estimation Errors

Linear fusion of estimates has been studied from the perspectives of known and unknown correlations of estimation errors. Whereas optimal linear combinations can be designed in the former case, a robust approach is usually chosen in the latter one. The loss of performance may be unacceptably high, which raises the need to find a middle ground. This paper reviews various approaches to information fusion, formulates the problem of random correlation and presents the solution. Monte Carlo verification of the results is discussed and an illustration is provided.