Application of a sensor fusion algorithm to nonlinear data containing nonstationary, multiplicative noise of unknown distribution

Three sensors make noisy angle-of-arrival measurements on an emitter whose position is to be estimated. Two angle measurements are triangulated to obtain one position measurement and a second pair is used to generate the other position measurement. With one sensor common to both triangulations, the two measurement vectors have a non-zero cross-covariance matrix. The distortions in the triangulation equations, coupled with nonGaussian angle errors, produce nonlinear measurement vectors containing nonstationary, multiplicative noise of unknown distribution. The optimum fusion algorithm, which is designed to take account of arbitrary cross-covariance in the data, is applied here under suboptimum conditions. Broadside and end-fire array emitter positions are considered