A robust two-stage multisensor fusion in contaminated Gaussian channel noise

To solve the problem of distributed multisensor fusion, conventional fusion methods can be efficiently used in Gaussian noise models. In practice, the channel noise is usually non-Gaussian making those methods fail. A robust two-stage fusion algorithm based on the preliminary rejection of outliers in the data with the subsequent application of the conventional fusion method to the rest of the data is proposed. This fusion algorithm exhibits both high robustness in heavy contaminated Gaussian channel noise and good efficiency in nearly Gaussian channel noise both with small and large numbers of sensors.