Multisensor constrained estimation with unscented transformation

Abstract

The paper focuses on multisensor constrained estimation. The possibility of an extension of a current unconstrained fusion technique is discussed for equality constrained estimates. Basic approaches to multisensor constrained estimation with fusion centre are proposed and discussed. The stress is laid on the constraint in or outside the estimation loop and on the constrained fusion of unconstrained or constrained estimates. The constraint is applied by the estimate projection method with the use of the unscented transformation which does not require the computation of derivatives and detour the problem of singular covariance matrices that are obtained by linearising the projection. A ground tracking example is given and the approaches are compared.