A nonlinear voter-estimator for redundant systems

A redundant system is one in which several different measurements are assumed to describe the same physical phenomena. This paper describes a class of voter-estimators which determines a discrete output given a group of discrete redundant measurements. The voterestimator is derived by combining least squares estimation with failure detection and isolation. This is accomplished by using weight factors which are themselves nonlinear functions of the measurements. The effect of this formulation is to filter random signal variations in individual channels while simultaneously discriminating against a failed channel. Also proposed is a method for mitigating performance degradation in the presence of successive failures. Despite the nonlinear character of the voting and isolation mechanism, the mathematics involved is of closed form and requires no iterative computations for implementation of the voter-estimator.