Extension of the midvalue selection technique for redundancy management of inertial sensors

Abstract

In many redundant inertial sensor applications, the primary goal is absolute tolerance to some specified number of failures. To achieve this tolerance, a sufficient performance margin must be provided to cover the worst-case degradation that can occur. The worst-case degradation depends upon the particular algorithm used. An optimal algorithm can be defined implicitly by taking that value which minimizes an appropriately chosen worst-case performance index as the estimate. In the important case of first-failure tolerance with five skewed sensors, a simple realization has been discovered which is the direct generalization of midvalue selection to three dimensions.