Abstract: A redundant sensor system must satisfy two requirements, one is that the effect of noise on system's output is small, the other is that the detection method is sensitive to failure. These requirements are achived by selecting sensor mounting configuration and parity vector. In this paper, the mounting configuration selection problem is regarded as an optimization problem according to the criterion that the effect of noise on system's output is minimal.The selection of parity vector is regarded as an optimization problem according to the criterion that the failure detection and failure identification capability is most powerful. The optimal condition is presented, an optimal parity vector scheme for detecting and identifying failure is based on this condition .This optimal scheme is compared with several other failure detection schemes.

Key words: inertial navigation, redundant sensor systems, failure detection, optimization