Derivative-free distributed filtering for integrity monitoring of AGV navigation sensors

Abstract

The paper proposes a new distributed filtering method, for integrity monitoring of navigation sensors in automatic ground vehicles (AGV). Unlike the Extended Information Filter (EIF), the proposed filter avoids approximation errors caused by the linearization of the AGV kinematic model and does not require the computation of Jacobians. The use of a statistical fault detection and isolation algorithm for processing the residuals generated by the proposed filtering method, can provide an indication about the condition of the navigation sensors and about failures that may have appeared. As an an application example the paper considers failure diagnosis for wheel encoders or IMU devices of an AGV.