Fault detection in a decentralized navigation architecture using parity space method

Estimation of position and velocity is a critical stage in the navigation of autonomous vehicle. Unexpended faults in sensors, models and implementation of estimation algorithm must be detected vigorously and dealt with, in order to have a robust navigation. In this paper, a multi-stage fault detection algorithm is presented for a land vehicle navigated by a decentralized information fusion algorithm using the Parity Space Method. The algorithm includes fault detection between several local loops and association of the different loop estimates in higher levels of fault detection. The capability of the algorithm in the navigation of land vehicles is demonstrated using the real data collected in a test run by Australian Center for Field Robotics.