The Kalman filter (KF), which recursively generates a relatively optimal estimate of underlying system state based upon a series of\nobserved measurements, has been widely used in integrated navigation system. Due to its dependence on the accuracy of system\nmodel and reliability of observation data, the precision of KF will degrade or even diverge, when using inaccurate model or trustless\ndata set. In this paper, a fault-tolerant adaptive Kalman filter (FTAKF) algorithm for the integrated navigation system composed\nof a strapdown inertial navigation system (SINS), a Doppler velocity log (DVL), and a magnetic compass (MCP) is proposed. The\nevolutionary artificial neural networks (EANN) are used in self-learning and training of the intelligent data fusion algorithm. The\nproposed algorithm can significantly out perform the traditional KF in providing estimation continuously with higher accuracy and\nsmoothing the KF outputs when observation data are inaccurate or unavailable for a short period.The experiments of the prototype\nverify the effectiveness of the proposed method.
Loading....