A Fault-Tolerant Filtering Algorithm for SINS/DVL/MCP Integrated Navigation System

Abstract

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