A Sparse Canonical Correlation Analysis Method for Fault Detection in GNSS/INS Integrated Navigation System

This paper presents an on-line canonical correlation analysis (CCA)-based framework for soft fault detection in global navigation satellite system (GNSS) and inertial navigation system (INS) integrated navigation system. A CCA model measuring the correlations of the innovation and the state error vector is firstly built based on Kalman filter (KF). In order to improve the real-time performance, a gradient-based algorithm with penalty functions is then used to introduce sparsity to the projection vectors of CCA. With the sparse representation, the T2 fault detection statistic is then defined for deciding whether a soft fault occurred in the integrated navigation system. After that, a fault isolation index is deduced through the contribution analysis methodology to identify the fault source. The physical simulations demonstrate that the proposed method has superior performance compared with the traditional fault detection methods.