A Fault-Tolerant Scheme Based on MSVR for Integrated Navigation

To enhance the reliability of integrated navigation systems, an intelligent fault-tolerant scheme based on multi-output support vector regression (MSVR) is proposed. Specifically, a novel fault detection function that is made up of the filtering innovations and the estimated ones is designed so that it can timely detect the faults that may occur. Meanwhile, MSVR is utilized to build the estimation model for more reliable estimations. The special construction of the fault detection function makes it very sensitive to faults. Additionally, a variable that symbolizes the confidence of measurement is defined and applied to realize fault isolation. The method effectively improves the navigation accuracy at the initial stage of the occurrence of a gradual fault. The results of experiments reveal that the proposed algorithm has a stronger fault-tolerant ability compared with the traditional algorithm. Concretely, when dealing with the artificial faults, the position precision in north and east directions are increased up to 70.06% and 67.78%, respectively, by applying the proposed fault-tolerant scheme, in comparison with the residual chi-square test.