A fault detection method for GNSS/INS integrated navigation system based on GARCH model

This paper concerns the integration of GNSS and INS for integrated navigation applications. INS can be integrated synergistically with GNSS so that the advantages of both INS and GNSS can be exploited. Since GNSS receiver may produce abnormal measurement faults which mainly caused by external factors such as signal attenuation, multipath and jamming, efficient fault detection methods are needed to avoid the performance deterioration of the GNSS/INS integrated system. In general, faults can be classified into two categories: abrupt changing faults and gradual changing faults. Traditional fault detection methods work well when detecting abrupt changing faults, but gradual changing faults cannot be detected if these faults do not reach a certain threshold. The limitation of the fault detection methods for gradual changing fault detection is the key problem to deal with. This paper presents a new method to effectively detect gradual changing fault in a GNSS/INS integrated system by using GARCH (Generalized Autoregressive Conditional Heteroskedasticity) model. GARCH model is widely used in finance, which can predict and analyze the volatility of a data series. Based on this characteristics, GARCH model is used to predict the variance of the measurement residual. The predictive variance is compared with a preset threshold, so that the gradual changing fault can be efficiently detected by the variance evaluation in advance. The parameters of GARCH model are updated online at each epoch, which can accurately reflect the variance of measurement residual changing with time. Experimental results demonstrate that compared with traditional residual chi-square detection method, the proposed fault detection method significantly improve the fault-tolerant ability and positioning accuracy of GNSS/INS integrated navigation system, especially for gradual changing faults.