A Cascaded Multi-fault Detection and Recovery Algorithm for Model-based Fault-tolerant Navigation in UGVs

This study proposes a fixed-interval entropy-based cascade multi-fault detection and recovery algorithm to enable fast and reliable responses to various GNSS and IMU sensors’ faults that may occur in UGVs. The proposed algorithm employs a two-stage (cascade) approach: primary fault detection is performed using dual-threshold Z-test on speed information from GNSS, INS, and model-based dead reckoning within a sliding window; additionally, secondary fault confirmation is achieved by combining EWMA and CUSUM of entropy values, together with absolute and relative thresholding of fixed-interval entropy. Entropy is estimated from small samples using the Kozachenko-Leonenko estimator. The proposed method shows quick, strong, and dependable detection of sudden and gradual faults in GNSS and IMU sensors, along with effective recovery from GNSS faults over diverse fault magnitudes, based on extensive Monte Carlo simulations and experimental data.