• This study derives the equivalence of multiple hypothesis solution separation (MHSS) method’s fault detection statistics across three directions under the single-fault assumption and constructs a novel fault detection statistic , improving the performance of fault detection algorithms. • The algorithm is based on the modified fault detection statistic, incorporating both a sliding window in the time domain and a check factor in the observation domain, enabling rapid and accurate detection of slowly growing error (SGE). • The effectiveness and applicability of the algorithm were verified through static simulation and urban kinematic experiments. As global navigation satellite system (GNSS) application scenarios become more complex, the probability of slowly growing error (SGE) increases dramatically due to factors such as multipath effects and atmospheric disturbances. Given the slow change in SGE over time, it takes some time to trigger the detection threshold, which causes a special challenge for fault detection methods. To address this issue, this study proposes a fault detection method combining observation, position, and time-domain information for the SGE. This method includes three parts. First, a new detection statistic is derived based on the equivalence of the multiple hypothesis solution separation (MHSS) method’s detection statistic across three coordinate components. Second, a sliding window in the time domain is introduced to ensure the detection statistic fully reflects the change characteristics of SGE, enabling rapid detection of SGE. Finally, residuals in the observation domain are utilized to calculate a check factor, ensuring consistency of information within the sliding window and enabling accurate detection of the end time of the SGE. To validate the effectiveness and applicability of the proposed method, we conducted experiments using static simulations and urban kinematic data. The results demonstrate that the proposed method enables rapid and accurate detection of SGE, effectively mitigating the threat posed by various outlier observations to positioning solutions in urban applications. The maximum positioning error of ENU was reduced by 32.7%, 38.4%, and 18.1%, respectively.
Li et al. (Fri,) studied this question.