A new method for fault identification in real-time integrity monitoring of autonomous vehicles positioning using PPP-RTK

Abstract

Autonomous vehicles require a real-time positioning system with in-lane accuracy. They also require an autonomous onboard integrity monitoring (IM) technique to verify the estimated positions at a pre-defined probability. This can be computationally demanding. PPP-RTK is a promising positioning technique that can serve this purpose. Since PPP-RTK is developed to process undifferenced and uncombined (UDUC) observations for both network and user sides, it provides the residuals of the individual measurements. This can be exploited to reduce the computational load consumed in the fault detection and exclusion (FDE) process, included in the IM task, without compromising the positioning availability. This research proposes filtering the faulty satellites by the network, then the hardware and location-dependent faults at the user end can be identified. This is achieved by calculating the ratio between the matching UDUC residuals of the user receiver and the nearest reference station observations. This ratio is used to rank the individual observations where the observation with the largest ratio is most likely to be the faulty one. Therefore, it is more likely to identify the faulty observation without generating and testing numerous subsets. In addition, the exclusion can be attempted per observation, which preserves observation availability, unlike the grouping techniques that perform the exclusion per satellite. The method was examined in two test cases where geodetic and commercial receivers were used. Results show that the computational load has been reduced significantly by about 85–99% compared to the solution separation and Chi-squared test methods that are commonly used for FDE.