|Abstract:||Precise Point Positioning (PPP) can provide cm to decimeter level accuracy using a single receiver in un-differenced mode. In recent years, the advent of real-time precise orbit and clock correction streams allows end users to shift from the traditional post-mission PPP processing to Real-Time PPP (RT-PPP) solution anywhere in the world. A serious concern in real-time PPP is the possibility of suffering a communication break in receiving the correction data that may occur, for instance, due to a temporary modem hardware failure, a weakness in the wireless mobile network that may be experienced in some areas, when using UAVs, or even when starting RT-PPP in an area with good mobile communication coverage and continuing work in an area with no or weak network. In such a case, severe decline of positioning accuracy may result leading to a degrading of positioning accuracy from the cm-decimeter level to the m level of the default single point positioning mode. This contribution presents a practical solution to this problem. Real-time precise orbit and clock corrections are predicted as time series and it is assumed for PPP with ambiguity resolution that the last received factional cycle biases and code biases are valid for a few hours and thus are not predicted. We restrict attention to GPS due to the fact that its real-time precise orbits and clock correction products are available, whereas products for other systems are mostly still in the experimental phase. We also focus on free open-access real-time PPP products. To determine the best prediction model, the clock corrections were analyzed using Allan variance and their frequency stability was investigated for different types of satellite blocks over a period of one month. To determine the number of points that can be used for building the prediction model, and noting that different users may work with different sampling rates, instead of defining a fixed number of points we conducted comprehensive autocorrelation analysis to estimate the significant correlation time length during which the number of sample points can be selected. Different prediction models were analytically compared and their predicted clock corrections were compared with a reference International GNSS Service (IGS) final products. Similarly, different methods for prediction of the precise orbits were compared. Suitable models are recommended. Another challenge in real-time PPP is its integrity monitoring (IM). In IM, the system automatically performs checks to detect faults, isolates faulty information or anomalous measurements, and raises an alarm in the event of the positioning system operation being considered unsafe for use. When no malfunction is detected, IM provides a protection level (PL) that bounds the true position error with a certain probability of risk. We propose here two significant additions to existing PPP integrity monitoring approaches. Firstly, current PPP methods treat the precise orbits and clock corrections as calibrated values. Therefore, any error in their estimation or if they were been tampered with by hackers will result in wrong positioning. We present here an augmented PPP model where the precise orbits and clock corrections are treated as additional ‘quasi’ observations. This is acceptable due to the fact that they are independent from the measurements of the user, given that these corrections are estimated at stations using measurements that are different from those of the user. The uncertainty of their estimation will also be included in the covariance matrix of the observations. Hence, possible faults in the precise orbits or clock corrections can be checked at the user level through the process of fault detection and exclusion (FDE). In addition, the probability of such faults can be included within the total probability budget of the integrity risk. Secondly, previous studies had computed the PL for PPP using only a fault-free mode assuming that all faults are perfectly isolated. We introduce a new model that considers both the fault-free case and the likelihood of undetected faults, including those which may arise from prediction of the precise orbits and clock corrections. For the observations that pass the FDE stage, a position error bound is created for each possible fault mode that might be miss-detected. We modeled two types of errors; step errors that may take place at any epoch; and ramping errors that accumulate with time, for instance, due to the use of a dynamic model when using a Kalman filter in processing PPP data. Testing of the proposed approach showed that it accomplished positioning accuracy of a few cm to a decimeter for both static and kinematic modes in land and marine environments. This accuracy was maintained for up to three hours of prediction of the precise orbits and clock corrections. When PPP was initialized without prediction, solution convergence took between 22 and 30 minutes on average for the tested data sets. When PPP is reset after the break, the initialization period almost doubled; however, positioning precision was maintained to less than a decimeter after solution convergence. The validity of the proposed integrity monitoring approach was demonstrated where the protection levels bounded the position error during the whole periods of the tests.|
Proceedings of the 29th International Technical Meeting of The Satellite Division of the Institute of Navigation (ION GNSS+ 2016)
September 12 - 16, 2016
Oregon Convention Center
|Pages:||3276 - 3294|
|Cite this article:||
El-Mowafy, Ahmed, "Facing Some Critical Challenges in Real-Time Precise Point Positioning," Proceedings of the 29th International Technical Meeting of The Satellite Division of the Institute of Navigation (ION GNSS+ 2016), Portland, Oregon, September 2016, pp. 3276-3294.
ION Members/Non-Members: 1 Download Credit