Reliable smartphone GNSS positioning in urban environments remains challenging due to limited observation availability, frequent carrier-phase interruptions, and the poor quality of pseudorange measurements. This study proposes a Doppler-based correction approach for code-enhanced precise point positioning (CEPPP), combined with inter-satellite differencing to eliminate discontinuities among observations, thereby reducing pseudorange noise and multipath effects while maximizing the use of available GNSS observations. Furthermore, a new weighting strategy is proposed by jointly incorporating carrier-to-noise density ratio (C/N₀) and a Doppler second difference-based temporal consistency factor, enabling adaptive down-weighting of measurements affected by multipath, signal loss and interruptions. The method is evaluated using urban smartphone GNSS datasets. For single-point positioning (SPP), the root-mean-square (rms) error is reduced from 3.3 to 2.8 m. For precise point positioning (PPP), the rms error is reduced from 2.0 to 1.2 m, corresponding to a 40.0
With the growing demand for high-integrity real-time Global Navigation Satellite System (GNSS) applications, the reliability of satellite orbit and clock corrections has become increasingly critical. However, existing studies often lack a systematic strategy that connects integrity monitoring on the provider-side with practical implementation on the user-side. To address this limitation, we propose a comprehensive end-to-end integrity monitoring algorithm towards reliable Precise Point Positioning (PPP) applications. The algorithm structures the integrity monitoring process into modular components on both the provider-side and user-side. On the provider-side, carrier-phase residuals are extracted using an uncombined PPP-AR (PPP with Ambiguity Resolution) model to form samples for integrity monitoring. A Gaussian Mixture Model (GMM) is used to characterize residuals and generate associated Quality Information (QI), while simulated faults injected in orbit or clock corrections are identified via an Extreme Gradient Boosting (XGBoost) algorithm. Additionally, the Health Information (HI) is derived through an observation entropy-weighted approach, which considers the observations’ quality for all monitoring stations. QI and HI jointly constitute an Integrity Support Message (ISM) and are delivered to the user. On the user-side, an ISM-aided PPP stochastic model is designed to enhance positioning performance. Experimental results demonstrate that the proposed methods not only enable accurate generation of QI and HI on the provider-side, but also enhance PPP performance with the aid of ISM on the user-side.
This study investigates the consistency and reliability of GNSS receiver antenna type-mean phase center corrections by analyzing both individual and type-mean Phase Center Correction (PCC) models. Using data from approximately 70 geodetic-grade antennas within the Polish ASG-EUPOS network, we quantify variations in calibration models across multiple GNSS signals. Differences between individual and type-mean calibrations are assessed in both the pattern and position domains. The analysis reveals signal-specific discrepancies, particularly for GPS L2 (G02), where deviations reach up to 5 mm in dPCC RMS and 30 mm in MAX values, impacting ionosphere-free (IF) linear-combinations. Position-domain validation via Precise Point Positioning further demonstrates vertical offsets exceeding 5 mm due to outdated or incomplete PCC models, with systematic biases arising from cross-system substitutions, e.g., GPS corrections used for GALILEO signals. The findings emphasize the need for consistent, up-to-date calibration models, and propose robust, signal-specific metrics to support the reliable use of GNSS data in regional and global reference frames.
Accurate regional ionospheric modeling is crucial for network real-time kinematic (NRTK), particularly for long baselines and in edge and cross baseline regions under active ionospheric conditions. Conventional methods often make limited use of network geometry and baseline redundancy, resulting in relatively large modeling errors. Building on the baseline neighborhood enhanced weighted lower-order surface model (BNE-WLSM), this study proposes its fully connected extension, termed FC-BNE-WLSM. The key advancement is to exploit neighborhood constraints from all baselines among the reference stations within each NRTK cell, extending baseline neighborhood enhancement to a unified multi-baseline and multi-directional constraint framework. For each NRTK star network cell, independent double-differenced ionospheric observations are mapped into a virtual fully connected baseline network. Geometry-dependent weights are propagated to the independent-observable domain for weighted surface estimation, while a joint weighting function based on baseline length and projected baseline length ratio regulates baseline contributions according to geometric reliability. Results from two Australian reference station datasets with an average baseline length of 120.6 km show that FC-BNE-WLSM achieves the lowest ionospheric modeling errors, with mean improvements of 7.91
GNSS real-time kinematic (RTK) deformation monitoring at fixed stations is vulnerable to site-specific observation degradation caused by reflection, diffraction, and obstruction. Conventional stochastic models based on elevation and carrier-to-noise density ratio (C/N0), together with residual-triggered robust estimation, may not fully account for direction-related degradation and persistent moderate errors. This study proposes a site-specific LightGBM framework for GNSS observation quality assessment and adaptive weighting. Residual-based labels are generated offline from coordinate-fixed RTK solutions, and conventional RTK features are used to train separate classifiers for each station, satellite system, and observation type. Predicted low-quality probabilities are used to adjust the double-difference measurement noise covariance before IGG3 robust weighting. The evaluation used chronologically separated training and test dates at JC01 and HHU2, two fixed stations with contrasting observation environments, including dual-frequency RTK validation at HHU2. The classifiers showed recall-oriented behavior, with recall for low-quality observations of 84.2–93.1 k_r=1.5 as a balanced common threshold. Epoch-level analyses of representative periods showed that probability-based adjustment down-weighted moderate errors that did not trigger IGG3. For the tested datasets, site-specific probability-based weighting complemented robust estimation and reduced RTK positioning errors while maintaining high ambiguity fix rates.
