MobileGNSS-SPP is an advanced, open-source Single Point Positioning (SPP) system built upon the foundation of RTKLIB, tailored specifically for optimizing GNSS performance on smartphone chipsets. While initially developed for a specific GNSS chipset, the optimization techniques and engineering approaches are designed with generality and adaptability in mind, making them applicable to a wide range of GNSS hardware. The project prioritizes practical engineering solutions and algorithmic robustness across diverse real-world scenarios, serving as a comprehensive roadmap for transitioning from open-source code to production-ready GNSS solutions.
- Smartphone-Optimized GNSS Processing: Optimize the algorithm for mobile devices with low-quality data to improve positioning accuracy.
- Robustness Across Scenarios: Extensively tested in diverse environments, from open-sky highways to urban canyons with heavy multipath interference.
- Comprehensive Testing Framework: Includes Python-based tools for batch processing, accuracy evaluation, and in-depth data analysis.
The project introduces several key enhancements to the RTKLIB codebase, primarily in the rtklib_src/pntpos.c file, to improve positioning accuracy and robustness:
- M-Estimation (Robust Estimation): Employs iterative least squares with a Huber kernel-based weight matrix (W) to mitigate outliers, incorporating truncation for large residuals to enhance stability.
- Zero-Velocity Correction: Improves positioning accuracy during static or low-motion scenarios, critical for smartphone use cases.
- SNR-Weighted Model: Applies Signal-to-Noise Ratio (SNR) weighting to both pseudorange and Doppler measurements, improving signal quality assessment.
- Multipath Error Compensation: Leverages empirical chipset testing to compensate for pseudorange residuals caused by multipath effects, enhancing performance in complex environments.
- Adaptive Q Matrix: Dynamically adjusts the velocity covariance in the Extended Kalman Filter (EKF) using the velocity difference between EKF predictions and Robust Weighted Least Squares (RWLS) estimates.
- Cost Minimization via Quadratic Programming: Implements a post-processing optimization technique based on quadratic programming to minimize positioning errors. This part is implemented using Python and is only for post-processing.
For more technical details, please refer to the online documentation.
MobileGNSS-SPP uses CMake for cross-platform build management, supporting compilation on Linux, macOS, and Windows. For optimal compatibility and ease of use, we recommend compiling on Windows using Microsoft Visual Studio (VS).
The rnx2rtkp application includes a pre-configured Visual Studio project file (msc) for seamless integration. Follow these steps to build the project:
- Open the
rnx2rtkpproject in Visual Studio. - Switch the build configuration to Release mode and set the platform to Win32 to avoid compatibility issues.
- Configure the command-line arguments in Visual Studio:
- Navigate to Configuration Properties > Debugging > Command Arguments.
- Ensure the configuration is set to Release.
- Add the following command-line arguments:
.\rnx2rtkp -x 0 -k ..\conf\rover.conf -o ..\..\..\data\01-opensky\data01\rover.pos ..\..\..\data\01-opensky\data01\rover.obs ..\..\..\data\01-opensky\data01\rover.nav- Build the solution to generate the executable.
MobileGNSS-SPP has been rigorously tested across diverse environments to ensure robustness and reliability. The following scenarios are supported for evaluation:
| Test Scenario | Description |
|---|---|
| Open-Sky (Highway) | High-speed testing on elevated highways or outer ring roads with unobstructed GNSS signals, ideal for evaluating performance in optimal conditions. |
| Urban Streets (Tree-Lined) | Testing in city streets with tree cover, introducing moderate GNSS signal occlusion for assessing performance under partial interference. |
| Complex Urban (Downtown) | Testing in dense urban environments with high-rise buildings, heavy multipath, and signal obstructions, challenging the system's robustness. |
| Occluded Environment (Underpass) | Testing near or under elevated structures (e.g., bridges), where GNSS signals face significant blockage and multipath effects. |
The project includes a suite of Python scripts to facilitate testing, evaluation, and data analysis:
\python
├── rnx2rtkp_batch.py : Batch processes rnx2rtkp for multiple test scenarios.
├── scores.py : Computes positioning accuracy metrics for algorithm evaluation.
├── scores_batch.py : Aggregates accuracy metrics across multiple test runs (run after rnx2rtkp_batch.py).
├── data_analysis : Tools for in-depth GNSS data analysis.
│ ├── 2.4#pr_doppler_corr.py : Analyzes pseudorange and Doppler clock drift (see Section 2.4 of the documentation).
│ └── 2.5#prr_tdcp.py : Examines Doppler and Time-Differenced Carrier Phase (TDCP) correlations (see Section 2.5).
├── mincost : Quadratic programming-based global optimizer for post-processing.
└── rtklipy : Python implementation of RTKLIB for additional flexibility.
- Prepare test data in the
datadirectory, organized by scenario (e.g.,data/01-opensky/). - Execute
rnx2rtkp_batch.pyto process test data across multiple scenarios. - Run
scores_batch.pyto evaluate positioning accuracy and generate performance reports. - Use the
data_analysisscripts for detailed insights into pseudorange, Doppler, and multipath effects.
MobileGNSS-SPP is licensed under the MIT License. See the LICENSE file for details.
- Built upon the robust foundation of RTKLIB.
- Gratitude to the participants of the Google Decimeter Challenge for their excellent code contributions: @taroz1461, @saitodevel01, @timeverett.
- Special thanks to the GNSS research community for providing valuable insights and test methodologies.
For questions, bug reports, or feature requests, please open an issue on the GitHub repository. For general inquiries, contact winchell.hu@outlook.com.