TL;DR
This paper investigates the performance of real-time GPS/Galileo Precise Point Positioning (PPP) integrated with an Inertial Navigation System (INS), utilizing uncombined bias products for improved ambiguity resolution. The study demonstrates that this integrated approach significantly enhances positioning accuracy, especially in challenging environments with GNSS signal interruptions, outperforming standalone PPP.
What Problem Does This Paper Address?
Precise Point Positioning (PPP) offers high-accuracy positioning using GNSS data from a single receiver but typically requires a lengthy convergence period. Integrating PPP with an Inertial Navigation System (INS) can improve robustness and resilience, particularly during GNSS signal outages. However, effectively combining these systems, especially with the latest GNSS advancements like Galileo and the use of uncombined bias products for faster ambiguity resolution, remains an area of active research. This paper aims to analyze the performance of a real-time GPS/Galileo PPP/INS integration that leverages uncombined bias products to accelerate ambiguity resolution and improve positioning accuracy in various scenarios, including challenging urban environments.
Methods
The core of this research lies in the integration of GPS/Galileo dual-frequency ionosphere-free (IF) PPP with an INS. The key novelty is the use of uncombined bias products provided by CNES. These products allow for the resolution of carrier phase ambiguities independently of the user-side PPP modeling, enabling faster and more robust ambiguity resolution (AR). The study evaluates two integration strategies:
- Loosely Coupled Integration (LCI): The GNSS receiver is treated as a "black box," and its position/velocity outputs are fused with INS data.
- Tightly Coupled Integration (TCI): GNSS measurements are directly integrated into the INS state estimation, offering better resilience to signal outages.
The paper details the mathematical models for both LCI and TCI, incorporating the uncombined bias products into the PPP observation equations. A discrete extended Kalman filter is employed for the PPP/INS integration.
Datasets
The study utilizes real-world data collected from two test scenarios:
- Train Test: Conducted on the roof of the Nottingham Geospatial Building, offering good satellite visibility and a circular path.
- Van Tests: Performed in a complex road and city center environment, characterized by frequent GNSS signal blockages due to bridges, vegetation, and high-rise buildings.
The GNSS data was collected using a LEICA GS10 receiver, and the inertial data was gathered from a NovAtel SPAN UIMU-LCI tactical-grade IMU. CNES real-time orbit, clock, and uncombined bias products were used as corrections.
Key Insights and Ideas
- Uncombined Bias Products are Key: The use of uncombined bias products significantly aids in resolving carrier phase ambiguities, which is crucial for achieving high-accuracy PPP, especially in dynamic scenarios.
- TCI Outperforms LCI in Challenging Environments: While LCI and TCI show similar performance in open-sky conditions, TCI demonstrates superior robustness and accuracy when GNSS signals are frequently interrupted, effectively mitigating positioning drift.
- INS Complements PPP During Outages: The INS plays a vital role in bridging GNSS signal gaps, preventing complete loss of positioning and maintaining a higher level of accuracy than standalone PPP during these periods.
- Ambiguity Resolution Boosts Accuracy: Fixing carrier phase ambiguities (AR) leads to substantial improvements in positioning accuracy, particularly in the east component, across all tested modes.
- Real-time Performance is Achievable: The study successfully demonstrates the real-time application of these advanced PPP/INS integration techniques, showing their practical viability.
Key Results
- Train Test (Open Sky):
- Ambiguity-float PPP, LCI, and TCI achieved similar accuracies (e.g., ~4.9 cm in the Up direction for TCI).
- After AR, the east error component saw significant improvements: 47% for PPP-AR, 40% for PPP-AR/INS LCI, and 38% for PPP-AR/INS TCI.
- Van Tests (Complex Road/City Center):
- Frequent signal interruptions led to re-convergence issues for PPP.
- TCI achieved the highest accuracies: 32 cm (N), 29 cm (E), and 41 cm (U).
- TCI effectively eliminated the solution re-convergence seen in PPP.
- The additional uncombined bias correction further improved accuracy, especially for the up component.
Limitations and Future Work
- Heading Alignment: The study notes that dramatic changes in heading during movement increased the difficulty of precise yaw angle determination, suggesting that external heading alignment might be required for improved performance.
- Convergence Time: While improved by AR, the dual-frequency IF PPP model may still require over 15 minutes to converge to high accuracy in some scenarios.
- Low-Cost IMU Integration: The paper suggests that further investigation is needed for IF PPP integration with low-cost IMUs.
- Multi-Frequency PPP: The current dual-frequency model could be further evaluated, and the potential of multi-frequency PPP/INS integration with uncombined bias products warrants more exploration.
Summary updated