Inertial-Navigation-Aided Single-Satellite Highly Dynamic Positioning Algorithm

Lingling Zhang, Chengkai Tang, Yi Zhang, Houbing SongSensors 2019 · DOI

TL;DR

This paper introduces a novel algorithm for highly dynamic positioning using a single satellite, crucial for situations where traditional GNSS is unavailable. By fusing inertial navigation system (INS) data with single-satellite measurements using an Unscented Kalman Filter (UKF), the algorithm achieves significantly improved accuracy and robustness compared to existing methods.

What Problem Does This Paper Address?

Global Navigation Satellite Systems (GNSS) like GPS are widely used for positioning, but they are vulnerable in environments where signals are denied, such as during wartime or in urban canyons. In such scenarios, a rapidly deployable single-satellite positioning system offers a promising alternative. However, existing single-satellite methods struggle with highly dynamic targets, often exhibiting low accuracy and rapid divergence. This paper tackles the challenge of achieving precise and stable positioning for fast-moving objects using only a single satellite, by effectively integrating INS data to compensate for the inherent limitations of single-satellite systems.

Methods

The proposed algorithm combines two key components: an initial coarse positioning step and a refined data fusion step.

  1. Initial Coarse Positioning:
    • The Inertial Navigation System (INS) is used to provide real-time displacement feedback for the highly dynamic target. This compensates for the target's motion.
  2. The algorithm then utilizes the continuity of position change and the high accuracy of the Unscented Kalman Filter (UKF).
  3. Integral Doppler pseudorange is used to construct an "equal-frequency cone surface," and height information from an altimeter establishes a "height surface." These surfaces are used to obtain a coarse positioning solution.
  4. A least-squares method, based on Taylor expansion, is employed to solve for the initial target position by iteratively refining the solution until a desired accuracy threshold is met.
  5. Navigation-Data Fusion (UKF):
    • The initial coarse position from the previous step is further refined using an Unscented Kalman Filter (UKF). The UKF is chosen over traditional Kalman filters because it handles nonlinear models more effectively, which is crucial for dynamic targets and the inherent nonlinearities in satellite measurements.
  6. The UKF approximates the probability distribution of the state, avoiding the need for Jacobi matrix calculations, which simplifies implementation and improves computational efficiency.
  7. The augmented state vector includes position, velocity, acceleration, and INS-related parameters, allowing the filter to simultaneously estimate and correct for these factors.
  8. The prediction and update steps of the UKF are iterated to achieve stable final positioning results with smooth transitions.

Datasets

The paper does not explicitly mention the use of pre-existing datasets for training. Instead, it relies on simulated data generated to mimic various error conditions and a real-world test using a DJI drone.

  • Simulated Data: The authors generated data to simulate several error sources:
  • Orbital errors (0-100 m)
    • Pseudorange errors (0-100 m)
    • Satellite-velocity errors (0-1 m/s)
    • Comprehensive error conditions combining multiple sources.
  • Real-Environment Testing: A DJI four-rotor unmanned aerial vehicle (UAV) equipped with a Hongyan satellite-enhancement system was used for real-world testing. The flight route was recorded, and the ideal path was obtained via GPS and wireless beacon.

Key Insights and Ideas

  • INS as a Motion Compensator: The paper highlights the critical role of INS in compensating for the rapid motion of targets, which is a major limitation for traditional single-satellite positioning. This allows the system to maintain a stable reference frame.
  • UKF for Nonlinearity: The choice of UKF is a key insight. It effectively handles the nonlinear relationships between satellite measurements (like pseudorange and Doppler) and the target's position, outperforming linear filters like EKF in dynamic scenarios.
  • Fusion of Multiple Data Sources: By combining INS-derived motion information with satellite measurements (pseudorange and height), the algorithm creates a more robust and accurate positioning solution than relying on any single source alone.
  • Iterative Refinement: The initial coarse positioning using least-squares followed by UKF refinement provides a stable and accurate final solution, demonstrating the benefit of a multi-stage approach.

Key Results

The paper presents simulation and real-world testing results comparing the proposed algorithm against two other methods: a least-squares Newton-iterative Doppler single-satellite positioning system (LS) and an EKF-based Doppler single-satellite positioning method (EKF).

  • Accuracy Improvement: The proposed UKF-based algorithm consistently outperformed both the LS and EKF methods across various error conditions.
  • In synthetic error conditions, the positioning error was 10% less than the other two algorithms.
  • Under orbital error, the proposed algorithm showed significantly better performance than LS, and was superior to EKF.
  • Under pseudorange error, the proposed algorithm was better than EKF and much better than LS.
  • Under satellite-velocity error, the proposed algorithm was also superior to EKF and better than LS.
  • In a comprehensive error environment, the mean positioning error was less than 50 m, with rapid convergence (within 10 s). The mean error was only 3.9% of the least-squares algorithm, and the EKF algorithm was 8.6% of the least-squares algorithm.
  • Real-Environment Testing: The average error in real-world testing was 46.6% of that of the EKF algorithm and significantly superior to the LS algorithm.
  • Robustness: The UKF-based algorithm operated smoothly with no evident divergence, even under comprehensive error conditions.

Limitations and Future Work

  • INS Errors: While the algorithm uses INS, it acknowledges that INS itself has inherent errors (bias, random walk). The UKF is designed to mitigate these, but significant INS errors could still impact performance.
  • Single-Satellite Dependency: The core limitation remains the reliance on a single satellite. If the satellite signal is lost entirely, positioning will cease.
  • Computational Load: While UKF is computationally efficient for nonlinear problems, highly dynamic scenarios with complex INS integration might still pose computational challenges for resource-constrained platforms.

The paper suggests that the proposed algorithm has strong practical value under GNSS rejection conditions. Future work could explore extending this to multi-satellite systems or investigating more advanced INS error modeling and compensation techniques.

Summary updated

BibTeX
@article{zhang2019inertial,
  author = {Lingling Zhang and Chengkai Tang and Yi Zhang and Houbing Song},
  title = {{Inertial-Navigation-Aided Single-Satellite Highly Dynamic Positioning Algorithm}},
  journal = {Sensors},
  year = {2019},
  doi = {10.3390/s19194196},
}
Figure