An Anti-Turbulence Self-Alignment Method for SINS under Unknown Latitude Conditions

Yubing Jiao, Jie Li, Xihong Ma, Kaiqiang Feng, Xiaoting Guo, Xiaokai Wei, Yujun Feng, Chenming Zhang, Jingqi WangSensors 2022 · DOI

TL;DR

This paper introduces a novel self-alignment method for Strapdown Inertial Navigation Systems (SINS) that can operate even when the system's latitude is unknown and it's subjected to significant turbulence. By combining an ant colony simulated annealing algorithm for optimizing gravity vector measurements with a polynomial fitting approach for simultaneous latitude estimation and attitude alignment, the method achieves robust performance and improved accuracy in challenging environments.

What Problem Does This Paper Address?

Strapdown Inertial Navigation Systems (SINS) are crucial for navigation in various applications, but their accuracy heavily relies on knowing the system's initial state, particularly its geographic latitude. Traditional alignment methods often require precise latitude information, which is not always available, especially in environments with poor satellite reception or significant interference. Furthermore, turbulence and vibrations can severely degrade the accuracy of inertial sensor readings, making alignment difficult. Existing methods for aligning SINS under unknown latitude conditions often struggle with interference, require significant computational resources, or don't fully utilize the available sensor data. This paper addresses the challenge of achieving accurate SINS self-alignment in the presence of unknown latitude and significant environmental disturbances like turbulence and vibration.

Methods

The proposed method combines two key innovations:

  • Ant Colony Simulated Annealing (ACSA) for Gravity Vector Optimization: The accuracy of SINS alignment is highly dependent on the quality of gravity measurements. In turbulent environments, accelerometer data is noisy and prone to errors. This paper employs an ACSA algorithm to optimize the gravitational apparent motion (GAM) vector. This approach leverages the positive feedback mechanism of ant colony optimization to explore the solution space and the simulated annealing algorithm to escape local optima and converge faster to a more accurate gravity vector. This is a novel combination aimed at improving the robustness of the gravity measurement against noise and interference.
  • Polynomial Fitting for Simultaneous Latitude Estimation and Self-Alignment: Once a more accurate gravity vector is obtained, the paper uses a polynomial fitting algorithm to simultaneously estimate the unknown latitude and align the SINS attitude. Instead of estimating latitude first and then performing alignment, this method integrates both processes. The polynomial fitting models the time-varying gravity vector, and by analyzing the coefficients of this polynomial, both the latitude and the attitude matrix can be determined in real-time. This approach aims to improve alignment speed and accuracy by utilizing the full inertial device output.

Datasets

Not discussed in the paper. The paper relies on simulations and onboard experiments using a vehicle equipped with an Inertial Measurement Unit (IMU) and a reference system (NovAtel SPAN-LCI) for ground truth.

Key Insights and Ideas

  • Integrated Latitude and Attitude Alignment: The most significant insight is the move away from a sequential estimation of latitude followed by attitude alignment. By integrating these two processes using polynomial fitting, the method can achieve better performance and potentially faster convergence.
  • ACSA for Robust Gravity Measurement: The application of ACSA to optimize the gravity vector is a key contribution. It demonstrates that metaheuristic algorithms can effectively denoise and improve the accuracy of inertial sensor data in challenging conditions, which is fundamental for subsequent alignment.
  • Improved Anti-Interference Capability: The paper highlights that the proposed method is significantly more robust to turbulence and vibrations compared to traditional methods like the Optimization-Based Alignment (OBA). This is crucial for SINS applications in dynamic and harsh environments.
  • Efficiency of Low-Order Polynomial Fitting: While higher-order polynomial fits were explored, the study found that a third-order polynomial provided the best balance between fitting accuracy and convergence for latitude estimation, suggesting that overly complex models are not always superior.

Key Results

  • Latitude Estimation Accuracy: The proposed method improved latitude estimation accuracy by six times compared to traditional methods. The RMSE for latitude error in the 201-300s interval was 0.2857°, significantly better than DV (3.6445°) and DW-VS (0.6718°).
  • Attitude Alignment Accuracy:
    • Yaw angle error RMSE was 0.1779° (compared to 1.3572° for OBA).
  • Roll angle error RMSE was 0.0131° (compared to 0.1394° for OBA).
  • Pitch angle error RMSE was 0.0082° (compared to 0.1140° for OBA).
  • The alignment error of the yaw angle ranged from 0.2° to 0.7°, and roll/pitch angle errors were within 0.1°.
  • Gravity Vector Optimization: The RMSE of the gravity vector solved by ACSA was 0.0014, an order of magnitude lower than raw data (0.0449) and significantly better than Newton's method (0.0064).
  • Computational Efficiency: The ACSA algorithm reduced the gravity vector optimization time to 4.77s, a significant speedup compared to Newton's method (15.86s) and ACC (13.03s).

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that while the proposed method is more robust, it's still susceptible to interference. The OBA method's performance degrades significantly under motion, while the proposed method shows only a small change, implying room for further improvement in extreme interference scenarios.
  • The study primarily focuses on the "wobble base" scenario, which simulates turbulence. Its performance in other types of complex environments (e.g., extreme temperature variations, strong magnetic fields) is not explicitly detailed.
  • Future Work:
    • The paper suggests that the method is "easy to apply in engineering practice," implying potential for real-world deployment.
  • Further research could explore extending this method to more complex and dynamic interference environments.
  • Investigating the impact of different types of inertial sensor biases and noise models on the ACSA and polynomial fitting algorithms could also be a valuable direction.

Summary updated

BibTeX
@article{jiao2022anti,
  author = {Yubing Jiao and Jie Li and Xihong Ma and Kaiqiang Feng and Xiaoting Guo and Xiaokai Wei and Yujun Feng and Chenming Zhang and Jingqi Wang},
  title = {{An Anti-Turbulence Self-Alignment Method for SINS under Unknown Latitude Conditions}},
  journal = {Sensors},
  year = {2022},
  doi = {10.3390/s22134686},
}

Loading related content...

Figure