TL;DR
This paper introduces "selective alignment," a novel analytic alignment method for Strapdown Inertial Navigation Systems (SINS) that uses fewer IMU outputs than traditional methods. It offers improved yaw accuracy and potential applications in self-alignment, fault detection, and self-calibration.
What Problem Does This Paper Address?
Strapdown Inertial Navigation Systems (SINS) require initial alignment to determine their orientation relative to the Earth. Analytic alignment (AA) is a method that uses the Earth's gravity and rotation vectors for this purpose, particularly effective when the SINS is stationary with limited vibrations. Traditional analytic coarse alignment methods typically utilize all six outputs from the Inertial Measurement Unit (IMU). However, this can be computationally intensive and may not be optimal. The challenge addressed here is to develop a more efficient and potentially more accurate analytic alignment method that requires fewer IMU inputs while maintaining or improving performance, especially in terms of yaw accuracy.
Methods
The paper proposes a "selective alignment" method. Unlike traditional analytic coarse alignment that uses all six IMU outputs, selective alignment leverages only three outputs from the IMU (e.g., accelerometer and gyroscope readings along specific axes) and uses "a few properties from the remaining outputs such as the sign and the approximate value" to determine the attitude. This method is presented as an extension of the TRIAD algorithm. The derivation involves geometrical analysis to determine the orientation of the body frame axes relative to the navigation frame. The method handles various cases based on which IMU outputs are selected (e.g., FWW for three gyroscope outputs, FWF for two gyroscope and one accelerometer output). The paper also details exception handling for scenarios like zero denominators or negative radicands that can arise from sensor errors or vibrations. A flowchart illustrates the algorithm's logic, including steps for determining the alpha and beta axes and selecting the correct solution from multiple possibilities.
Datasets
Not discussed in the paper. The paper relies on simulations and experimental data collected by the authors.
Key Insights and Ideas
- Reduced IMU Dependency: The core idea is that not all six IMU outputs are strictly necessary for accurate analytic alignment. By selectively using key outputs and inferring information from others, efficiency can be gained.
- Improved Yaw Accuracy: The selective alignment method demonstrated improved yaw accuracy compared to traditional analytic coarse alignment in vehicle experiments.
- Extension of TRIAD: Selective alignment builds upon the TRIAD algorithm, suggesting that the TRIAD algorithm might be overdetermined for certain alignment problems, and fewer vector pairs can suffice.
- Robustness through Exception Handling: The method incorporates specific strategies to handle potential issues like division by zero or imaginary results, making it more robust in the presence of sensor noise and vibrations.
- Versatility: The method shows promise for applications beyond initial alignment, including fault detection and self-calibration.
Key Results
- Simulations and experiments showed that the selective alignment method can achieve comparable or better alignment accuracy than traditional analytic coarse alignment.
- In static alignment simulations, selective alignment yielded results very close to the true attitude, with some cases showing slightly better precision than analytic coarse alignment.
- In a vehicle experiment with vibration, selective alignment using specific combinations of IMU outputs (e.g., $f_x, w_z, f_x$) produced the best results, demonstrating its effectiveness in dynamic conditions.
- The method was shown to be effective in identifying faulty gyroscope outputs in a fault detection simulation.
- In self-calibration tests, selective alignment successfully compensated for errors in lower-accuracy gyroscopes.
Limitations and Future Work
- Vibration Sensitivity: While the method handles some vibrations, the paper notes that in cases of significant vibration, Inertial Navigation Computational Alignment (INCA) might still be necessary, although AA offers lower computational complexity and less alignment time.
- Exception Handling Complexity: While exception handling is implemented, the paper acknowledges that constructing algebraic rules to prevent negative radicands can be complicated.
- Redundancy of Vectors: The authors suggest that the principle of selective alignment, showing that two vectors might be redundant for determining the transformation matrix, could be further studied in other attitude determination problems where the TRIAD algorithm is applied, such as spacecraft attitude estimation. This implies potential for further optimization and application in related fields.
Summary updated