TL;DR
This paper introduces a novel 30-position calibration scheme for rotational inertial navigation systems (RINS) that significantly improves the accuracy of inertial measurement units (IMUs). By employing a 30-dimensional Kalman filter and a carefully designed rotation sequence, the method achieves higher observability of sensor error parameters, leading to more precise calibration and reduced navigation errors compared to existing techniques.
What Problem Does This Paper Address?
Inertial Navigation Systems (INS), particularly those using Rotational Inertial Navigation Systems (RINS), are crucial for navigation. However, their accuracy is heavily dependent on the quality of the Inertial Measurement Unit (IMU). IMUs are prone to various errors, such as constant biases, scale factor inaccuracies, and installation misalignments. While separate calibration of IMU sensors is common, it's time-consuming and impractical for systems that cannot be easily disassembled or returned to a lab. System-level calibration offers a solution by estimating these errors directly within the navigation system. The challenge lies in designing a calibration scheme that can effectively excite and isolate all relevant error parameters, ensuring high observability and accurate estimation, especially for the complex error models of modern IMUs. Existing system-level calibration methods often struggle to achieve comprehensive error identification, leading to suboptimal navigation performance.
Methods
The core of this paper is the development of an improved system-level calibration scheme for RINS. The authors propose a 30-position rotation calibration scheme designed to maximize the observability of IMU error parameters. This scheme is integrated with a 30-dimensional Kalman filter to estimate a comprehensive set of error parameters.
Key methodological aspects include:
- Error Model Establishment: The paper first defines a detailed error model for IMUs, encompassing constant errors (bias), scale factor errors, and installation errors for both gyroscopes and accelerometers.
- 30-Dimensional Kalman Filter: A Kalman filter is designed to estimate these error parameters by observing the velocity and position errors of the navigation system. The state vector includes attitude, velocity, and position errors, along with all the identified IMU sensor error parameters.
- 30-Position Rotation Scheme: A specific sequence of 30 distinct rotational maneuvers is designed. This sequence is crucial for providing the necessary excitation to differentiate and estimate the various error parameters. The paper notes that only a portion of this scheme is presented, with the remaining maneuvers being opposite in direction to ensure comprehensive excitation.
- Observability Analysis: The authors employ covariance analysis to quantitatively assess the observability of the error parameters for their proposed scheme and compare it against a traditional 18-position scheme. This analysis helps demonstrate the superiority of their approach in exciting and identifying error sources.
Datasets
Not discussed in the paper. The paper relies on simulation experiments and experimental data from a specific IMU model for verification.
Key Insights and Ideas
- The Power of a Well-Designed Trajectory: The study highlights that the specific sequence of rotations (the calibration scheme) is paramount for effective system-level calibration. A carefully crafted trajectory can significantly improve the ability to distinguish and estimate multiple error parameters simultaneously.
- 30-D Kalman Filter for Comprehensive Calibration: Moving to a higher-dimensional state space (30 dimensions) allows for the simultaneous estimation of a more complete set of IMU error parameters, including constant, scale factor, and installation errors for both gyroscopes and accelerometers.
- Observability is Key: The paper strongly emphasizes the link between the calibration scheme's design and the observability of error parameters. Higher observability directly translates to more accurate error estimation and, consequently, better navigation performance.
- Practical Value Demonstrated: Beyond simulations, the authors validate their approach with real-world experimental data from an IMU, showing good convergence of error parameter estimates, which underscores the practical applicability of their method.
Key Results
- Improved Observability: The proposed 30-position scheme demonstrates significantly higher observability for most error parameters compared to a traditional 18-position scheme.
- Reduced Residual Errors: Simulations show that the 30-position scheme results in smaller residual errors for the estimated sensor parameters after calibration.
- Enhanced Navigation Accuracy: In simulations, the 30-position scheme leads to a substantial reduction in navigation positioning error, with a maximum position error reduction of 3639 m compared to the 18-position scheme after 48 hours.
- Experimental Convergence: Real-world experiments with a specific IMU model show that the error parameter estimation curves converge well, indicating the practical value and effectiveness of the proposed calibration method.
Limitations and Future Work
- Computational Complexity: While not explicitly detailed as a limitation, a 30-dimensional Kalman filter inherently increases computational load compared to lower-dimensional filters.
- Assumptions: The error model assumes that random errors are negligible and that the installation errors are small.
- Future Work: The paper implicitly suggests that further optimization of calibration schemes and exploration of more advanced filtering techniques could lead to even greater improvements in calibration accuracy and efficiency. The authors also mention that the scheme can be applied to other types of IMUs, implying potential for broader application.
Summary updated