Fast Field Calibration of MIMU Based on the Powell Algorithm

Lin Ma, Wanwan Chen, Bin Li, Zheng You, Zhigang Chen2014 · DOI

TL;DR

This paper introduces a novel, fast, and cost-effective field calibration method for Micro Inertial Measurement Units (MIMUs) using the Powell algorithm. This approach significantly reduces calibration time and complexity compared to traditional laboratory methods, making it more practical for real-world applications, especially in scenarios where precise equipment is unavailable.

What Problem Does This Paper Address?

Micro Inertial Measurement Units (MIMUs) are crucial components in modern navigation systems, offering advantages like small size and low cost. However, their accuracy is heavily impacted by inherent sensor errors such as bias, scale factor, and misalignment. Traditional calibration methods, often performed in laboratories using specialized equipment like three-axis turntables, are time-consuming, expensive, and require strict alignment, limiting their applicability in field scenarios, particularly in demanding environments like battlefields or remote locations. The paper addresses the need for a faster, more convenient, and less equipment-intensive calibration method that can be performed in the field to ensure reliable navigation performance.

Methods

The core of this paper is the development of a "fast field self-calibration method" for MIMUs. The method leverages the Powell algorithm, a robust numerical optimization technique, to solve a set of nonlinear error equations derived from a MIMU error model.

Key aspects of the method include:

  • Multi-Position Calibration Theory: Unlike traditional methods that require precise alignment to gravity and Earth's rotation, this method relies on the fundamental principles that the norm of the accelerometer's output vector equals the local gravity magnitude, and the norm of the gyroscope's output vector equals the applied rotation rate. This allows for calibration without precise alignment to a local level frame.
  • Powell Algorithm: This algorithm is chosen for its ability to solve unconstrained optimization problems without requiring derivative calculations, making it suitable for the nonlinear equations involved in MIMU calibration. It iteratively searches for the optimal parameters by adjusting search directions.
  • Simplified Equipment Requirements: The method utilizes a single-axis turntable for providing angular rates for gyroscope calibration, which is significantly less complex and more portable than the three-axis turntables used in traditional methods. This makes it suitable for field applications.
  • Error Model: The paper utilizes standard error models for accelerometers and gyroscopes, accounting for biases, scale factors, and misalignments.

Datasets

The paper describes two types of data used:

  • Synthetically Generated Data: The proposed fast field calibration method was initially validated using synthetically generated data through MATLAB. This data was designed to simulate actual MIMU outputs for 18 different positions, allowing for a controlled comparison of the algorithm's performance against known parameters.
  • Experimental Data: The calibrated MIMU was then tested in a real-world vehicle test. The navigation performance was evaluated against a high-accuracy GPS/INS device, serving as the "true ground" reference. This involved collecting data during straight-line and circular trajectories.

Key Insights and Ideas

  • Field Calibration is Feasible and Effective: The most significant insight is that accurate MIMU calibration can be achieved in a field setting with significantly less specialized equipment and time than traditional laboratory methods.
  • Powell Algorithm's Suitability: The Powell algorithm proves to be an effective tool for solving the complex nonlinear equations inherent in MIMU error modeling, offering a practical alternative to other optimization techniques.
  • Reduced Alignment Sensitivity: The multi-position calibration theory, by focusing on the norms of sensor outputs rather than precise orientations, drastically reduces the need for meticulous alignment, a major bottleneck in traditional methods.
  • Cost and Time Savings: The combination of simplified equipment and a faster calibration procedure leads to substantial cost and time savings, making MIMU calibration more accessible.

Key Results

  • Comparable Accuracy: The calibration results obtained using the proposed multi-position method (biases and scale factors) were found to be comparable to those achieved with traditional laboratory calibration methods.
  • Navigation Performance: Vehicle tests demonstrated that the navigation performance of the MIMU calibrated with the proposed method was close to that of the classical method, with only slightly higher navigation errors (e.g., eastward absolute mean error of 7.92 m vs. 7.54 m in straight-line tests). These differences were deemed acceptable.
  • Time Efficiency: The paper emphasizes that the proposed method saves significant time compared to traditional calibration. While specific time savings are not quantified numerically, the description of the process implies a substantial reduction.
  • Relative Errors: For the synthetically generated data, the relative errors between assumed and simulated parameters were less than 1000, demonstrating the effectiveness of the algorithm.

Limitations and Future Work

  • Temperature Independence: The current calibration method does not account for temperature variations, which can affect sensor performance. The authors acknowledge this limitation and suggest it as an area for future research.
  • Triad Misalignment: The paper notes that the approach may not achieve precise calibration when the accelerometer and gyroscope triads are significantly misaligned. This is also identified as a topic for further investigation.
  • Further Exploration: The authors suggest that future studies could explore these limitations to further enhance the calibration accuracy and applicability of the method.

Summary updated

BibTeX
@misc{ma2014fast,
  author = {Lin Ma and Wanwan Chen and Bin Li and Zheng You and Zhigang Chen},
  title = {{Fast Field Calibration of MIMU Based on the Powell Algorithm}},
  year = {2014},
  doi = {10.3390/s140916062},
}
Figure