Scholarlink

<p>Global Navigation Satellite System (GNSS) enables absolute positioning and attitude determination, essential for applications like autonomous driving, structural health monitoring, and aviation. A GNSS antenna array with a known geometry, mounted on the object, enables simultaneous estimation of position and attitude parameters in a nonlinear dynamic system. Integrated GNSS positioning and attitude determination is typically a complex nonlinear estimation problem, often solved using the Extended Kalman Filter (EKF), which employs local linearization to ensure reliable convergence. We developed an approach with Unscented Kalman Filter (UKF) to improve the convergence robustness and estimation accuracy of integrated positioning and attitude determination especially in scenarios where the initial attitude error is large. This approach can well handle the nonlinearity of the dynamic system, which is less sensitive to initial attitude accuracy, allowing rapid convergence even with significant initial attitude errors. We parameterized the attitude as a unit quaternion and detailed the approaches based on both EKF and UKF. This study explores UKF's performance in the integrated positioning and attitude determination. Experimental results indicate that the accuracy and reliability of the EKF-based approach significantly deteriorate when the initial attitude error exceeds 5°, and divergence may occur when it exceeds 10°. Under an initial attitude error of 10°, the EKF-based method yields a maximum attitude Root Mean Square Errors (RMSE) of 6.634°, with horizontal and vertical positioning RMSE errors of 0.095&#xa0;m and 0.371&#xa0;m, respectively. Its horizontal convergence time exceeds 900&#xa0;s. In contrast, the UKF-based approach demonstrates substantially better performance under the same conditions, achieving a maximum attitude RMSE of just 0.026°, horizontal and vertical positioning RMSE of 0.062&#xa0;m and 0.082&#xa0;m, and a shorter horizontal convergence time of approximately 660&#xa0;s. It maintains accuracy and convergence even with initial attitude errors up to 15 degrees. These findings highlight UKF's superiority over EKF for simultaneous positioning and attitude determination, especially when initial attitude accuracy is not guaranteed.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integrated GNSS positioning and attitude determination with unscented Kalman filter

  • 期刊论文

  • 发表日期：

    2025年10月8日

查看全文
  • Xiangdong An,
  • Xiaolin Meng,
  • Yilin Xie,
  • Fan Zhang,
  • Liangliang Hu,
  • Rui Shang

摘要

Global Navigation Satellite System (GNSS) enables absolute positioning and attitude determination, essential for applications like autonomous driving, structural health monitoring, and aviation. A GNSS antenna array with a known geometry, mounted on the object, enables simultaneous estimation of position and attitude parameters in a nonlinear dynamic system. Integrated GNSS positioning and attitude determination is typically a complex nonlinear estimation problem, often solved using the Extended Kalman Filter (EKF), which employs local linearization to ensure reliable convergence. We developed an approach with Unscented Kalman Filter (UKF) to improve the convergence robustness and estimation accuracy of integrated positioning and attitude determination especially in scenarios where the initial attitude error is large. This approach can well handle the nonlinearity of the dynamic system, which is less sensitive to initial attitude accuracy, allowing rapid convergence even with significant initial attitude errors. We parameterized the attitude as a unit quaternion and detailed the approaches based on both EKF and UKF. This study explores UKF's performance in the integrated positioning and attitude determination. Experimental results indicate that the accuracy and reliability of the EKF-based approach significantly deteriorate when the initial attitude error exceeds 5°, and divergence may occur when it exceeds 10°. Under an initial attitude error of 10°, the EKF-based method yields a maximum attitude Root Mean Square Errors (RMSE) of 6.634°, with horizontal and vertical positioning RMSE errors of 0.095 m and 0.371 m, respectively. Its horizontal convergence time exceeds 900 s. In contrast, the UKF-based approach demonstrates substantially better performance under the same conditions, achieving a maximum attitude RMSE of just 0.026°, horizontal and vertical positioning RMSE of 0.062 m and 0.082 m, and a shorter horizontal convergence time of approximately 660 s. It maintains accuracy and convergence even with initial attitude errors up to 15 degrees. These findings highlight UKF's superiority over EKF for simultaneous positioning and attitude determination, especially when initial attitude accuracy is not guaranteed.