Scholarlink

<p>The development of the Internet of Things and autonomous driving has significantly increased the demand for high-precision and robust vehicle positioning in modern intelligent transport systems. Nowadays, precise point positioning (PPP) technology is recognized as an important approach for worldwide accurate positioning services with Global Navigation Satellite Systems (GNSS). However, traditional PPP methods, which typically rely on the extended Kalman filter (EKF) for state estimation at a single time step, struggle to deliver satisfactory performance in urban canyons, where GNSS measurements are often affected by multipath and non-line-of-sight signals. In this contribution, we propose a PPP method based on a factor graph optimization (FGO) framework associated with robust regression, aiming to enhance the positioning accuracy and robustness of vehicular navigation in urban areas. The FGO estimator fully exploit the historical information by incorporating multi-epoch raw code and phase observations in a sliding window and correlates multi-epoch state parameters using a marginalization strategy. In particular, to mitigate the impact of outliers in urban environments, the Huber regression model, which combines the advantages of the mean squared error (MSE) and mean absolute error models, is introduced in parameter estimation. Under the FGO framework, the impact of multi-window processing and Huber regression on parameter estimation is carefully investigated and discussed. In scenarios with generally open conditions, the proposed FGO-based PPP algorithm demonstrates a gradual reduction in positioning errors during the PPP convergence process, with an increase in window size. When the window size reaches 20&#xa0;s, this method achieves a 46.6% improvement compared to the EKF-based PPP method. In obstructed environments, the Huber regression plays an important role in the suppression of outliers, and the 3D RMSE of positioning errors is improved from 1.22 to 0.51&#xa0;m when employing the proposed FGO-based PPP method.</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

Decimeter-level vehicular positioning in urban environments with factor graph-based PPP and robust regression

  • 期刊论文

  • 发表日期：

    2025年6月13日

查看全文
  • Xin Li,
  • Feiyang Wu,
  • Hanyu Chang,
  • Xingxing Li,
  • Yuxuan Tan,
  • Zhiheng Shen,
  • Xiaohong Zhang

摘要

The development of the Internet of Things and autonomous driving has significantly increased the demand for high-precision and robust vehicle positioning in modern intelligent transport systems. Nowadays, precise point positioning (PPP) technology is recognized as an important approach for worldwide accurate positioning services with Global Navigation Satellite Systems (GNSS). However, traditional PPP methods, which typically rely on the extended Kalman filter (EKF) for state estimation at a single time step, struggle to deliver satisfactory performance in urban canyons, where GNSS measurements are often affected by multipath and non-line-of-sight signals. In this contribution, we propose a PPP method based on a factor graph optimization (FGO) framework associated with robust regression, aiming to enhance the positioning accuracy and robustness of vehicular navigation in urban areas. The FGO estimator fully exploit the historical information by incorporating multi-epoch raw code and phase observations in a sliding window and correlates multi-epoch state parameters using a marginalization strategy. In particular, to mitigate the impact of outliers in urban environments, the Huber regression model, which combines the advantages of the mean squared error (MSE) and mean absolute error models, is introduced in parameter estimation. Under the FGO framework, the impact of multi-window processing and Huber regression on parameter estimation is carefully investigated and discussed. In scenarios with generally open conditions, the proposed FGO-based PPP algorithm demonstrates a gradual reduction in positioning errors during the PPP convergence process, with an increase in window size. When the window size reaches 20 s, this method achieves a 46.6% improvement compared to the EKF-based PPP method. In obstructed environments, the Huber regression plays an important role in the suppression of outliers, and the 3D RMSE of positioning errors is improved from 1.22 to 0.51 m when employing the proposed FGO-based PPP method.