Scholarlink

<p>With the increasing complexity of the transportation environment, ensuring the credibility of positioning information is a key challenge for the safe application of PNT (Positioning, Navigation, and Timing) services. To explore the feasibility of credible positioning from the client side, this paper presents a credible positioning model based on the Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) network-based tight integration of Real Time Kinematic (RTK) and Inertial navigation system (INS). In this model, factors that affect the positioning error of the RTK/INS tight integration are considered as inputs, and a customized loss function is designed. By using such a method, a credible factor of positioning solution is generated. To evaluate the capability of this credible factor, several sets of vehicle-borne data in open-sky and urban environments are processed and analyzed. Results illustrate that the credible factor can envelop 95.31% and 86.06% of the real positioning errors in horizontal and vertical directions at envelope levels of 2.0&#xa0;cm and 2.5&#xa0;cm in the open-sky environment. In the urban environment, the envelope level of credible factors in horizontal and vertical directions can reach centimeter-level and 20-cm level with the corresponding probabilities of 87.21% and 85.39%. Compared with the protection level calculated using&#xa0;the Multiple Hypothesis Solution Separations (MHSS) Advanced Autonomous Integrity Monitoring (ARAIM) solution, the credible factor provides a more precise and efficient representation of positioning errors.</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

An initial credible positioning model: the CNN-LSTM aided RTK/INS tight integration

  • 期刊论文

  • 发表日期：

    2025年7月8日

查看全文
  • Qiaozhuang Xu,
  • Zhouzheng Gao,
  • Hongzhou Yang,
  • Cheng Yang,
  • Guan Wang,
  • Jie Lv

摘要

With the increasing complexity of the transportation environment, ensuring the credibility of positioning information is a key challenge for the safe application of PNT (Positioning, Navigation, and Timing) services. To explore the feasibility of credible positioning from the client side, this paper presents a credible positioning model based on the Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) network-based tight integration of Real Time Kinematic (RTK) and Inertial navigation system (INS). In this model, factors that affect the positioning error of the RTK/INS tight integration are considered as inputs, and a customized loss function is designed. By using such a method, a credible factor of positioning solution is generated. To evaluate the capability of this credible factor, several sets of vehicle-borne data in open-sky and urban environments are processed and analyzed. Results illustrate that the credible factor can envelop 95.31% and 86.06% of the real positioning errors in horizontal and vertical directions at envelope levels of 2.0 cm and 2.5 cm in the open-sky environment. In the urban environment, the envelope level of credible factors in horizontal and vertical directions can reach centimeter-level and 20-cm level with the corresponding probabilities of 87.21% and 85.39%. Compared with the protection level calculated using the Multiple Hypothesis Solution Separations (MHSS) Advanced Autonomous Integrity Monitoring (ARAIM) solution, the credible factor provides a more precise and efficient representation of positioning errors.