Accurate localization is paramount for unmanned aerial vehicles (UAVs) spanning various technical and industrial domains, necessitating a comprehensive assessment of global navigation satellite system (GNSS) precision...Accurate localization is paramount for unmanned aerial vehicles (UAVs) spanning various technical and industrial domains, necessitating a comprehensive assessment of global navigation satellite system (GNSS) precision. This study investigates the performance of distinct GNSS constellations in determining the precise location of a building utilizing a high-precision GNSS receiver. The receiver, incorporating advanced multi-frequency and full-constellation positioning capabilities, was integrated with a smartphone via Bluetooth to enable the UAV’s acquisition of centimeter-level positioning data. Sequential utilization of single satellite systems—such as GPS-only, GLONASS-only, Galileo-only, SBAS-only, and BeiDou-only—facilitated the documentation of latitude and longitude coordinates for the designated building. Subsequent comparison of these coordinates with a specialized Geographic Information System (GIS) was conducted to evaluate their positional accuracy. The comparative analysis underscores significant variability in the precision offered by each satellite constellation, providing valuable insights for optimizing UAV navigation across GIS, IoT, construction, and other sectors requiring high-precision localization. This research underscores the significance of high-precision GNSS receivers in enhancing UAV-based geospatial assessments, emphasizing the critical selection of appropriate satellite systems for tailored localization tasks. The study contributes to advancing UAV navigation strategies, ensuring robust and accurate geospatial data collection within diverse operational frameworks.展开更多
GNSS观测时间序列包含复杂的非线性构造运动,如地面质量荷载、模型残差、周围环境因素等。由于环境因素的复杂性,季节性信号可能具备准周期时变的特征,传统的时间序列分析模型很难模型化。因此,可以采用一种双向长短期记忆(Bidirectiona...GNSS观测时间序列包含复杂的非线性构造运动,如地面质量荷载、模型残差、周围环境因素等。由于环境因素的复杂性,季节性信号可能具备准周期时变的特征,传统的时间序列分析模型很难模型化。因此,可以采用一种双向长短期记忆(Bidirectional Long Short-Term Memory,BiLSTM)循环神经网络与变分模态分解(Variational Mode Decomposition,VMD)联合的信号重构方法。首先利用VMD强大的分解能力将GNSS信号进行频域剖分并将其分为多项子信号和噪声项,再基于BiLSTM强大的学习能力对GNSS信号进行训练建模。结果表明,BiLSTM+VMD模型能充分挖掘信号的时频域特征,提高信号重构的精度和稳定性,GNSS N、E、U三分量重构结果均方根误差(Root Mean Squared Error,RMSE)都表现出不同程度的降低,尤其水平方向效果更为显著,相比EMD与VMD方法,E方向离散度分别降低了61%和19%,N方向离散度分别降低了20%和14%。这为GNSS观测时间序列中信号提取与模型参数估计提供了一个有价值的模型。展开更多
文摘Accurate localization is paramount for unmanned aerial vehicles (UAVs) spanning various technical and industrial domains, necessitating a comprehensive assessment of global navigation satellite system (GNSS) precision. This study investigates the performance of distinct GNSS constellations in determining the precise location of a building utilizing a high-precision GNSS receiver. The receiver, incorporating advanced multi-frequency and full-constellation positioning capabilities, was integrated with a smartphone via Bluetooth to enable the UAV’s acquisition of centimeter-level positioning data. Sequential utilization of single satellite systems—such as GPS-only, GLONASS-only, Galileo-only, SBAS-only, and BeiDou-only—facilitated the documentation of latitude and longitude coordinates for the designated building. Subsequent comparison of these coordinates with a specialized Geographic Information System (GIS) was conducted to evaluate their positional accuracy. The comparative analysis underscores significant variability in the precision offered by each satellite constellation, providing valuable insights for optimizing UAV navigation across GIS, IoT, construction, and other sectors requiring high-precision localization. This research underscores the significance of high-precision GNSS receivers in enhancing UAV-based geospatial assessments, emphasizing the critical selection of appropriate satellite systems for tailored localization tasks. The study contributes to advancing UAV navigation strategies, ensuring robust and accurate geospatial data collection within diverse operational frameworks.
文摘GNSS观测时间序列包含复杂的非线性构造运动,如地面质量荷载、模型残差、周围环境因素等。由于环境因素的复杂性,季节性信号可能具备准周期时变的特征,传统的时间序列分析模型很难模型化。因此,可以采用一种双向长短期记忆(Bidirectional Long Short-Term Memory,BiLSTM)循环神经网络与变分模态分解(Variational Mode Decomposition,VMD)联合的信号重构方法。首先利用VMD强大的分解能力将GNSS信号进行频域剖分并将其分为多项子信号和噪声项,再基于BiLSTM强大的学习能力对GNSS信号进行训练建模。结果表明,BiLSTM+VMD模型能充分挖掘信号的时频域特征,提高信号重构的精度和稳定性,GNSS N、E、U三分量重构结果均方根误差(Root Mean Squared Error,RMSE)都表现出不同程度的降低,尤其水平方向效果更为显著,相比EMD与VMD方法,E方向离散度分别降低了61%和19%,N方向离散度分别降低了20%和14%。这为GNSS观测时间序列中信号提取与模型参数估计提供了一个有价值的模型。