Scholarlink

<p>In complex urban environments, where GNSS positioning is severely degraded by multipath interference and non-line-of-sight reception, data-driven methods offer a promising solution by effectively modeling complex non-Gaussian errors from sufficient data for positioning correction. However, existing methods primarily focus on either spatial or temporal features of GNSS observation in isolation, failing to fully exploit the complex multidimensional spatial–temporal correlations inherent in GNSS observation. Moreover, dynamic changes in real-world environments induce data distribution shift between training and testing, requiring generalization capability for the data-driven model in unseen scenarios. To address these limitations, we propose DRL-RSTR, a deep reinforcement learning model with robust spatial–temporal representation for GNSS positioning correction. Specifically, to jointly capture the spatial–temporal representation, we construct a multi-observation GCN-transformer (MOGT), where spatial geometric relationships among constellations are modeled by a graph convolutional network and temporal dependencies are captured by transformer. Furthermore, the spatial–temporal representation is fused through summation, and a cross-attention network is employed to model the interactions among multi-observations to obtain a comprehensive environmental representation. Additionally, to enhance generalization capability against data distribution shifts, a self-supervised pretext task (SST) is introduced to improve the robustness of spatial–temporal representation through consistency regularization across non-augmented and augmented observations. We conduct extensive experiments on the public GSDC and built GZGNSS datasets, results show that DRL-RSTR achieves superior positioning accuracy and generalization compared to the model-based and learning-based state-of-the-art methods, with improvements of 51.2% and 41.4% on the GZGNSS dataset and 6.5% compared with kalman filters on the GSDC dataset in terms of positioning accuracy.</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

Deep reinforcement learning with robust spatial–temporal representation for improving GNSS positioning correction

  • 期刊论文

  • 发表日期：

    2026年6月16日

查看全文
  • Zhenni Li,
  • Peili Li,
  • Jianhao Tang,
  • Yulong Song,
  • Liji Chen,
  • Yiting Cai,
  • Shengli Xie

摘要

In complex urban environments, where GNSS positioning is severely degraded by multipath interference and non-line-of-sight reception, data-driven methods offer a promising solution by effectively modeling complex non-Gaussian errors from sufficient data for positioning correction. However, existing methods primarily focus on either spatial or temporal features of GNSS observation in isolation, failing to fully exploit the complex multidimensional spatial–temporal correlations inherent in GNSS observation. Moreover, dynamic changes in real-world environments induce data distribution shift between training and testing, requiring generalization capability for the data-driven model in unseen scenarios. To address these limitations, we propose DRL-RSTR, a deep reinforcement learning model with robust spatial–temporal representation for GNSS positioning correction. Specifically, to jointly capture the spatial–temporal representation, we construct a multi-observation GCN-transformer (MOGT), where spatial geometric relationships among constellations are modeled by a graph convolutional network and temporal dependencies are captured by transformer. Furthermore, the spatial–temporal representation is fused through summation, and a cross-attention network is employed to model the interactions among multi-observations to obtain a comprehensive environmental representation. Additionally, to enhance generalization capability against data distribution shifts, a self-supervised pretext task (SST) is introduced to improve the robustness of spatial–temporal representation through consistency regularization across non-augmented and augmented observations. We conduct extensive experiments on the public GSDC and built GZGNSS datasets, results show that DRL-RSTR achieves superior positioning accuracy and generalization compared to the model-based and learning-based state-of-the-art methods, with improvements of 51.2% and 41.4% on the GZGNSS dataset and 6.5% compared with kalman filters on the GSDC dataset in terms of positioning accuracy.