Scholarlink

In response to the escalating demand for Indoor Positioning Services (IPS) everywhere, this study explores techniques to enhance accuracy and adaptability in a dynamic environment. Leveraging the WiFi fingerprinting and the Channel State Information (CSI) phase of a dataset collected inside the University of Passau in Germany, this research introduces a groundbreaking approach by integrating Dueling Q-Network (Dueling QN) within Deep Reinforcement Learning (DRL). The trained agent achieves precise localization within a remarkable 0 cm distance error, even without prior knowledge of the floor plan. Experimental results compare the position predicted using CSI phase and CSI amplitude, providing insights into their respective contributions. The promising findings emphasize the potential of DRL as a robust solution, opening new avenues for exploration and application in 3D WiFi datasets.

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

CSI Phase Fingerprinting for Indoor Positioning Services Using Deep Reinforcement Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Wiem Fekih Hassen,
  • Haifa Ben Salem

摘要

In response to the escalating demand for Indoor Positioning Services (IPS) everywhere, this study explores techniques to enhance accuracy and adaptability in a dynamic environment. Leveraging the WiFi fingerprinting and the Channel State Information (CSI) phase of a dataset collected inside the University of Passau in Germany, this research introduces a groundbreaking approach by integrating Dueling Q-Network (Dueling QN) within Deep Reinforcement Learning (DRL). The trained agent achieves precise localization within a remarkable 0 cm distance error, even without prior knowledge of the floor plan. Experimental results compare the position predicted using CSI phase and CSI amplitude, providing insights into their respective contributions. The promising findings emphasize the potential of DRL as a robust solution, opening new avenues for exploration and application in 3D WiFi datasets.