Deep Adaptive Network for WiFi-Based Indoor Localization

@InProceedings{10.1007/978-3-031-43699-4_38,
  author="Ahmad, Afnan
  and Sohn, Gunho",
  editor="Kolbe, Thomas H.
  and Donaubauer, Andreas
  and Beil, Christof",
  title="Deep Adaptive Network for WiFi-Based Indoor Localization",
  booktitle="Recent Advances in 3D Geoinformation Science",
  year="2024",
  publisher="Springer Nature Switzerland",
  address="Cham",
  pages="617--631",
  abstract="A growing trend for indoor localization is relying on the existing WiFi signal strength. However, this method faces challenges due to WiFi's received signal strength (RSS) being susceptible to multipath, signal attenuation, and environmental variations, making it an unreliable measure of signal strength. To address this issue, a study was conducted which combined WiFi signals from various locations to create a localization system with high accuracy within a few meters. The study utilized WiFi propagation characteristics as a type of location fingerprinting, providing a new method for indoor localization using Wi-Fi RSSI fingerprinting. To adapt to new surroundings, the system employed a Variational Autoencoder to distribute WiFi signal properties, an LSTM network to analyze temporal relations of Wi-Fi signals, and a feature backpropagating refinement module to adjust neural network weights during inference. These tools were instrumental in achieving the system's main objective of domain adaptability. The accuracy of localization improved by approximately 18{\%} compared to the baseline neural network.",
  isbn="978-3-031-43699-4"
  }