EnvAwareLoc: Precision Localization Network Based on the Introduction of Environmental Information

Deep learning methods for Bluetooth-based fingerprints have demonstrated promising results in recent years. Deep learning models can capture the features of signals and map them to the corresponding locations. Most deep learning-based fingerprint methods for localization performance improvement mainly focus on data quality and quantity, which is undeniably a more critical point. Nevertheless, the primary focus of this paper lies in introducing environmental information to investigate its impact on improving fingerprint localization performance. We introduce EnvAwareLoc, a deep-learning network composed of two key components. The primary segment focuses on learning the mapping relationship between input signals and locations, while the secondary segment extracts insights from the input environmental data. The final location estimation is derived by amalgamating the features acquired from both segments. The experimental results show that the average localization error of our proposed method is reduced by about 30%, and the localization accuracy is greatly improved compared to that of the network without introducing environmental information. Additionally, by utilizing environmental information as input for our proposed method, we mitigate its sensitivity to environmental variations, consequently enhancing its robustness.