In recent years, extensive research efforts have been devoted to developing high-accuracy and reliable indoor localization techniques. Among the various approaches, Wi-Fi fingerprinting has gained widespread adoption due to the ubiquity and availability of Wi-Fi hardware. In this paper, we propose an indoor positioning system that operates in two stages: an offline training phase and an online position estimation phase. In the offline phase, Received Signal Strength (RSS) data are collected from multiple access points (APs) across different locations within the building. These measurements are then processed by a deep learning network based on an autoencoder equipped with non-smooth regularizers. The autoencoder extracts low-dimensional representations of the RSS data and builds a fingerprint database that stores the mapping between learned representations and their corresponding physical locations. During the online estimation phase, RSS values are measured at an unknown location and passed through the trained autoencoder to obtain their low-dimensional representation. The system then searches for the nearest match within the fingerprint database, and the associated location is returned as the estimated position. To enhance localization accuracy, different types of regularizers are incorporated into the autoencoder structure. Simulation results demonstrate that the proposed deep learning-based indoor localization system with non-smooth regularizers significantly outperforms baseline methods without regularization or with smooth regularizers. For example, when the regularization parameter is set to 0.0003, the Root Mean Square Error (RMSE) decreases from 44.82 m (without regularization) to 24.03 m with smooth regularizers, and further to 7.09 m with non-smooth regularizers, validating the superiority of the proposed framework.
Karimi, A., & R.Danaee, M. (2025). Enhancing Single-User Localization in WiFi Networks Using Deep Learning-Based Autoencoder Architecture with Non-Smooth Regularizers. , 3(4), -.
MLA
ali Karimi; Meysam R.Danaee. "Enhancing Single-User Localization in WiFi Networks Using Deep Learning-Based Autoencoder Architecture with Non-Smooth Regularizers", , 3, 4, 2025, -.
HARVARD
Karimi, A., R.Danaee, M. (2025). 'Enhancing Single-User Localization in WiFi Networks Using Deep Learning-Based Autoencoder Architecture with Non-Smooth Regularizers', , 3(4), pp. -.
VANCOUVER
Karimi, A., R.Danaee, M. Enhancing Single-User Localization in WiFi Networks Using Deep Learning-Based Autoencoder Architecture with Non-Smooth Regularizers. , 2025; 3(4): -.
)