Exploring Indoor Localization with Transformer-Based Models : A CNN-Transformer Hybrid Approach for WiFi Fingerprinting

Indoor localization has become a target for many researchers due to its vast range of applications. Due to signal fading and scattering, conventional GPS-based techniques are impractical for indoor localization. However, state-of-the-art deep learning models have shown promising results in this field. The method for indoor localization presented in this research makes use of a transformer-based model and Received Signal Strength (RSS) measurements. The proposed model will be assessed in both regression tasks: predicting X and Y coordinates, and classification tasks: floor classification. The results of this research aim to contribute to the advancement of indoor localization systems by providing evidence that transformer-based models might be a good direction to follow for enhancing localization accuracy.