Indoor Localization with GravNetConv and Dynamic Graphs

Traditional indoor localization methods often rely on Received Signal Strength Indicator (RSSI) measurements from Wi-Fi access points (APs) and utilize techniques like fingerprinting. However, these methods face challenges due to signal variability and environmental factors, leading to limited accuracy. To address these issues, we propose a novel approach leveraging Graph Neural Networks (GNNs) with a dynamic graph generation process using GravNetConv. This method processes input data in a learned lower-dimensional space, using a Gaussian function to emphasize proximate points, thereby capturing complex spatial relationships and improving localization accuracy through more effective local neighbor capture and fine-grained detail recognition. Our experiments show that the GravNetConv-based method outperforms conventional indoor localization approaches, particularly on larger, more diverse datasets, leading to more reliable and precise positioning in dynamic environments.