UE-GLoc: Uncertainty-Aware Graph Neural Networks for Robust and Accurate WiFi Indoor Localization

WiFi fingerprinting is a cost-effective solution for indoor localization, but traditional methods struggle with environmental variability and limited adaptability. Recent deep learning approaches, especially Graph Neural Networks (GNNs), improve accuracy by modeling spatial relationships, yet often lack robustness to uncertainty and long-range dependencies. This paper presents UE-GLoc, a GNN-based localization framework that integrates an environmental uncertainty estimation (EUE) and a graph outer attention (GOA) mechanism. By capturing both local and global dependencies, UE-GLoc blue aims to enhance feature representation and blue improve model reliability under varying conditions. Experiments on public datasets show that UE-GLoc achieves higher accuracy and generally better performance compared to several state-of-the-art methods, with potential adaptability to new environments.