Indoor navigation remains a significant challenge due to the limitations of Global Navigation Satellite Systems (GNSS) in enclosed environments. This paper presents a novel approach to indoor positioning by leveraging WiFi fingerprinting enhanced with a Transformer-based model. The proposed model processes sequential WiFi signal measurements to capture temporal dependencies and improve positioning accuracy. A key innovation is the replacement of traditional layer normalization with Dirichlet distribution-based normalization, which enhances the model’s generalization and robustness. Experimental results demonstrate that our approach significantly reduces the average positioning error compared to existing methods, showcasing its potential for practical deployment in complex indoor environments.