Transformer-Based Resource-Efficient Indoor Localization for 5G/6G NLOS-Rich Environments

Ongoing research in 6G wireless technology envisions transformative advancements, such as achieving centimeter-level positioning accuracy in real time, to enable highly precise and context-aware applications. This paper addresses the critical challenge of indoor localization in highly non-line-of-sight (NLOS) environments, where achieving accurate positioning remains a significant challenge. Although deep learning (DL) holds promise for 6G systems, its high computational complexity—measured in floating-point operations (FLOPs)—limits its suitability for resource-constrained and real-time applications. Transformer models have achieved remarkable success in various domains. However, their use in indoor localization remains nascent, and directly applying Transformers for indoor localization can be both computationally intensive and exhibit limitations in accuracy. To address these challenges, we introduce a novel tokenization approach, referred to as Sensor Snapshot Tokenization (SST), which preserves variable-specific representations of power delay profile (PDP) and enhances attention mechanisms by effectively capturing multi-variate correlation. The SST method reduces the Transformer’s computational burden and reliance on large datasets, enhancing its applicability to resource-constrained scenarios. This approach achieves a 90th percentile positioning error of 0.489 m in a highly NLOS indoor factory, surpassing conventional tokenization methods. Furthermore, it achieves a 19.72% performance improvement over a model 14.1 times larger, demonstrating superior computational efficiency.