Indoor localization using transformer ensemble regression and LED signals

Tóm tắt

Improving the localization accuracy under the influence of multipath reflections and signal interference in indoor localization systems using visible light signals remains a complex and challenging task. In this study, we propose utilizing the transformer architecture in a localization system based on light intensity signals of 16 LEDs. By leveraging the self-attention mechanism, the model can detect and focus on the locations with the highest relevance. The predictions are aggregated thanks to an ensemble strategy. Simulation results show that the proposed method achieves a Root Mean Square Error of approximately 0.334 m for the entire room (5x5 m), 0.14 m for the central region (3×3 m), and 0.39 m for the remaining areas near the walls and corners.