Transformer-Based Environment-Aware Localization in the NLoS Scenarios

In the era of 6G communication, the demand for accurate localization is ever-increasing to support a wide range of applications and devices. Due to the densely distributed obstacles in urban areas, it is difficult to locate the target accurately without knowing the reflection point in non-line-of-sight (NLoS) propagation. In such cases, the layout of the urban area can provide additional information in localization. In this work, we introduce a Transformer-based localization technique, referred as Map Embedded Localization Transformer (MELT), using the environment-awareness. Specifically, MELT learns the geometric correlations between wireless geometry presented in layout image and channel parameters by employing attention mechanism of Transformer. By utilizing the correlations, MELT estimates the target location accurately and robustly. Our simulation results demonstrate the effectiveness of the proposed scheme for localization in NLoS environments in terms of root mean square error (RMSE) and the coverage.