Towards Grid-Free Positioning for Near-Field Communications

This paper proposes a grid-free positioning algorithm for near-field (NF) communications based on machine learning. Due to the NF effects of large-aperture array antenna, beam training measurements are affected by the angle and the distance of the user, which makes it possible to be used for user positioning. Therefore, unlike conventional beam management, the proposed algorithm is designed to learn the relationship between beam training measurements and the user position, enabling not only the establishment of communication links but also the additional estimation of the user's position. Throughout simulations, we compare the proposed algorithm with the traditional beam training algorithms. The simulation results confirm that the proposed algorithm achieves higher user positioning accuracy compared to beam training.