Near/Far-Field State Detection for 6G Terahertz Communications Systems

One of the principal specifics of future 6G systems operating in the subterahertz and terahertz frequency bands (0.1–0.3 THz and 0.3–3 THz) is that a part of the base station (BS) coverage will be located in the near-field region. As in this region, the signal received power (SRP) heavily depends not only on the separation distance between user equipment (UE) and BS but on concrete coordinates of UE, more comprehensive wavefront engineering methods than beamforming need to be used. To enable seamless switching between the antenna’s operational regimes, reliable algorithms determining the location of UE with respect to the BS are needed. In this paper, we utilize the tools of machine learning and propose a simple yet reliable state detection algorithm and test it using two scenarios: (i) UE moves along guiding rail, and (ii) UE moves freely in hand of a user. Our results show that a long short-term memory (LSTM) network provides the best performance having detection accuracy higher than 90% even in the case of UE in hand scenario and reaching unity for UE on rail. Simple tree-based algorithms can detect the change between near- and far-fields with accuracy reaching 80-85% and 95-97% for these scenarios. Thus, we recommend utilizing LSTM algorithm with detection window size of 100– 300 allowing for fast (200–400 ms) state detection and facilitates online implementation.