Beam Training Algorithm for Robust LoS-MIMO Systems With Opportunistic Utilization of Reflective Paths

This paper proposes beam training algorithms for robust line-of-sight (LoS) multi-input and multi-output (MIMO) systems with subarray-based beamforming functionality at both the transmitter (TX) and receiver (RX). In general, a conventional beam training algorithm is executed to maximize the received signal power of each pair of the TX and RX subarray. As a result, each pair of subarrays directs its beam towards each other. In LoS-MIMO regime, however, the selected beams are not always adequate to obtain better channel capacity due to its spatial correlation dependency on a distance among TX and RX. To address the issue, we suggest utilizing non-line-of-sight (NLoS) paths opportunistically based on the environmental situation, using estimated angles of arrival (AoA) and departure (AoD) derived from beam measurement data, to ensure stable higher-layer MIMO transmission and propose an efficient beam training algorithm to operate that concept in real situation without knowledge about the distance. Moreover, we propose three other algorithms reducing the overhead of the beam training while maintaining the channel capacity performance. The first low overhead beam training algorithm reduces the number of exhaustive search by utilizing the condition number of the channel matrix. In the second one, the overhead is reduced by approximating the AoAs and AoDs of paths for all subarrays. The third one reduces the overhead by utilizing the condition number and avoiding the exhaustive search. The robustness of the proposed algorithms against changes in distance among TX and RX, horizontal misalignment of subarrays and antenna height change is shown by the computer simulations. Finally, four beam training algorithms are compared in terms of the channel capacity and computational complexity and the best beam training algorithm is clarified for each scenario.