Bi-Directional Training for Wideband Systems

Abstract

To further improve the spectrum efficiency of 5G massive MIMO networks, bi-directional training (BiT) was developed for TDD systems to maximize the downlink weighted sum rate. However, the previous work was limited to narrowband systems. In this paper, we extend BiT for 5G wideband systems. A global, centralized optimization problem is first formulated for a wideband system. The (sub-optimal) solution is then distributed across the base stations and user equipment (UE), resulting into a wideband BiT algorithm that iteratively adapts transmission and reception filters for each base station and each UE with only local information. The algorithm may be seen as a narrowband BiT operating on an optimal narrowband representation of a group of subcarriers each with a different channel, and the optimal narrowband representation maintains the first and second moments of all the channels. Simulation results are provided to evaluate the performance of the algorithm in a wideband system.