Coordinated Beam Selection for Training Overhead Reduction in FDD Massive MIMO

While acquiring channel state information at the transmitter (CSIT) in time division duplexing systems can exploit channel reciprocity, acquiring accurate CSIT in frequency division duplexing massive multiple-input multiple-output systems is not trivial. The two main difficulties in these systems are the scalability of the downlink reference signals and the overhead associated with the required uplink feedback. Although several approaches for ensuring scalability and reducing overhead by leveraging some presumed channel properties have been studied, existing schemes do not offer a fully satisfactory solution. In this work, we propose a novel cooperative method which exploits low-rate beam-related information exchange between the mobile terminals, reduces the overhead under the assumption of the so-called grid-of-beams design, and strikes a balance between CSIT acquisition overhead, user spatial separability and coordination complexity.