Bi-Directional Beamformer Training for Dynamic TDD Networks

Abstract

In dynamic time-division-duplexing networks, the available resources per cell can be freely allocated to either uplink (UL) or downlink (DL) depending on the instantaneous traffic demand. Hence, complicated UL–DL and DL–UL interference scenarios arise due to simultaneous UL and DL data transmission in adjacent cells. In this paper, decentralized iterative beamformer designs are obtained for several traffic aware network optimization objectives such that only minimal information exchange is required among the coordinated base stations (BS) and user equipment (UE). Bi-directional forward–backward training via spatially precoded over-the-air pilot signaling is used to facilitate coordinated beamforming. This allows BSs and UEs to iteratively optimize their respective transmitters/receivers based on only locally measured reverse link pilot measurements. Novel bi-directional beamformer training strategies and methods for direct estimation (DE) of the stream specific beamformers are developed for each intermediate beamformer update in a limited and noisy pilot environment. The proposed signaling and DE schemes allow for non-orthogonal and overlapping pilots, which considerably reduces the resource coordination effort. Also, the decontamination ability of the proposed strategies are analyzed with limited pilot resources. The numerical examples illustrate the superior system performance of the proposed training and estimation framework in comparison to both the traditional stream-specific channel estimation method and an uncoordinated system.