Linear Interference Alignment in Full-Duplex MIMO Networks With Imperfect CSI
Linear Interference Alignment in Full-Duplex MIMO Networks With Imperfect CSI
复制标题
DOI：
10.1109/tcomm.2017.2744647
复制
发表时间：
2017-12-01
影响因子：
8.3
通讯作者：
Ratnarajah, Tharmalingam
中科院分区：
文献类型：
--
作者：
Aquilina, Paula;Ratnarajah, Tharmalingam
在本文中，我们考虑了一个系统，其中全双工（FD）基站（BS）与半双工（HD）的下行链路（DL）和上行链路（UL）的用户在多用户多小区网络，其中所有节点都配备了多个天线进行通信。FD BS的引入提供了提高频谱效率的潜力，然而，与HD网络对应物相比，它也导致干扰链路的数量激增。在这里，我们应用线性干扰对齐（IA）在这个网络中的不完美的信道状态信息（CSI）下管理干扰。首先，我们描述的性能损失相对于可实现的总和速率和自由度（DoF）。结果表明，性能损失的一般趋势主要取决于误差如何与信噪比（SNR）成比例。特别地，当信道误差至少与SNR成反比时，即使在不完美的CSI下也可以实现完全的UL和DL DoF。此外，在这种情况下，总速率损失总是有限的，并且要么趋于零，要么由导出值限定上限。其次，我们设计了两个线性IA算法适用于所考虑的系统。这些都是基于最小化的均方误差（MMSE）和最大化的信号与干扰加噪声比，并考虑增加的鲁棒性的CSI误差的统计知识。所提出的算法遵循特定的设计原则，在各种波束形成器之间分配不同的干扰分量，并导致酉接收器和预编码器。此外，我们表明，在一定条件下，这两种设计结果相同的波束形成的解决方案，即使MMSE算法具有较低的计算复杂度。最后，我们还得到了适当的条件IA的可行性，在考虑中的多细胞系统。
In this paper, we consider a system where full-duplex (FD) base-stations (BSs) communicate with half-duplex (HD) downlink (DL) and uplink (UL) users in a multiuser multi-cell network, where all nodes are equipped with multiple antennas. The introduction of FD BSs offers potential to increase spectral efficiency, however, it also causes a surge in the number of interference links compared with the HD network counterpart. Here, we apply linear interference alignment (IA) to manage interference in this network under imperfect channel state information (CSI). First, we characterize the performance losses incurred with respect to the achievable sum rate and degrees of freedom (DoFs). Results show that the general trend in performance loss is mainly determined by how the error scales with the signal-to-noise ratio (SNR). In particular, full UL and DL DoF can be achieved even under imperfect CSI when the channel error is at least inversely proportional to SNR. Moreover, in such cases the sum rate loss is always finite, and either goes to zero or is upper bounded by a derived value. Second, we design two linear IA algorithms applicable to the system under consideration. These are based on minimizing the mean square error (MMSE) and maximizing the signal-to-interference-plus-noise ratio, and consider statistical knowledge of the CSI error for added robustness. The proposed algorithms follow specific design principles that distribute the different interference components amongst the various beamformers and result in unitary receivers and precoders. In addition, we show that under certain conditions both designs result in identical beamforming solutions, even though the MMSE algorithm has lower computational complexity. Finally, we also derive the proper condition for IA feasibility in the multi-cell system under consideration.