用于下行链路协调基站系统的鲁棒收发机Optimization:分布式算法
Robust Transceiver Optimization for Downlink Coordinated Base Station Systems: Distributed Algorithm
摘要:
本文考虑具有协调基站(BS)的下行链路多用户多输入单输出(MISO)系统的联合收发器设计,其中BS和移动台(MS)处提供不完美的信道状态信息(CSI)。通过在BS处结合天线相关性并考虑信道估计误差,我们解决了两个鲁棒设计问题:1)在每个BS天线功率约束下最小化均方误差(MSE)的加权和,以及2)在每个用户MSE目标和每个BS天线功率约束下最小化所有BS的总功率。这些问题的解决方法如下。首先,对于固定的接收器,我们提出了集中式和新的计算效率的分布式算法,共同优化所有用户的预编码器。我们的集中式算法采用二阶锥规划(SOCP)的方法,而我们的新的分布式算法使用拉格朗日对偶分解,修改的矩阵分数最小化和迭代法。第二,对于固定BS预编码器,接收器通过最小均方误差(MMSE)准则更新。重复这两个步骤,直到实现收敛。在我们所有的模拟结果中,我们已经观察到,所提出的分布式算法实现了相同的性能的集中式算法。此外,计算机模拟验证了所提出的鲁棒设计的鲁棒性相比,非鲁棒/天真的设计。
英文摘要:
This paper considers the joint transceiver design for downlink multiuser multiple-input single-output (MISO) systems with coordinated base stations (BSs) where imperfect channel state information (CSI) is available at the BSs and mobile stations (MSs). By incorporating antenna correlation at the BSs and taking channel estimation errors into account, we solve two robust design problems: 1) minimizing the weighted sum of meansquare- error (MSE) with per BS antenna power constraint, and 2) minimizing the total power of all BSs with per user MSE target and per BS antenna power constraints. These problems are solved as follows. First, for fixed receivers, we propose centralized and novel computationally efficient distributed algorithms to jointly optimize the precoders of all users. Our centralized algorithms employ the second-order-cone programming (SOCP) approach, whereas, our novel distributed algorithms use the Lagrangian dual decomposition, modified matrix fractional minimization and an iterative method. Second, for fixed BS precoders, the receivers are updated by the minimum mean-square-error (MMSE) criterion. These two steps are repeated until convergence is achieved. In all of our simulation results, we have observed that the proposed distributed algorithms achieve the same performance as that of the centralized algorithms. Moreover, computer simulations verify the robustness of the proposed robust designs compared to the non-robust/naive designs.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。