提出了一种基于匙孔信道的旋转四元素准正交空时分组码(QQOSTBC-CR,Constellation Rotation QuaternionQuasi-Orthogonal Space Time Block Code),该码可以通过极化天线进行发射和接收,设计出发射天线数为8的QOSTPBC-CR,并对N=8的情况进行成对译码,最后与匙孔信道下传统旋转准正交空时分组码(QOSTBC-CR,Constellation Rotation Quasi-OrthogonalSpace Time Block Code)、准正交空时分组码(QOSTBC,Quasi-Orthogonal Space Time Block Code)以及瑞利信道下QQOSTBC-CR进行了仿真比较。结果表明:对4个发射天线的情况,当BER=10^(-3)时,BPSK和QPSK调制下本文所提QQOSTBC-CR分别比QOSTBC-CR有4.5dB和7dB的增益,分别比瑞利信道下QQOSTBC-CR有-4dB和-3dB的增益。
Due to the high complexity of the pairwise decoding algorithm and the poor performance of zero forcing（ ZF） /minimum mean square error（ MMSE） decoding algorithm, two low-complexity suboptimal decoding algorithms, called pairwisequasi-ZF and pairwise-quasi-MMSE decoders, are proposed. First,two transmit signals are detected by the quasi-ZF or the quasiMMSE algorithm at the receiver. Then, the two detected signals as the decoding results are substituted into the two pairwise decoding algorithm expressions to detect the other two transmit signals. The bit error rate（ BER） performance of the proposed algorithms is compared with that of the current known decoding algorithms.Also, the number of calculations of ZF, MMSE, quasi-ZF and quasi-MMSE algorithms is compared with each other. Simulation results showthat the BER performance of the proposed algorithms is substantially improved in comparison to the quasi-ZF and quasiMMSE algorithms. The BER performance of the pairwise-quasiZF（ pairwise-quasi-MMSE） decoder is equivalent to the pairwiseZF（ pairwise-MMSE） decoder, while the computational complexity is significantly reduced.
在用户密集分布的蜂窝网络中,功率分配是决定系统性能和通信质量的重要因素之一。由于现有的功率分配算法往往达不到理想效果,而且泛化能力较差。在此基础上,提出一种基于D3QN(dueling double deep Q network)的功率分配算法来优化系统的传输速率。D3QN采用双神经网络和竞争网络优化神经网络的结构,通过解耦动作的选择和价值的评估,解决了DQN中出现的高估问题。仿真结果表明,该算法能够获得的平均速率比DQN高7.14%,在收敛速度和稳定性方面也有较好的表现,且泛化能力较强,可适用于不同实际场景。
