西南大学学报 (自然科学版)  2020, Vol. 42 Issue (2): 129-136.  DOI: 10.13718/j.cnki.xdzk.2020.02.016
0
Article Options
  • PDF
  • Abstract
  • Figures
  • References
  • 扩展功能
    Email Alert
    RSS
    本文作者相关文章
    刘康
    刘璐
    杨崇海
    汤玮
    冯文江
    刘欢
    欢迎关注西南大学期刊社
     

  • 全双工双通路中继网络干扰处理与信号检测算法    [PDF全文]
    刘康1, 刘璐1, 杨崇海2, 汤玮1, 冯文江2, 刘欢1    
    1. 贵州电网有限责任公司 电力调度控制中心, 贵阳 550002;
    2. 重庆大学 微电子与通信工程学院, 重庆 400044
    摘要:全双工双通路连续中继(FD-TPSR)网络频谱效率高,状态控制简单,但全双工中继存在残留自干扰(RSI),发射中继对接收中继也会形成中继间干扰(IRI),会导致通信性能下降.这里提出一种中继间干扰处理方法,中继节点消除部分IRI,以提高端到端信干噪比;保留部分IRI,在目的节点构成延迟转发编码结构,以提供分集增益.同时提出一种基于并行软干扰消除的匹配滤波(MF-PSIC)信号检测算法,匹配滤波器结构简单,实现复杂度低,基于软输出的并行干扰消除能同时检测所有时隙的符号,处理时延小.仿真结果表明,中继间干扰处理方法兼顾了分集增益和累积干扰与噪声因素,相比于无IRI消除和完全IRI消除,误比特率最低;MF-PSIC信号检测算法实现复杂度低,相比于ML信号检测算法仅有很少的性能损失.
    关键词中继网络    双通路连续中继网络    全双工    中继间干扰    信号检测    

    双通路连续中继(two-path successive relaying,TPSR)网络[1]以钻石中继(diamond relaying)结构为基础,两个中继节点轮流接收和转发源节点发射符号,目的节点在每个时隙能接收并解码来自源节点的发射符号,频谱效率高[2].如果中继节点采用全双工(full-duplex,FD)通信还能简化状态控制[3].但在全双工TPSR(FD-TPSR)网络中,一方面,全双工中继存在残留自干扰(residual self-interference,RSI),会导致中继节点和目的节点的信干噪比下降.另一方面,发射中继会对接收中继形成中继间干扰(inter-relay interference,IRI).在基于放大转发(amplify-and-forward,AF)的FD-TPSR网络中,虽然目的节点可以根据历史信息消除IRI,但会降低信干噪比[4-5];而在采用译码转发(decode-and-forward,DF)的FD-TPSR网络中,中继节点利用连续干扰消除(successive interference cancellation,SIC)首先解码IRI并从接收信号中剔除,然后转发无IRI的信号至目的节点[6],但如果中继之间的信道质量差,IRI解码差错会向目的节点传播[7].文献[8]针对解调转发(demodulate-and-forward,DmF)TPSR网络提出一种混合策略,依据中继间信道质量,在差分解调和直接解调间切换.文献[9]将多天线和波束赋形应用于TPSR网络,设计了一种基于信噪比的中继选择协议,分析了分布式空时编码传输方案的性能.文献[10]针对多天线双向中继网络提出一种中继方案和用户组选择方案,同时优化中继预编码器和接收检测器,以降低中继间干扰和接收端自干扰.信号检测主要关注可靠性和实时性,基于极大似然(maximum likelihood,ML)准则的信号检测是最优检测[11],但处理复杂度高.文献[12]采用最小均方误差-判决反馈均衡器执行信号检测,能获得满分集增益,但若排序出错,判决反馈均衡器在执行SIC时会造成差错传播,导致性能恶化.此外,采用逐符号判决反馈,目的节点必须完成所有源节点信号接收后才能处理,难以满足端到端时延要求.

    本文针对全双工双通路连续中继网络,提出一种IRI处理方法,中继节点消除部分IRI,以提高端到端信干噪比;保留部分IRI,在目的节点构成延迟转发编码结构,以提供时间分集.提出一种基于并行软干扰消除的匹配滤波(MF-PSIC)算法,匹配滤波器结构简单,实现复杂度低,基于软输出的并行干扰消除能并行检测所有时隙的符号,处理时延小,且具有与ML检测算法相近的性能.

    1 系统模型

    全双工双通路连续中继网络(TD-TPSR)由一个半双工源节点S,两个全双工中继节点R1R2和一个半双工目的节点D构成,如图 1所示.记源节点S到中继节点R1R2的信道系数分别为f1f2,中继节点R1R2到目的节点D的信道系数分别为g1g2,中继节点R1R2之间的信道系数为h.假设信道为准静态瑞利衰落信道,所有信道系数服从均值为0和方差为1的实高斯分布. SD之间无直达链路,只能通过R1R2中继通信.

    图 1 FD-TPSR网络结构图
    2 中继间干扰处理方法 2.1 信道状态信息获取

    FD-TPSR网络采用基于导频的信道估计获得信道状态信息,其导频集合为{p1p2p3},获取步骤如下:

    在信道估计第1时隙,R1R2分别发送导频p1p2S发送导频p3. D基于导频p1p2估计获得R1R2D的信道系数g1g2R1R2基于导频p3估计获得SR1R2的信道系数f1f2R1R2基于导频p1p2估计获得R1R2之间的信道系数h.

    在信道估计第2时隙,R1R2分别利用同相和正交载波将f1f2发送至D,此时,R1R2分别获得了部分CSI(f1hf2h),而D获得了全局CSI(f1f2g1g2h).

    2.2 编码调制符号传输

    源节点S将发送信息比特编码调制后生成长度为T的发送符号s=[s(1),…,s(T)].在信号传输第1时隙,SR1R2广播发送符号s(1),R1R2的接收信号分别为

    $ \begin{array}{*{20}{c}} {{y_{R1}}\left( 1 \right) = {f_1}s\left( 1 \right) + {n_1}\left( 1 \right)}\\ {{y_{R2}}\left( 1 \right) = {f_2}s\left( 1 \right) + {n_2}\left( 1 \right)} \end{array} $ (1)

    其中,n1(1)和n2(1)分别为R1R2处的加性高斯白噪声,ni(t)~CN(0,1),i=1,2;t=1,…,T.在信号传输第2时隙,R1R2分别对信号传输第1时隙的接收信号放大转发,其转发信号分别为

    $ \begin{array}{*{20}{c}} {{x_1}\left( 2 \right) = b{y_{{{R}}1}}\left( 1 \right)}\\ {{x_2}\left( 2 \right) = b{y_{{{R}}2}}\left( 1 \right)} \end{array} $ (2)

    其中,$b = \sqrt {\left( {\sqrt {{A^2} + 6A + 1} - A - 1} \right)/2} $R1R2的放大系数,A=P/(βPλ+1),0<β<1表示线性自干扰抑制度,0<λ<1表示指数自干扰抑制度,PR1R2的发射功率. D的接收信号为

    $ \begin{array}{l} {y_D}\left( 2 \right) = {g_1}{x_1}\left( 2 \right) + {g_2}{x_2}\left( 2 \right) + {n_D}\left( 2 \right) = \\ \;\;\;\;\;\;\;\;\;\;\;\;b\left( {{g_1}{f_1} + {g_2}{f_2}} \right)s\left( 1 \right) + w\left( 2 \right) \end{array} $ (3)

    其中,w(2)=b[g1n1(1)+g2n2(1)]+nD(2)表示在信号传输第2时隙目的节点D的接收等效干扰和噪声,nD(t)~CN(0,1),t=2,…,T+1.同时,R1(R2)接收S在该时隙的发送符号s(2)以及R2(R1)的转发信号x2(2)(x1(2)). R1R2的接收信号分别为

    $ \begin{array}{*{20}{c}} {{y_{R1}}\left( 2 \right) = {f_1}s\left( 2 \right) + bh\left[ {{f_2}s\left( 1 \right) + {n_2}\left( 1 \right)} \right] + {c_1}\left( 2 \right) + {n_1}\left( 2 \right)}\\ {{y_{R2}}\left( 2 \right) = {f_2}s\left( 2 \right) + bh\left[ {{f_1}s\left( 1 \right) + {n_1}\left( 1 \right)} \right] + {c_2}\left( 2 \right) + {n_2}\left( 2 \right)} \end{array} $ (4)

    其中,c1(2),c2(2)分别为R1R2处的残留自干扰,ci(t)~CN(0,βPλ),i=1,2;t=2,…,T.在信号传输第3时隙,R1R2分别对信号传输第2时隙的接收信号放大转发,其转发信号分别为

    $ \begin{array}{*{20}{c}} {{x_1}\left( 3 \right) = b{y_{R1}}\left( 2 \right)}\\ {{x_2}\left( 3 \right) = b{y_{R2}}\left( 2 \right)} \end{array} $ (5)

    目的节点D的接收信号为

    $ \begin{array}{l} {y_D}\left( 3 \right) = {g_1}{x_1}\left( 3 \right) + {g_2}{x_2}\left( 3 \right) + {n_D}\left( 3 \right) = \\ \;\;\;\;\;\;\;\;\;\;\;b\left( {{g_1}{f_1} + {g_2}{f_2}} \right)s\left( 2 \right) + {b^2}h\left( {{g_1}{f_2} + {g_2}{f_1}} \right)s\left( 1 \right) + w\left( 3 \right) \end{array} $ (6)

    其中,w(3)=$\tilde c $(3)+$\tilde n $(3)+nD(3)表示在信号传输第3时隙目的节点D的接收等效干扰和噪声,

    $ \tilde c\left( 3 \right) = b\left[ {{g_1}\left( {{c_1}\left( 2 \right) + {g_2}{c_2}\left( 2 \right)} \right]} \right. $
    $ \tilde n\left( 3 \right) = {b^2}h\left[ {{g_1}{n_2}\left( 1 \right) + {g_2}{n_1}\left( 1 \right)} \right] + b\left[ {{g_1}{n_1}\left( 2 \right) + {g_2}{n_2}\left( 2 \right)} \right] $

    同时,R1(R2)接收S在第3时隙的发送符号s(3)以及R2(R1)的转发信号x2(3)(x1(3)),R1R2的接收信号分别为

    $ \begin{array}{*{20}{c}} {{y_{R1}}\left( 3 \right) = {f_1}s\left( 3 \right) + {a_1}\left( 3 \right) + {c_1}\left( 3 \right) + {n_1}\left( 3 \right)}\\ {{y_{R2}}\left( 3 \right) = {f_2}s\left( 3 \right) + {a_2}\left( 3 \right) + {c_2}\left( 3 \right) + {n_2}\left( 3 \right)} \end{array} $ (7)

    其中,

    $ {a_1}\left( 3 \right) = bh\left[ {{f_2}s\left( 2 \right) + bh{y_{R1}}\left( 1 \right) + {c_2}\left( 2 \right) + {n_2}\left( 2 \right)} \right] $
    $ {a_2}\left( 3 \right) = bh\left[ {{f_1}s\left( 2 \right) + bh{y_{R2}}\left( 1 \right) + {c_1}\left( 2 \right) + {n_1}\left( 2 \right)} \right] $

    R1R2利用历史接收信号消除信号传输第1时隙接收信号对信号传输第3时隙接收信号形成的干扰,保留信号传输第2时隙接收信号形成的干扰,即

    $ \begin{array}{l} {{y'}_{R1}}\left( 3 \right) = {y_{R1}}\left( 3 \right) - {\left( {^bh} \right)^2}{y_{R1}}\left( 1 \right) = \\ \;\;\;\;\;\;\;\;\;\;\;\;{f_1}s\left( 3 \right) + {{a'}_1}\left( 3 \right) + {c_1}\left( 3 \right) + {n_1}\left( 3 \right)\\ {{y'}_{R2}}\left( 3 \right) = {y_{R2}}\left( 3 \right) - {\left( {^bh} \right)^2}{y_{R2}}\left( 1 \right) = \\ \;\;\;\;\;\;\;\;\;\;\;\;{f_2}s\left( 3 \right) + {{a'}_2}\left( 3 \right) + {c_2}\left( 3 \right) + {n_2}\left( 3 \right) \end{array} $ (8)

    其中,

    $ {{a'}_1}\left( 3 \right) = bh\left[ {{f_2}s\left( 2 \right) + {c_2}\left( 2 \right) + {n_2}\left( 2 \right)} \right] $
    $ {{a'}_2}\left( 3 \right) = bh\left[ {{f_1}s\left( 2 \right) + {c_1}\left( 2 \right) + {n_1}\left( 2 \right)} \right] $

    在信号传输第t≥4时隙,R1R2分别对信号传输第t-1时隙消除了部分干扰的接收信号执行放大转发,其转发信号分别为

    $ \begin{array}{l} {x_1}\left( t \right) = b{{y'}_{R1}}\left( {t - 1} \right)\\ {x_2}\left( t \right) = b{{y'}_{R2}}\left( {t - 1} \right) \end{array} $ (9)

    目的节点D的接收信号为

    $ \begin{array}{l} {y_D}\left( t \right) = {g_1}{x_1}\left( t \right) + {g_2}{x_2}\left( t \right) + {n_D}\left( t \right) = \\ \;\;\;\;\;\;\;\;\;\;\;b\left( {{g_1}{f_1} + {g_2}{f_2}} \right)s\left( {t - 1} \right) + {b^2}h\left( {{g_1}{f_2} + {g_2}{f_1}} \right)s\left( {t - 2} \right) + w\left( t \right) \end{array} $ (10)

    其中,w(t)=$\tilde c $(t)+$\tilde n $(t)+nD(t)表示在信号传输第t时隙目的节点D的接收等效干扰和噪声,

    $ \tilde c\left( t \right) = {b^2}h\left[ {{g_1}{c_2}\left( {t - 2} \right) + {g_2}{c_1}\left( {t - 2} \right)} \right] + b\left[ {{g_1}{c_1}\left( t \right) + {g_2}{c_2}\left( t \right)} \right] $
    $ \tilde n\left( t \right) = {b^2}h\left[ {{g_1}{n_2}\left( {t - 2} \right) + {g_2}{n_1}\left( {t - 2} \right)} \right] + b\left[ {{g_1}{n_1}\left( t \right) + {g_2}{n_2}\left( t \right)} \right] $

    同时,R1(R2)接收S在第t时隙的发送符号s(t)以及R2(R1)的转发信号x2(t)(x1(t)),R1R2的接收信号分别为

    $ \begin{array}{*{20}{c}} {{y_{R1}}\left( t \right) = {f_1}s\left( t \right) + {a_1}\left( t \right) + {c_1}\left( t \right) + {n_1}\left( t \right)}\\ {{y_{R2}}\left( t \right) = {f_2}s\left( t \right) + {a_2}\left( t \right) + {c_2}\left( t \right) + {n_2}\left( t \right)} \end{array} $ (11)

    其中,

    $ {a_1}\left( t \right) = bh\left[ {{f_2}s\left( {t - 1} \right) + bh{y_{R1}}\left( {t - 2} \right) + {c_2}\left( {t - 1} \right) + {n_2}\left( {t - 1} \right)} \right] $
    $ {a_2}\left( t \right) = bh\left[ {{f_1}s\left( {t - 1} \right) + bh{y_{R2}}\left( {t - 2} \right) + {c_1}\left( {t - 1} \right) + {n_1}\left( {t - 1} \right)} \right] $

    R1R2利用历史信号消除信号传输第t-2时隙接收信号对信号传输第t时隙接收信号造成的干扰,保留信号传输第t-1时隙接收信号形成的干扰,即:

    $ \begin{array}{l} {{y'}_{R1}}\left( t \right) = {y_{R1}}\left( t \right) - {\left( {^bh} \right)^2}{y_{R1}}\left( {t - 2} \right) = \\ \;\;\;\;\;\;\;\;\;\;\;{f_1}s\left( t \right) + {{a'}_1}\left( t \right) + {c_1}\left( t \right) + {n_1}\left( t \right)\\ {{y'}_{R2}}\left( t \right) = {y_{R2}}\left( t \right) - {\left( {^bh} \right)^2}{y_{R2}}\left( {t - 2} \right) = \\ \;\;\;\;\;\;\;\;\;\;\;{f_2}s\left( t \right) + {{a'}_2}\left( t \right) + {c_2}\left( t \right) + {n_2}\left( t \right) \end{array} $ (12)

    其中,

    $ {{a'}_1}\left( t \right) = bh\left[ {{f_2}s\left( {t - 1} \right) + {c_2}\left( {t - 1} \right) + {n_2}\left( {t - 1} \right)} \right] $
    $ {{a'}_2}\left( t \right) = bh\left[ {{f_1}s\left( {t - 1} \right) + {c_1}\left( {t - 1} \right) + {n_1}\left( {t - 1} \right)} \right] $

    R1R2依次对各信号传输时隙的接收信号处理,消除部分干扰后转发至D,直到t=T+1,编码调制符号传输结束.

    3 信号检测算法

    目的节点D将所有时隙的接收信号构成向量形式:

    $ \mathit{\boldsymbol{y}} = \mathit{\boldsymbol{Hs}} + \mathit{\boldsymbol{w}} $ (13)

    其中,y=[yD(2),…,yD(T+1)]T表示目的节点D的接收信号向量,s=[s(1),…,s(T)]T表示源节点S的发送信号向量,w=[w(2),…,w(T+1)]T表示目的节点D的接收等效干扰和噪声向量,

    $ \mathit{\boldsymbol{H}} = \left[ {\begin{array}{*{20}{c}} {{h_1}}&0&0& \cdots &0\\ {{h_2}}&{{h_1}}&0& \cdots &0\\ 0&{{h_2}}& \ddots & \ddots & \vdots \\ \vdots & \vdots & \vdots &{{h_1}}&0\\ 0&0& \cdots &{{h_2}}&{{h_1}} \end{array}} \right] $

    其中,

    $ \begin{array}{*{20}{c}} {{h_1} = b\left( {{g_1}{f_1} + {g_2}{f_2}} \right)}&{{h_2} = b{h^2}\left( {{g_1}{f_2} + {g_2}{f_1}} \right)} \end{array} $

    为了执行信号检测,提出一种基于并行软干扰消除的匹配滤波(MF-PSIC)信号检测算法,匹配滤波器结构简单,复杂度低,基于软输出的并行干扰消除能同时检测所有时隙的符号[13],处理时延小. MF-PSIC算法结构如图 2所示,其基本思想如下:

    图 2 MF-PSIC结构框图

    计算匹配滤波加权矩阵K

    $ \mathit{\boldsymbol{K}} = {\left( {{\mathit{\boldsymbol{H}}^{\rm{T}}}\mathit{\boldsymbol{H}}} \right)^{ - 1}}{\mathit{\boldsymbol{H}}^{\rm{T}}} $ (14)

    利用加权矩阵K执行源节点发送符号的粗估计$\tilde s $

    $ \tilde s = \mathit{\boldsymbol{Ky}} $ (15)

    t时隙源节点发送符号的粗估计为

    $ \tilde s\left( t \right) = {\mathit{\boldsymbol{w}}_t}\mathit{\boldsymbol{y}} $ (16)

    其中,kt表示加权矩阵K的第t行向量.

    发送符号s(t)的对数似然比(logarithmic likelihood ratio,LLR):

    $ \eta \left( t \right) = \ln \left( {\frac{{pr\left( {s'\left( t \right)\left| {s\left( t \right) = + 1} \right.} \right)}}{{pr\left( {s'\left( t \right)\left| {s\left( t \right) = - 1} \right.} \right)}}} \right) $ (17)

    根据LLR计算该符号的软输出:

    $ s'\left( t \right) = \tanh \left( {\eta \left( t \right)/2} \right) $ (18)

    最后,第t时隙源节点发送符号的精估计:

    $ \hat s\left( t \right) = \tilde s\left( t \right) - \sum\limits_{i = 1,i \ne t}^T {{\mathit{\boldsymbol{k}}_t}{\mathit{\boldsymbol{h}}_t}s'\left( i \right)} $ (19)

    其中,ht表示H的第t列向量.完成各时隙符号估计后,再对其解调解码,重构源节点信息比特,从而完成接收信号并行检测.

    4 差错性能分析

    一般地,成对差错概率(pairwise error probability,PEP)无闭合表达式,在分析差错性能时,首先求取其上界,然后分析分集增益.将式(13)改写为:

    $ \mathit{\boldsymbol{y}} = \mathit{\boldsymbol{Sfg}} + \mathit{\boldsymbol{w}} $ (20)

    其中,

    $ \mathit{\boldsymbol{S}} = \left[ {\begin{array}{*{20}{c}} {s\left( 1 \right)}&0&0& \cdots &0\\ {{h^2}s\left( 1 \right)}&{s\left( 2 \right)}&0& \cdots &0\\ 0&{{h^2}s\left( 2 \right)}& \ddots & \ddots & \vdots \\ \vdots & \vdots & \vdots &{s\left( {T - 1} \right)}&0\\ 0&0& \cdots &{{h^2}s\left( {T - 1} \right)}&{s\left( T \right)} \end{array}} \right] $
    $ \begin{array}{*{20}{c}} {\mathit{\boldsymbol{f}} = \left[ {\begin{array}{*{20}{c}} {{f_1}}&{{f_2}}\\ {{f_2}}&{{f_1}} \end{array}} \right]}&{\mathit{\boldsymbol{g}} = \left[ {\begin{array}{*{20}{c}} {{g_1}}\\ {{g_2}} \end{array}} \right]} \end{array} $

    根据式(20),两个不同码字SS之间的欧式距离:

    $ {d^2}\left( {\mathit{\boldsymbol{S}},\mathit{\boldsymbol{S'}}} \right) = {\mathit{\boldsymbol{g}}^H}{\mathit{\boldsymbol{f}}^H}\Delta {\mathit{\boldsymbol{S}}^H}{\mathit{\boldsymbol{ \boldsymbol{\varSigma} }}^{ - 1}}\Delta \mathit{\boldsymbol{Sfg}} $ (21)

    其中,ΔS=S-SΣ为等效干扰和噪声w的自相关矩阵.因此,FD-TPSR网络的成对差错概率可表示为

    $ PEP = Q\left( {\frac{{\sqrt {{d^2}\left( {\mathit{\boldsymbol{S}},\mathit{\boldsymbol{S'}}} \right)} }}{2}} \right) $ (22)

    其中,Q(·)为高斯Q函数.然而,高斯Q函数无法用初等函数表示,为此,求式(22)的数学期望即切尔诺夫界(Chernoff bound):

    $ \overline {PEP} = \mathop E\limits_{f,g} \left\{ {\exp \left( { - \frac{{{\mathit{\boldsymbol{g}}^H}{\mathit{\boldsymbol{f}}^H}\Delta {\mathit{\boldsymbol{S}}^H}{\mathit{\boldsymbol{ \boldsymbol{\varSigma} }}^{ - 1}}\Delta \mathit{\boldsymbol{Sfg}}}}{4}} \right)} \right\} $ (23)

    由于fg是互相独立的复高斯随机向量,求对g的数学期望:

    $ \overline {PEP} = \mathop E\limits_f \left\{ {{{\det }^{ - 1}}\left( {\mathit{\boldsymbol{I}} + \frac{{{\mathit{\boldsymbol{f}}^H}\left( {\Delta {\mathit{\boldsymbol{S}}^H}\Delta \mathit{\boldsymbol{S}}} \right)\mathit{\boldsymbol{f}}}}{{4{\rm{tr}}\left( \Sigma \right)}}} \right)} \right\} $ (24)

    上式利用了文献[14]的结论Σ≤tr(ΣI.根据分集增益定义,分集阶数取决于矩阵的秩.显然,单位矩阵I满秩;f的元素取自连续的随机过程,因此以概率1满秩. ΔS的秩较为复杂,分为两种情形分别讨论:

    情形1  假设源节点发送的连续T个码字中仅有一个符号与其他符号不同,称该符号为“异常符号”.若该符号为s(T),则ΔS的秩为1.

    情形2  若“异常符号”为s(1),…,s(T-1)中的任意一个,则ΔS的秩为2.

    综上:在源节点发送的所有符号中,s(T)取得1阶分集,其他符号均取得2阶分集.

    接下来考察等效干扰和噪声的功率特征,根据式(13),tr(Σ)满足如下性质:

    $ {\rm{tr}}\left( \mathit{\boldsymbol{ \boldsymbol{\varSigma} }} \right) \sim O\left( {{P^\lambda }} \right) $ (25)

    由此可知,在FD-TPSR网络中,由于全双工中继存在RSI,其等效干扰和噪声功率是发射功率的幂函数,即每阶分集增益为1-λ.

    5 仿真实验与结果分析

    采用蒙特卡洛法仿真评价中继间干扰处理方法和信号检测算法的性能.针对图 1所示的FD-TPSR网络模型,源节点采用QPSK调制,中继节点采用AF转发协议,RSI参数设置为λ=0.05,β=10-3,每次仿真中,信道参数为独立实现的高斯随机变量.

    图 3所示为无IRI消除、部分IRI消除和完全IRI消除方法的误码性能对比.由图可知,部分IRI消除方法的误比特率最低,这是因为完全消除IRI后,FD-TPSR网络退化为2跳全双工单中继网络,不能获得分集增益,而无IRI消除方法虽然保留了分集增益的获取能力,但由于AF转发固有的噪声和残留自干扰累积,导致源节点发送符号的端到端信干噪比降低,而部分IRI消除方法兼顾了分集增益和累积干扰与噪声因素,误比特率最低.

    图 3 不同IRI处理方法的误码性能对比

    图 4所示为极大似然(ML)信号检测算法、最小均方误差(MMSE)信号检测算法、MF-PSIC信号检测算法的误比特率性能对比.由图可知,由于MMSE信号检测算法无法处理由信道记忆造成的符号间干扰,其误码率曲线斜率逐渐趋于平缓,表明线性接收机无法获得分集增益,而ML信号检测算法和MF-PSIC信号检测算法的性能接近,表明二者均能保证分集增益可达,且MF-PSIC信号检测算法相比于ML信号检测算法,仅有很少的性能损失.

    图 4 不同信号检测算法的误码性能对比

    图 5所示为ML信号检测算法、MMSE信号检测算法和MF-PSIC信号检测算法的计算复杂度对比曲线.显然,ML信号检测算法的计算复杂度随发送符号数的增加快速增加.其原因在于:多符号联合ML检测问题等效为格空间内最小二乘问题,已被证明为NP难,因此ML信号检测算法具有指数级计算复杂度. MMSE信号检测算法的复杂度为O(T2),运算量主要表现为矩阵求逆.相比而言,MF检测仅需进行矩阵相乘运算,且LLR计算可用查表等快速方法实现,因此MF-PSIC信号检测算法的复杂度最低.

    图 5 不同信号检测算法的计算复杂度对比
    6 结论

    针对全双工双通路连续中继网络,本文提出了一种中继间干扰处理方法及信号检测算法.在中继间干扰处理中,中继节点消除部分中继间干扰,以提高端到端信干噪比,而目的节点利用保留的部分中继间干扰构成延迟转发编码,旨在获得分集增益.在信号检测部分,采用匹配滤波—软干扰消除的多符号并行检测,匹配滤波器结构简单,基于软输出的并行干扰消除能同时检测所有时隙的符号.仿真结果表明,本文提出的中继间干扰处理方法兼顾了分集增益和累积干扰与噪声因素,相比于无IRI消除和完全IRI消除,误比特率最低. MF-PSIC检测算法实现复杂度低,相比于ML检测算法仅有很少的性能损失.

    参考文献
    [1]
    OECHTERING T J, SCHNURR C, BJELAKOVIC I, et al. Broadcast Capacity Region of Two-Phase Bidirectional Relaying[J]. IEEE Transactions on Information Theory, 2008, 54(1): 454-458.
    [2]
    LIAU Q Y, LEOW C Y. Study of Relay Position in Two-Path Successive Relaying with Interference Cancellation[C]//IEEE Asia Pacific Conference on Wireless and Mobile Autust 28-30, 2014. Bali, Indonesia, New York, USA: IEEE, 2014.
    [3]
    LIU G, YU F R, JI H, et al. In-Band Full-duplex Relaying:A Survey, Research Issues and Challenges[J]. IEEE Communications Surveys & Tutorials, 2015, 17(2): 500-524.
    [4]
    ZHANG R. On Achievable Rates of Two-Path Successive Relaying[J]. IEEE Transactions on Communications, 2009, 57(10): 2914-2917. DOI:10.1109/TCOMM.2009.10.080113
    [5]
    DOU G Q, DENG R, HE X W, et al. Precoding-Based Inter-Relay Interference Cancelation for Amplify- and-Forward Two-Path Successive Relay Networks[J]. IEEE Signal Processing Letters, 2018, 25(2): 229-233.
    [6]
    LUO C B, GONG Y, ZHANG F C. Full Interference Cancellation for Two-Path Relay Cooperative Networks[J]. IEEE Transactions on Vehicular Technology, 2011, 60(1): 343-347. DOI:10.1109/TVT.2010.2090676
    [7]
    TIAN F, ZHANG W, MA W K, et al. An Effective Distributed Space-Time Code for Two-Path Successive Relay Network[J]. IEEE Transactions on Communications, 2011, 59(8): 2254-2263. DOI:10.1109/TCOMM.2011.060911.100686
    [8]
    GONG Y, LUO C B, CHEN Z. Two-Path Successive Relaying With Hybrid Demodulate and Forward[J]. IEEE Transactions on Vehicular Technology, 2012, 61(5): 2044-2053. DOI:10.1109/TVT.2012.2191807
    [9]
    GILAN M S, OLFAT A. On the Performance of Distributed Space-time Coding with Selection Relaying and Beamforming for Two-Path Successive Relay Network[C]//2017 IEEE International Black Sea Conference on Communications and Networking (BlackSeaCom), June 5-8, 2017. Istanbul. New York, USA: IEEE, 2017. IEEE Press, 2017: 1-5.
    [10]
    陈绍东. 基于二次约束及用户组选择的双向中继网络[J]. 西南师范大学学报(自然科学版), 2018, 43(1): 25-30.
    [11]
    YOON E. Maximum Likelihood Detection with a Closed-Form Solution for the Square QAM Constellation[J]. IEEE Communications Letters, 2017, 21(4): 829-832. DOI:10.1109/LCOMM.2016.2642924
    [12]
    LIU Y, XIA X G, ZHANG H L. Distributed Linear Convolutional Space-time Coding for Two-relay Full-duplex Asynchronous Cooperative Networks[J]. IEEE Transactions on Wireless Communications, 2013, 12(12): 6406-6417. DOI:10.1109/TWC.2013.102313.130541
    [13]
    ZHANG M X, AHMED S, KIM S. Iterative MMSE-based Soft MIMO Detection with Parallel Interference Cancellation[J]. IET communications, 2017, 11(11): 1775-1781.
    [14]
    JI Y C, HAN C C, WANG A G, et al. Partial Inter-relay Interference Cancellation in Two-Path Successive Relay Network[J]. IEEE Communications Letters, 2014, 18(3): 451-454. DOI:10.1109/LCOMM.2014.011714.140016
    An IRI Processing and Signal Detection Algorithm for Full-Duplex Two-Path Successive Relay Networks
    LIU Kang1, LIU Lu1, YANG Chong-hai2, TANG Wei1, FENG Wen-jiang2, LIU Huan1    
    1. Power Distribution and Control Center, Guizhou Power Grid Co., Ltd., Guiyang 550002, China;
    2. College of Microelectronics and Communication Engineering, Chongqing University, Chongqing 400044, China
    Abstract: Full-duplex two-path successive relay (FD-TPSR) networks have high spectral efficiency and simple state control. However, there is residual self-interference (RSI) in full duplex relay, and there will also be inter-relay interference (IRI) between transmitting relay and receiving relay, which will lead to the degradation of communication performance. In this paper, an IRI processing method is proposed, in which the relay nodes eliminate partial IRI to improve the end-to-end signal-to-interference-to-noise ratio (SINR) and reserve partial IRI to form a delayed forwarding coding structure at the destination node to provide time diversity gain. A matched-filter with parallel soft interference cancellation (MF-PSIC) algorithm is proposed, in which the structure of the matched filter is simple and the implementation complexity is low, and the parallel interference cancellation based on soft output can simultaneously detect all time slot symbols in parallel, and the processing delay is small. The simulation results show that the proposed IRI processing method takes into account diversity gain, cumulative interference and noise effects and, compared with non-IRI cancellation and full IRI cancellation, its bit error rate is the lowest. The implementation complexity of MF-PSIC detection algorithm is low, and there is little performance loss compared with ML detection algorithm.
    Key words: relay network    two-path successive relay network    full-duplex    inter-relay interference    signal detection    
    X