CN105049154B - A kind of multi-user's cognition network precoding optimization method based on MIMO VFDM - Google Patents

Abstract

The invention provides a kind of multi-user's cognition network precoding optimization method based on MIMO VFDM, on the basis of CTP suppresses interlayer interference, interfering between SC internal users is controlled by ITP, pass through the ITP encoder matrixs after the selection to kernel and rotation structure optimization, SC channel capacity can be made to reach maximum, so as to while the interference between suppressing double layer network multi-user, lift the capacity of cognition network.

Description

A kind of multi-user's cognition network precoding optimization method based on MIMO-VFDM

Technical field

The present invention relates to one kind using multiple-input and multiple-output (Multiple-Input Multiple-Output, MIMO) and Vandermonde subspace frequency division multiplexing (Vandermonde-subspace Frequency Division Multiplexing, VFDM) the optimizing design scheme of two layers of cognition network precoding of multi-user of transmission, for suppressing user in two layers of cognition network Between interference and maximize the channel capacity of double layer network.

Background technology

The sustainable growth of wireless traffic result in that frequency domain resource is more and more rare, therefore high spectrum utilization is wirelessly transferred As the study hotspot of modern communicationses.The deployment of the two-tier networks such as cognition network is a kind of feasible program for solving frequency spectrum resource. In the two-tier network communication system based on cognitive radio technology, cell (small-cells, SCs) is used as second layer net Network is disposed in the periphery of macrocell (macro-cell, MC) and shared spectrum transmissions.MC has preferential to shared frequency spectrum The right to use, double layer network opportunistic, which accesses frequency spectrum and must assure that, does not produce interference or caused interference to MC in tolerable scope It is interior.Therefore interference of the SCs to MC how is controlled to turn into the subject matter studied in the implementation of two layers of cognition network.

VFDM technologies make use of block transmission system as a kind of emerging method for solving cognition network interference control problem The frequency band redundancy that cyclic prefix provides in (such as ofdm system) establishes new transmission link and without interference with the normal of MC for SCs Communication.After VFDM, the signal that two layers of emitter are sent passes through interlayer precoding (Cross-Tier Precoder, CTP) quilt Double layer network is tied to in the kernel of a layer network interference channel.Afterwards, using precoding (Intra-Tier in layer Precoder, ITP) control interfering between SC internal users.

" the Channel estimation impact for lte small cells based on mu-vfdm of document 1 [Wireless Communications and Networking Conference(WCNC),IEEE,2012:2560- 2565] multi-user VFDM concept " is introduced during CTP is designed, wherein one layer of communication link forms MC, double layered communication Link forms SCs.

" the Spatial-Frequency Signal Alignment for Opportunistic of document 2 Transmission[IEEE Transactions on Signal Processing,2014,62:1561-1575] " will be more User VFDM research extend to mimo system.

" the Zero-forcing methods for downlink spatial multiplexing in of document 3 multiuser MIMO channels.[Signal Processing,IEEE Transactions on,2004,52(2): 461-471] propose a kind of diagonal ZF of block (Block-Diagonal Zero-Forcing, BD-ZF) precoding algorithms to Solves the interference problem between user in multiuser MIMO downlink.BD-ZF precoding algorithms have design conveniently, simple in construction The advantages that, it is highly suitable for the design of ITP pre-coding matrixes between multi-user.

The content of the invention

For overcome the deficiencies in the prior art, the present invention proposes that one kind is based on BD- on the basis of existing precoding algorithms ZF double layer network precoding optimization method., can by the ITP encoder matrixs after the selection to kernel and rotation structure optimization So that SC channel capacity reaches maximum, so as to which while the interference between suppressing double layer network multi-user, lifting recognizes net The capacity of network.

The technical solution adopted for the present invention to solve the technical problems comprises the following steps：

Step 1, to SC base stations SBS to MC users MU overall interlayer channel matrix H_SMSingular value decomposition is done to obtainWherein,WithIt is unitary matrice,For diagonal matrix,Wherein,It is one Individual diagonal matrix, diagonal entry are H_SMSingular value；

Step 2, subdivision matrix V_ctq=(V₁, V₂), wherein V₂For H_SMOne group of orthonormal basis of kernel, design CTP Matrix Cs=V₂, draw the dimension J=(N of available kernel_T-1)K+ N_TL, wherein, N_TFor SBS antenna numbers, K is MC significant character lengths, and L is MC circulating prefix-lengths；

Step 3, by equivalent channel matrix of the channel matrix after CTP precodings in SC layersBy line splitting into N_SUIndividual K ×[(N_T-1)K+N_TL] dimension submatrixN=1 ..., N_SU, N_SUSC numbers of users are represented, i.e.,WhereinBy(n-1) K+1 rows to nK rows member Element is formed, and represents the equivalent channel matrix of SBS to n-th SU in SC；

Step 4, by distracter w_SIt is divided into N_SUThe submatrix w that individual K × 1 is tieed up_S[i], i=1 ..., N_SU, i.e. w_S=(w_S[1 ]^H, w_S[2]^H..., w_S[NSU]^H)^H, wherein w_s[i] is by w_S(i-1) K+1 rows to iK rows element form, represent i-th of SU The distracter received；

Step 5, m=1 is initialized；

Step 6, one group of orthonormal basis V of matrix kernel corresponding to m-th of SU user ITP is calculated_null[m], including Following steps：

I. equivalent interference channel matrix in m-th of SU layer is constructed

Wherein,It is [(a N_SU-1)K]×[(N_T-1)K+N_TL] dimension matrix, at least one N_null[m]=N_TL +(N_T-N_SU) K dimension kernel；

Ii. it is rightCarry out singular value decompositionTry to achieveKernel standard Orthogonal basis V_null[m], wherein,With It is unitary matrice,For diagonal matrix, its diagonal entry is(N_SU- 1) K singular value；ThenOne group of orthonormal basis V of kernel_null[m]=V_SS[m](：, (N_SU-1)K+1：(N_T-1)K+ N_TL)；

Step 7, post processing matrix is calculatedWherein,It is w_S The autocorrelation matrix of [m]；

Step 8, solving and choose and rotation process matrix T [m] so that m-th of SU channel capacity reaches maximum, including with Lower step：

I. calculating matrixTo X^H[m] X [m] does Eigenvalues Decomposition：

Wherein,For unitary matrice,For to angular moment Battle array, its diagonal entry is matrix X^H[m] X [m] characteristic value；

Ii. matrix T [m] is by Q_xxPreceding D [m] row of [m] are formed, i.e. T [m]=Q_xx[m](：, 1：D [m]), wherein, D [m] is Dimension is sent corresponding to SU, whenWhen, D [m]=N_null[m], whenWhen D [m]=K；

Step 9, m-th of SU optimization ITP matrix Us are sought_BD-ZF_opt[m]=V_null[m]T[m]；

Step 10, m-th of SU channel capacity is calculated

Wherein, I_KThe unit matrix of K × K dimensions is represented,Expression is sent to M SU signal power allocation matrix, P [m] is obtained by water injection power algorithm；

Step 11, m adds 1, repeat step 6~11, until m=N_SUOptimization design is completed.

The beneficial effects of the invention are as follows：While inter-user interference is effectively suppressed, the biography of double layer network can be maximized Defeated capacity.

Brief description of the drawings

Fig. 1 is the MC based on OFDMA and multi-user's SC system model schematic diagrames based on MIMO-VFDM；

Fig. 2 is N_SUThe ITP based on BD-ZF can use dimension region and N when fixed_TCorresponding relation schematic diagram；

Fig. 3 is SC channel capacity and number of users N_SURelation schematic diagram；

Fig. 4 is SC channel capacity and number of users N_TRelation schematic diagram；

Fig. 5 is the channel capacity schematic diagram of SC under different signal to noise ratio.

Embodiment

The present invention is further described with reference to the accompanying drawings and examples, and the present invention includes but are not limited to following implementations Example.

The system model of two layers of cognition network is as shown in Figure 1.Cognition network is made up of one layer of MC and two layer of SC, and MC is adopted Communicated with LTE orthogonal frequency-time multiple access modes, SC is communicated using MIMO-VFDM modes.MC and SC shares spectral bandwidth, but MC There is preferential right to shared frequency spectrum.The MC of one layer network is by a MC base station (MC Base Station, MBS) and one MC user (MC User, MU) forms, and the SC of double layer network is by a SC base station (SC Base Station, SBS) and N_SUIndividual SC User (SC User, SU) forms.Assuming that all nodes perfect can obtain required channel condition information, and to system model Make following setting：MBS and all MU uses single antenna in one layer network MC, and SBS is led to using multiple antennas in double layer network SC Letter, SBS antenna numbers are N_T, all SU use single antenna.MC uses significant character length as K, and circulating prefix-length is L's OFDM transmission mode.

With reference to system model, definitionSymbolic vector is sent for MBS,For received by MU Signal vector,N=1,2 ..., N_TThe signal vector sent for SBS n-th antenna,M=1,2 ..., N_SUThe signal vector received for m-th of SU,Received for all SU Overall noise signal.Channel matrix is in MC layerRepresent the channel frequency domain response matrix from MBS to MU. Channel matrix is in SC layerRepresent the overall channel matrix of SBS to each SU in SC.Interlayer Channel matrix includes：WithRepresent SBS to MU and MBS to all SU's respectively Overall interlayer channel matrix.

On the basis of CTP suppresses interlayer interference, interfering between SC internal users is controlled by ITP.The present invention will Two parts are divided to be described, i.e. CTP designs and ITP optimization design.

I.CTP design

OrderFor SBS CTP coding before transmission signal vector,For CTP encoder matrixs, Wherein J represents the maximum transmission dimensions of SBS.It is dry that the design of CTP matrixes must ensure that communication of the SC transmission signal to MC is not formed Disturb, meet H_SMCs_S=0, i.e.,

H_SMC=0 (1)

Meet that the CTP matrixes of formula (1) fall in matrix H_SMKernel on, then CTP specific design scheme is as follows：

1) to channel matrix H_SMSingular value decomposition is done to obtainWherein,WithIt is unitary matrice,For diagonal matrix, its expression formula isWherein,It is a diagonal matrix, diagonal entry is H_SMIt is strange Different value.

2) to matrix V_ctpDo following segmentation：

V_ctp=(V₁, V₂) (2)

WhereinEasily prove V₂Meet H_SMV₂=0, because This V₂For H_SMOne group of orthonormal basis of kernel.Design CTP Matrix Cs=V₂, it can be deduced that J=(N_T-1)K+N_TL。

II.ITP optimization design

ITP design principle is similar to CTP, and its main purpose is to eliminate interfering for signal in SC, is maximized simultaneously SC channel capacities.Equivalent channel matrix of the channel matrix after CTP precodings in SC layersIt is represented by SU receives external disturbance item and noise isITP optimization design step is as follows：

1) by matrixBy line splitting into Nx_UIndividual K × [(N_T-1)K+N_TL] dimension submatrixN=1 ..., N_SU, I.e.WhereinBy(n-1) K+1 rows to nK rows Element is formed, and represents the equivalent channel matrix of SBS to n-th SU in SC.

2) by distracter w_SIt is divided into N_SUThe submatrix w that individual K × 1 is tieed up_S[i], i=1 ..., N_SU, i.e. w_S=(w_S[1]^H, w_S [2]^H..., w_S[NSU]^H)^H, wherein w_s[i] is by w_SThe element of (i-1) K+1 rows to iK rows form, represent that i-th SU is received The distracter arrived.

3) m=0 is initialized.

4) one group of orthonormal basis V of matrix kernel corresponding to m-th of SU user ITP is calculated_null[m]。

Iii. equivalent interference channel matrix in m-th of SU layer is constructedIt is as follows：

Wherein,It is [(a N_SU-1)×[(N_T-1)K+N_TL] dimension matrix, at least one N_null[m]=N_TL+ (N_T-N_SU) K dimension kernel；

Iv. it is rightSingular value decomposition is carried out, is tried to achieveKernel orthonormal basis V_null[m]。It is unusual Value is decomposed into：

Wherein,With For unitary matrice,For diagonal matrix, its diagonal entry is(N_SU-1)K Individual singular value.

ThenOne group of orthonormal basis of kernel can be represented by the formula：

V_null[m]=V_SS[m](：, (N_SU-1)K+1：(N_T-1)K+N_TL) (5)

5) post processing matrix Q [m] is calculated.In order to maximize SC channel capacities, post processing is used at each SU receivers The signal vector that matrix receives to m-th of SUAlbefaction is carried out, eliminates the interference from MC and noise signal. Q [m] calculation formula is：

Wherein,It is w_SThe autocorrelation matrix of [m].

6) solve and choose and rotation process matrix T [m] so that m-th of SU channel capacity reaches maximum.Then T [m] is asked Solution preocess is as follows：

Ii. calculating matrixTo X^H[m] X [m] does Eigenvalues Decomposition：

Wherein,For unitary matrice,For diagonal matrix, Its diagonal entry is matrix X^H[m] X [m] characteristic value.

Iii. matrix T [m] is by Q_xxPreceding D [m] row of [m] are formed, i.e. T [m]=Q_xx[m](：, 1：D [m]), wherein, D [m] It is that dimension is sent corresponding to SU, its selection rule is provided by Fig. 2：WhenWhen, D [m]=N_null[m], whenWhen D [m]=K.

7) m-th of SU optimization ITP matrixes are soughtCalculation formula is as follows：

8) m-th of SU channel capacity C [m] is calculated, its calculation formula is as follows：

Its In, I_KThe unit matrix of K × K dimensions is represented,Expression is sent to m-th of SU letter Number power distribution matrix, P [m] is obtained by water injection power algorithm.

9) m adds 1, repeats 4)~9) until m=N_SUOptimization design is completed.

One layer network of embodiment uses K=64, L=16, the OFDM transmission with a width of 1.92MHz；Channel impulse sound exists OFDM symbol is indeclinable when transmitting, and obeys multiple Gauss random distribution.

Fig. 2 illustrates N_SUFor definite value when the ITP based on BD-ZF can use dimension region and N_TRelation.Dash area in figure Represent the available transmission dimension region using SBS to m-th SU link in the case of CTP.Observation is found：I) whenWhen, the upper bound of available dimension of the dimension equal to SU of kernel；Ii) whenWhen, zero Spatial Dimension is more than required dimension.Situation i) need to be to V_null[m] carries out certain rotation to buildMatrix, make SC channel capacity is maximum；Situation ii) need to V_null[m] choose with rotation process to buildMatrix. SC channel capacities are made to reach maximum.To V_nullTwo kinds of operations of [m] can be realized by choosing with spin matrix T [m].

Fig. 3 is given in N_T=8, signal to noise ratio uses optimization and is not optimised under conditions of being SNR=10dB and SNR=10dB BD-ZF ITP matrixes when SC channel capacity and N_SUCorresponding relation.It can be seen that N_SUThe present invention is carried during ＜ 9 Optimization ITP algorithm performances be significantly better than and be not optimised ITP algorithms, N_SUTwo kinds of algorithm performances are consistent when=9.It could be observed that With N_SUIncrease, the SC channel capacities of optimization ITP algorithms proposed by the invention are not monotonic increase, but are first increased After reduce.This number of users illustrated in cell is not The more the better, but cell channel capacity is reached most in the presence of one Excellent number of users.

Fig. 4 is given in N_SU=2, signal to noise ratio uses optimization and is not optimised under conditions of being SNR=10dB and SNR=30dB BD-ZF ITP matrixes when SC channel capacity and N_TCorresponding relation.As seen from Figure 4, with transmitting antenna number N_TIncreasing Add, the ITP algorithm for design performances of optimization, which are substantially better than, is not optimised ITP algorithm for designs.

Fig. 5, which is illustrated, to be optimized ITP algorithms and is not optimised ITP algorithm SC transmission capacity comparing results under different signal to noise ratio. In Fig. 5, SBS transmitting antenna number is set as N_T=8, give SC numbers of users N_SU=4 and N_SU=9 two kinds of situations.Can be with by Fig. 5 Find out, work as N_SUWhen=4, the channel capacity performance of optimization ITP algorithms, which is substantially better than, is not optimised ITP algorithms, works as N_SUWhen=9, optimization For ITP algorithm for designs with being not optimised that algorithm performance is consistent, this is consistent with the result shown in Fig. 3.

Claims (1)

1. a kind of multi-user's cognition network precoding optimization method based on MIMO-VFDM, VFDM is vandermonde subspace frequency division Multiplexing, it is characterised in that comprise the steps：

Step 1, to SC base stations SBS to MC users MU overall interlayer channel matrix H_SMSingular value decomposition is done to obtainWherein,WithIt is unitary matrice,For diagonal matrix,Wherein,It is one Diagonal matrix, diagonal entry are H_SMSingular value；

Step 2, subdivision matrix V_ctp=(V₁, V₂), wherein V₂For H_SMOne group of orthonormal basis of kernel, design interlayer precoding CTP Matrix Cs=V₂, draw the maximum transmission dimensions of SBS J=(N_T-1)K+N_TL, wherein, N_TFor SBS antenna numbers, K is MC significant character lengths, and L is MC circulating prefix-lengths；

Step 3, the equivalent channel matrix by channel matrix in SC layers after interlayer precoding CTP precodingsBy line splitting into N_SUIndividual K × [(N_T-1)K+N_TL] dimension submatrixN=1 ..., N_SU, N_SUSC numbers of users are represented, i.e.,WhereinBy(n-1) K+1 rows to nK rows member Element is formed, and represents the equivalent channel matrix of SBS to n-th SU in SC；

Step 4, by distracter w_SIt is divided into N_SUThe submatrix w that individual K × 1 is tieed up_S[i], i=1 ..., N_SU, i.e. w_S=(w_S[1]^H, w_S [2]^H..., w_S[N_SU]^H)^H, wherein w_s[i] is by w_S(i-1) K+1 rows to iK rows element form, represent i-th of SU receive Distracter；

Step 5, m=1 is initialized；

Step 6, one group of orthonormal basis V of matrix kernel corresponding to precoding ITP in m-th of SU client layer is calculated_null[m], Comprise the following steps：

I. equivalent interference channel matrix in m-th of SU layer is constructed

Wherein,It is [(a N_SU-1)K]×[(N_T-1)K+N_TL] dimension matrix, at least one N_nullThe zero of [m] dimension is empty Between, wherein N_null[m]=N_TL+(N_T-N_SU)K；

Ii. it is rightCarry out singular value decompositionTry to achieveKernel standard just Hand over base V_null[m], wherein,With It is unitary matrice,For diagonal matrix, its diagonal entry is(N_SU- 1) K singular value；ThenOne group of orthonormal basis V of kernel_null[m]=V_SS[m](：, (N_SU-1)K+1：(N_T-1)K+ N_TL)；

Step 7, post processing matrix is calculatedWherein,It is w_S[m] from Correlation matrix；

Step 8, solve and choose and rotation process matrix T [m] so that m-th of SU channel capacity reaches maximum, including following step Suddenly：

I. calculating matrixTo X^H[m] X [m] does Eigenvalues Decomposition：

<mrow> <msup> <mi>X</mi> <mi>H</mi> </msup> <mo>&amp;lsqb;</mo> <mi>m</mi> <mo>&amp;rsqb;</mo> <mi>X</mi> <mo>&amp;lsqb;</mo> <mi>m</mi> <mo>&amp;rsqb;</mo> <mo>=</mo> <msub> <mi>Q</mi> <mrow> <mi>x</mi> <mi>x</mi> </mrow> </msub> <mo>&amp;lsqb;</mo> <mi>m</mi> <mo>&amp;rsqb;</mo> <msub> <mi>&amp;Lambda;</mi> <mrow> <mi>x</mi> <mi>x</mi> </mrow> </msub> <mo>&amp;lsqb;</mo> <mi>m</mi> <mo>&amp;rsqb;</mo> <msubsup> <mi>Q</mi> <mrow> <mi>x</mi> <mi>x</mi> </mrow> <mi>H</mi> </msubsup> <mo>&amp;lsqb;</mo> <mi>m</mi> <mo>&amp;rsqb;</mo> <mo>;</mo> </mrow>

Wherein,For unitary matrice,For diagonal matrix, its is right Diagonal element is matrix X^H[m] X [m] characteristic value；

Ii. matrix T [m] is by Q_xxPreceding D [m] row of [m] are formed, i.e. T [m]=Q_xx[m](：, 1：D [m]), wherein, D [m] is SU pairs The transmission dimension answered, whenWhen, D [m]=N_null[m], whenWhen D [m]=K；

Step 9, precoding ITP matrixes in m-th of SU optimization layer are sought

Step 10, m-th of SU channel capacity is calculated

<mrow> <mi>C</mi> <mo>&amp;lsqb;</mo> <mi>m</mi> <mo>&amp;rsqb;</mo> <mo>=</mo> <mfrac> <mn>1</mn> <mrow> <mi>K</mi> <mo>+</mo> <mi>L</mi> </mrow> </mfrac> <msub> <mi>log</mi> <mn>2</mn> </msub> <mo>|</mo> <msub> <mi>I</mi> <mi>K</mi> </msub> <mo>+</mo> <mi>Q</mi> <mo>&amp;lsqb;</mo> <mi>m</mi> <mo>&amp;rsqb;</mo> <msub> <mover> <mi>H</mi> <mo>&amp;OverBar;</mo> </mover> <mrow> <mi>S</mi> <mi>S</mi> </mrow> </msub> <mo>&amp;lsqb;</mo> <mi>m</mi> <mo>&amp;rsqb;</mo> <msub> <mi>V</mi> <mrow> <mi>n</mi> <mi>u</mi> <mi>l</mi> <mi>l</mi> </mrow> </msub> <mo>&amp;lsqb;</mo> <mi>m</mi> <mo>&amp;rsqb;</mo> <mi>T</mi> <mo>&amp;lsqb;</mo> <mi>m</mi> <mo>&amp;rsqb;</mo> <mi>P</mi> <mo>&amp;lsqb;</mo> <mi>m</mi> <mo>&amp;rsqb;</mo> <msup> <mi>T</mi> <mi>H</mi> </msup> <mo>&amp;lsqb;</mo> <mi>m</mi> <mo>&amp;rsqb;</mo> <msubsup> <mi>V</mi> <mrow> <mi>n</mi> <mi>u</mi> <mi>l</mi> <mi>l</mi> </mrow> <mi>H</mi> </msubsup> <mo>&amp;lsqb;</mo> <mi>m</mi> <mo>&amp;rsqb;</mo> <msubsup> <mover> <mi>H</mi> <mo>&amp;OverBar;</mo> </mover> <mrow> <mi>S</mi> <mi>S</mi> </mrow> <mi>H</mi> </msubsup> <mo>&amp;lsqb;</mo> <mi>m</mi> <mo>&amp;rsqb;</mo> <mi>Q</mi> <msup> <mrow> <mo>&amp;lsqb;</mo> <mi>m</mi> <mo>&amp;rsqb;</mo> </mrow> <mi>H</mi> </msup> <mo>|</mo> <mo>,</mo> </mrow>

Wherein, I_KThe unit matrix of K × K dimensions is represented,Expression is sent to m-th SU signal power allocation matrix, P [m] is obtained by water injection power algorithm；

Step 11, m adds 1, repeat step 6~11, until m=N_SUOptimization design is completed.