CN106961297A - A kind of interference alignment schemes distributed based on water injection power - Google Patents

Abstract

The invention discloses a kind of interference alignment schemes distributed based on water injection power, base station random broadcast pre-coding matrix first, user is to minimize interference leakage power summation as target, decoding matrix is designed along the direction that transmission rate rises using the gradient descent algorithm based on Grassmann manifold, then communication direction is inverted, the decoding matrix of renewal is sent to base station by user, and base station equally updates pre-coding matrix, convergence is reached after iteration, the final effect for obtaining interference alignment.Then base station is responsible for carrying out power distribution, first each data flow mean allocation power is given according to base station maximum available power set in advance, then according to the channel gain of different data streams, power distribution is iterated to the different data streams for being sent to user using the power distribution mode of water filling.The present invention can be obtained on the premise of complexity is not dramatically increased while reaching interference alignment -- higher transmission rate.

Description

A kind of interference alignment schemes distributed based on water injection power

Technical field

The present invention relates to mobile radio telecommunications technical field, particularly a kind of interference alignment side distributed based on water injection power Method.

Background technology

With the extensive use of multi-antenna technology, interference has become the principal element of limitation wireless network performance lifting. Traditional disturbance restraining method handle interference is as noise processed or orthogonalization, but these methods have respective weak point, The method for example handled disturbing as noise is although simple and easy to apply, but can not the larger interference of process signal intensity.

Interference alignment techniques are one of newest interference mitigation technologies, and its basic thought is the interference signal for compressing receiving terminal The Spatial Dimension of occupancy, ensures that the subspace dimension of desired signal is sufficiently large as far as possible.The appearance of interference alignment techniques changes people Before the understanding limited by interference of the capacity to wireless network.Because interference alignment techniques have transmission rate high, it is easy to real The advantages of not limited with power system capacity by number of users now, is one of technology of great potential in next-generation communication network.

Traditional interference alignment can be divided into enclosed resolving Algorithm and distributed iterative interference alignment algorithm, wherein distributed calculate Method is revealed as target with the interference for minimizing receiving terminal, using the reciprocity of channel, and the interference of alternative optimization transmitting-receiving node is revealed Power, finally when it is zero to disturb leakage, that is, reaches perfect interference alignment, compared with enclosed resolving Algorithm, it is only necessary to know every The local channel state information of individual node, and greatly reduce feedback overhead.Also there are many optimized algorithms in distribution On the basis of iterative algorithm, according to different object functions come design update precoding and decoding matrix, it can reach higher System velocity.But these algorithms are all to each data flow mean allocation power, in order to more fully utilize the work(of system Rate, the co-design interference alignment of some scholars proposition in recent years and power distribution algorithm, although what is had can reduce the error code of system Rate, but to sacrifice a part of transmission rate as cost, what is had then may be only available for certain applications scene, there is certain limitation Property, some algorithm complexes are higher, it is difficult to be applied in actual engineering.

The content of the invention

The technical problems to be solved by the invention are to overcome the deficiencies in the prior art and provide a kind of based on water injection power point The interference alignment schemes matched somebody with somebody, the present invention can be obtained on the premise of complexity is not dramatically increased while reaching interference alignment Higher system transfer rate.

The present invention uses following technical scheme to solve above-mentioned technical problem：

According to a kind of interference alignment schemes distributed based on water injection power proposed by the present invention, comprise the following steps：

Step 1, initialization iterations q=1, initialize step-length t, initialize the pre-coding matrix V of k-th of base station_k,0； Base station is distributed to the data flow of user by maximum available power averaging；

Step 2, k-th of base station randomly generate M × d dimension pre-coding matrixes V_k,q-1To the user of this cell, and meetWherein, I is unit matrix, and M is antenna for base station number, and d is sent to the data flow of user for base station Number, subscript H be conjugate transposition, k=1,2,3 ..., K, K be base station total number；

L-th of user that k-th of base station is serviced is according to minimizing when interference leakage criterion calculates the q-1 times iteration the The decoding matrix U of l user_l,q-1,

Wherein, H_klRepresent k-th of base station to the channel gain matrix of l-th of user, l=1,2,3 ..., K, channel coefficients It is zero to obey average, and variance is distributed for 1 independent identically distributed multiple Gauss, V_k,q-1Represent during the q-1 times iteration k-th base station Pre-coding matrix,The corresponding characteristic vector of d characteristic value minimum representing matrix *, P_kFor the power point of k-th of base station With matrix；

Step 3, by U_l,q-1The direction increased along geodesic curve to transmission rate is moved, and step-length t is used for adjusting U_l,q-1It is mobile away from From size, obtain the decoding matrix U of l-th of user during the q times iteration_l,q；

Step 4, reversion communication direction, k-th of base station calculates the pre-coding matrix V updated during the q times iteration_k,q,

Step 5, renewal step-length t=t × 0.90；

Step 6, it is not more than default maximum interference leakage power when the maximum interference leakage power summation that user is subject to, or When person's iterations has reached maximum iteration, the V now obtained_k,q, U_l,qMake transmission rate performance best respectively Pre-coding matrixAnd decoding matrixInterference leakage power suffered by user now is zero, that is, has reached interference alignment； Otherwise q=q+1, repeat step 2 to 5 are made；

Step 7, the power water filling value λ for going out using dichotomizing search k-th of base station_kValue；

Step 8, minimize interference leakage criterion under using maximize useful signal spatial power as target reach interference pair On the basis of neat, according to the λ of acquisition_k, the power p of the different data streams sent by way of water filling to base station_kjDivided Match somebody with somebody.

It is used as a kind of further prioritization scheme of interference alignment schemes distributed based on water injection power of the present invention, step It is 0.90 that step-length t is initialized in 1.

It is used as a kind of further prioritization scheme of interference alignment schemes distributed based on water injection power of the present invention, step U is obtained in 3_l,qDetailed process it is as follows：

First obtain the power in useful signal spaceIt is right U_l,q-1Derivative Q_klU_l,q-1, wherein,Then ladders of the P along Grassmann manifold is obtained Degree descent direction is Z_k,And to Z_kCarry out Eigenvalues Decomposition and obtain G Σ F^H, its Middle G, F are unitary matrice, and Σ is diagonal matrix；

The decoding matrix then updated is U_l,q=(U_l,q-1F(cosΣt)F^H)+(G(sinΣt)F^H)。

It is used as a kind of further prioritization scheme of interference alignment schemes distributed based on water injection power of the present invention, step Maximum iteration is 5000 times in 6, and default maximum interference leakage power is 10^-6Watt.

It is used as a kind of further prioritization scheme of interference alignment schemes distributed based on water injection power of the present invention, step The power p of each data flow in 8_kjObtained from below equation：

Wherein, parameter t_kjRepresent interference of j-th of the data flow of k-th of base station to other data flows, I_ljIt is l-th of use Summation, σ are disturbed between inter-user interference and data flow that family is subject to²For the variance of multiple Gauss noise,For k-th base station J-th of data flows to the efficient channel gain of j-th of data flow of l-th of user, (*)⁺=max (*, 0),The jth row of subscript [* j] representing matrix.

The present invention uses above technical scheme compared with prior art, with following technique effect：

(1) interference alignment and power distributing technique are combined, is designed first by maximizing with speed interference alignment schemes Pre-coding matrix and decoding matrix, i.e., move matrix along the direction that transmission rate increases, and is reached in base station and user terminal dry Disturb after alignment, the performance of further lifting system is combined with water injection power distribution method, due to the power distribution algorithm not only Consider the interference between user, it is also contemplated that the interference between data flow, therefore base station can be with to the transmission rate performance of user Obtain good lifting；

(2) to maximize power as object function, using the gradient descent algorithm based on Grassmann manifold along transmission The direction design decoding matrix that speed rises, can simplify derivative operation, reduce complexity；The method of the invention is applied to Wireless data transmission scene in MIMO interference channels.

Brief description of the drawings

Fig. 1 is the equivalent interference channel figure of the applicable K cell K custom systems of the present invention.

Fig. 2 is the equivalent scene graph for the user of 3 cell 3 that present example is applicable.

Fig. 3 is flow chart of the invention.

Fig. 4 is the comparison figure of the invention with the system average transmission rate performance of traditional minimum interference leakage method.

Embodiment

Technical scheme is described in further detail below in conjunction with the accompanying drawings：

Fig. 1 is the equivalent interference channel figure of the applicable K cell K custom systems of the present invention, and present case uses 3 cells, often Individual one user of cell serves, base station and user are equipped with the downlink scenario of 5 antennas, and each base station is sent to user 2 Individual data flow.System works in the downlink.Each base station can produce interference, especially minizone to the user of non-cell Edge customer can be by larger interference, system model as shown in Fig. 2 flow is as shown in figure 3, specific implementation steps are：

(1) iterations q=1, is initialized, initialization step-length t, k-th of Base station initialization 5 × 2 ties up pre-coding matrix V_k,0, k=1,2,3, maximum available transmission power is averagely allocated to the data flow of user by base station.

(2) k-th of the dimension pre-coding matrix of Base Transmitter 5 × 2 V_k,q-1To the user of this cell, and meetWherein, I is unit matrix, k=1,2,3.

L-th of user that k-th of base station is serviced designs decoding matrix U according to interference leakage criterion is minimized_l,q-1,

Wherein, H_klK-th of base station is represented to the channel gain matrix of l-th of user, it is zero that channel coefficients, which obey average, side Difference is distributed for 1 independent identically distributed multiple Gauss, V_k,q-1The pre-coding matrix of k-th of base station during the q-1 times iteration is represented,The corresponding characteristic vector of d characteristic value minimum representing matrix *, P_kFor the power distribution matrix of k-th of base station, k=1, 2,3。

(3) power in useful signal space is first obtained To U_l,q-1Derivative Q_klU_l,q-1, wherein,H_klRepresent k-th of base station to l-th of use The channel gain matrix at family, V_k,q-1The pre-coding matrix of k-th of base station, U during for the q-1 times iteration_l,q-1It is the q-1 times iteration When l-th of user decoding matrix, P_kFor the power distribution matrix of k-th of base station；Then ladders of the P along Grassmann manifold is obtained Degree descent direction is Z_k,And to Z_kCarry out Eigenvalues Decomposition and obtain G Σ F^H, its Middle G, F are unitary matrice, and Σ is diagonal matrix；

The decoding matrix then updated is U_l,q=(U_l,q-1F(cos Σt)F^H)+(G(sinΣt)F^H), l=1,2,3.

By the decoding matrix U of l-th of user during the q times iteration_l,q-1The direction increased along geodesic curve to transmission rate is moved, Step-length t is used for adjusting the size of matrix displacement, can obtain the decoding matrix of the renewal of l-th of user during the q times iteration U_l,q。

(4) communication direction is inverted, k-th of base station calculates the pre-coding matrix updated when obtaining the q times iteration

(5) step-length t=t × 0.90 is updated.

(6) when the maximum interference leakage power summation that user is subject to is not more than default leakage power, or reach During maximum iteration, V now_k,q, U_l,qIt is the pre-coding matrix for making transmission rate performance best respectivelyWith decoding square Battle arrayInterference leakage power suffered by user now is zero, that is, has reached interference alignment；Otherwise q=q+1, repeat step are made 2 to 5.

(7) the power water filling value λ of k-th of base station is gone out using dichotomizing search_kValue, λ_kValue from below equation Obtain：

Wherein, P_ktIt is P for the maximum available power of k-th of base station_kt, σ²For the variance of multiple Gauss noise, parameter t_kjRepresent Interference of j-th of the data flow of k-th of base station to other data flows, I_ljIt is inter-user interference and data that l-th of user is subject to Summation is disturbed between stream.p_kiFor the power of i-th of data flow distribution of k-th of base station, p_niFor i-th of data of nth base station The power of distribution is flowed,The efficient channel for j-th of data flow for flowing to l-th of user for i-th of data of nth base station increases Benefit,For k-th of base station i-th of data flow to l-th of user j-th of data flow efficient channel gain,For m F-th of data of individual base station flow to the efficient channel gain of i-th of data flow of l-th of user,For the i-th of k-th of base station Individual data flow to the efficient channel gain of i-th of data flow of l-th of user,J-th of data for k-th of base station are flowed to The efficient channel gain of j-th of data flow of l-th of user.

(8) minimize interference leakage criterion under using maximize useful signal spatial power as target reach interference alignment On the basis of, according to the λ of acquisition_k, the power p of the different data streams sent by way of water filling to base station_kjDivided successively Match somebody with somebody.

Wherein, (*)⁺=max (*, 0), J-th of data for k-th of base station are flowed to The efficient channel gain of j-th of data flow of l-th of user, the jth row of subscript [* j] representing matrix.

In order to evaluate the performance of the present invention, the average transmission rate performance of system is done using the method for Monte Carlo simulation Emulation.It is zero that channel matrix, which obeys average, and variance is distributed for 1 Cyclic Symmetry multiple Gauss, initializes pre-coding matrix, initially Change power distribution matrix and step-length sets all same, maximum iteration is 5000 steps, takes the average value of 100 simulation results.

Fig. 4 is compared the interference alignment algorithm distributed based on water injection power and the least interference that constant power is distributed and reveals algorithm Average and speed with signal to noise ratio variation relation.As can be seen that based on water injection power control interference alignment schemes performance compared with The former averagely obtains 10.3bps/Hz rate gain.

It should be appreciated that for those of ordinary skills, can according to the above description be improved or converted, And all these modifications and variations should all belong to the protection domain of appended claims of the present invention.

Claims (5)

1. a kind of interference alignment schemes distributed based on water injection power, it is characterised in that comprise the following steps：

Step 1, initialization iterations q=1, initialize step-length t, initialize the pre-coding matrix V of k-th of base station_k,0；Base station Maximum available power averaging is distributed to the data flow of user；

Step 2, k-th of base station randomly generate M × d dimension pre-coding matrixes V_k,q-1To the user of this cell, and meetWherein, I is unit matrix, and M is antenna for base station number, and d is sent to the data flow of user for base station Number, subscript H be conjugate transposition, k=1,2,3 ..., K, K be base station total number；

L-th of user that k-th of base station is serviced is revealed when criterion calculates the q-1 times iteration l-th according to interference is minimized The decoding matrix U of user_l,q-1,

$U_{l,q-1}=v_{\min }^d\left(\underset{k\≠l}{\Σ}{H}_{kl}{V}_{k,q-1}{P}_k{P}_k^H{V}_{k,q-1}^H{H}_{kl}^H\right),$

Wherein, H_klRepresent k-th of base station to the channel gain matrix of l-th of user, l=1,2,3 ..., K, channel coefficients obedience Average is zero, and variance is distributed for 1 independent identically distributed multiple Gauss, V_k,q-1Represent during the q-1 times iteration prelisting for k-th base station Code matrix,The corresponding characteristic vector of d characteristic value minimum representing matrix *, P_kFor the power distribution square of k-th of base station Battle array；

Step 3, by U_l,q-1The direction increased along geodesic curve to transmission rate is moved, and step-length t is used for adjusting U_l,q-1Displacement Size, obtains the decoding matrix U of l-th of user during the q times iteration_l,q；

Step 4, reversion communication direction, k-th of base station calculates the pre-coding matrix V updated during the q times iteration_k,q,

Step 5, renewal step-length t=t × 0.90；

Step 6, it is not more than default maximum interference leakage power when the maximum interference leakage power summation that user is subject to, or repeatedly When generation number has reached maximum iteration, the V now obtained_k,q, U_l,qIt is prelisting of making transmission rate performance best respectively Code matrixAnd decoding matrixInterference leakage power suffered by user now is zero, that is, has reached interference alignment；Otherwise Make q=q+1, repeat step 2 to 5；

Step 7, the power water filling value λ for going out using dichotomizing search k-th of base station_kValue；

Step 8, minimize interference leakage criterion under using maximize useful signal spatial power as target reach interference alignment On the basis of, according to the λ of acquisition_k, the power p of the different data streams sent by way of water filling to base station_kjIt is allocated.

2. a kind of interference alignment schemes distributed based on water injection power according to claim 1, it is characterised in that step 1 Middle initialization step-length t is 0.90.

3. a kind of interference alignment schemes distributed based on water injection power according to claim 1, it is characterised in that step 3 In obtain U_l,qDetailed process it is as follows：

First obtain the power in useful signal spaceTo U_l,q-1's Derivative Q_klU_l,q-1, wherein,Then gradients of the P along Grassmann manifold is obtained to decline Direction is Z_k,And to Z_kCarry out Eigenvalues Decomposition and obtain G Σ F^H, wherein G, F is equal For unitary matrice, Σ is diagonal matrix；

The decoding matrix then updated is U_l,q=(U_l,q-1F(cosΣt)F^H)+(G(sinΣt)F^H)。

4. a kind of interference alignment schemes distributed based on water injection power according to claim 1, it is characterised in that step 6 Middle maximum iteration is 5000 times, and default maximum interference leakage power is 10^-6Watt.

5. a kind of interference alignment schemes distributed based on water injection power according to claim 1, it is characterised in that step 8 In each data flow power p_kjObtained from below equation：

$p_{kj}={(\frac{1}{t_{kj}+{\mathrm{\λ}}_k}-\frac{I_{lj}+{\mathrm{\σ}}^2}{\left|h_{kl}^{{\mathrm{jj}}^2}\right|})}^{+}$

Wherein, parameter t_kjRepresent interference of j-th of the data flow of k-th of base station to other data flows, I_ljL-th of user by Inter-user interference and data flow between disturb summation, σ²For the variance of multiple Gauss noise,For j-th of number of k-th of base station According to the efficient channel gain for j-th of data flow for flowing to l-th of user, (*)⁺=max (*, 0), The jth row of subscript [* j] representing matrix.