CN104539339B - Resource allocation methods based on SLNR multi-user's dual-stream beamforming - Google Patents

Abstract

Based on the resource allocation method of SLNR multi-user dual-stream beamforming, the invention aims to achieve the optimal resource under the condition of uncertain interference in the resource allocation of multi-user multi-stream beamforming in the TD‑LTE‑A system distribute. The method: step 1, obtain the channel quality of the user under the condition of not calculating the co-channel interference through the initial channel information, and determine the priority for the user to be allocated according to the channel quality; step 2, the base station assigns RB to the user according to the user priority, and On the RB, determine the grouping by calculating the spatial orientation coefficient η _l,k,n , and comprehensively considering the interference between users and the user channel condition; step 3, after completing the allocation of a certain RB, update and calculate the user rate, and judge whether the target requirement is met , no more resources will be allocated to users who have reached the target rate; step 4, repeat the above two steps until the allocation is complete.

Description

Resource allocation method based on SLNR multi-user dual-stream beamforming

technical field

The invention relates to a resource allocation method based on SLNR multi-user beamforming.

Background technique

In November 2004, related concepts of LTE were proposed in the 3GPP long-term evolution plan working conference in Toronto, Canada. After a series of improvements by 3GPP, the current 3G communication standard system is composed of the three major standards of WCDMA, CDMA2000, and TD-SCDMA and the 802.16 WiMax standard added later. In order to further meet the growing demand for mobile communication and welcome the arrival of the 4G mobile communication era, 3GPP has carried out LTE-FDD and TD-LTE for the two existing 3G standards of WCDMA and TD-SCDMA respectively during the gradual development of LTE. standardized work. In both 3G and 4G systems, MIMO technology is an essential key technology. MIMO technology uses multiple antennas at the receiver and transmitter, combined with some advanced signal processing techniques. In standard versions after R10, there are nine MIMO transmission modes. At present, MIMO transmission modes are mainly divided into three categories: transmit diversity, spatial multiplexing, and beamforming. Traditional beamforming is applied to small-pitch antenna arrays. The strong correlation between antenna elements is used to form beams pointing in the direction of users, thereby improving the signal-to-noise ratio. The currently used beamforming technology is not limited to traditional beamforming , it is not required that there must be a strong correlation between the antenna elements, but the beamforming algorithm is implemented by the precoding beamforming algorithm at the transmitting end to obtain the multiplexing gain. The main beamforming algorithms include ZF (zero forcing), MMSE (minimizing mean square error), BD (block diagonalization), SLNR (signal-to-leakage-to-noise ratio), etc. Among them, the SLNR algorithm comprehensively considers noise and interference factors, effectively suppresses the inter-stream interference of multiple streams, and introduces the concept of "leakage" to represent the interference of a certain user to other users, which greatly reduces the complexity of the algorithm compared with the calculation of SINR. degree, and thus has been the most widely used.

In multi-user beamforming, resource scheduling and allocation is also an important factor that determines system performance. The TD-LTE-A physical layer defines RB (Physical Resource Block) as the division of uplink and downlink physical resources, and is the smallest unit of air interface physical resource allocation. One RB includes 12 consecutive subcarriers, occupies a frequency band of 180 kHz in the frequency domain, and includes 7 consecutive OFDM symbols in the time domain, with a time length of 0.5 ms. If an extended cyclic prefix is used, it is 6 OFDM symbols. Multiple users of beamforming will transmit simultaneously on the same RB to gain gain. Most resource allocation schemes are aimed at optimizing overall system throughput. After multiple users in a cell experience frequency selective fading channels, the fading on each RB is different, and the mutual interference of different users on the same RB is also different. In addition, because multiple users are involved, the interference between users It will also have an impact on the system, and these issues need to be considered in the system-level optimization process. RBs with deep fading of the user's transmission signal and users in the same group with greater interference will affect the user's transmission rate, thereby affecting the overall throughput of the system. Fairness is another important metric in resource allocation algorithms. Fairness refers to the difference between user transmission rates. As a system consideration, the rate difference between different users should not be too large under the premise of meeting user requirements. In addition, the complexity of the algorithm is also an issue that needs to be considered in the scheme design process, and the algorithm must be practically feasible.

Existing algorithms related to the present invention: Most of the current algorithms are mainly based on single-stream beamforming scenarios, which were published in Document 1 in 2009: "Efficient and low-complexity user selection for the multiuser MISO downlink" and Document 2 in 2014 : "Low-Complexity User Selection for Rate Maximization in MIMO Broadcast Channels with Downlink Beamforming" mainly provides a scheme for user selection and resource allocation for multi-purpose single-stream beamforming, and its essence is capacity optimization allocation based on MISO. This algorithm has high system capacity, but lacks consideration of system fairness.

Document 3: "User selection and resource allocation algorithm with fairness in MISO-OFDMA" proposes user grouping to reduce complexity, starting from reducing interference between users, and on this basis, considering user fairness to allocate resources, taking into account throughput. However, this algorithm does not consider the user's channel conditions, and because the restriction on fairness is too strict, there will be a problem that the throughput will decrease when the number of users in the cell increases. At the same time, literatures [1]-[3] still have a common problem, that is, these schemes are all based on the single-stream beamforming scenario, and all use the ZF beamforming algorithm. In the multi-stream scenario, there will be lead to performance degradation. Document 4: "Capacity maximization for zero-forcing MIMO-OFDMA downlink systems with multiuser diversity" considers the multi-stream beamforming scenario, determines the number of streams by receiving space selection, and implements power allocation and user selection according to the design of intermediate variable iterations The rightmost scheme of , and finally give the resource allocation method to maximize the throughput. However, this algorithm ignores the fairness among users in the system. In the actual system, some users will always be in the low-speed transmission state, which will affect the user experience. In addition, the complexity of this scheme is high, and there are great difficulties in realization.

Contents of the invention

The purpose of the present invention is to achieve optimal resource allocation under the condition that the interference cannot be known in the resource allocation of multi-user multi-stream beamforming in the TD-LTE-A system, thereby providing a multi-user dual-stream beamforming based on SLNR Resource allocation method for beamforming.

A resource allocation method based on SLNR multi-user dual-stream beamforming, which is implemented by the following steps:

Step 1. Initialize the user set U and the carrier set Nn, and calculate the average signal-to-noise ratio SINR of each user;

Step 2. Arrange the SNR values of each user in descending order as the priority of users to be allocated, and arrange all users into a queue Ui to be allocated according to the priority;

Step 3: The current user k with the highest priority traverses all resource blocks (RBs), and selects the nth RB such that Add the user k to the user allocation set A _n of RBn, so that A _n = {k}, and remove k from the queue Ui to be allocated, at this time: Ui = Ui-{k};

Step 4. According to the formula on the nth RB:

Obtain the spatial correlation coefficient η _{l, k, n} of other users and existing users on the nth RB, where: η _{l, k, n} represent the correlation coefficient between user k and user l on RBn; H _{l, n} on RBn is the channel matrix of user l, H _{k, n} on RBn is the channel matrix of user k channel matrix;

Step five, according to the formula:

Calculate the average correlation coefficient between other users and existing users on the nth RB Take out the top L average correlation coefficients The smallest user; L is a positive integer; wherein: η _{l, m, n} represent on RBn, the correlation coefficient between user m and user l;

Step six, in the L average correlation coefficient Among the smallest users, judge whether there is a user who can increase the sum rate on the RBn, if the judgment result is yes, then perform step 61; if the judgment result is no, then perform step 62;

Step 61: Select the user m who can increase the sum rate on RBn the most, add user m to the RB, A _n =A _n ∪{m}, and remove user m from the allocation queue, at this time Ui=Ui -{m}; Execute step seven;

Step 62: Terminate the allocation of the RB, and remove the RB from the RB set, Nn=Nn-{n}, use the current An as the final result of user allocation on the RB, and perform step 8;

Step 7. Repeat steps 4 to 6 until the number of users allocated to n reaches the upper limit, and end the allocation of the RB, Nn=Nn-{n}; the number of users allocated on each RB satisfies: In the formula: N _t is the number of transmitting antennas, S is the number of data streams transmitted by a single user;

Step 8: Update the average signal-to-noise ratio SINR value of the user, and calculate the current rate obtained by all users. If there is a user j that satisfies R _j ≥ γ _j , where R _j represents the current rate obtained by all users, and γ _j represents the rate of user j. target rate, remove user j from user set U, U=U-{j};

Step 9: According to the updated user credit U in step 8, re-sort the allocation queue Ui according to the criterion of the maximum average SINR, and repeat 3 to 8 until the RB allocation or user allocation is completed.

The rate obtained by user k on RBn is:

r _{k, n} = (nsym-ncsym) × Qm _{k, n} × nsubcar × coderate _{k, n}

Among them: nsym and ncsym respectively represent the total number of Orthogonal Frequency Division Multiplexing (OFDM) symbols transmitted on one RB and the number of OFDM symbols used for control information in one transmission time interval (TTI), coderate _{k, n} and Qm _{k, n} are the symbol rate obtained by user k on RBn and the number of bits modulated per symbol; nsubcar is the number of subcarriers on one RB.

The present invention aims at multi-user and multi-stream beamforming in an LTE-A system, and proposes a resource allocation scheme under the criterion of minimizing inter-user interference, and comprehensively considers user throughput and fairness. By introducing the spatial correlation coefficient, the scheme of the present invention proposes a reasonable grouping method, effectively solves the co-channel interference processing under the condition that the user interference is unknown, and improves the throughput of the system on the premise of ensuring that the fairness meets the requirements.

Description of drawings

Fig. 1 is a schematic diagram of a multi-user beamforming model;

Fig. 2 is a schematic flow chart of the joint method of grouping and resource allocation in the present invention;

Fig. 3 is a schematic diagram of the throughput comparison between the present invention and the polling scheme that does not consider user grouping, and the user grouping resource allocation scheme based on fairness;

4 is a schematic diagram of a comparison between a method for considering user grouping based on throughput and a method for considering user grouping based on fairness;

FIG. 5 is a schematic diagram of a comparison between a method for considering user grouping based on throughput and a method for considering user grouping based on fairness under different numbers of users;

detailed description

Specific embodiments 1. This specific embodiment is described in conjunction with FIG. 1 and FIG. 2. The resource allocation method based on SLNR multi-user dual-stream beamforming includes the following steps:

Step 1. Obtain the channel quality of the user under the condition of not calculating reasonable interference through the initial channel information, and determine the priority for the user to be allocated according to the channel quality;

Step 2, the base station assigns RBs to the users according to the user priority, and determines the grouping by calculating the spatial orientation coefficients η _{l, k, n} on the RBs, and comprehensively considering the inter-user interference and user channel conditions;

Step 3. After completing the allocation of a certain RB, update and calculate the user rate to determine whether the target requirement is met, and no longer allocate resources to users who have reached the target rate, so as to ensure that more users can have the opportunity to obtain resources and improve the fairness of the system sex;

Step 4. Repeat the above two steps until the allocation is completed, update the user's allocation result, and output the final overall allocation plan.

The present invention conducts research on the resource allocation method based on multi-user multi-stream beamforming under the LTE-A system. The applicable scenarios include not only multi-stream, but also single-stream as a special case of multi-stream. This solution can also be used for scheduling, solving the problem of Most of the current solutions are only suitable for the drawbacks of single-stream beamforming.

The present invention proposes an effective grouping method based on minimizing inter-user interference, and subtly replaces the original inter-user interference calculation by using the spatial correlation coefficient. This method not only effectively reduces the computational complexity, but also solves the problem that the base station calculation is difficult because the co-channel interference cannot be known in the feedback.

Glossary:

MIMO: Multiple-Input Multiple-Output, multiple input multiple output antenna system;

UE: User Equipment, user equipment;

TTI: Transmission Time Interval, transmission time interval;

CQI: Channel Quality Indicator, channel quality indicator;

PMI: Precoder Type Indicator, precoding type indication;

OFDM: Orthogonal Frequency Division Multiplexing, Orthogonal Frequency Division Multiplexing;

SINR: Signal to Interference and Noise Ratio, signal to interference and noise ratio;

SLNR: Signal to Leakage and Noise Ratio, Signal to Leakage and Noise Ratio;

MMSE: Minimum Mean Square Error, minimum mean square error;

ZF: Zero Forcing, zero forcing;

RB: Resource Block, resource block;

TDD: Time Division Duplex, time division duplex;

MCS: Modulation and Coding Scheme, modulation and coding strategy;

Principle: The schematic diagram of multi-user multi-stream beamforming system is shown in Figure 1.

Suppose there are K users in the system, each user is equipped with N _r transmitting antennas, the base station is equipped with N _t transmitting antennas, and the number of data streams transmitted by a single user is S. For user k, its beamforming matrix on RBn can be expressed as

Where w _k,s is the transmission weight of the sth stream data of the kth user at the transmitter. At the base station, the signal undergoes serial-to-parallel conversion and is transmitted through precoding beamforming, and then reaches the receiving end after going through the transmission channel. In the case of equal power distribution, the signal received by user k at the receiving end can be expressed as

is the original transmission data of user k. And H _k is the channel matrix of user k.

Observation formula (2), the first item is user k’s own transmission data, the second item It means the data originally transmitted to other users but received by user k, that is, the interference from other users. This is the new problem faced by multi-user beamforming compared to single-user beamforming, inter-user interference. In the case of a single user, since only one user is transmitting on the same time-frequency resource, there will be no co-channel interference between users. In order to obtain multiplexing gain, multiple users need to transmit data of multiple users on the same frequency. Frequency interference will inevitably appear. Generally speaking, the user will filter the received signal at the receiving end, using the filter matrix:

The rate obtained by user k on RBn can be expressed as

r _k,n =(nsym-ncsym)×Qm _k,n ×nsubcar×coderate _k,n (4)

Among them, nsym and ncsym represent the total number of OFDM symbols transmitted on one RB and the number of OFDM symbols used for control information in one TTI, respectively, and coderate _k,n and Qm _k,n are obtained by user k on RBn The symbol rate and the number of bits modulated per symbol are determined by the MCS (coded modulation strategy) of user k on RBn. nsubcar is the number of subcarriers on one RB, and in the TD-LTE-A system, nsubcar=12.

Here we use the sum rate of users in the system as the throughput of the system. During the time of T simulation subframes, the average rate obtained by user k is

ρ _k,n in the formula is a resource allocation indicator. ρ _k,n =1 indicates that user k is allocated on RBn, and ρ _k,n =0 indicates that user k is not allocated on RBn. The throughput of the system can be expressed as

Although our main goal is to improve the throughput of the system, fairness is another important indicator that cannot be ignored when evaluating system scheduling algorithms. The present invention adopts the Jain's index of the user rate as an index for measuring fairness, and its expression is as follows

The value range of Jain's index F _p is When F _p is closer to 1, it indicates that the fairness of the system is better.

For the Jain's index, there is no strict evaluation standard. In order to describe the fairness requirements of the scheme with a clearer standard, Table 1 shows the requirements of 3GPP for CDF (cumulative distribution function).

Table 1 CDF curve reference point

In the CDF graph, the horizontal axis represents the user's normalized throughput, and the vertical axis represents the cumulative probability distribution value, that is, the proportion of users in the system that is less than or equal to the normalized throughput value. The position where the horizontal axis is 1 represents the average throughput of the system. If the growth of the curve is concentrated around 1, it can be considered that the fairness is better. Taking the data set (0.1, 0.1) in the table as an example, the users whose throughput is less than or equal to the average throughput of the system cannot exceed one-tenth of the total number of users.

It has been mentioned in the analysis of (4) that the number of modulated bits and the symbol rate of transmission are determined by the coding and modulation strategy, and the coding and modulation strategy is related to CQI, and CQI corresponds to SINR. On the one hand, a larger SINR corresponds to a larger SINR. It also leads to larger coderate _k,n and Qm _k,n values, thereby obtaining a larger transmission rate.

At the level of the actual system, when user k is assigned to RBn, its receiving end SINR can be expressed as:

In the formula, Ps represents the power of the useful signal received by user k, ^σ2 is the Gaussian white noise power, _Oc is the interference of neighboring cell base stations, _Ic is the interference between users, and their calculation formulas are:

Among them: E _k is the energy allocated by the base station to user k. Since this chapter does not consider the power allocation problem, the strategy of equal user power allocation is adopted. α _k is the large-scale fading of user k, including shadow fading and path loss in practical systems. G _k,n and W _k,n are the user's receiving matrix and beamforming matrix, respectively. U={1,2,...,K} represents a user set composed of all K users. In particular, when a user transmits more than one-class data, there will also be inter-stream interference between the user's own data streams, that is, I _l in Equation (9). For the s-th stream data of user k, the inter-stream interference The formula for calculating interference is

When the SLNR algorithm is used, the interference and noise of multi-stream beamforming will be better suppressed. For the improvement of SINR in the scheduling process, since O _c and σ ² have little relationship with the scheduling algorithm, and the inter-stream interference I _l It also exists in the user and cannot be eliminated alone, so we start with reducing _Ic and increasing _Ps . Through the allocation relationship between resources and user groups, the interference between users can be reduced; on different RBs, the user's H _k,n has different values, and its F norm will also be different. When the channel condition is good, The user's H _{k and n} have a large norm value, so the value of P _s will also be large. Although this will invisibly amplify the interference, but because the SLNR algorithm itself has the function of suppressing interference, the final P _s caused by RB allocation The impact brought by the increase will be far greater than the impact brought by the increase in interference.

However, in the allocation process, the interference between users cannot be known under the condition that the allocation scheme is not determined, which brings difficulty to the determination of the scheme. The optimal scheme can be obtained through traversal search, but this algorithm has too much computation and no practical operation possibility. To this end, we introduce the concept of the number of spatial correlations between users:

η _{l, k, n} represent the correlation coefficient between user k and user l on RBn. The spatial correlation coefficient represents the strength of mutual interference between user spatial channels. The larger the spatial correlation coefficient, the stronger the interference between users, the greater the power required to eliminate interference, and the obtained The lower the data rate. The introduction of η subtly solves the problem of difficult calculation of inter-user interference. Inter-user interference can be well suppressed by minimizing η, which greatly reduces the complexity compared to exhaustive search. Under this simple grouping strategy, combined with the resource allocation scheme of the maximum carrier-to-interference ratio idea, the SINR at the receiving end can be greatly improved, and its simplicity is much better than the power iteration scheme in Document 4.

Based on the above analysis, the entire flow chart of our scheme is given, as shown in Figure 2.

Among them, γ _k represents the target rate of user k, if the rate of a certain user has reached the requirement of its target rate, it will stop allocating resources for this user. In the algorithm, the initial user set U={1,2,...K}, the resource block set Nn={1,2,...N} are firstly required, the initial SINR is calculated, and the users are prioritized. The main purpose of optimization is to improve throughput, so sorting is performed according to the user's channel conditions. Here we express the channel condition in the following form

That is, the average signal-to-interference-noise ratio of k on all RBs is used to characterize the user channel quality. Since the final user is allocated to a specific resource block, the SINR at the receiving end is a specific value, but the average SINR value of the user on the RB used can effectively represent the overall channel condition of the user. inferior. Although this method is not absolutely accurate in principle, due to the large differences in large-scale fading coefficients among different users, the SINR difference of the same user on different RBs is much smaller than the SINR difference of different users on the same RB. It means that users with better average channel conditions will have higher SINR on each RB. This sorting priority calculation method avoids the complexity of exhaustion, and as a suboptimal method, it can measure the channel conditions more accurately. The subsequent simulation results also prove that this method is feasible.

After determining the index of channel quality indication, the next step is the joint algorithm of grouping and resource allocation. The specific process is as follows:

(1) sort the users in U, and arrange all users into a waiting queue Ui according to the order of priority;

(2) The first user in Ui, that is, user k with the best channel conditions, traverses all RBs first, and picks out the nth RB such that Add k to the user allocation set A _n of RBn, so that A _n = {k}, and remove k from the queue Ui to be allocated, Ui = Ui-{k};

(3) Calculate the spatial correlation coefficients between other users and existing users on n according to (12)

(4) Calculate the average correlation coefficient between other users and existing users on n Take out the first L smallest user;

(5) Among the L users, select the user m who can increase the sum rate on the RBn the most, add m to the RB, A _n =A _n ∪{m}, and remove m from the allocation queue, Ui=Ui-{m}, but if no user can be found and the rate on this RB continues to increase, then the allocation of this RB is terminated and removed from the RB set, Nn=Nn-{n}, the current An is the final result of user allocation on the RB, skip (6) and go directly to step (7);

(6) Repeat (3) to (5) until the number of users allocated for n reaches the upper limit, and the number of users allocated on each RB should satisfy After the number of users reaches the upper limit, end the allocation of the RB, Nn=Nn-{n};

(7) Update the SINR value of the user, and calculate the current rate obtained by all users. If there is R _j ≥ γ _j , where R _j represents the current rate obtained by all users, and γ _j represents the target rate of user j, then U=U -{j}, note that j is removed from the total user set U;

(8) According to the new U, reorder the allocation queue Ui according to the criterion of the maximum average SINR, and repeat the steps (2)-(7) until the RBs are allocated or the users are allocated.

The setting of L is to balance the consideration of channel conditions and interference during the grouping process. Although the general principle of grouping is to minimize mutual interference, sometimes users with less interference are not the best for improving RB and rate, because although the interference is small, their channel conditions may be poor, and their own rate is not very high . In order to avoid this situation, slack is set when grouping. In formula (4), instead of directly selecting users with the least correlation, L users with very little interference are selected, and then we can see which user joins It can bring the highest RB and rate, which effectively improves the system throughput. There are different methods for the setting of L in different literatures, and the following form is used here:

L=min{card(U _i ), N _t /S} (14)

In this method, the grouping and resource allocation are combined, through the mutual traversal of RBs and users, not only the interference between users is noticed in the grouping process, but also the channel conditions of users on RBs are considered, plus the priority based on channel conditions Sort, so that the allocation method can achieve better throughput. At the same time, we set that users will no longer participate in resource allocation after reaching the target rate, which ensures the fairness of the system to a certain extent.

Beneficial effects brought by the technical solution of the present invention: the present invention is aimed at multi-user multi-stream beamforming in the LTE-A system, and proposes a resource allocation scheme under the criterion of minimizing inter-user interference, comprehensively considering user throughput quantity and fairness. By introducing the spatial correlation coefficient, the scheme of the present invention designs a reasonable grouping method, effectively solves the co-channel interference processing under the condition that the user interference is unknown, and improves the throughput of the system on the premise of ensuring that the fairness meets the requirements.

Figure 3 shows the throughput of the scheme. It can be seen from the figure that the method proposed in the present invention has obvious advantages in throughput. When the number of users is small, the throughput of the three algorithms is almost the same; as the number of users increases, the throughput of the method proposed by the present invention is always the highest, and the advantages are more obvious, and the throughput of the polling algorithm is always lower than that of the present invention method, and the algorithm given in Document 3 does not consider the user channel conditions, only pays attention to the fairness to study the user grouping method, and then keeps the lowest in throughput. It has almost reached 2 times of the algorithm based on fairness grouping. At the same time, it is found that when the number of users continues to increase, the throughput of the algorithm in Document 3 decreases instead, and the method we invented effectively solves this problem.

Figure 4 shows the Jain's index comparison of fairness. It can be found that the fairness of the proposed scheme is excellent in most scenarios. The Jain's index is higher than 0.85. When the number of users exceeds 80, it is lower than 0.85 in some scenarios, but it is still higher than 0.8. Fairness Good, and we combine the CDF curve in Figure 5, we can see that under the conditions of 80 and 90 users, the fairness of the proposed scheme meets the fairness requirements given by 3GPP. Therefore, it can be said that under the premise of excellent throughput performance, this solution still ensures fairness that meets the requirements.

The advantages of the present invention also include greatly reducing algorithm complexity. If the traversal search scheme is adopted, the final calculation amount is O(K ^(Nt/S)N ), and if the solution of Document 4 is adopted, the calculation amount is O(NK ^(Nt/S) ). The present invention is directly based on the optimal user group allocation on RBs, directly calculates the appropriate user combination for each RB by calculating the spatial correlation coefficient, avoids other combination attempts, and effectively reduces complexity. The complexity of the present invention is only O(NK(N _t /S)).

Claims (2)

1. based on the resource allocation method of SLNR multi-user dual-stream beamforming, it is characterized in that: it is realized by the following steps: Step 1. Initialize the user set U and the carrier set Nn, and calculate the average signal-to-noise ratio SINR of each user; Step 2. Arrange the SNR values of each user in descending order as the priority of users to be allocated, and arrange all users into a queue Ui to be allocated according to the priority; Step 3: User k with the highest priority currently traverses all resource block RBs, and selects the nth RB such that Add the user k to the user allocation set A _n of RBn, so that A _n = {k}, and remove k from the queue Ui to be allocated, at this time: Ui = Ui-{k}; Step 4. According to the formula on the nth RB: <mrow><msub><mi>&amp;eta;</mi><mrow><mi>l</mi><mo>,</mo><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>=</mo><mfrac><mrow><mo>|</mo><msub><mi>H</mi><mrow><mi>l</mi><mo>,</mo><mi>n</mi></mrow></msub><msup><mrow><mo>(</mo><msub><mi>H</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>)</mo></mrow><mi>H</mi></msup><mo>|</mo></mrow><mrow><mo>|</mo><mo>|</mo><msub><mi>H</mi><mrow><mi>l</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>|</mo><msub><mo>|</mo><mi>F</mi></msub><mo>|</mo><mo>|</mo><msub><mi>H</mi><mrow><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>|</mo><msub><mo>|</mo><mi>F</mi></msub></mrow></mfrac><mo>,</mo><mn>0</mn><mo>&amp;le;</mo><msub><mi>&amp;eta;</mi><mrow><mi>l</mi><mo>,</mo><mi>k</mi><mo>,</mo><mi>n</mi></mrow></msub><mo>&amp;le;</mo><mn>1</mn></mrow> Obtain the spatial correlation coefficient η _l,k,n of other users and existing users on the nth RB, where: Indicates the correlation coefficient between user k and user l on RBn; H _{l, n} on RBn is the channel matrix of user l, and H _{k, n} on RBn is the channel matrix of user k; Step five, according to the formula: <mrow><mover><msub><mi>C</mi><mi>m</mi></msub><mo>&amp;OverBar;</mo></mover><mo>=</mo><mfrac><mrow><munder><mi>&amp;Sigma;</mi><mrow><mi>l</mi><mo>&amp;Element;</mo><msub><mi>A</mi><mi>n</mi></msub></mrow></munder><msub><mi>&amp;eta;</mi><mrow><mi>l</mi><mo>,</mo><mi>m</mi><mo>,</mo><mi>n</mi></mrow></msub></mrow><mrow><mi>c</mi><mi>a</mi><mi>r</mi><mi>d</mi><mrow><mo>(</mo><msub><mi>A</mi><mi>n</mi></msub><mo>)</mo></mrow></mrow></mfrac></mrow> Calculate the average correlation coefficient between other users and existing users on the nth RB Take out the top L average correlation coefficients The smallest user; L is a positive integer; wherein: η _{l, m, n} represent on RBn, the correlation coefficient between user m and user l; The setting of L is to balance the consideration of channel conditions and interference during the grouping process; the setting of L takes the following form: L=min{card(U _i ), N _t /S} In the formula: N _t is the number of transmitting antennas, S is the number of data streams transmitted by a single user; In this method, the grouping and resource allocation are combined, through the mutual traversal of RBs and users, not only the interference between users is noticed in the grouping process, but also the channel conditions of users on RBs are considered, plus the priority based on channel conditions Sorting, so that the allocation method can achieve better throughput; at the same time, set the user to no longer participate in resource allocation after reaching the target rate; Step six, in the L average correlation coefficient Among the smallest users, judge whether there is a user who can increase the sum rate on the RBn, if the judgment result is yes, then perform step 61; if the judgment result is no, then perform step 62; Step 61: Select the user m who can increase the sum rate on RBn the most, add user m to the RB, A _n =A _n ∪{m}, and remove user m from the allocation queue, at this time Ui=Ui -{m}; Execute step seven; Step 62: Terminate the allocation of the RB, and remove the RB from the RB set, Nn=Nn-{n}, use the current An as the final result of user allocation on the RB, and perform step 8; Step 7. Repeat steps 4 to 6 until the number of users allocated to n reaches the upper limit, and end the allocation of the RB, Nn=Nn-{n}; the number of users allocated on each RB satisfies: Step 8: Update the average signal-to-noise ratio SINR value of the user, and calculate the current rate obtained by all users. If there is a user j that satisfies R _j ≥ γ _j , where R _j represents the current rate obtained by all users, and γ _j represents the rate of user j. target rate, remove user j from user set U, U=U-{j}; Step 9: According to the updated user credit U in step 8, re-sort the allocation queue Ui according to the criterion of the maximum average SINR, and repeat 3 to 8 until the RB allocation or user allocation is completed. 2. The resource allocation method based on SLNR multi-user dual-stream beamforming according to claim 1, wherein the rate that user k obtains on RBn is: r _{k, n} = (nsym-ncsym) × Qm _{k, n} × nsubcar × coderate _{k, n} Among them: nsym and ncsym respectively represent the total number of OFDM symbols transmitted on one RB and the number of OFDM symbols used for control information within a transmission time interval TTI, coderate _{k, n} and Qm _{k , n} are the symbol rate obtained by user k on RBn and the number of bits modulated per symbol; nsubcar is the number of subcarriers on one RB.