CN103873197B - The 3D MIMO Limited Feedback overhead reduction methods that spatial coherence is combined with sub-clustering - Google Patents

Abstract

The invention relates to the technical field of wireless communication, and requests protection of a 3D MIMO limited feedback overhead reduction method combining channel space correlation and clustering. In the 3D MIMO limited feedback system, the base station adopts Uniform antenna array, the user end uses a linear array, and the user end feeds back the optimal precoding matrix index (PMI) for the horizontal and vertical dimensions based on the spatial correlation of the channel and the co-phase characteristics of the vertical channel of the 2D antenna. When each vertical dimension channel has the same horizontal phase, only one PMI is fed back in the horizontal dimension, and the vertical dimension is clustered, and each cluster feeds back a PMI. When feeding back, an auxiliary bit is added to the horizontal dimension and vertical dimension PMI respectively, and the base station finds Corresponding vertical dimension and horizontal dimension precoding matrix, the horizontal dimension and vertical dimension precoding matrix are expanded and dot-multiplied to obtain a 3D precoding matrix. The method can effectively reduce the feedback overhead, has low complexity and is easy to implement.

Description

3D MIMO Finite Feedback Overhead Reduction Method Combining Spatial Correlation and Clustering

technical field

The present invention relates to the technical field of wireless communication, and relates to the technical field of long-term evolution of mobile communication (LTE-Advanced).

Background technique

In wireless mobile communication, the next generation 4G communication system is required to provide higher data rate and better service quality. In order to improve system capacity and spectrum efficiency, companies such as Nokia Siemens Networks, Aran, DOCOMO, Qualcomm, and Vodafone established the ARTIST4G working group to study the advanced radio interface technology (Advanced Radio Interface Technologies for 4G SysTems, ARTIST4G) of 4G systems. Among them, 3D multi-antenna Technology is the key technology to improve system capacity and spectrum efficiency, which has attracted the industry's deep attention. Compared with traditional 2D MIMO, 3D MIMO adds an available dimension in the vertical dimension, and adopts a 2D array antenna structure at the base station. In this way, the spectral efficiency of the horizontal dimension and the vertical dimension is effectively improved.

In the 3D MIMO limited feedback system, by constructing a horizontal dimension and vertical dimension codebook known by both the transmitting end and the receiving end, the receiving end uses the horizontal dimension and vertical dimension channel state information (CSI), according to a certain optimization criterion in the pre-designed The optimal precoding matrix is selected in the codebook, and then the horizontal dimension and vertical dimension PMI are respectively fed back to the sending end, and the sending end performs some calculation processing according to the received horizontal dimension and vertical dimension PMI, and finally forms a 3D precoding matrix, which is Fundamentals of codebook-based finite feedback in 3D MIMO systems.

In a limited feedback system, due to the limited bandwidth of the uplink feedback link, 3D MIMO increases the available dimension in the vertical dimension, so that the feedback overhead will increase compared with 2D MIMO. Therefore, how to effectively Reducing the feedback overhead of 3D MIMO becomes the key to the commercialization of 3D MIMO technology.

Contents of the invention

In view of the above technical problems, the present invention proposes a 3D MIMO limited feedback overhead reduction method that combines spatial correlation and clustering. As long as the vertical channel is clustered within the coherent distance, it can not only effectively reduce the feedback overhead , and the system performance loss is small.

In the spatial multipath fading channel, the spatial correlation has a great influence on the performance of the system. There are two main factors affecting the strength of the spatial correlation: the antenna spacing and the size of the angle spread (AS). When the angle spread is fixed, the larger the antenna spacing, the smaller the correlation, and the smaller the antenna spacing, the larger the correlation. In general, the coherence distance D _c can be used to measure the correlation between antennas. The coherence distance generally takes the value when the correlation coefficient is 0.5. If the distance between the antennas is smaller than the coherence distance, the antennas are considered to be correlated. If the distance between the antennas is greater than the coherence distance, it is considered Antennas are not correlated. To this end, spatial correlation can be used to improve the spectral efficiency of the 3D MIMO system.

A 3D MIMO limited feedback overhead reduction method combining spatial correlation and clustering, the main implementation steps are as follows:

Step 1: The base station uses an N*M antenna array, and the receiving end uses a uniform linear array to feed back the horizontal and vertical dimensions of 3D MIMO respectively. Since the vertical dimension channel has the same horizontal phase, the horizontal dimension channel selects an optimal precoding matrix with the maximum SNR in the horizontal dimension according to the SNR maximization criterion, and records the corresponding optimal precoding matrix index PMI( horizontal dimension PMI);

Step 2: Group the channels in the vertical dimension into clusters within the coherence distance, and each cluster selects an optimal precoding matrix with the minimum bit error rate in the vertical dimension according to the criterion of minimizing the bit error rate, and records the corresponding optimal precoding matrix Index PMI (vertical dimension PMI);

Step 3: Add an auxiliary bit to the horizontal dimension PMI and the vertical dimension PMI respectively, and feed it back to the base station;

Step 4: The base station restores the optimal precoding matrix in the horizontal dimension and the vertical dimension according to the PMI fed back in the horizontal dimension and the vertical dimension, and then expands them respectively, and then performs a dot multiplication operation on the expanded matrix to obtain a 3D precoding matrix;

Both vertical and horizontal dimensions of 3D MIMO use independent limited feedback mechanisms. When feeding back the PMI, auxiliary bits need to be added to distinguish the vertical dimension from the horizontal dimension. According to the received horizontal dimension and vertical dimension PMI, the base station finds out the corresponding horizontal dimension and vertical dimension precoding matrix in the codebook set through decoding, and expands it.

According to the formula: Select the optimal precoding matrix for the k-th cluster of the vertical dimension channel, where represents the i-th vertical channel information, Ω={w ₁ ,w ₂ ,...w _s ...w _S } represents the codebook set, p represents the total power of the transmitting antenna, and N ₀ represents the noise power. Add bit 0 before the horizontal dimension PMI, and feed it back to the base station. Divide the m vertical dimension channels into one cluster, add bit 1 before the vertical dimension PMI, and feed it back to the base station. When the base station uses an 8*8 antenna array and the receiver uses 2 antennas, the coherence distance at the base station is 2.2λ, and the 4 vertical channels can be divided into a cluster.

In the traditional MIMO system, the precoding technology is only aimed at the horizontal direction, and the present invention comprehensively utilizes the channel information of the electromagnetic wave in the horizontal direction and the vertical direction. In this way, both the horizontal dimension and the vertical dimension gains can be utilized simultaneously to improve the system utility. In the vertical direction, since the channel responses have a strong correlation when the antenna spacing is small, the present invention proposes a 3D MIMO limited feedback overhead reduction method that combines spatial correlation and clustering. , cluster the vertical channel, and then each cluster feeds back a PMI, which can effectively reduce the feedback overhead.

Description of drawings

Fig. 1 is a system block diagram of the 3D MIMO codebook-based limited feedback method of the present invention;

Fig. 2 is a schematic diagram of the method for reducing the limited feedback overhead of the present invention;

Fig. 3 simulation of base station end spatial correlation under indoor NLOS environment of the present invention;

Fig. 4 is the comparison diagram of the number of feedback bits of the present invention;

Fig. 5 is a comparison chart of system performance simulation of the present invention.

detailed description

In the limited feedback scheme of the 3D MIMO system, the optimal precoding index is fed back in the horizontal dimension and the vertical dimension respectively, and the base station uses the fed-back index to start the codebook set Ω={w ₁ ,w ₂ ,...w _s ...w _S } to find the corresponding optimal precoding matrix, after expansion and dot multiplication, the 3D precoding matrix can be obtained, which can effectively reduce the system overhead.

FIG. 1 is a system block diagram of the codebook-based limited feedback method for 3D MIMO proposed by the present invention. Assume that the base station uses N*M array antennas, where N represents the number of rows of the antenna array, M represents the number of columns of the antenna array, and the number of antennas at the receiving end is N _r . The input data stream is divided into L parallel For the convenience of calculation, we assume that L=1. Then, L parallel substreams are matched to the base station array antenna through precoding and sent out. The output signal can be expressed as:

where G _ZF represents the zero-forcing detection matrix, p represents the total power of the transmitting antenna, H represents the 3D channel matrix, W _3D represents the corresponding 3D precoding matrix, x represents the parallel data stream, n ₀ follows the mean value to zero, and the variance is N ₀ additive white Gaussian noise.

The receiving end estimates the channel to obtain channel information H:

Further, decompose H to obtain the channel information of the jth horizontal dimension:

and the i-th vertical dimension channel information:

After receiving the signal, the receiving end adopts the zero-forcing detection algorithm, and the corresponding zero-forcing detection linear transformation matrix is:

The corresponding received signal-to-noise ratio can be expressed as:

In a finite feedback system, the codebook set Ω includes a total of S codewords {w ₁ ,w ₂ ,...w _s ...w _S }, and the receiving end utilizes the SNR maximization criterion according to the horizontal dimension CSI at Select the optimal precoding matrix from the codebook set, use the bit error rate minimization criterion according to the vertical dimension CSI, each cluster selects the optimal precoding matrix from the codebook set, and then pass the horizontal dimension and vertical dimension PMI through additional auxiliary After the bits are fed back to the base station, the base station restores the optimal precoding matrix in the horizontal and vertical dimensions according to the PMI fed back in the horizontal and vertical dimensions, and then expands them respectively, and then uses the point multiplication operation of the expanded matrix to obtain a 3D precoding matrix. Coding matrix W _3D .

As shown in Figure 2, it is a schematic diagram of the limited feedback overhead reduction method proposed by the present invention, and its specific steps are as follows:

(1) According to the horizontal dimension CSI, search the optimal precoding matrix w ^h satisfying the maximization of SNR in the codebook set Ω={w ₁ ,w ₂ ,...w _s ...w _S },

in Indicates the jth (j=1, 2...N) horizontal dimension channel information.

(2) Cluster the vertical-dimensional channels according to the spatial correlation. Within the coherence distance, the channels have a strong correlation. Therefore, according to the spatial correlation, the m vertical-dimensional channels are divided into one cluster. The larger the value of m , the smaller the feedback overhead, but the cost is the loss of system performance, the smaller the value of m, the better the system performance, but the cost is the increase of feedback overhead, so when clustering, a compromise should be made between the two .

(3) Select the optimal precoding matrix for the kth cluster of the vertical dimension channel according to the bit error rate minimization criterion:

in, Indicates the i-th vertical channel information, Ω={w ₁ ,w ₂ ,...w _s ...w _S } represents the codebook set, p represents the total power of the transmitting antenna, N ₀ represents the noise power, m is the The number of vertical dimension channels in a family.

(4) Add an auxiliary bit to the horizontal dimension PMI and vertical dimension PMI respectively, and feed it back to the base station, for example, add bit 0 before the horizontal dimension PMI, and add bit 1 before the vertical dimension PMI.

(5) The base station restores the optimal precoding matrix in the horizontal and vertical dimensions according to the PMI fed back in the horizontal and vertical dimensions, and then expands them respectively: W _h =[w ^h ;w ^h ;...w ^h ] _{(N *M)*L} . Among them, w ^h represents the horizontal dimension precoding matrix. The vertical dimension precoding matrix can be expanded into the following form according to the number of channel information contained in the cluster. The vertical dimension precoding matrix includes the number of precoding matrices of each group equal to the number m of vertical dimension channels in each group. which is:

When m=2:

When m=4:

Among them, W _v represents the vertical dimension precoding matrix, Represents the precoding matrix of the first cluster in the vertical dimension, Represents the precoding matrix of the second cluster in the vertical dimension, Represents the precoding matrix of the M/2th cluster in the vertical dimension, Represents the precoding matrix of the M/4th cluster in the vertical dimension, and L is the number of substreams of the input data.

(5) Apply dot multiplication to the extended horizontal and vertical optimal precoding matrices to obtain a 3D precoding matrix: W _3D =W _h .*W _v

FIG. 3 shows the spatial correlation simulation of the present invention in an indoor NLOS environment. In order to facilitate the analysis of the influence of the antenna spacing on the spatial correlation, the angle extension in the present invention is set to a fixed value, i.e. AS=5°. During the simulation, it is assumed that the antenna spacing of the base station and the receiving end is 0.5λ, and the base station uses 8*8 The antenna array uses 2 receiving antennas at the receiving end. The normalized spatial correlation between different antennas at the base station is expressed as:

Among them, ρ represents the spatial correlation, Δd represents the vertical antenna spacing at the base station, Δd _n represents the horizontal antenna separation at the base station, Δd _u represents the antenna separation at the receiving end, and τ represents the time delay between different antennas, with Represent the channel response of the two antennas at the receiving end, and σ represents the standard deviation of the channel response. We can see from the figure that the coherence distance at the base station is about 2.2λ. Therefore, the 4 vertical dimensional channels can be divided into a cluster, that is, m≤4 is optimal.

FIG. 4 is a comparison diagram of the number of feedback bits in the present invention, where the number of feedback frames is 100. It can be seen from the figure that within the coherence distance, the feedback overhead when m=4 and m=2 is much less than the feedback overhead of the vertical dimension ideal feedback, and at the same time, the feedback overhead when m=4 is lower than that of m=2 Therefore, within the coherent distance, the larger the value of m, the smaller the feedback overhead.

FIG. 5 is a comparison diagram of system performance simulation of the present invention. It can be seen from the figure that the smaller the value of m is, the closer the bit error rate performance is to the vertical-dimensional ideal full feedback. When m=4, the performance of the bit error rate is slightly worse than that of m=2, but it can be seen from FIG. 4 that the corresponding feedback overhead is greatly reduced, although the performance is lost. For practical applications, due to the limited bandwidth of the uplink feedback link, this performance loss is worthwhile. The present invention proposes a 3D MIMO limited feedback overhead reduction method combining channel space correlation and clustering. The method can effectively reduce feedback overhead, has low complexity, and is relatively easy to implement.

Claims (6)

1. A3D MIMO limited feedback overhead reduction method combining spatial correlation and clustering is characterized by comprising the following steps: the base station end adopts an N-M antenna array, the receiving end adopts a uniform linear array, and the horizontal dimension and the vertical dimension of the 3D MIMO are respectively fed back; selecting a horizontal dimension optimal precoding matrix when the signal-to-noise ratio is maximum, and recording a corresponding horizontal dimension PMI; clustering vertical dimension channels within a coherent distance, selecting an optimal precoding matrix in each cluster according to a bit error rate minimization criterion, and recording a corresponding vertical dimension PMI; respectively adding an auxiliary bit to the horizontal dimension PMI and the vertical dimension PMI, and feeding back the auxiliary bits to a base station end; and the base station side recovers the optimal precoding matrix of the horizontal dimension and the vertical dimension according to the fed back horizontal dimension and vertical dimension PMI, and performs point multiplication operation after expanding the optimal precoding matrix respectively to obtain a 3D precoding matrix.

2. The method of claim 1, wherein in the horizontal dimension, according to the formula:in codebook set Ω ═ w₁,w₂,...w_s...w_SSearching for the optimal precoding matrix w meeting the maximization of the signal-to-noise ratio^hWhereinrepresents the j (j ═ 1,2.. N) th horizontal-dimensional channel information.

3. The method according to claim 1, wherein selecting the optimal precoding matrix according to the criterion of minimizing the error rate specifically comprises: according to the formula:selecting an optimal precoding matrix of a kth cluster of a vertical-dimension channel; wherein,representing the ith vertical channel information, p representing the total power of the transmitting antenna, N₀Representing noise power, m is the number of channels in each cluster in vertical dimension, and Ω ═ w₁,w₂,...w_s...w_SDenotes a codebook set.

4. The method of claim 1, wherein adding one auxiliary bit to each of the horizontal dimension PMI and the vertical dimension PMI is specifically: bit 0 is added before the PMI in the horizontal dimension, m vertical dimension channels are divided into a cluster, and bit 1 is added before the PMI in the vertical dimension.

5. The method of claim 1, wherein the expanding the optimal precoding matrices in the horizontal and vertical dimensions respectively comprises: according to the formula W_h＝[w^h；w^h；...w^h]_(N*M)*LExtending a horizontal dimension optimal precoding matrix, wherein w^hRepresenting a horizontal dimension precoding matrix; the vertical dimension precoding matrix is: when m is 2:when m is 4:wherein L represents the number of parallel sub-streams; w_vA vertical-dimensional precoding matrix is represented,a precoding matrix representing the 1 st cluster in the vertical dimension,a precoding matrix representing the 2 nd cluster in the vertical dimension,a precoding matrix representing the M/2 th cluster in the vertical dimension,and representing the precoding matrix of the M/4 th cluster in the vertical dimension, wherein M is the number of channels in the vertical dimension in each cluster.

6. The method of claim 3, wherein when the base station uses 8 x 8 antenna array, the receiving end uses 2 antennas, and the coherence distance of the base station is 2.2 λ, the 4 vertical channels are divided into a cluster.