CN112737650B - MIMO system transmitting antenna selection method based on machine learning - Google Patents

Abstract

The invention discloses a machine learning-based MIMO system transmitting antenna selection method, which relates to the technical field of MIMO and solves the technical problems of high complexity and low efficiency of the existing antenna selection algorithm, and the method comprises the following steps: s1, collecting a plurality of training channel matrixes H_mAnd is combined with H_mNormalizing to obtain X_m(ii) a S2, evaluating each training channel matrix H_mAbout

The receiving end signal-to-noise ratio (or bit error rate, spectrum efficiency, energy efficiency and the like) performance of each antenna combination is obtained, the antenna combination serial number with the optimal receiving end performance is obtained, N is the total number of transmitting antennas, and K is the number of selected antennas; s3, constructing J two types of classifiers, wherein each classifier calculates the probability of selecting the jth antenna combination (J is more than or equal to 1 and less than or equal to J)

If p is_jAnd if the antenna combination is more than or equal to 0.5, selecting the jth antenna combination. The invention can effectively reduce the inference deviation and improve the performance of the signal-to-noise ratio (or bit error rate, spectrum efficiency, energy efficiency and the like) of the receiving end, and has the advantages of lower algorithm complexity, less required training input data and high efficiency.

Description

MIMO system transmitting antenna selection method based on machine learning

Technical Field

The invention relates to the technical field of MIMO, in particular to a method for selecting a transmitting antenna of an MIMO system based on machine learning.

Background

Signals transmitted and received by each antenna in a MIMO (Multi-Input Multi-Output) system need to be processed through one radio frequency link, and as the number of antennas increases, the configuration of the radio frequency link increases, which increases the hardware cost of the MIMO system. The antenna selection algorithm can reduce the complexity and economic cost of equipment and ensure the diversity and multiplexing gains of the MIMO system by selecting the part of the antenna subsets with the best performance from the available antennas for transmitting and receiving signals.

The optimal algorithm is an exhaustive method, which selects the optimal subset by calculating the performance such as channel capacity or bit error rate of all possible antenna subset combinations, and the computational complexity of the algorithm increases dramatically as the number of antennas increases. In real-time channels, the antenna selection algorithm needs to adapt quickly to the environment where the channel state changes frequently, so that some low-complexity antenna selection algorithms are needed.

The MOLISCH provides a rapid selection algorithm based on channel correlation, so that the calculation complexity is greatly reduced, and the channel capacity loss is large; LU et al propose a rapid selection algorithm based on priority, but the multi-parameter and cross-mutation operation of genetic algorithms leaves the algorithm to be improved in terms of balance of complexity and performance.

The common disadvantages of the prior art methods are: (1) the algorithm has higher complexity and low efficiency; (2) the performance of the algorithm is greatly affected by the initial value setting.

Disclosure of Invention

The invention aims to solve the technical problem of the prior art, and aims to provide a machine learning-based MIMO system transmitting antenna selection method with low algorithm complexity and high efficiency.

The technical scheme of the invention is as follows: a MIMO system transmitting antenna selection method based on machine learning comprises the following steps:

s1, collecting a plurality of training channel matrixes H_mM is more than or equal to 1 and less than or equal to M, and H_mNormalizing to obtain X_m；

S2, evaluating each training channel matrix H_mAbout

The receiving end performance of each antenna combination obtains the antenna combination serial number with the optimal receiving end performance, N is the total number of transmitting antennas, and K is the number of selected antennas;

s3, constructing J two types of classifiers, wherein each classifier calculates the probability of selecting the jth antenna combination (J is more than or equal to 1 and less than or equal to J)

If p is_jIf the antenna combination is more than or equal to 0.5, selecting the jth antenna combination,

wherein the decision function f_j(X)＝A_jK^TKernel function K ═ K (X, X)_i)]，K(X,X_i)＝exp{-||X-X_i||²}。

As a further improvement, the coefficient A_jThe specific calculation steps are as follows:

s31, in the 1 st cycle of training the jth classifier

D＝{X_m(m＝1,2,…,M)},

For X_nE, calculating an objective function by using the C \ C:

wherein y ═ y₁,…y_m,…y_M]^TIf training the channel matrix H_mThe antenna combination serial number corresponding to the receiving end with the optimal performance is j, then y_m1, otherwise y_m＝0；

K₁＝[K(X_m,X_p)],X_m∈D,X_p∈C∪{X_n},

K₂＝[K(X_p,X_q)],X_q∈C∪{X_n},

In the k-th cycle:

W＝diag[p_j(X_m)(1-p_j(X_m))],

p＝[p_j(X₁),p_j(X₂),…p_j(X_M)]^T；

s32, finding so that H (X)_n) Minimum size

Order to

S33, calculating H obtained by two continuous circulations_kRelative rate of change of

If R is<δ, stopping the cycle; otherwise, steps S31, S32 are repeated.

Further, δ is 0.001.

Further, the performance of the receiving end is any one of a signal-to-noise ratio, a bit error rate, a spectrum efficiency or an energy efficiency.

Advantageous effects

Compared with the prior art, the invention has the advantages that:

the invention realizes the multi-class classification for transmitting antenna selection by designing the two-class classifier, can effectively reduce the inference deviation and improve the signal-to-noise ratio (or bit error rate, spectrum efficiency, energy efficiency and the like) performance of a receiving end, and has the advantages of lower algorithm complexity, less required training input data and high efficiency.

Drawings

FIG. 1 is a flow chart of the present invention.

Detailed Description

The invention will be further described with reference to specific embodiments shown in the drawings.

Referring to fig. 1, a method for selecting transmit antennas of a MIMO system based on machine learning includes the following steps:

s1, collecting a plurality of training lettersTrack matrix H_mM is more than or equal to 1 and less than or equal to M, and H_mNormalizing to obtain X_m；

S2, evaluating each training channel matrix H_mAbout

The receiving end performance of each antenna combination obtains the antenna combination serial number with the optimal receiving end performance, N is the total number of transmitting antennas, and K is the number of selected antennas;

s3, constructing J two types of classifiers, and calculating the probability of each classifier selecting the jth (J is more than or equal to 1 and less than or equal to J) antenna combination

If p is_jIf the antenna combination is more than or equal to 0.5, selecting the jth antenna combination,

wherein the decision function f_j(X)＝A_jX^TKernel function K ═ K (X, X)_i)]，K(X,X_i)＝exp{-||X-X_i||²}。

Coefficient A_jThe specific calculation steps are as follows:

s31, in the 1 st cycle of training the jth classifier

D＝{X_m(m＝1,2,…,M)},

For X_nE, calculating an objective function by using the C \ C:

wherein y ═ y₁,…y_m,…y_M]^TIf training the channel matrix H_mThe antenna combination serial number corresponding to the receiving end with the optimal performance is j, then y_m1, otherwise y_m＝0；

K₁＝[K(X_m,X_p)],X_m∈D,X_p∈C∪{X_n},

K₂＝[K(X_p,X_q)],X_q∈C∪{X_n},

In the k-th cycle:

W＝diag[p_j(X_m)(1-p_j(X_m))],

p＝[p_j(X₁),p_j(X₂),…p_j(X_M)]^T；

s32, finding so that H (X)_n) Minimum size

Order to

S33, calculating H obtained by two continuous circulations_kRelative rate of change of

If R is<δ, stopping the cycle; otherwise, steps S31, S32 are repeated.

In this embodiment, δ is 0.001. The performance of the receiving end is any one of signal-to-noise ratio, bit error rate, spectrum efficiency or energy efficiency.

The invention realizes the multi-class classification for transmitting antenna selection by designing the two-class classifier, can effectively reduce the inference deviation and improve the signal-to-noise ratio (or bit error rate, spectrum efficiency, energy efficiency and the like) performance of a receiving end, and has the advantages of lower algorithm complexity, less required training input data and high efficiency.

The above is only a preferred embodiment of the present invention, and it should be noted that it is obvious to those skilled in the art that several variations and modifications can be made without departing from the structure of the present invention, which will not affect the effect of the implementation of the present invention and the utility of the patent.

Claims (1)

1. A MIMO system transmitting antenna selection method based on machine learning is characterized by comprising the following steps:

s1, collecting a plurality of training channel matrixes H_mM is more than or equal to 1 and less than or equal to M, and H_mNormalizing to obtain X_m；

S2, evaluating each training channel matrix H_mAbout

The receiving end performance of each antenna combination obtains the antenna combination serial number with the optimal receiving end performance, N is the total number of transmitting antennas, and K is the number of selected antennas;

s3, constructing J two types of classifiers, wherein each classifier calculates the probability of selecting the jth antenna combination (J is more than or equal to 1 and less than or equal to J)

If p is_jIf the antenna combination is more than or equal to 0.5, selecting the jth antenna combination,

wherein the decision function f_j(X)＝A_jK^TKernel function K ═ K (X, X)_i)]，K(X,X_i)＝exp{-||X-X_i||²}；

The coefficient A_jThe specific calculation steps are as follows:

s31, in the 1 st cycle of training the jth classifier

D＝{X_m(m＝1,2,…,M)},

For X_nE.g. D \ C calculation purposeThe standard function is:

wherein y ═ y₁,…y_m,…y_M]^TIf training the channel matrix H_mThe antenna combination serial number corresponding to the receiving end with the optimal performance is j, then y_m1, otherwise y_m＝0；

K₁＝[K(X_m,X_p)],X_m∈D,X_p∈C∪{X_n},

K₂＝[K(X_p,X_q)],X_q∈C∪{X_n},

In the k-th cycle:

W＝diag[p_j(X_m)(1-p_j(X_m))],

p＝[p_j(X₁),p_j(X₂),…p_j(X_M)]^T；

s32, finding so that H (X)_n) Minimum size

Order to

S33, calculating H obtained by two continuous circulations_kRelative rate of change of

If R is<δ, stopping the cycle; otherwise, repeating the steps S31 and S32;

the performance of the receiving end is any one of bit error rate, spectrum efficiency or energy efficiency;

the delta is 0.001.

Images