KR102228091B1 - Apparatus and method for hybrid beamforming of millimeter wave massive mimo systems - Google Patents

Abstract

The present technology discloses a hybrid beamforming apparatus and method for a large-scale MIMO system based on millimeter wave. According to a specific example of the present technology, singular value decomposition of a low-dimensional effective channel matrix is performed in the baseband domain to derive a singular vector of a data stream, and a digital beam is generated and combined with the derived singular vector, and incremental continuous in the RF domain. As a data stream is restored by retrieving a subset of the array response vectors from the dictionary matrix through a selection technique, generating and combining analog beams with the retrieved array response vectors, hybrid beamforming is performed by separating into analog domains and digital domains. The complexity can maximize the spectral efficiency, thereby improving the performance of the system.

Description

Hybrid beamforming device and method for millimeter wave-based large-scale MIMO system {APPARATUS AND METHOD FOR HYBRID BEAMFORMING OF MILLIMETER WAVE MASSIVE MIMO SYSTEMS}

The present invention relates to a hybrid beamforming apparatus and method for a millimeter wave-based large-scale MIMO system, and more particularly, by separating a digital domain and an analog domain to perform beamforming, it is possible to improve spectral efficiency with low computational complexity. It is about a technology that can improve the performance of the system.

Since the frequency resources of 6 GHz or less currently used in 4G cellular systems are insufficient to handle the exponentially increasing data traffic, millimeter wave (30 GHz to 300 GHz band) communication technology that can provide broadband frequency resources is the next generation mobile. It is in the spotlight as one of the candidate technologies for communication systems.

Since the millimeter wave channel suffers high path attenuation due to its frequency characteristics, a beamforming technique that can obtain a signal-to-noise ratio (SNR) gain in a desired direction using a large number of antennas is required. It is possible to integrate the antenna.

Therefore, in the next-generation mobile communication using millimeter wave channels such as 5G mobile communication, hybrid beamforming that combines digital beamforming and analog beamforming is used. In this case, the flexibility of digital beamforming and multi-layer transmission and analog beamforming are used. By combining the ease of implementation, the number of antennas is efficiently increased to implement a large-scale multi-antenna system.

In order to perform hybrid beamforming in such a millimeter wave-based large-scale MIMO system, an optimal RF precoder and RF combiner are designed to select an array response vector corresponding to a channel propagation path having the highest gain from a dictionary matrix.

At this time, in order to improve the performance of the system in the transmission of multiple data streams, an orthogonal matching pursuit (OMP)-based algorithm has been recently developed. The OMP-based algorithm selects an array response vector that minimizes the Frobenius norm of the difference between the optimal precoder and combiner and the entire hybrid precoder and combiner.

However, the system using the OMP-based algorithm still has high computational complexity since it is necessary to select the array response vector through a complete search technique for all possible combinations of the array response vectors.

Accordingly, in terms of spectral efficiency, high computational complexity and relatively low performance are encountered, and further improvement is required.

Accordingly, the applicant performs hybrid beamforming by separating the analog domain and the digital domain, but forms an analog beam with the array response vector of the prior matrix matching the index of the effective channel path having the maximum spectral efficiency in the RF domain. Singular value decomposition (SVD) for the low-dimensional effective channel derived by the product of the channel matrix and the matrix of the RF precoder in the band domain to derive a singular vector and form a digital beam with the derived singular vector. I would like to propose a technology.

The present invention for solving this problem is that in performing hybrid beamforming by separating the analog domain and the digital domain, it is possible to derive a solution to the problem of maximizing spectral efficiency with low computational complexity, thereby improving the performance of the system. It is an object of the present invention to provide a hybrid beamforming apparatus and method for a large-scale MIMO system based on millimeter wave that can be used.

The object of the present invention is not limited to the above-mentioned object, and other objects and advantages of the present invention that are not mentioned can be understood by the following description, and will be more clearly understood by the embodiments of the present invention. In addition, it will be easily understood that the objects and advantages of the present invention can be realized by the means shown in the claims and combinations thereof.

Hybrid beamforming apparatus of a millimeter wave-based large-scale MIMO system according to an embodiment of the present invention for achieving the above object,

A transmitter for beamforming by forming a digital beam for the data stream based on a baseband precode and then forming an analog beam through an RF precoder; And

Including a receiver for restoring the data stream by analog-combining the beam transmitted from the transmitter based on an RF combiner and then digitally combining the beam received from the transmitter through a baseband combiner,

The transmitter is

It is provided to generate an analog beam and a digital beam by separating it into an RF domain and a baseband domain,

The receiver

The generated beam may be separated into an RF domain and a baseband domain to perform analog combining and digital combining to restore a data stream.

Preferably the transmitter,

A digital beam is generated as a singular vector by performing singular value decomposition of the low-dimensional effective channel matrix in the baseband domain,

It may be provided to retrieve a subset of the array response vectors from the dictionary matrix through an incremental continuous selection technique in the RF domain and to generate an analog beam with the retrieved array response vectors.

Preferably the receiver,

In the RF domain, a subset of the array response vector is retrieved from the dictionary matrix through an incremental continuous selection technique, and the received beams are analog-combined with the retrieved array response vector,

It may be provided to derive a singular vector of a data stream by performing singular value decomposition of a low-dimensional effective channel matrix in the baseband domain, and to digitally combine the derived singular vector.

Preferably the RF precoder is

Set the antenna array vector as the spectral capacity for each of a predetermined number of channel paths,

Selecting the index of the array response vector that has contributed the most to the spectral capacity, adding the selected index to the set of pre-selected indices, and removing the selected index from the entire set is repeated,

When the number of repetitions reaches the number of RF chains of the transmitter, a column of a dictionary matrix corresponding to the selected index may be selected as an array response vector, and an analog beam may be formed using the selected array response vector.

Preferably the RF combiner,

The effective channel matrix is derived by the product of the channel matrix and the matrix of the RF precoder,

Set the antenna array vector of the derived effective channel matrix as the spectral capacity,

After selecting the index of the array response vector that has contributed the most to the spectral capacity, adding the selected index to the set of pre-selected indices and removing the selected index from the entire set is repeated,

When the number of repetitions reaches the number of RF chains of the receiver, a column of a dictionary matrix corresponding to the selected index may be selected as an array response vector, and the beams received as the selected array response vector may be analog-combined.

Preferably the baseband precoder,

The low-dimensional effective channel matrix is derived by multiplying the channel matrix and the matrix of the RF precode and the Hermitian transpose matrix of the matrix of the RF combiner,

It may be provided to generate a digital beam with a singular vector derived through a singular value decomposition operation on the derived low-dimensional effective channel matrix.

Preferably the baseband coupler,

The low-dimensional effective channel matrix is derived by multiplying the channel matrix and the matrix of the RF combiner and the matrix of the RF precoder by the Hermitian transpose matrix.

It may be provided to combine the digital beam with a singular vector derived through a singular value decomposition operation on the derived low-dimensional effective channel matrix.

According to another embodiment of the present invention, a hybrid beamforming method of a millimeter wave-based large-scale MIMO system,

Deriving a singular vector by performing singular value decomposition of a low-dimensional effective channel matrix in a baseband domain by a transmitter, and generating a digital beam from the derived singular vector;

Including the step of performing beamforming by searching for an array response vector from a dictionary matrix through an incremental continuous selection technique in the RF domain of the transmitter and forming an analog beam for the digital beam with the searched array response vector,

Deriving a singular vector by performing singular value decomposition of a low-dimensional effective channel matrix in a baseband domain by a receiver, and digitally combining the received beam with the derived singular vector; And

It may be provided to include the step of retrieving an array response vector from the dictionary matrix through an incremental continuous selection technique in the RF domain of the receiver and combining the analog beams with the retrieved array response vector.

Preferably the analog beam forming step,

Set antenna array vectors of a predetermined number of effective channel paths as spectral capacity,

Selecting the index of the array response vector that has contributed the most to the spectral capacity, adding the selected index to the set of pre-selected indices, and removing the selected index from the entire set is repeated,

When the number of repetitions reaches the number of RF chains of the transmitter, a column of a dictionary matrix corresponding to the selected index may be selected as an array response vector, and an analog beam may be formed using the selected array response vector.

Preferably the analog beam combining step,

The combined channel matrix is derived by the product of the received channel matrix and the matrix of the RF precoder,

Set the antenna array vector of the derived combined channel matrix as the spectral capacity,

After selecting the index of the array response vector that has contributed the most to the spectral capacity, adding the selected index to the set of pre-selected indices, and then repeating the process of removing the selected indexer from the entire set,

When the number of repetitions reaches the number of RF chains of the receiver, a column of a dictionary matrix corresponding to the selected index may be selected as an array response vector, and the beams received as the selected array response vector may be analog-combined.

According to an embodiment, a singular value decomposition of a low-dimensional effective channel matrix is performed in baseband domains of a transmitter and a receiver of a hybrid beamforming apparatus to derive a singular vector of a data stream, and a digital beam is generated and combined with the derived singular vector. In the RF domain, a subset of the array response vector is retrieved from the dictionary matrix through an incremental continuous selection technique, and an analog beam is generated and combined with the retrieved array response vector to restore the data stream, thereby separating the data stream into an analog domain and a digital domain. By performing forming, spectral efficiency can be maximized with low computational complexity, thereby improving system performance.

The following drawings attached in the present specification illustrate preferred embodiments of the present invention, and serve to further understand the technical idea of the present invention together with the detailed description of the present invention to be described later, so the present invention is described in such drawings. It is limited to and should not be interpreted.
1 is a schematic diagram showing the configuration of a system according to an embodiment.
2 is a flowchart illustrating an operation of an RF precoder in the system according to an embodiment.
3 is a flowchart illustrating an operation of an RF combiner of the system according to an exemplary embodiment.
4 is an exemplary diagram of SNR versus spectral efficiency in one embodiment.
5 is an exemplary diagram of antenna size versus spectral efficiency in an embodiment.
6 is an exemplary diagram of the number of propagation paths versus spectral efficiency in one embodiment.
7 is an exemplary diagram showing the number of antennas versus the number of FLOPs according to an embodiment.
8 is an exemplary diagram showing the number of propagation paths versus the number of FLOPs according to an embodiment.

Hereinafter, embodiments of the present invention will be described in more detail with reference to the drawings.

Advantages and features of the present invention, and a method of achieving them will be apparent with reference to the embodiments described later together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in a variety of different forms, and only these embodiments make the disclosure of the present invention complete, and are common knowledge in the technical field to which the present invention pertains. It is provided to completely inform the scope of the invention to those who have, and the invention is only defined by the scope of the claims.

The terms used in the present specification will be briefly described, and the present invention will be described in detail.

Terms used in the present invention have selected general terms that are currently widely used as possible while taking functions of the present invention into consideration, but this may vary according to the intention or precedent of a technician working in the field, the emergence of new technologies, and the like. In addition, in certain cases, there are terms arbitrarily selected by the applicant, and in this case, the meaning of the terms will be described in detail in the description of the corresponding invention. Therefore, the terms used in the present invention should be defined based on the meaning of the term and the overall contents of the present invention, not a simple name of the term.

When a part of the specification is said to "include" a certain component, it means that other components may be further included rather than excluding other components unless specifically stated to the contrary. In addition, the term "unit" used in the specification refers to a hardware component such as software, FPGA, or ASIC, and "unit" performs certain roles. However, "unit" is not meant to be limited to software or hardware. The “unit” may be configured to be in an addressable storage medium or may be configured to reproduce one or more processors.

Thus, as an example, "unit" refers to components such as software components, object-oriented software components, class components, and task components, processes, functions, properties, procedures, Includes subroutines, segments of program code, drivers, firmware, microcode, circuitry, data, database, data structures, tables, arrays and variables. The functions provided within the components and "units" may be combined into a smaller number of components and "units" or may be further separated into additional components and "units".

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those of ordinary skill in the art may easily implement the present invention. In the drawings, parts not related to the description are omitted in order to clearly describe the present invention.

The communication system to which the exemplary embodiment is applied may include arbitrary dogs for each component in any appropriate configuration. In general, computing and communication systems appear in a wide variety of configurations, and the drawings do not limit the scope of the present disclosure to any particular configuration. Although the figure shows one operating environment in which the various features disclosed in this patent document may be used, such features may be used in any other suitable system.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

The embodiments relate to a technology for performing beamforming for a receiver that falls within the coverage of a transmitter in a millimeter-wave channel environment. When the number of RF chains is less than or equal to the number of antennas of the transmitter, multiple receivers simultaneously receive digital It relates to a hybrid beamforming technology that performs beamforming and analog beamforming. In this specification, it is assumed that the number of RF chains of the transmitter is less than or equal to the number of antennas. That is, the number of RF chains may be less than or equal to the number of antennas of the transmitter.

In addition, in an embodiment, a pre-designed dictionary matrix may be shared between the transmitter and the receiver to be stored and maintained. Based on the pre-stored dictionary matrix, the transmitter may form and combine an analog beam targeting the receiver. Here, the dictionary matrix refers to a matrix composed of array response vectors (beamforming vectors) previously promised between a transmitter and a receiver.

In addition, in some embodiments, hybrid beamforming may be represented by beamforming a digital beam performed at a baseband end and an analog beam performed at an RF end.

In some embodiments, an analog beamforming is performed based on a prior matrix of the transmitter, but the receiver can also restore the original signal by receiving a signal transmitted by beamforming by the transmitter. Components corresponding to digital beamforming and analog beamforming may be included in order to restore a signal.

The transmitter of an embodiment forms a digital beam from an array response vector derived using singular value decomposition of a low-dimensional effective channel matrix in the digital domain, and a sub of the array response vector retrieved from the dictionary matrix through an incremental continuous selection technique in the analog domain. An analog beam is formed as a set and transmitted to a receiver through an antenna.

The receiver of an embodiment searches for a subset of the array response vector from the dictionary matrix through the incremental continuous selection technique in the analog domain, and performs analog coupling to the beam received through the antenna with the searched array response vector, and performs a low-dimensional effective channel in the digital domain. A data stream is output by digitally combining the analog-combined beam with a singular vector derived using the singular value decomposition of the matrix.

FIG. 1 is a diagram showing the configuration of a single-user millimeter-based large-scale MIMO system to which an embodiment is applied. Referring to FIG. 1, in an embodiment, an analog domain and a digital domain are separated for beamforming, and an analog domain and a digital domain are formed. It may be implemented to recover the data stream by combining the separately received beams.

Accordingly, the single-user millimeter wave-based large-scale MIMO system S to which an embodiment is applied may include a transmitter 100 and a receiver 200. Here, the transmitter 100 has N _t antennas 110 and M _t RF chain (130) Including, Receiver 200 has N _r Antenna 210 and M _r RF chain 230 Can include.

이고,

라고 가정하고, 송신기(100)는 데이터 스트림 N _s 을 디지털 처리하기 위해

의 행렬

의 기저대역 프리코더(140)와 위상 시프트에 의해 구현되는

의 행렬

의 RF 프리코더(120)를 더 포함할 수 있다.In addition, the data stream N _s is sent to the receiver 200 in the form of a beam,

ego,

Assuming that, the transmitter 100 digitally processes the data stream N _s

Procession of

Implemented by the baseband precoder 140 and the phase shift of

Procession of

It may further include an RF precoder 120 of.

의 행렬

의 RF 결합기(220)와

의 행렬

의 기저대역 결합기(240)를 더 포함하여 수신 신호에 순차적으로 적용하도록 구비될 수 있다.Meanwhile, the receiver 200

Procession of

RF combiner 220 and

Procession of

It may be provided to sequentially apply to the received signal by further including a baseband combiner 240 of.

는 하기 식 1로 주어진다.In this narrowband block fading propagation channel model, the received signal

Is given by Equation 1 below.

[Equation 1]

는 평균 전송 신호 전력이고,

는 하이브리드 결합 후 수신 신호 벡터이며,

은

이 되도록 정규화된 신호 전력을 가지는 전송 신호 벡터이다. 또한,

은

를 만족하는 채널행렬이며,

는

을 포함하는 독립적이고 동일하게 분포된 부가 백색 가우시안 잡음 벡터이다. here,

Is the average transmitted signal power,

Is the received signal vector after hybrid combining,

silver

It is a transmission signal vector having a signal power normalized to be. Also,

silver

Is a channel matrix that satisfies

Is

Is an independent and equally distributed additive white Gaussian noise vector including.

을 최대화하기 위한 기저대역 프리코더(140)의 행렬

및 RF 프리코더(120)의 행렬

와 RF 결합기(220)의 행렬

및 기저대역 결합기(240)의 행렬

를 설계하여야 한다.One embodiment is spectral efficiency in a millimeter-based large-scale MIMO system

Matrix of baseband precoder 140 to maximize

And the matrix of the RF precoder 120

And the matrix of the RF combiner 220

And matrix of baseband combiner 240

Should be designed.

는 다음 식 2a 내지 식 2c로 주어진다.This is spectral efficiency

Is given by the following equations 2a to 2c.

[Equation 2a]

[Equation 2b]

[Equation 2c]

와

이다. 식 2b는 RF 도메인에서의 위상 시프트에 의거 도출되고, 식 2c는 송신기(100)의 총 전력 제한으로 도출된다. here,

Wow

to be. Equation 2b is derived based on the phase shift in the RF domain, and Equation 2c is derived as the total power limit of the transmitter 100.

In addition, one embodiment adopts a geometric channel model due to the limited scattering characteristics of the millimeter wave channel, and each antenna 110 and 210 of the transmitter 100 and the receiver 200 includes a low noise power amplifier PA and a low noise power amplifier LPA, respectively. It is provided with a uniform linear antenna array.

는 다음 식 3a 및 식 3b를 만족한다.This is the physical channel matrix

Satisfies the following equations 3a and 3b.

[Equation 3a]

[Equation 3b]

로 가정하면, 경로 이득 행렬은

로 나타낼 수 있다. Where L is To the effective channel path corresponding to the limited scattering characteristics

Assuming as, the path gain matrix is

It can be expressed as

는

번째 경로의 복소수 이득이고,

와

는 어레이 응답 행렬(사전 행렬)이다. 여기서, 어레이 응답 행렬의 열은 송신기(100) 및 수신기(200)에서 방위각 AoD와 AoA에 대응되는 어레이 응답 벡터이고, 정규화된 어레이 응답 벡터는 다음 식 4 및 식 5로 각각 나타낸다.here,

Is

Is the complex gain of the th path,

Wow

Is the array response matrix (the dictionary matrix). Here, the columns of the array response matrix are array response vectors corresponding to the azimuth angles AoD and AoA in the transmitter 100 and the receiver 200, and the normalized array response vectors are represented by the following Equations 4 and 5, respectively.

[Equation 4]

[Equation 5]

와

는

번째 경로의 방위각 AoD(angle of departure)와 AoA(angle of arrival)이고, d 는 안테나 간격이며, λ는 캐리어 신호의 파장이다.here,

Wow

Is

Is the azimuth angle of departure (AoD) and angle of arrival (AoA) of the first route, and d is Is the antenna spacing, and λ is the wavelength of the carrier signal.

의 특이값 분해의 첫번째 데이터 스트림 N _s 오른쪽/왼쪽 특이(singular) 벡터를 취함으로써 제공될 수 있다. The optimal hybrid beamforming device is a channel matrix

The first data stream of the singular decomposition of N _s can be provided by taking the right/left singular vectors.

,

각각의 선형 결합으로 나타낼 수 있다.In addition, the optimal precoder and combiner columns are the array response vectors.

,

It can be represented by each linear combination.

를 최대화하고 시스템의 성능을 개선하며 공동 RF / 기저대역 최적화에 대한 연산 복잡성을 줄이기 위해, 일 실시 예의 하이브리드 빔포밍장치는 아날로그 및 디지털 도메인을 분리하여 디지털 스트림을 빔포밍하고, 아날로그 및 디지털 도메인을 분리하여 수신된 빔을 결합하여 복원할 수 있다. 즉, 하이브리드 빔장치는 아날로그 도메인(RF 도메인)에서 사전의 어레이 응답 벡터의 서브세트를 사용하여 RF 프리코더(120) 및 RF 결합기(220) 행렬을 설계하고, 디지털 도메인(기저대역 도메인)에서 저차원 유효 채널 행렬의 특이값 분해를 이용하여 도출된 특이 벡터로 기저대역 프리코더(140) 및 기저대역 결합기(240) 행렬을 생성할 수 있다I.e. spectral efficiency

In order to maximize the performance of the system, improve the performance of the system, and reduce the computational complexity for joint RF/baseband optimization, the hybrid beamforming apparatus of an embodiment separates the analog and digital domains to beamforming a digital stream, and separates the analog and digital domains. The separated and received beams can be combined and restored. That is, the hybrid beam apparatus designs the RF precoder 120 and RF combiner 220 matrices using a subset of the array response vectors in advance in the analog domain (RF domain), and saves them in the digital domain (baseband domain). The baseband precoder 140 and the baseband combiner 240 may generate matrices with singular vectors derived using singular value decomposition of the dimensional effective channel matrix.

을 최대화하기 위해, 일 실시 예에서 송신기(100)의 기저대역 프리코드(140)의 행렬

및 기저대역 프리코드(140)의 행렬

에 대한 Hermitian 전치 행렬

는

을 만족한다고 가정하고, 기저대역 결합기(240)의 행렬

및 기저대역 결합기(240)의 행렬

의 Hermitian 전치 행렬

는

를 만족한다고 가정하면, RF 프리코더(120)의 행렬

와 RF 결합기(220)의 행렬

는 스펙트랄 효율의 최대화를 위해, 사전 행렬

,

의 L 어레이의 응답 벡터 중 RF 체인(130)의 M _t 벡터 및 RF 체인(230)의 M _r 벡터의 선택하여 설계된다. 여기서

는 N_s X N_s 의 단위 행렬이다. Spectral efficiency

In order to maximize, in one embodiment, the matrix of the baseband precode 140 of the transmitter 100

And a matrix of baseband precodes 140

Hermitian transpose matrix for

Is

Assuming that is satisfied, the matrix of the baseband combiner 240

And matrix of baseband combiner 240

Hermitian transpose matrix

Is

Assuming that is satisfied, the matrix of the RF precoder 120

And the matrix of the RF combiner 220

Is the prior matrix for maximization of spectral efficiency.

,

M _t vector of the RF chain 130 among the response vectors of the L array of And the selection of the M _r vector of the RF chain 230. here

Is the N _s XN _s It is an identity matrix.

와 기저대역 결합기(240)의 행렬

와 RF 프리코더(120)의 행렬

와 RF 결합기(220)의 행렬

의 최적화는 스펙트랄 효율

의 최대화 문제의 해로 도출될 수 있고, 스펙트랄 효율

의 최대화 문제는 하기 식 6a 및 식 6b로 각각 공식화될 수 있다.Matrix of baseband precodes (140)

And matrix of baseband combiner 240

And the matrix of the RF precoder 120

And the matrix of the RF combiner 220

Optimization of the spectral efficiency

Can be derived as a solution to the maximization problem of spectral efficiency

The maximization problem of can be formulated by the following equations 6a and 6b, respectively.

[Equation 6a]

[Equation 6b]

이다.here,

to be.

및 RF 결합기(220)의 행렬

와 기저대역 프리코더(140)의 행렬

및 기저대역 결합기(240)의 행렬

의 최적화는 스펙트랄 효율

최대화 문제의 해로 도출될 수 있다.Also, the matrix of the fixed RF precoder 120

And the matrix of the RF combiner 220

And matrix of baseband precoder (140)

And matrix of baseband combiner 240

Optimization of the spectral efficiency

It can be derived as a solution to the maximization problem.

의 요소를 가지며, 위상 시프트에 의해 구현될 수 있다.And, each of the RF precoder 120 and the RF combiner 220 is a fixed-size array response vector derived from the azimuth AoD and AoA

It has an element of, and can be implemented by phase shift.

,

의 최적화를 도출한다. 즉, RF 결합기의 행렬은

을 만족하고 대형 안테나 어레이가 유지된다고 가정하에 RF 프리코더(120)의 행렬

는 설계되어야 한다.First, the matrix of the RF precoder 120 and the RF combiner 220 to derive a solution to the problem of maximizing spectral efficiency with low computational complexity.

,

To derive the optimization of. In other words, the matrix of the RF combiner is

And the matrix of the RF precoder 120 on the assumption that a large antenna array is maintained.

Must be designed.

는

의 결합 채널 행렬

을 적용하여 설계된다. 각각 RF 프리코더(120) 및 RF 결합기(220)의 준최적화 문제는 다음 식 7a 및 식 7b로 주어진다.That is, the matrix of the RF combiner 220

Is

Combined channel matrix of

It is designed by applying. The sub-optimization problem of the RF precoder 120 and RF combiner 220, respectively, is given by the following equations 7a and 7b.

[Equation 7a]

[Equation 7b]

이다. 여기서,

는 평균 전송 신호 전력이고, N _s 는 데이터 심볼이며,

는 분산이다.here,

to be. here,

Is the average transmitted signal power, N _s is the data symbol,

Is the variance.

In addition, in order to select an optimal array response vector with low computational complexity, the RF precoder 120 and the RF combiner 220 of an embodiment select a subset of the array response vectors defined in advance using an incremental continuous selection algorithm. .

의

번째 열은 RF 프리코더(120)의 행렬

의 m 번째 열로 선택된다. 또한 RF 프리코더(120)의 부분 행렬

은 사전 행렬

의 m 번째 열로 선택되고, m 번째 반복에 대응되는 스펙트럼 용량

,

은 하기 식 8로 나타낸다.That is, a dictionary matrix to provide the maximum spectral capacity (channel capacity) for any m-th iteration

of

The first column is the matrix of the RF precoder 120

Is selected as the m th column of. Also, the partial matrix of the RF precoder 120

Is a dictionary matrix

Spectral capacity selected as the m- th column of and corresponding to the m-th iteration

,

Is represented by the following formula (8).

[Equation 8]

이다. here,

to be.

의 부분 행렬

을 고려하면, m+1 번째 반복에 대한 사전 행렬

에서 RF 프리코더(120)는 m+1 번째 반복에서의 스펙트럼 용량

을 설정하고, 설정된 m+1 번째 반복에서의 스펙트럼 용량

는 하기 식 9로 주어진다.Dictionary matrix

Partial matrix of

Considering, the lexicographic matrix for the m+1th iteration

In the RF precoder (120) is the spectral capacity at m+1th iteration

And the spectral capacity at the set m+1th iteration

Is given by Equation 9 below.

[Equation 9]

를 적용하면, 다음 식 10을 도출할 수 있다.Sherman-Morrision formula in Equation 9 and

If is applied, the following Equation 10 can be derived.

[Equation 10]

로 정의하면 식 10은 하기 식 11로 정리된다.And, the positive semidefinite matrix

If defined as, Equation 10 is summarized in Equation 11 below.

[Equation 11]

은

에 대한 양의 수이고, 스페트럼 용량

의 m 의 증가 함수이다.Where, the increment value in Equation 11

silver

Is the number of positives for the spectral capacity

Is the increasing function of m.

에 대한

번째 어레이 응답 벡터의 기여

는 다음 식 12에 의거 결정된다.Spectral capacity selected from m+1 iterations from Equation 11

for

The contribution of the first array response vector

Is determined according to Equation 12 below.

[Equation 12]

은 다음 식 13으로 정리된다.Spectral capacity selected in m+ 1th iteration from Equation 11 and Equation 12

Is summarized in Equation 13 below.

[Equation 13]

번째 경로는

,

이고, 최대 스펙트럼 용량과 동일한

번째 안테나 어레이 벡터는 하기 식 14로 얻어진다.here,

The second path is

,

And is equal to the maximum spectral capacity

The antenna array vector is obtained by Equation 14 below.

[Equation 14]

은

으로 업데이트되며 업데이트된 양의 반정 부호 행렬

는 하기 식 15와 같다.Positive half-definite matrix for each iteration

silver

And updated positive half-definite matrix

Is as shown in Equation 15 below.

[Equation 15]

Here, (a) can be derived by applying the Sherman-Morrison formula, and (a) is obtained by Equation 16 below.

[Equation 16]

If Equation 16 is substituted for Equation 15 and summarized, it is derived by Equation 17 below.

[Equation 17]

번째 어레이 응답 벡터의 기여

는 식 19와 같이 초기화될 수 있다.Therefore, in order to reduce the computational complexity

The contribution of the first array response vector

Can be initialized as in Equation 19.

[Equation 19]

를 구성하기 위해 사전 행렬

의

번째 열 중

의 어레이 응답 벡터를 선택하는 알고리즘을 수행한다.The RF precoder 120 of an embodiment is a matrix

Dictionary matrix to construct

of

Out of the second column

Perform an algorithm to select an array response vector of.

를 도출하는 일련의 과정과 각 단계 별 연산 복잡도를 보인 도면으로서, 도 2를 참조하면, RF 프리코더(120)는 양의 반정 부호 행렬

,

, 및

번째 어레이 응답 벡터의 기여

를 각각 연산하며, 이때 연산 복잡도는 도 2에 도시된 바와 같다. 2 is a matrix of the RF precoder 120

As a diagram showing a series of processes for deriving and computational complexity for each step, referring to FIG. 2, the RF precoder 120 is a positive half-definite code matrix

,

, And

The contribution of the first array response vector

Each is calculated, and the calculation complexity is as shown in FIG. 2.

, 선택된 어레이 응답 벡터들의 인덱스 집합

, 전체집합

의 어레이 응답 벡터를 초기화한다. 단계 2 내지 4에서 일 실시 예는 1부터 L을 포함하는 유효 채널 경로

의 각 안테나 어레이 벡터의 기여를

로 초기화한다.That is, in step 1, the RF precoder 120 is a positive half-definite code matrix

, The set of indices of the selected array response vectors

, Whole set

Initialize the array response vector of. In steps 2 to 4, one embodiment includes L from 1 Effective channel path

The contribution of each antenna array vector of

Initialize with

를 선택한 다음 전체집합

에서 선택된 어레이 응답 벡터의 인덱스

를 제거하는 반면, 기 선택된 인덱스들의 집합인

에 어레이 응답 벡터의 인덱스

를 추가한다. 단계 9, 10, 11 및 12에서, 일 실시 예는 반복 횟수 m이 RF 체인(130)의 수 M _t 보다 작은 경우 식 16 내지 18에 의거 양의 반정 부호 행렬

를

으로 업데이트한다. 반복 횟수 m이 RF 체인(130)의 수 M _t 에 도달할 때까지 단계 9 내지 11, 및 12 내지 14를 반복 수행한다.And, in step 5 to step 8, one embodiment is the index of the array response vector contributed to the maximum of the spectral capacity

Select and then set the whole

The index of the array response vector selected from

Is removed, whereas the set of pre-selected indices

The index of the array response vector to

Add In steps 9, 10, 11 and 12, in one embodiment, the number of repetitions m is greater than the number M _{t of the RF chain 130} For small cases, a positive half-code matrix according to Equations 16 to 18

To

Update to. Steps 9 to 11 and 12 to 14 are repeated until the number of repetitions m reaches the number M _{t of the RF chains 130.}

를 생성할 수 있다.Meanwhile, the matrix of the RF combiner 220 is obtained by transforming Equations 8 to 19 to derive the solution of the sub-optimization problem of Equation 7b.

Can be created.

를 생성하는 과정과 각 단계 별 연산 복잡도를 나타낸 도면으로서, 도 3을 참조하면, RF 결합기(220)는 사전 행렬

의 L 열에서 M _r 의 어레이의 응답 벡터를 선택하는 과정을 통해 RF 결합기(220)의 행렬

을 도출할 수 있다. 3 is a matrix of the RF combiner 220

As a diagram showing the process of generating and the computational complexity for each step, referring to FIG. 3, the RF combiner 220 is a dictionary matrix

Of M _r in column L of Matrix of the RF combiner 220 through the process of selecting the response vector of the array

Can be derived.

를 도출하는 일련의 과정은 RF 프리코드(120)의 행렬

를 도출하는 일련의 과정과 동일 또는 유사하므로 이에 대한 구체적인 설명은 생략한다.Here, the matrix of the RF combiner 220

A series of processes for deriving is the matrix of the RF precode 120

Since it is the same or similar to a series of processes for deriving a, a detailed description thereof will be omitted.

,

를 결정할 수 있고, 저차원 유효 채널

은 다음 식 20으로 정의된다.Meanwhile, in an embodiment, each matrix of the baseband precoder 140 and the baseband combiner 240 is a solution to the spectral efficiency maximization problem.

,

Can determine the low-dimensional effective channel

Is defined by Equation 20 below.

[Equation 20]

에 의해 행렬

,

를 각각 생성할 수 있다. The baseband precoder 140 and the baseband combiner 240 calculate the singular value decomposition of the low-dimensional effective channel matrix

Procession by

,

Each can be created.

,

에 대해

및

의 첫번째 데이터 스트림 N _s 열을 각각 획득할 수 있다.That is, in one embodiment, each matrix

,

About

And

Each of the first data streams N _{s of} may be obtained.

,

을 생성하기 위해 유효 채널 행렬

의 특이값 분해를 수행하면, 채널 행렬

의 크기보다 유효 채널 행렬

크기가 작기 때문에 일 실시 예의 연산 복잡도는 크게 감소된다. Each matrix

,

Effective channel matrix to generate

Performing the singular value decomposition of, the channel matrix

Effective channel matrix than the size of

Since the size is small, the computational complexity of an embodiment is greatly reduced.

Simulation results for system performance

이고, 안테나 간격

이며, 송신기(100)와 수신기(200)의 RF 체인(130)의 수가 동일하며, 최소 소비 에너지를 가지는 멀티 스트림 전송을 위해 RF 체인(130)의 수와 데이터 스트림 N _s 동일하다고 가정하자. 즉

,

. 여기서, 신호대 잡음비(SNR: signal-to-noise ratio)는

으로, 분산과 평균 전송신호 전력의 비로 나타낸다. 그리고, 데이트 스트림 당 동일한 전력 할당을 가지는 각 프리코더에는 동일한 총 전력 제약이 적용된다. In a single-user millimeter wave-based large-scale MIMO system according to an embodiment, uniformly distributed random azimuth angles AoD and AoA are obtained in a channel environment with an effective channel path L=15.

Is, and the antenna spacing

It is assumed that the number of RF chains 130 of the transmitter 100 and the receiver 200 is the same, and that the number of RF chains 130 and the data stream N _{s are the same for multi-stream transmission having the minimum energy consumption.} In other words

,

. Here, the signal-to-noise ratio (SNR) is

It is expressed as the ratio of the variance and the average transmission signal power. In addition, the same total power constraint is applied to each precoder having the same power allocation per data stream.

,

,

,

를 생성되므로, 신호대 잡음비(SNR) 범위에서 기존의 알고리즘(OMP)과 비교하여 높은 시스템의 성능이 달성됨을 알 수 있다. FIG. 4 is a diagram comparing spectral efficiencies that can be achieved for various signal-to-noise ratios in a uniform linear array (ULA) system with Nt = Nr = 64 and M = Ns = 4, and FIG. 4 Referring to, an embodiment is a matrix as a solution to the spectral efficiency maximization problem

,

,

,

As compared to the conventional algorithm (OMP) in the signal-to-noise ratio (SNR) range, it can be seen that high system performance is achieved.

5 is a diagram showing the spectral efficiency achievable for each antenna array size of Nt = Nr = {32, 64, 128, 256, 512} in a system with M = Ns = 4 and SNR = 10 dB. Referring to 5, since the array response vector converges to an optimal precoding vector and a combination vector at the limit of a large antenna array, it can be seen that the performance of the system is superior to that of the conventional OMP algorithm according to an embodiment.

6 is a diagram showing the effect of the number of propagation paths of a channel on the spectral efficiency achievable in a system with Nt = Nr = 64, M = Ns = 4, and SNR = 10 dB. It can be seen that in the scattering channel environment such as path L = 40, an exemplary embodiment achieves a spectral efficiency gain of about 7.4% compared to the conventional OMP algorithm.

7 is a diagram showing the number of floating-point operations (FLOPs) for each size of various antenna arrays. Referring to FIG. 7, Nt = Nr = {32, 64, 128, 256, 512} and M = Ns = 4 It can be seen that in the linear array system, the computational complexity of an embodiment is reduced compared to the conventional OMP algorithm. For example, in the case of Nt = Nr = 32 and Nt = Nr = 512, the reduction ratio of the computational complexity of an embodiment compared to the conventional OMP algorithm is 78.8% and 98.7%, respectively.

FIG. 8 is a diagram showing computational complexity when the number of propagation paths changes in a uniform linear array system with Nt = Nr = 64 and M = Ns = 4. Referring to FIG. 8, an effective channel path L = 4 and L = 40 When is, it can be seen that the computational complexity according to an embodiment is 96.5% and 74.3% lower than that of the existing OMP algorithm, respectively.

Accordingly, one embodiment generates an RF precoder and an RF combiner by searching for a subset of the prior array response vectors in the analog domain through an incremental continuous selection technique, and using singular value decomposition of the low-dimensional effective channel matrix in the digital domain. By generating a baseband precoder and a baseband combiner, spectral efficiency can be maximized with low computational complexity, and performance of a millimeter-based large-scale MIMO system can be improved.

According to another aspect of an embodiment, the hybrid beamforming method of a millimeter wave-based large-scale MIMO system is performed by decomposing a singular value of a low-dimensional effective channel matrix in a baseband domain by a transmitter to derive a singular vector of a data stream. Generating a digital beam from a vector; Retrieving a subset of the array response vectors from the dictionary matrix through an incremental continuous selection technique in the RF domain, and forming an analog beam for the digital beam with the retrieved array response vector, and is valid in a baseband domain by a receiver. Deriving a singular vector of the data stream by performing singular value decomposition of the channel matrix, and digitally combining the received beam with the derived singular vector; And retrieving a subset of the array response vectors from the dictionary matrix through an incremental continuous selection technique in the RF domain and combining the analog beams with the retrieved array response vectors.

Here, in the step of forming the analog beam, the antenna array vectors of a predetermined number of effective channel paths are set as the spectral capacity, the index of the array response vector having the maximum spectral capacity is selected, and the selected index is added to the set of pre-selected indices. , Iteratively performs the process of removing the selected index from the entire set, and when the number of iterations reaches the number of RF chains of the transmitter, the column of the dictionary matrix corresponding to the selected index is selected as the array response vector, and the selected array response vector is analog. It may be provided to form a beam.

In addition, in the step of combining the analog beam, a combined channel matrix is derived by a product of a received channel matrix and a matrix of an RF precoder, and an antenna array vector of the derived combined channel matrix is set as a spectral capacity, and has a maximum spectral capacity. After selecting the index of the array response vector, the selected index is added to the set of pre-selected indices, and the process of removing the selected index from the entire set is repeated. When the number of repetitions reaches the number of RF chains of the receiver, the selected index is It may be provided to select a column of a corresponding dictionary matrix as an array response vector and analogly combine a beam received with the selected array response vector.

Meanwhile, in the digital beam forming step, a low-dimensional effective channel matrix is derived by multiplying a Hermitian transpose matrix for a matrix of an RF combiner and a channel matrix and a matrix of an RF precode, and singular values for the derived low-dimensional effective channel matrix It may be provided to generate a digital beam with a singular vector derived through a decomposition operation.

The digital beam combining step is to derive a low-dimensional effective channel matrix by multiplying the Hermitian transpose matrix for the matrix of the RF precoder and the channel matrix and the matrix of the RF RF combiner to derive a singular value decomposition operation on the derived low-dimensional effective channel matrix. It may be provided to combine the digital beam with the singular vector derived through, and each step of this hybrid beamforming method includes the baseband precoder 140, the RF precoder 120 and the receiver ( This is a function performed by the RF combiner 220 and the baseband combiner 240 of 200), and detailed reference will be omitted.

As described above, although the embodiments have been described by the limited embodiments and drawings, various modifications and variations are possible from the above description to those of ordinary skill in the art. For example, the described techniques are performed in a different order from the described method, and/or components such as the described system, structure, device, circuit, etc. are combined or combined in a form different from the described method, or have different configurations. Appropriate results can be achieved even if substituted or substituted by elements or equivalents. Therefore, the scope of the present invention is limited to the described embodiments and should not be defined, but should be defined by the claims and equivalents as well as the claims to be described later.

The singular value decomposition of the low-dimensional effective channel matrix is performed in the baseband domain to derive the singular vector of the data stream, the digital beam is generated and combined with the derived singular vector, and the array from the dictionary matrix through the incremental continuous selection technique in the RF domain The problem of maximizing spectral efficiency with low computational complexity by performing hybrid beamforming by separating into analog and digital domains by retrieving a subset of response vectors and restoring the data stream by generating and combining analog beams with the retrieved array response vectors. In terms of accuracy and reliability of the operation of the hybrid beamforming device and method of a millimeter wave-based large-scale MIMO system that can improve the performance of the system, and further improve the performance of the system, it will bring a great progress in terms of performance efficiency. It is an invention that has industrial applicability because it is not only possible to market or sell a system providing a wireless communication service sufficiently, but also to a degree that can be implemented clearly in reality.

Claims (10)

delete A transmitter for beamforming by forming a digital beam for the data stream based on a baseband precode and then forming an analog beam through an RF precoder; And
Including a receiver for restoring the data stream by analog-combining the beam transmitted from the transmitter based on an RF combiner and then digitally combining the beam received from the transmitter through a baseband combiner,
The transmitter is
It is provided to generate an analog beam and a digital beam by separating it into an RF domain and a baseband domain,
The receiver
The generated beam is separated into an RF domain and a baseband domain to perform analog combining and digital combining to restore a data stream,
The transmitter is
A digital beam is generated as a singular vector by performing singular value decomposition of the low-dimensional effective channel matrix in the baseband domain,
A hybrid beamforming apparatus for a millimeter wave-based large-scale MIMO system, comprising: searching a subset of array response vectors from a dictionary matrix through an incremental continuous selection technique in an RF domain and generating an analog beam using the searched array response vectors. The method of claim 2, wherein the receiver
In the RF domain, a subset of the array response vector is retrieved from the dictionary matrix through an incremental continuous selection technique, and the received beams are analog-combined with the retrieved array response vector,
A hybrid beamforming apparatus for a millimeter wave-based large-scale MIMO system, comprising deriving a singular vector of a data stream by performing singular value decomposition of a low-dimensional effective channel matrix in a baseband domain, and digitally combining the derived singular vector. The method of claim 2, wherein the RF precoder
Set the antenna array vector as the spectral capacity for each of a predetermined number of channel paths,
Selecting the index of the array response vector that has contributed the most to the spectral capacity, adding the selected index to the set of pre-selected indices, and removing the selected index from the entire set is repeated,
When the number of repetitions reaches the number of RF chains of the transmitter, millimeter wave-based large-scale MIMO, characterized in that it is provided to select a column of a dictionary matrix corresponding to the selected index as an array response vector and to form an analog beam with the selected array response vector. Hybrid beamforming device of the system. The method of claim 4, wherein the RF combiner,
The effective channel matrix is derived by the product of the channel matrix and the matrix of the RF precoder,
Set the antenna array vector of the derived effective channel matrix as the spectral capacity,
After selecting the index of the array response vector that has contributed the most to the spectral capacity, the process of adding the selected index to the set of pre-selected indices and removing the selected index from the entire set is repeated,
When the number of repetitions reaches the number of RF chains of the receiver, a column of a dictionary matrix corresponding to the selected index is selected as an array response vector, and the beams received with the selected array response vector are analog-combined. Hybrid beamforming device for large-scale MIMO system. The method of claim 2, wherein the baseband precoder,
The low-dimensional effective channel matrix is derived by multiplying the channel matrix and the matrix of the RF precode and the Hermitian transpose matrix of the matrix of the RF combiner,
A hybrid beamforming apparatus for a millimeter wave-based large-scale MIMO system, comprising generating a digital beam with a singular vector derived through a singular value decomposition operation on the derived low-dimensional effective channel matrix. The method of claim 6, wherein the baseband coupler,
The low-dimensional effective channel matrix is derived by multiplying the channel matrix and the matrix of the RF combiner and the matrix of the RF precoder by the Hermitian transpose matrix.
A hybrid beamforming apparatus for a millimeter wave-based large-scale MIMO system, comprising: combining digital beams with singular vectors derived through singular value decomposition operations on the derived low-dimensional effective channel matrix. Deriving a singular vector by performing singular value decomposition of a low-dimensional effective channel matrix in a baseband domain by a transmitter, and generating a digital beam from the derived singular vector;
Including the step of performing beamforming by searching for an array response vector from a dictionary matrix through an incremental continuous selection technique in the RF domain of the transmitter and forming an analog beam for the digital beam with the searched array response vector,
Deriving a singular vector by performing singular value decomposition of a low-dimensional effective channel matrix in a baseband domain by a receiver, and digitally combining the received beam with the derived singular vector; And
Hybrid beamforming of a millimeter wave-based large-scale MIMO system, comprising the step of searching for an array response vector from a dictionary matrix through an incremental continuous selection technique in the RF domain of the receiver and combining the analog beams with the searched array response vector. Way. The method of claim 8, wherein forming the analog beam,
Set antenna array vectors of a predetermined number of effective channel paths as spectral capacity,
After selecting the index of the array response vector that has contributed the most to the spectral capacity, adding the selected index to the set of pre-selected indices, and repeating the process of removing the index from the entire set,
When the number of repetitions reaches the number of RF chains of the transmitter, millimeter wave-based large-scale MIMO, characterized in that it is provided to select a column of a dictionary matrix corresponding to the selected index as an array response vector and to form an analog beam with the selected array response vector. Hybrid beamforming method of the system. The method of claim 9, wherein combining the analog beams,
The effective channel matrix is derived by the product of the received channel matrix and the matrix of the RF precoder,
Set the antenna array vector of the derived effective channel matrix as the spectral capacity,
After selecting the index of the array response vector that contributed the most to the spectral capacity, adding the selected index to the set of pre-selected indices, and repeatedly performing the process of removing the selected index from the entire set of indices,
When the number of repetitions reaches the number of RF chains of the receiver, a column of a dictionary matrix corresponding to the selected index is selected as an array response vector, and the beams received with the selected array response vector are analog-combined. Hybrid beamforming method of large-scale MIMO system.

Images