KR102027674B1 - Method and system for estimating i/q imbalance parameter of transceiver based on reinforcement learning - Google Patents

Abstract

A method and system for estimating I / Q imbalance parameters of a multi-antenna transceiver based on enhanced learning is presented. According to an embodiment, an I / Q imbalance parameter estimation method of an enhanced learning-based multi-antenna transceiver includes arbitrarily setting an initial value of each parameter in a multi-antenna channel in which at least one I / Q imbalance occurs between a transmitter and a receiver. Doing; Measuring an effective channel including an I / Q imbalance value; Calculating an estimated value of the I / Q imbalance value by increasing or decreasing the transmitter / receiver I / Q imbalance value estimated in the previous step; Calculating an effective channel compensated for I / Q imbalance by using the estimated value of the calculated I / Q imbalance value; Measuring a cost indicating a degree to which an I / Q imbalance is compensated for by using an effective channel compensated for the I / Q imbalance; Calculating a reward in proportion to the cost; And setting a next step size based on the compensation.

Description

I / Q imbalance parameter estimation method and system for reinforcement learning-based multi-antenna transceivers {METHOD AND SYSTEM FOR ESTIMATING I / Q IMBALANCE PARAMETER OF TRANSCEIVER BASED ON REINFORCEMENT LEARNING}

The following embodiments relate to a method and system for estimating I / Q imbalance parameters of a multi-antenna transceiver based on reinforcement learning. More specifically, the present invention relates to an I / Q imbalance parameter estimation method in a multi-antenna system in which an I / Q imbalance exists in a transceiver. / Q unbalance parameter estimation method and system.

Massive Multi Input Multi Output (MIMO) technology, which uses a large number of base station antennas, has recently been in the spotlight as a technology for eliminating signal interference between users and providing high frequency efficiency.

In order to use a large number of antennas, low-cost devices must be used, and the in-phase and quadrature phase imbalances (I / Q imbalance) occurring in these RF devices are the in-phase components of the carrier wave. Inconsistency in the amplitude and phase of the orthogonal phase component degrades orthogonality and interference, which significantly degrades performance.

Conventionally, a technique for compensating the influence of the I / Q imbalance has been proposed, but for such a technique, it is necessary to estimate the I / Q imbalance parameter.

Korean Patent Laid-Open No. 10-2009-0089531 relates to an apparatus and method for estimating an eye / cue unbalance parameter in such an orthogonal frequency division multiplexing receiver, and to a method and apparatus for efficiently managing a file stored in an external memory of a portable terminal. It describes the technology.

Korean Patent Publication No. 10-2009-0089531

Embodiments describe a method and system for estimating I / Q imbalance parameters of a multi-antenna transceiver based on reinforcement learning, and more specifically, I / Q using a reinforcement learning technique in a multi-antenna system in which an I / Q imbalance exists in the transceiver. Provide a technique for estimating unbalanced parameters.

Embodiments are a method and system for estimating the I / Q imbalance parameter of a reinforcement learning-based multi-antenna transceiver having a characteristic that measures the degree to which the I / Q imbalance is compensated, converges to a value corrected over time, and is robust to parameter change. To provide.

According to an embodiment, an I / Q imbalance parameter estimation method of an enhanced learning-based multi-antenna transceiver includes arbitrarily setting an initial value of each parameter in a multi-antenna channel in which at least one I / Q imbalance occurs between a transmitter and a receiver. Doing; Measuring an effective channel including an I / Q imbalance value; Calculating an estimated value of the I / Q imbalance value by increasing or decreasing the transmitter / receiver I / Q imbalance value estimated in the previous step; Calculating an effective channel compensated for I / Q imbalance by using the estimated value of the calculated I / Q imbalance value; Measuring a cost indicating a degree to which an I / Q imbalance is compensated for by using an effective channel compensated for the I / Q imbalance; Calculating a reward in proportion to the cost; And setting a next step size based on the compensation.

Selecting a higher rewarding action among the rewards compensated at the transmitting end or the receiving end; And updating the I / Q imbalance based on the higher rewarding action selected, and repeating the step for each parameter.

The calculating of the effective channel compensated for I / Q imbalance using the estimated value of the I / Q imbalance value may include: calculating an effective I / Q imbalance compensated for using the estimated value of the I / Q imbalance value through a compensator. The channel can be calculated and passed to the compensation generator.

Measuring the cost representing the degree to which the I / Q imbalance is compensated by using the effective channel compensated for the I / Q imbalance, and receiving a valid channel matrix compensated for the I / Q imbalance from the compensation generator The cost indicating the degree to which the I / Q imbalance is compensated for is measured, and can be calculated by comparing the degree corresponding to the real part and the imaginary part in the augmented channel notation. .

In the step of setting a next step size based on the compensation, if the compensation is large because the step size is set in proportion to the cost, the estimated value is close to the actual value, so that the step size is reduced. In the case where the compensation is small, the step size may be set to be large because the estimated value is different from the actual value.

The transmitting end is a user terminal and the receiving end is a base station, and the I / Q imbalance value may be a base station gain imbalance value, a base station phase imbalance value, a user terminal amplitude imbalance value, and a user terminal phase imbalance value. have.

According to another embodiment, an I / Q imbalance parameter estimation system of a reinforcement learning based multiple antenna transceiver stage may arbitrarily set an initial value of each parameter in a multi-antenna channel in which at least one I / Q imbalance occurs between a transmitter and a receiver. And measuring an effective channel including an I / Q imbalance value, and calculating an estimated value of the I / Q imbalance value by increasing or decreasing the transmitter or receiver I / Q imbalance value estimated in the previous step. And a receiver that calculates an effective channel compensated for I / Q imbalance and measures a cost representing the degree to which the I / Q imbalance is compensated, and calculates a reward in proportion to the cost. Can be.

The receiving end may select a higher rewarding action among the rewarded rewards to update the I / Q imbalance based on the selected higher rewarding action.

The transmitting end is a user terminal and the receiving end is a base station, and the I / Q imbalance value may be a base station gain imbalance value, a base station phase imbalance value, a user terminal amplitude imbalance value, and a user terminal phase imbalance value. have.

According to another embodiment, an I / Q imbalance parameter estimating system of an enhanced learning-based multi-antenna transceiver includes: a compensator for estimating an effective channel by receiving a signal from a transmitting end; And receiving a valid channel matrix compensated for I / Q imbalance from the compensator, measuring a cost representing a degree to which the I / Q imbalance is compensated, and calculating a reward in proportion to the cost. A compensator, wherein the compensator is a multi-antenna channel in which at least one of I / Q imbalance occurs in at least one of a transmitter and a receiver, arbitrarily sets an initial value of each parameter, and includes an effective channel including an I / Q imbalance value. It is possible to calculate the estimated value of the I / Q imbalance value by calculating the estimated value of the I / Q imbalance value by increasing or decreasing the I / Q imbalance value of the transmitter or the receiver estimated in the previous step. have.

The receiving end selects a higher rewarding action among the compensated rewards, and the compensator may update the I / Q imbalance based on the selected higher rewarding action.

The compensation generator receives an effective channel matrix compensated for the I / Q imbalance and measures a cost indicating a degree to which the I / Q imbalance is compensated in the compensation generator, and an augmented channel notation Can be calculated by comparing the degree corresponding to the real part and the part corresponding to the imaginary part.

The compensator sets a next step size based on the compensation, and if the compensation is large by setting the step size in proportion to the cost, the estimated value is close to the actual value. When the compensation is small, the step size can be set large because the estimated value is a large difference from the actual value.

The transmitting end is a user terminal and the receiving end is a base station, and the I / Q imbalance value may be a base station gain imbalance value, a base station phase imbalance value, a user terminal amplitude imbalance value, and a user terminal phase imbalance value. have.

Iterative estimation is performed by randomly estimating and compensating I / Q imbalance parameters from the effective channel measured based on reinforcement learning, and calculating compensation based on the degree of estimation of the I / Q imbalance. The convergence speed and convergence accuracy can be adjusted by adaptively adjusting the size of.

According to embodiments, a method for estimating I / Q imbalance parameters of a reinforcement learning-based multi-antenna transceiver having a characteristic of measuring I / Q imbalance compensated and converged to a value corrected over time and robust to parameter changes And a system.

In addition, according to embodiments, since the size of the next behavior is determined based on the existing compensation value, an I / A generated in a communication system using a large number of antennas, which has a fast convergence speed, high accuracy of convergence, and the like, may be used. Can be used to resolve Q imbalance.

1 is a block diagram illustrating an I / Q imbalance parameter estimation system of an enhanced learning based multiple antenna transceiver according to an embodiment.
2 is a flowchart illustrating a method for estimating an I / Q imbalance parameter of a reinforcement learning based multi-antenna transceiver according to an embodiment.
3 illustrates an amplitude imbalance parameter estimation MSE according to an embodiment.
4 is a diagram illustrating a phase imbalance parameter estimation MSE according to an embodiment.

Hereinafter, exemplary embodiments will be described with reference to the accompanying drawings. However, the described embodiments may be modified in many different forms, and the scope of the present invention is not limited to the embodiments described below. In addition, various embodiments are provided to more fully describe the present invention to those skilled in the art. Shape and size of the elements in the drawings may be exaggerated for more clear description.

In the present specification, embodiments of the present invention will be described based on a relationship between data transmission and reception between a base station and a terminal. Here, the base station has a meaning as a terminal node of the network that directly communicates with the terminal. The specific operation described below to be performed by the base station may be performed by an upper node of the base station in some cases.

That is, it is obvious that various operations performed for communication with a terminal in a network composed of a plurality of network nodes including a base station may be performed by the base station or other network nodes other than the base station. A 'base station (BS)' may be replaced by terms such as a fixed station, a Node B, an eNode B (eNB), an access point (AP), and the like. In addition, the term base station may be used below as a concept including a cell or a sector. On the other hand, the repeater may be replaced by terms such as Relay Node (RN), Relay Station (RS). The term “terminal” may be replaced with terms such as user equipment (UE), mobile station (MS), mobile subscriber station (MSS), and subscriber station (SS).

Specific terms used in the following description are provided to help the understanding of the present invention, and the use of such specific terms may be changed to other forms without departing from the technical spirit of the present invention.

In some instances, well-known structures and devices may be omitted or shown in block diagram form centering on the core functions of the structures and devices in order to avoid obscuring the concepts of the present invention. In addition, the same components will be described with the same reference numerals throughout the present specification.

In this embodiment, a system in which I / Q imbalance occurs in a transmitting end and a receiving end of a multi-antenna communication system is considered. Conventionally, a technique for compensating the effects of the I / Q imbalance has been proposed, but for this technique, an estimation of the I / Q imbalance parameter is required. In addition, when the I / Q imbalance occurs at both the receiving end and the transmitting end, the number of variables to be obtained is greater than the number of equations so that direct calculation is impossible.

In a massive MIMO system where the number of antennas increases, supervised learning techniques that provide labels for all situations require that the number of training data required be increased because the number of I / Q imbalance parameters that must be estimated increases in proportion to the number of antennas. Not suitable for cursor use

On the other hand, reinforcement learning does not require training data, and the agent proceeds with the action in the direction of increasing the reward, which is suitable for the proposed problem.

In the following embodiment, an algorithm for estimating I / Q imbalance parameters using reinforcement learning is proposed in a multi-antenna system in which I / Q imbalance exists in a transceiver. The proposed algorithm is designed so that the system adaptively reduces the estimation error of parameters over time and is robust to changes in the system.

I / Q imbalances in the RF elements of multiple antennas interfere with in-phase and quadrature signals, significantly reducing the transmission speed. In order to compensate for this, an accurate I / Q imbalance parameter needs to be estimated. According to the present embodiment, since an accurate I / Q imbalance parameter can be estimated, wireless communication having a higher transmission speed is possible.

1 is a block diagram illustrating an I / Q imbalance parameter estimation system of an enhanced learning based multiple antenna transceiver according to an embodiment.

Referring to FIG. 1, an I / Q imbalance parameter estimating system of an enhanced learning based multi-antenna transceiver according to an embodiment includes a plurality of antennas based on enhanced learning based on I / Q at a base station (BS) and a user terminal (UT) of a system. The Q imbalance can be estimated.

According to an exemplary embodiment, an I / Q imbalance parameter estimating system 100 of a reinforcement learning based multi-antenna transceiver includes a compensator 120 and a compensation generator between a user terminal UT 110 and a base station BS 140. reward generator, 130).

The user terminal 110 may transmit a pilot signal to allow the base station 140 to estimate an effective channel.

The base station 140 increases or decreases the I / Q imbalance value by the existing step size based on the previously estimated I / Q imbalance parameter based on the received channel information, thereby determining the I / Q imbalance matrix of the corresponding value. The estimated value may be calculated to compensate for both the base station 140 and the user terminal 110.

The compensated channel information is transmitted to the input of the compensation generator 130, and the compensation generator 130 measures the degree to which the I / Q imbalance is compensated from the input matrix value. This is calculated by comparing the degree corresponding to the real part and the imaginary part in the augmented channel notation.

The reward is calculated in proportion to the amount compensated.

The base station 140 selects an action that gives a higher reward among the compensated rewards. In this case, the action is taken in such a way that the compensation increases during the increase or decrease of the parameter.

Set the next step size based on the reward received. If the existing compensation is large, the step size is reduced because the estimated value is close to the actual value. If the compensation is small, the step size is large because the estimated value is large from the actual value. The above procedure is repeated for all parameters.

Hereinafter, an I / Q imbalance parameter estimation system of an enhanced learning based multi-antenna transceiver according to an embodiment will be described in more detail with an example.

The I / Q imbalance parameter estimating system 100 of the reinforcement learning based multi-antenna transceiver according to an embodiment may include a compensator 120 and a compensation generator 130. Meanwhile, the transmitting end is the user terminal 110 and the receiving end is the base station 140, and the I / Q imbalance value is the amplitude imbalance value of the base station 140, the phase imbalance value of the base station 140, and the user terminal 110. ) May be an amplitude imbalance value and a user terminal 110 phase imbalance value.

The compensator 120 may receive a signal from the transmitter to estimate the effective channel. In addition, the compensator 120 arbitrarily sets an initial value of each parameter and measures an effective channel including an I / Q imbalance value in a multi-antenna channel in which at least one I / Q imbalance occurs in at least one of a transmitter and a receiver. In addition, the estimated value of the I / Q imbalance value can be calculated by calculating the I / Q imbalance value by increasing or decreasing the transmitter / receiver I / Q imbalance value estimated in the previous step and calculating the effective channel compensated for the I / Q imbalance value.

On the other hand, the receiving end selects an action that gives a higher reward among the rewarded rewards, and the compensator 120 selects a higher reward that is selected as the higher reward among the rewarded rewards is selected. The note can update the I / Q imbalance based on the behavior.

The compensator 120 sets the next step size based on the compensation. If the compensation is large by setting the step size proportional to the cost, the estimated value is close to the actual value, and thus the step size is reduced. For example, if the compensation is small, the estimated value is a large difference from the actual value, so that the step size can be set large.

The compensation generator 130 receives an effective channel matrix in which the I / Q imbalance is compensated from the compensator 120, measures a cost indicating the degree to which the I / Q imbalance is compensated, and compensates in proportion to the cost. (reward) can be calculated.

In addition, the compensation generator 130 receives an effective channel matrix compensated for I / Q imbalance, and measures a cost indicating the degree to which the I / Q imbalance is compensated by the compensation generator 130, and the channel notation It can be calculated by comparing the degree of the portion corresponding to the real part and the imaginary part in the augmented channel notation.

Iterative estimation is performed by randomly estimating and compensating the I / Q imbalance parameter from the effective channel measured based on reinforcement learning, and calculating the compensation according to the estimation degree of the I / Q imbalance. Can be adjusted adaptively to adjust the convergence speed and convergence accuracy.

According to the embodiments, the compensation calculation measures the degree to which the I / Q imbalance is compensated, converges to a value corrected over time, and has a strong characteristic against parameter changes. Since the size of the next action is determined based on the existing compensation value, the convergence speed is initially high and the accuracy of convergence is high. And it can be used to solve the I / Q imbalance that occurs in a communication system using a large number of antennas.

According to another embodiment, an I / Q imbalance parameter estimation system of an enhanced learning-based multi-antenna transceiver may arbitrarily set an initial value of each parameter in a multi-antenna channel having at least one I / Q imbalance between a transmitter and a receiver. We measure the effective channel that contains the I / Q imbalance value, and calculate the estimated value of the I / Q imbalance value by increasing or decreasing the transmitter or receiver I / Q imbalance value estimated in the previous step. The receiver may be configured to calculate a valid channel compensated for an imbalance, measure a cost representing a degree to which the I / Q imbalance is compensated, and calculate a reward in proportion to the cost.

The receiving end may select a higher rewarding action among the rewarded rewards to update the I / Q imbalance based on the selected higher rewarding action.

For example, the receiving end may include a compensator and a compensation generator, and this feature is omitted in the overlapping description of the I / Q imbalance parameter estimation system of the enhanced learning-based multi-antenna transceiver according to the above-described embodiment.

Meanwhile, the transmitting end is a user terminal and the receiving end is a base station, and the I / Q imbalance value may be a base station gain imbalance value, a base station phase imbalance value, a user terminal amplitude imbalance value, and a user terminal phase imbalance value.

2 is a flowchart illustrating a method for estimating an I / Q imbalance parameter of a reinforcement learning based multi-antenna transceiver according to an embodiment.

Referring to FIG. 2, in the I / Q imbalance parameter estimation method of an enhanced learning-based multi-antenna transceiver according to an embodiment, each parameter in a multi-antenna channel in which at least one I / Q imbalance occurs between a transmitter and a receiver is generated. In step 210, an initial value of is arbitrarily set, a step of measuring an effective channel including an I / Q imbalance value 220, and an increase or decrease of the transmitter or receiver I / Q imbalance value estimated in the previous step Calculating an estimated value of the I / Q imbalance value (230), calculating an effective channel compensated for I / Q imbalance using an estimated value of the calculated I / Q imbalance value (240), and I / Q Measuring a cost indicative of the degree to which the I / Q imbalance is compensated for using an effective channel compensated for an imbalance, calculating a reward in proportion to the cost (260), And step size based on the reward It may be made, including the step of setting (270).

Selecting (280) a higher reward action among the rewards compensated at the transmitting or receiving end and updating the I / Q imbalance based on the selected higher reward action (290), The above steps can be calculated repeatedly for each parameter.

According to the embodiments, iteratively estimates by a method of arbitrarily estimating and compensating I / Q imbalance parameters from the measured channel based on reinforcement learning, and calculating compensation based on the degree of estimation of the current I / Q imbalance. By adaptively adjusting the size of the action, the convergence speed and convergence accuracy can be controlled.

Hereinafter, a method for estimating an I / Q imbalance parameter of a reinforcement learning based multi-antenna transceiver according to an embodiment will be described in more detail.

An I / Q imbalance parameter estimation method of an enhanced learning based multi-antenna transceiver according to an embodiment is performed by using an I / Q imbalance parameter estimation system of an enhanced learning-based multi-antenna transceiver according to an embodiment described with reference to FIG. 1. For example, it can be explained. Here, the I / Q imbalance parameter estimation system of the reinforcement learning based multi-antenna transceiver according to an embodiment may be composed of a compensator and a reward generator between the user terminal (UT) and the base station (BS). Can be.

Step 210 may arbitrarily set an initial value of each parameter in a multi-antenna channel in which at least one I / Q imbalance occurs in at least one of a transmitter and a receiver. Meanwhile, the transmitting end may be a user terminal and the receiving end may be a base station, and the I / Q imbalance value may be a base station gain imbalance value, a base station phase imbalance value, a user terminal amplitude imbalance value, and a user terminal phase imbalance value. Can be.

In operation 220, an effective channel including an I / Q imbalance value may be measured.

In step 220, the estimated value of the I / Q imbalance value may be calculated by increasing or decreasing the transmitter or receiver I / Q imbalance value estimated in the previous step.

In operation 240, an estimated value of the calculated I / Q imbalance value may be used to calculate an effective channel compensated for I / Q imbalance. In this case, an effective channel compensated for I / Q imbalance is calculated using the estimated value of the I / Q imbalance value through the compensator, and then transferred to the compensation generator.

In step 250, the effective channel compensated for the I / Q imbalance may be used to measure the cost indicating the degree to which the I / Q imbalance is compensated. Here, a valid channel matrix compensated for I / Q imbalance is input to measure a cost representing the degree to which the I / Q imbalance is compensated by the compensation generator, and the real part is displayed in the augmented channel notation. It can be calculated by comparing the degree of the corresponding part with the imaginary part.

In step 260, a reward may be calculated in proportion to the cost.

In step 270, a next step size may be set based on the reward. Here, if the compensation is large by setting the step size in proportion to the cost, the step size is reduced because the estimated value is close to the actual value, and if the compensation is small, the step size is because the estimated value is different from the actual value. Can be set large.

Further, in step 280, an action that gives a higher reward among rewards compensated at the transmitting end or the receiving end may be selected.

In step 290, the I / Q imbalance may be updated based on the selected higher reward behavior. In this case, step 280 may be omitted, and the I / Q imbalance may be updated based on the higher rewarding action selected by selecting a higher rewarding action among the rewards compensated at the transmitting end or the receiving end. .

Hereinafter, a method and system for estimating I / Q imbalance parameters of an enhanced learning-based multi-antenna transceiver according to an embodiment will be described in more detail with one example.

First, a multi-antenna received signal model in which the transmit / receive end I / Q imbalance occurs will be described.

In a multi-antenna system in which a base station having N antennas and a single user antenna have K users, the downlink reception signal may be represented as follows in the case of I / Q imbalance at both a transmitting and receiving end. Here, N is the number of antennas of the base station (BS), K represents the number of user terminals (UT).

[Equation 1]

는 오그멘티드(augmented) 수신 심볼 벡터,

이고,

는 오그멘티드(augmented) 백색잡음(AWGN) 벡터,

이며,

는 오그멘티드(augmented) 송신 심볼 벡터,

이다.

는 오그멘티드(augmented) 채널 행렬,

이고,

는 프리코더(precoder) 행렬,

이다. 또한,

는 UT I/Q 불균형 행렬,

이고,

는 BS I/Q 불균형 행렬,

이며,

는 잡음 I/Q 불균형 행렬,

이다. here,

Is an augmented received symbol vector,

ego,

Is an augmented white noise (AWGN) vector,

Is,

Is an augmented transmission symbol vector,

to be.

Is an augmented channel matrix,

ego,

Is the precoder matrix,

to be. Also,

Is the UT I / Q imbalance matrix,

ego,

BS I / Q imbalance matrix,

Is,

Is the noise I / Q imbalance matrix,

to be.

은 다음 식과 같이 모델링이 가능하다.And,

Can be modeled as follows.

[Equation 2]

는 치환(permutation) 행렬,

이고,

는

을 위한 정규화 상수(normalize constant)이며, 기지국 치환 행렬(permutation matrix)

는 다음 식과 같은 성분을 가진다.

는 추적 연산자(Trace operator)이다. here,

Is the permutation matrix,

ego,

Is

Normalization constant for the base station permutation matrix (permutation matrix)

Has the following components.

Is a trace operator.

[Equation 3]

는 I 성분과 Q 성분이 분리되어 있는 오그멘티드 벡터(augmented vector)를 RF 체인(chain)별로 순서를 바꾸는 역할을 한다.

는 다음 식과 같이 정의될 수 있다.here,

Is used to change the order of the augmented vector in which the I and Q components are separated for each RF chain.

Can be defined as

[Equation 4]

는 n 번째 RF 체인의 I/Q 불균형 성분 행렬이며, 다음 식과 같이 나타낼 수 있다.bracket

Is the I / Q imbalance component matrix of the nth RF chain and can be expressed as follows.

[Equation 5]

는 n 번째 BS의 진폭 불균형이고,

는 n 번째 BS의 위상 불균형이다.here,

Is the amplitude imbalance of the nth BS,

Is the phase imbalance of the nth BS.

The I / Q imbalance matrix generated in the user terminal is modeled in a structure similar to the base station I / Q imbalance matrix, and can be expressed as follows.

[Equation 6]

는

을 위한 정규화 상수이며, 사용자 단말 치환 행렬

는 다음 식과 같은 성분을 가질 수 있다. here,

Is

Normalization constant for, user terminal substitution matrix

May have a component such as

[Equation 7]

는 다음 식과 같이 정의될 수 있다.here,

Can be defined as

[Equation 8]

는 k 번째 RF 체인의 I/Q 불균형 성분 행렬이며, 다음 식과 같이 나타낼 수 있다.bracket

Is the I / Q imbalance component matrix of the k- th RF chain and can be expressed as follows.

[Equation 9]

는 k 번째 UT의 진폭 불균형이고,

는 k 번째 UT의 위상 불균형이다. here,

Is the amplitude imbalance of the k th UT,

Is the phase imbalance of the k th UT.

는 다음 식과 같이 모델링될 수 있다.I / Q imbalances in the user terminal RF chain also affect noise. In an ideal RF chain, there is no correlation between noise generated because the carriers are orthogonal to each other. However, if I / Q imbalance occurs, the orthogonality of the carrier is broken, so that correlation between in-phase noise and quadrature phase noise occurs. Matrix representing this

Can be modeled as follows.

[Equation 10]

는

을 위한 정규화 상수이며,

는 다음 식과 같이 정의될 수 있다.here,

Is

Normalization constant for

Can be defined as

[Equation 11]

는 k 번째 RF 체인의 I/Q 불균형 성분 행렬이며, 다음 식과 같이 나타낼 수 있다.bracket

Is the I / Q imbalance component matrix of the k- th RF chain and can be expressed as follows.

[Equation 12]

Next, the reinforcement learning based I / Q imbalance estimation algorithm will be described.

In a multi-antenna channel in which I / Q imbalance occurs at both the base station and the user terminal, an effective channel matrix including the I / Q imbalance matrix may be defined as follows.

[Equation 13]

는 유효 채널 행렬,

이다.here,

Is the effective channel matrix,

to be.

,

를 추정하면, 이를 유효 채널의 양쪽에서 역행렬을 곱하는 방법으로 보상할 수 있다. 보상된 유효 채널

는 다음 식과 같이 정의할 수 있다.I / Q imbalance matrix between base station and user terminal

,

We can compensate for this by multiplying the inverse of both effective channels. Compensated Effective Channel

Can be defined as

[Equation 14]

는 I/Q 불균형을 보상된 유효 채널 행렬,

이고,

는 추정된 UT I/Q 불균형 행렬,

이며,

는 추정된 BS I/Q 불균형 행렬,

이다. here,

Is the effective channel matrix compensated for I / Q imbalance,

ego,

Is the estimated UT I / Q imbalance matrix,

Is,

Is the estimated BS I / Q imbalance matrix,

to be.

와

는 각각 보상된 I/Q 불균형 행렬을 나타낸다. I/Q 불균형을 완전히 추정한 경우,

가 성립하여

가 성립한다. 여기서,

는

단위 행렬(Identity matrix)이고,

는 복합 공액 전치(conjugate transpose)이다.

Wow

Denotes the compensated I / Q imbalance matrix, respectively. If you fully estimate the I / Q imbalance,

Is established

Is established. here,

Is

Is an identity matrix,

Is a conjugate conjugate.

From this, a cost representing the degree to which each parameter of the I / Q imbalance is compensated can be defined as follows.

[Equation 15]

Where U (a, b) represents an equal probability distribution having a minimum value a and a maximum value b.

에서 해당하는 파라미터가 얼마나 보상되었는지 나타내며, I/Q 불균형이 없는 이상적인 경우,

이 성립한다.
Each cost is

Indicates how much of the corresponding parameter is compensated for in the ideal case without I / Q imbalance,

This holds true.

The above properties can be applied to reinforcement learning to estimate the I / Q imbalance.

In this case, the agent is a communication system, and the environment is a channel environment. The agent's action is set to increase or decrease the estimation parameter. The reward, which is the measure by which the agent decides the next action, is set in inverse proportion to the cost, and the agent is designed to receive a higher reward as it approaches the correct value.

In order to increase convergence and convergence speed, the step size is set proportionally, so that the initial step is large, the step size is increased to increase the convergence speed, and the smaller the step size is, the more accurate the value is estimated to converge. It was.

Based on the above, we propose the following algorithm. That is, an I / Q imbalance parameter estimation method and system of a speech learning based multi-antenna transceiver according to an embodiment may be described.

In a multiple antenna channel where I / Q imbalance occurs at both the base station and the user terminal, an initial value of each parameter is arbitrarily set.

The effective channel including the I / Q imbalance matrix can be measured and expressed as follows.

[Equation 16]

는 유효 채널 행렬,

이다.
here,

Is the effective channel matrix,

to be.

BS amplitude gain imbalance update

i에 대해 다음을 반복한다.All base station antenna indexes

Repeat for i

The I / Q imbalance estimation matrix can be calculated when the base station amplitude imbalance estimated in the previous step is increased or decreased, and can be expressed as follows. The amplitude and phase imbalances of the other indices can be reflected as previously estimated.

Formula 17

는 진폭 불균형 추정 단계 크기(step size)이다. here,

Is the amplitude imbalance estimation step size.

Compensation for each case can be calculated with the reward function defined above and can be expressed as the following equation. Meanwhile, the amplitude / phase imbalance of the user terminal may reflect a previously estimated value.

[Equation 18]

The next step size can be set from the cost and can be expressed as follows.

[Equation 19]

The most rewarding case is selected and can be expressed as the following equation.

[Equation 20]

The base station amplitude imbalance can be updated with the corresponding value, and can be expressed as follows.

Formula 21

UT amplitude gain imbalance update

The same process may be repeated by changing the base station described above to a user terminal.

Base station (BS) phase imbalance update

i에 대해 다음을 반복한다.All base station antenna indexes

Repeat for i

The I / Q imbalance estimation matrix can be calculated when the base station amplitude imbalance estimated at the previous step is increased or decreased, and can be expressed as follows. The amplitude and phase imbalances of the other indices are reflected as previously estimated.

Formula 22

는 위상 불균형 추정 단계 크기(step size)이다. here,

Is the phase imbalance estimation step size.

Compensation for each case can be calculated with the compensation function defined above, and can be expressed as the following equation. The amplitude imbalance and the phase imbalance of the user terminal UT reflect a previously estimated value.

Formula 23

The next step size can be set from the cost and can be expressed as follows.

Formula 24

The most rewarding case can be selected and can be expressed as the following equation.

[Equation 25]

The base station amplitude imbalance can be updated with the corresponding value, which can be expressed as the following equation.

Formula 26

User terminal (UT) phase imbalance update

The same process may be repeated by changing the base station described above to a user terminal. The effective channel measurement step can then be returned.

The simulation environment may assume that I / Q imbalance occurs in both the base station and the user terminal in the MIMO system.

는 레일리 페이딩 채널(Rayleigh fading channel)이고,

는 백색잡음으로 발생할 수 있다.

Is a Rayleigh fading channel,

Can occur as white noise.

The initial estimated value can be expressed as follows.

[Equation 27]

The initial stage size can be expressed as

[Equation 28]

로 설정할 수 있다.

Can be set to

Experiment with pilot SNR for channel estimation from 0dB to 15dB.

3 illustrates an amplitude imbalance parameter estimation MSE according to an embodiment.

4 is a diagram illustrating a phase imbalance parameter estimation MSE according to an embodiment.

3 and 4, the performance of the method and system for estimating the I / Q imbalance parameter of the enhanced learning-based multi-antenna transceiver according to an embodiment may be checked. Performance was measured according to the pilot SNR, and the mean square error (MSE) between the estimated value and the actual parameter value, which is an estimated I / Q imbalance parameter value, was measured.

It was confirmed that the convergence to a low MSE over time, and even after the iteration proceeds, it can be seen that the MSE does not change significantly. An I / Q imbalance parameter estimation method and system of an enhanced learning-based multi-antenna transceiver according to an embodiment may be used to estimate and compensate an accurate value in an I / Q imbalance system by iterating an accurate value through iteration. Can be.

According to the embodiments, a method for solving an I / Q imbalance problem that can easily occur from a large number of antennas in a large array multi-antenna technology, which is noted as a core technology of future wireless communication, and the future of using wireless communication technology and multiple antennas It is expected to be applied to various fields such as vehicle communication. In particular, according to the embodiments, it is possible to solve a problem occurring in an RF device that solves a high device cost problem that is commonly encountered in a large antenna array technology, which is attracting attention as a core technology of future wireless communication.

The apparatus described above may be implemented as a hardware component, a software component, and / or a combination of hardware components and software components. For example, the devices and components described in the embodiments include, for example, processors, controllers, arithmetic logic units (ALUs), digital signal processors, microcomputers, field programmable arrays (FPAs), It may be implemented using one or more general purpose or special purpose computers, such as a programmable logic unit (PLU), microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to the execution of the software. For convenience of explanation, one processing device may be described as being used, but one of ordinary skill in the art will appreciate that the processing device includes a plurality of processing elements and / or a plurality of types of processing elements. It can be seen that it may include. For example, the processing device may include a plurality of processors or one processor and one controller. In addition, other processing configurations are possible, such as parallel processors.

The software may include a computer program, code, instructions, or a combination of one or more of the above, and configure the processing device to operate as desired, or process it independently or collectively. You can command the device. Software and / or data may be any type of machine, component, physical device, virtual equipment, computer storage medium or device in order to be interpreted by or to provide instructions or data to the processing device. It can be embodied in. The software may be distributed over networked computer systems so that they may be stored or executed in a distributed manner. Software and data may be stored on one or more computer readable recording media.

The method according to the embodiment may be embodied in the form of program instructions that can be executed by various computer means and recorded in a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the media may be those specially designed and constructed for the purposes of the embodiments, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like.

Although the embodiments have been described by the limited embodiments and the drawings as described above, various modifications and variations are possible to those skilled in the art from the above description. For example, the described techniques may be performed in a different order than the described method, and / or components of the described systems, structures, devices, circuits, etc. may be combined or combined in a different form than the described method, or other components. Or even if replaced or substituted by equivalents, an appropriate result can be achieved.

Therefore, other implementations, other embodiments, and equivalents to the claims are within the scope of the claims that follow.

Claims (15)

Arbitrarily setting an initial value of each parameter in a multi-antenna channel in which at least one I / Q imbalance occurs in at least one of a transmitting end and a receiving end;
Measuring an effective channel including an I / Q imbalance value;
Calculating an estimated value of the I / Q imbalance value by increasing or decreasing the transmitter / receiver I / Q imbalance value estimated in the previous step;
Calculating an effective channel compensated for I / Q imbalance by using the estimated value of the calculated I / Q imbalance value;
Measuring a cost indicating a degree to which an I / Q imbalance is compensated for by using an effective channel compensated for the I / Q imbalance;
Calculating a reward in proportion to the cost;
Setting a next step size based on the compensation;
Selecting a higher rewarding action among the rewards compensated at the transmitting end or the receiving end; And
Updating the I / Q imbalance based on the higher rewarding behavior selected
Including,
Setting a next step size based on the compensation,
If the compensation is large because the step size is set proportionally to the cost, the estimated value is close to the actual value, and the step size is reduced. If the compensation is small, the estimated value is different from the actual value. Set the step size to large,
Iteratively calculating the above steps for each parameter, iteratively estimating by means of randomly estimating and compensating I / Q imbalance parameters from the effective channels measured based on reinforcement learning, and calculating the I / Q imbalance. Adjusting convergence speed and convergence accuracy by adaptively adjusting the magnitude of the action by calculating the compensation according to the estimation degree
A method for estimating I / Q imbalance parameters of a reinforcement learning based multiple antenna transceiver. delete The method of claim 1,
Computing an effective channel compensated for I / Q imbalance using the estimated value of the I / Q imbalance value,
Computing an effective channel compensated for I / Q imbalance using the estimated value of the I / Q imbalance value through a compensator and passing it to a compensation generator
A method for estimating I / Q imbalance parameters of a reinforcement learning based multiple antenna transceiver. The method of claim 1,
Measuring a cost representing the degree to which the I / Q imbalance is compensated using the effective channel compensated for the I / Q imbalance,
A valid channel matrix compensated for the I / Q imbalance is input to measure a cost representing the degree to which the I / Q imbalance is compensated by a compensation generator, and corresponds to a real part in an augmented channel notation. To calculate the degree of comparison between the part corresponding to the imaginary part
A method for estimating I / Q imbalance parameters of a reinforcement learning based multiple antenna transceiver. delete The method of claim 1,
The transmitting end is a user terminal and the receiving end is a base station,
The I / Q imbalance value is
Base station gain unbalance value, base station phase unbalance value, user terminal amplitude unbalance value, and user terminal phase unbalance value
A method for estimating I / Q imbalance parameters of a reinforcement learning based multiple antenna transceiver. In a multi-antenna channel in which at least one of the transmitting and receiving I / Q imbalances occurs, the initial value of each parameter is arbitrarily set, the effective channel including the I / Q imbalance value is measured, and in the previous step The estimated value of the I / Q imbalance value is calculated by increasing or decreasing the estimated I / Q imbalance value of the transmitting end or the receiving end, and the effective channel compensated for the I / Q imbalance is calculated to indicate the degree to which the I / Q imbalance is compensated. A receiver that measures a cost and calculates a reward in proportion to the cost
Including,
The receiving end,
Selecting a higher rewarding action among the rewarded rewards to update the I / Q imbalance based on the higher rewarding action selected, and based on the reward to determine the next step size. Setting the step size in proportion to the cost, if the compensation is large, the estimated value is close to the actual value, and decreasing the step size; if the compensation is small, the estimated value is different from the actual value. Is large, so set the above step size to be large,
Iterative estimation is performed by randomly estimating and compensating I / Q imbalance parameters from the effective channel measured based on reinforcement learning, and calculating compensation based on the degree of estimation of the I / Q imbalance. Adjusting convergence speed and convergence accuracy by adaptively scaling
I / Q imbalance parameter estimation system of the reinforcement learning-based multiple antenna transceiver. delete The method of claim 7, wherein
The transmitting end is a user terminal and the receiving end is a base station,
The I / Q imbalance value is
Base station gain unbalance value, base station phase unbalance value, user terminal amplitude unbalance value, and user terminal phase unbalance value
I / Q imbalance parameter estimation system of the reinforcement learning-based multiple antenna transceiver. A compensator for receiving a signal from a transmitter and estimating an effective channel; And
Compensation that receives a valid channel matrix compensated for I / Q imbalance from the compensator, measures a cost representing the degree to which the I / Q imbalance is compensated, and calculates a reward in proportion to the cost. Generator
Including,
The compensator,
In a multi-antenna channel in which at least one of the transmitting and receiving I / Q imbalances occurs, the initial value of each parameter is arbitrarily set, the effective channel including the I / Q imbalance value is measured, and in the previous step Increasing or decreasing the estimated I / Q imbalance value of the transmitter or receiver to calculate an estimated value of the I / Q imbalance value and calculating an effective channel compensated for I / Q imbalance,
The receiving end,
Select an action that gives a higher reward among the rewarded rewards,
The compensator,
Update the I / Q imbalance based on the higher rewarding action selected, set a next step size based on the reward, and set the step size proportionately to the reward If the value is large, the step size is reduced because the estimated value is close to the actual value, and if the compensation is small, the step size is set large because the estimated value is large from the actual value,
Iterative estimation is performed by randomly estimating and compensating I / Q imbalance parameters from the effective channel measured based on reinforcement learning, and calculating compensation based on the degree of estimation of the I / Q imbalance. Adjusting convergence speed and convergence accuracy by adaptively scaling
I / Q imbalance parameter estimation system of the reinforcement learning-based multiple antenna transceiver. delete The method of claim 10,
The compensation generator,
A valid channel matrix compensated for the I / Q imbalance is input to measure a cost representing the degree to which the I / Q imbalance is compensated by a compensation generator, and corresponds to a real part in an augmented channel notation. To calculate the degree of comparison between the part corresponding to the imaginary part
I / Q imbalance parameter estimation system of the reinforcement learning-based multiple antenna transceiver. delete The method of claim 10,
The transmitting end is a user terminal and the receiving end is a base station,
The I / Q imbalance value is
Base station gain unbalance value, base station phase unbalance value, user terminal amplitude unbalance value, and user terminal phase unbalance value
I / Q imbalance parameter estimation system of the reinforcement learning-based multiple antenna transceiver. delete

Images