KR102518990B1 - Device for controlling transmission power considering imperfect channel state information - Google Patents

Abstract

The present invention relates to an apparatus for controlling transmission power considering inaccurate channel state information. The present invention is based on a first information module for inputting first channel state information (CSI), a second information module for inputting second channel state information (CSI), and the first channel state information (CSI). Using a DNN control module to control the transmission power of D2D user equipments (DUE) transceiver pairs (TPs) between devices, transmission power of the DNN control module and second channel state information (CSI) and a determination module for determining a loss function and updating parameters of the DNN module according to the loss function.

Description

Transmission power control device considering inaccurate channel state information {Device for controlling transmission power considering imperfect channel state information}

The present invention relates to a wireless communication system, and more particularly, to a transmission power control device considering inaccurate channel state information.

In underlay D2D (underlay D2D) communication, when radio resources of cellular user equipments (CUEs) are shared by user equipments (DUEs) between multiple devices, transmission of user equipments (DUEs) between devices can cause great interference to both user equipments (DUEs) between other devices, so the transmission power of user equipments (DUEs) between devices must be carefully determined.

Accurate channel state information (CSI) must be known by device-to-device user equipment (DUE) in order to optimally adjust transmit power. In practice, it is difficult to perfectly obtain channel state information (CSI) due to time-varying characteristics of a channel, quantization error of feedback for channel state information (CSI), and channel state information (CSI) estimation error.

Therefore, it is important to consider the channel state information (CSI) for the actual design of the transmit power control algorithm. Technology considering the influence of inaccurate channel state information (CSI) using stochastic geometry in a recent wireless communication system (WCS), technology for maximizing energy efficiency (EE) through channel uncertainty, and cellular user equipment Technology that maximizes the total speed of user equipment (DUE) between devices while guaranteeing QoS of (CUE), using a Deep Neural Network (DNN) structure, so that the DNN structure is directly based on the wireless channel without a theoretical formula. Research on D2D communication using channel state information (CSI), such as a technique for controlling the operation of a D2D communication, is being actively conducted. Among them, deep neural network (DNN)-based transmission power control technology is drawing great attention because it can approximate an optimal strategy with a DNN model that can be executed with a short calculation time.

However, conventional deep neural network (DNN)-based transmit power control technology can approach optimal transmit power control by using DNN-based Weighted Minimum Mean Square Error (WMMSE), , each user can control on/off transmission power based on DNN, which determines whether or not to distribute transmission. However, the conventional DNN-based transmit power control technology considers an environment in which channel state information (CSI) is accurate and does not consider an inaccurate environment. That is, the conventional technology has a problem in that it is not suitable for actual use because channel information may become inaccurate according to various factors in an actual wireless system.

The present invention is to solve the above-mentioned problems of the prior art, and an object of the present invention is to provide a transmission power control device that determines optimal transmission power even in an environment where channel information is inaccurate by using deep learning.

An object of the present invention is to provide a reliable transmission power control device for channel errors by calculating a loss value through a channel to which random noise is randomly added to a loss function while adjusting transmission power using an input channel value.

Objects of the present invention are not limited to those mentioned above, and other objects not mentioned above will be clearly understood by those skilled in the art from the description below.

The present invention for achieving the above object is a first information module for inputting first channel state information (CSI), a second information module for inputting second channel state information (CSI), the first A DNN control module for controlling transmit power of D2D user equipments (DUE) transceiver pairs (TPs) between devices based on channel state information (CSI), transmit power of the DNN control module and a second channel Provides transmit power control considering inaccurate channel state information including a determination module for determining a loss function using state information (CSI) and updating parameters of the DNN module according to the loss function.

The first channel state information is channel state information estimated in an inaccurate environment.

The second channel state information is arbitrarily distorted channel state information.

The DNN control module may obtain optimal transmission power of user equipment (DUE) transceiver pairs (TPs) using the updated parameters of the DNN module.

The first information module includes a normalizer for normalizing transmission power of the device-to-device user equipment (DUE) transceiver pairs (TPs) based on the first channel state information (CSI), the device-to-device user equipment (DUE) transceiver A reconstruction unit for reconstructing the transmission power of the pairs (TPs) from 3D to 1D is further included.

The DNN control module determines a first DNN unit for approximating the transmit power of the user equipment (DUE) transceiver pairs (TPs) between devices, and a ratio of transmit power of each user equipment (DUE) transceiver pair (TP) between devices. A second DNN unit for approximation is included.

The first DNN unit is characterized in that an output layer is a sigmoid layer.

The second DNN unit is characterized in that an output layer is a softmax layer.

The determination module is characterized in that it learns to minimize the value of the loss function.

The determination module calculates a transmission power ratio of the first DNN unit, a transmission power ratio of the second DNN unit, and a maximum transmission power of user equipment (DUE) transceiver pairs (TPs) between devices to determine a set of transmission power characterized by

The determination module may determine a loss function using the set of transmit powers and the second channel state information (CSI).

According to the transmit power control apparatus considering inaccurate channel state information of the present invention as described above, it is possible to determine and control optimal transmit power even in an environment where channel state information is inaccurate by using deep learning. In addition, transmission power of D2D user equipment (DUE) is optimized in each channel to maximize average spectral efficiency (SE).

The present invention maintains quality of service (QoS) of cellular user equipment (CUE) when there is an error in the channel state information (CSI) and uses artificially distorted CSI to inaccurate channel state information. By compensating for the effect of (CSI), reliable transmission power control for channel errors can be achieved.

The effects obtainable in the present invention are not limited to the effects mentioned above, and other effects not mentioned can be clearly understood by those skilled in the art from the description below. will be.

)에 대한 스펙트럼 효율(SE)의 그래프
도 4a 내지 4c는 본 발명에 따라 가변 간섭의 임계(

)에 대한 QoS 위반 확률의 그래프
도 5는 스펙트럼 효율(SE)에 대한 비교 그래프
도 6은 QoS 위반 확률에 대한 비교 그래프
도 7은 장치간 사용자 장비(DUE) 트랜스시버 쌍들(TPs)의 수(N)에 따른 스펙트럼 효율(SE)의 그래프
도 8은 장치간 사용자 장비(DUE) 트랜스시버 쌍들(TPs)의 수(N)에 따른 계산시간의 그래프1 is a block diagram for explaining an apparatus for controlling transmission power considering inaccurate channel state information according to the present invention.
Figure 2 is a block diagram for explaining the DNN control module and decision module of Figure 1
3a to 3c show the variable interference threshold (

Graph of spectral efficiency (SE) versus )
4a to 4c show the variable interference threshold (

) graph of QoS violation probability
5 is a comparison graph for spectral efficiency (SE)
6 is a comparison graph of QoS violation probability
7 is a graph of spectral efficiency (SE) as a function of number (N) of user equipment (DUE) transceiver pairs (TPs) between devices;
8 is a graph of calculation time as a function of the number (N) of user equipment (DUE) transceiver pairs (TPs) between devices;

Objects and effects of the present invention, and technical configurations for achieving them will become clear with reference to embodiments described later in detail in conjunction with the accompanying drawings. In describing the present invention, if it is determined that a detailed description of a known function or configuration may unnecessarily obscure the gist of the present invention, the detailed description will be omitted.

In addition, terms to be described later are terms defined in consideration of functions in the present invention, which may vary according to the intention or custom of a user or operator.

However, the present invention is not limited to the embodiments disclosed below and may be implemented in a variety of different forms. Only these embodiments are provided to complete the disclosure of the present invention and to fully inform those skilled in the art of the scope of the invention, and the present invention is defined by the scope of the claims. It only becomes. Therefore, the definition should be made based on the contents throughout this specification.

Although the present invention is described by the limited embodiments and drawings, the present invention is not limited thereto, and is described below and the technical spirit of the present invention by those skilled in the art to which the present invention belongs. Of course, various modifications and variations are possible within the scope of equivalence of the claims to be made.

Hereinafter, the present invention will be described in more detail based on the embodiments shown in the drawings. However, the present application is not limited to these examples and drawings.

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

1 is a block diagram for explaining an apparatus 1000 for controlling transmission power considering inaccurate channel state information according to the present invention.

As shown in FIG. 1, the transmission power control apparatus 1000 considering inaccurate channel state information according to the present invention includes a first information module 100 for inputting first channel state information (CSI), A second information module 200 for inputting second channel state information (CSI), a DNN control module 300 for controlling transmission power, determining a loss function and updating parameters of the DNN module according to the loss function module 400.

The first information module 100 receives first channel state information, which is estimated channel state information (CSI) estimated in an imperfect environment.

The first information module 100 normalizes the transmission power of D2D user equipments (DUE) transceiver pairs (TPs) between devices based on the first channel state information (CSI). ) and a reconfiguration unit 120 for reconstructing transmission power of user equipment (DUE) transceiver pairs (TPs) between the devices from three dimensions to one dimension.

The second information module 200 receives second channel state information, which is arbitrarily distorted CSI. Here, the second channel state information may be calculated based on the first channel state information of the first information module 100 .

The DNN control module 300 determines to control transmission power of user equipment (DUE) transceiver pairs (TPs) between devices using a DNN method, and the first DNN unit 310 and the second DNN unit 210 includes The first DNN unit 112 and the second DNN unit 114 are based on a feed-forward neural network (FFNN, FNN) in which operations are developed from the input layer to the output layer to calculate continuous outputs. can be done with

The first DNN unit 310 approximates the transmit power of individual device-to-device user equipment (DUE) transceiver pairs (TPs), and the second DNN unit 320 approximates the transmit power of device-to-device user equipment (DUE) transceiver pairs (TPs) in K channels. ) approximates the ratio of the transmit power of

The decision module 400 is responsible for determining a loss function, updating the parameters of the DNN module according to the loss function, and learning to minimize the value of the loss function.

)을 연산하여 전송 전력의 집합(

)을 결정한다. 그리고, 전송 전력의 집합(

)과 상기 제2 정보모듈(200)의 임의로 왜곡된 채널 상태 정보(arbitrarily distorted CSI)인 제2 채널 상태 정보를 이용하여 손실함수(loss function) 결정하게 된다.The determination module 400 determines the ratio of the transmit power of the first DNN unit 310 to the transmit power of the second DNN unit 320 and the maximum transmit power of the device-to-device user equipment (DUE) transceiver pairs (TPs) (

) to set the transmission power (

) to determine And, a set of transmission power (

) and second channel state information, which is arbitrarily distorted CSI of the second information module 200, to determine a loss function.

FIG. 2 is a block diagram for explaining the DNN control module 300 and decision module 400 of FIG. 1 .

The DNN control module 300 includes a first DNN unit 310 and a second DNN unit 320, and the first DNN unit 310 and the second DNN unit 320 are fully connected layers. ), a sigmoid layer or a softmax layer may be configured alone or in combination thereof.

According to an embodiment of the present invention, the first DNN unit 310 and the second DNN unit 320 include a fully connected layer and a plurality of Rectified Linear Unit (ReLU) layers serially connected to the fully connected layer, The output layer of the first DNN unit 310 may be a sigmoid layer 311, and the output layer of the second DNN unit 114 may be a softmax layer 321.

)와 숨겨진 노드 수(

)가 설정되고, 완전 연결 레이어(FC)에서 출력된 출력값을 ReLU 레이어로 전달한다. 이때, 제1 및 제2 DNN부(310, 320)는 ReLU 레이어로 전달될 출력값이 음수일 경우 차단할 수 있다.The first and second DNN units 310 and 320 have the number of layers in the fully connected layer in the operating mode (

) and the number of hidden nodes (

) is set, and the output value output from the fully connected layer (FC) is transferred to the ReLU layer. In this case, the first and second DNN units 310 and 320 may block when an output value to be delivered to the ReLU layer is a negative number.

이 입력될 경우

을 출력할 수 있다. 이는 DNN의 비선형성을 보여주기 위함이다.The ReLU layer is

If is entered

can output This is to show the nonlinearity of DNN.

)에 대해 출력값(

)으로 나타낼 수 있다. 제1 DNN부(310)의 출력 수는 N개의 장치간 사용자 장비(DUE) 트랜스시버 쌍(TP)의 정규화된 전송 전력이 적절하게 모델링될 수 있도록 N으로 설정된다.The first DNN unit 310 may use a sigmoid layer 311 as an output layer, and the sigmoid function is an input value (

) for the output value (

) can be expressed as The number of outputs of the first DNN unit 310 is set to N so that the normalized transmit power of N device-to-device user equipment (DUE) transceiver pairs (TPs) can be appropriately modeled.

)에 대해 출력값(

)을 가지게 된다. The second DNN unit 320 may use a softmax layer 321 as an output layer, and the softmax function is an input value (

) for the output value (

) will have

Meanwhile, the first DNN unit 310 and the second DNN unit 320 of the DNN control module 300 may perform unsupervised learning to approximate. Of course, it can also be learned through a supervised learning-based system. This is to evaluate the loss function using the second channel state information, which is arbitrarily distorted CSI, so that the decision module 400 can consider the inconsistency of the channel state information (CSI) during DNN training. This is for the DNN control module 300 to learn reliable transmission power control for channel errors.

의 송신기와

의 수신기 사이의 채널 이득은

라 할 때, 제1 정보모듈(100)의 추정된 채널 상태 정보(estimated CSI)인 제1 채널 상태 정보를

, 2 정보모듈(200)의 임의로 왜곡된 채널 상태 정보(arbitrarily distorted CSI)인 제2 채널 상태 정보(

)라 가정한다.First, for channel k

with the transmitter of

The channel gain between the receivers of

When , the first channel state information that is the estimated channel state information (estimated CSI) of the first information module 100

, 2nd channel state information that is arbitrarily distorted CSI of the information module 200 (

) is assumed.

)는 채널 추정 오류 및 무선 채널의 시간 변화 특성으로 인해 실제 값(

)과 차이가 있을 수 있으므로, 제1 채널 상태 정보(

)와 실제 값(

)을 하기 (수학식 1)과 같이 모델링할 수 있다.In this case, the channel gain includes both distance-related path loss and multi-path fading, and index 0 is assigned to the transmitter and receiver of legacy cellular transmission, that is, CUE and cellular BS. And first channel state information (

) is the actual value (

), the first channel state information (

) and the actual value (

) can be modeled as follows (Equation 1).

는 채널 이득의 실제값,

는 제1 정보모듈(100)의 제1 채널 상태 정보,

는 채널 불확실을 나타낸다.here

is the actual value of the channel gain,

Is the first channel state information of the first information module 100,

represents the channel uncertainty.

)의 매트릭스를

으로, 제1 정보모듈(100)의 제1 채널 상태 정보(

)에 대한 매트릭스를

라 가정한다.Also, the actual value (

) matrix of

, the first channel state information of the first information module 100 (

) matrix for

Assume that

)을 근사화하고, 제2 DNN부(320)가 각 채널에 할당된 전송 전력의 비율(

)을 근사화한다.As shown in FIG. 2, in the DNN control module 300 receiving the information of the first information module 100, the first DNN unit 310 normalizes user equipment (DUE) transceiver pairs (TP) between individual devices. total transmitted power (

), and the ratio of transmit power allocated to each channel by the second DNN unit 320 (

) approximates

는 제1 채널 상태 정보에 대한 DUE TP i의 전송 전력이 기입된 함수이며,

와 같이 표현될 수 있다.

는

와 같은 DNN 모델의 파라미터 집합이며,

는 장치간 사용자 장비(DUE) 트랜스시버 쌍들(TPs)의 최대 전송 전력이다.here,

Is a function in which the transmit power of DUE TP i for the first channel state information is written,

can be expressed as

Is

A set of parameters of a DNN model such as

is the maximum transmit power of user equipment (DUE) transceiver pairs (TPs) between devices.

)과 제2 DNN부(320)에서 출력된 전송 전력의 비율(

)에 장치간 사용자 장비(DUE) 트랜스시버 쌍들(TPs)의 최대 전송 전력(

)을 곱셈 연산하여 채널 k에 할당된 DUE TP i의 전송 전력의 집합(

)을 결정한다.The determination module 400 determines the total transmit power output from the first DNN unit 310 (

) and the ratio of the transmission power output from the second DNN unit 320 (

) to the maximum transmit power of user equipment (DUE) transceiver pairs (TPs) between devices (

) to obtain a set of transmission powers of DUE TP i assigned to channel k (

) to determine

라 하고, 각 DUE TP의 최대 전송 전력(

)은

조건을 만족하도록 설정할 수 있다. The set of transmit powers of DUE TP i is

, and the maximum transmission power of each DUE TP (

)silver

It can be set to satisfy the condition.

) 과 제2 정보모듈(200)의 임의로 왜곡된 채널 상태 정보(arbitrarily distorted CSI)인 제2 채널 상태 정보를 이용하여 손실함수를 결정한다.The determination module 400 sets the transmission power (

) and second channel state information, which is arbitrarily distorted CSI of the second information module 200, to determine a loss function.

)는 부정확한 환경에서 추정된 채널 상태 정보(

)의 영향을 보상하기 위해 하기 (수학식 2)와 같이 추정된 채널 상태 정보인 제1 채널 상태 정보(

)에 무작위로 생성된 채널 상태 정보(

)를 더하는 연산으로

를 생성할 수 있다. Second channel state information (which is arbitrarily distorted CSI)

) is channel state information (estimated in an inaccurate environment)

In order to compensate for the influence of ), the first channel state information (estimated channel state information as shown in Equation 2) (

) randomly generated channel state information (

) by adding

can create

)를 기반으로 하기 (수학식 3)의 손실함수를 사용하고, 하기 (수학식 4) 및 (수학식 5)를 이용하여 DNN 제어모듈(300)의 파라미터(

)를 업데이트한다.Then, the determination module 400 determines the second channel state information (

), using the loss function of (Equation 3) below, and using the following (Equation 4) and (Equation 5), the parameters of the DNN control module 300 (

) to update.

는 제2 정보모듈(200)의 제2 채널 상태 정보(

)에 대한 매트릭스,

는

로 계산된 달성 가능한 스펙트럼 효율(SE; spectral efficiency),

는 간섭에 대한 임계값,

와

는 제어하기 위한 파라미터에 해당한다.here

Is the second channel state information of the second information module 200 (

) for the matrix,

Is

The achievable spectral efficiency (SE) calculated as

is the threshold for interference,

and

corresponds to a parameter for controlling.

미만으로 보장하면서 손실함수의 값을 최소화하도록 DNN 제어모듈(300)의 파라미터(

)를 업데이트할 수 있다.The determination module 400 maximizes the spectral efficiency (SE) of user equipment (DUE) transceiver pairs (TPs) between devices through (Equation 3) above while minimizing interference with cellular transmission.

Parameters of the DNN control module 300 to minimize the value of the loss function while ensuring that it is less than

) can be updated.

는 잡음 스펙트럼 밀도,

는 대역폭,

는 셀룰러 사용자 단말(CUE)가 각 채널에 대한 전송 전력에 해당한다.here

is the noise spectral density,

is the bandwidth,

Corresponds to the transmit power of the cellular user equipment (CUE) for each channel.

)를 업데이트할 수 있다.The determination module 400 uses the gradient descent algorithm of (Equation 5) to determine the parameters of the DNN control module 300 (

) can be updated.

) 미만으로 유지하면서 장치간 사용자 장비(DUE) 트랜스시버 쌍들(TPs)의 평균 스펙트럼 효율(SE)를 최대화할 수 있다.According to an embodiment of the present invention, the determination module 400 determines the level of interference occurring in the legacy cellular network based on the inaccurate estimated CSI as a threshold value (

), while maximizing the average spectral efficiency (SE) of user equipment (DUE) transceiver pairs (TPs) between devices.

)는 -174dBm/Hz, 장치간 사용자 장비(DUE) 트랜스시버 쌍들(TPs)의 전송 전력 집합(

)과 채널(

)은 3, 최대 전송 전력(

)은 20dBm, 간섭의 임계(

)은 -55dBm, 경로 손실 계수(path loss coefficient)가

, 경로 손실 지수(path loss exponent)가 3.8로 설정하고, 장치간 사용자 단말(DUE)과 셀룰러 사용자 단말(CUE)가 30m X 30m의 영역에 무작위로 분포되어 있다고 가정한다. 3A to 8 show a bandwidth (W) of 10 MHz and a noise spectral density (

) is -174 dBm/Hz, the transmit power set of user equipment (DUE) transceiver pairs (TPs) between devices (

) and channels (

) is 3, the maximum transmit power (

) is 20 dBm, the threshold of interference (

) is -55 dBm, the path loss coefficient is

, and the path loss exponent is set to 3.8, and it is assumed that device-to-device user equipments (DUEs) and cellular user equipments (CUEs) are randomly distributed in an area of 30 m × 30 m.

)가 7개, 숨겨진 노드 수(

)가 100개 일 때이다.The structure of the DNN control module 300 is the number of layers (

) is 7, the number of hidden nodes (

) is 100.

For the performance metric, we consider the average spectral efficiency (SE) of device-to-device user equipment (DUE) transceiver pairs (TPs) and the QoS violation probability. This is because violating the QoS constraint sets the spectral efficiency (SE) of device-to-device user equipment (DUE) transceiver pairs (TPs) to zero, penalizing their performance.

으로, QoS 위반 확률은

으로 표현할 수 있다.The average spectral efficiency (SE) is

, the probability of QoS violation is

can be expressed as

)에 대한 스펙트럼 효율(SE)의 그래프이며, 도 4a 내지 4c는 본 발명에 따라 가변 간섭의 임계(

)에 대한 QoS 위반 확률의 그래프이다.3a to 3c show the variable interference threshold (

) is a graph of spectral efficiency (SE), and FIGS. 4a to 4c show the threshold of variable interference according to the present invention (

) is a graph of QoS violation probability.

It is shown that the average spectral efficiency (SE) decreases and QoS violations increase with decreasing Ithr. This is because the transmission power of user equipment (DUE) transceiver pairs (TPs) between devices is more tightly limited and QoS constraints are more difficult to meet.

According to an embodiment of the present invention, it can be seen that the present invention (corresponding to Proposed in the drawing) provides optimal performance with a negligibly low data QoS violation probability. The random scheme (corresponding to random in the figure) shows a low spectral efficiency (SE), but has an excessively high QoS violation. This is data showing the need for appropriate transmission power control.

으로 채널 상태 정보(CSI)가 정확할 때, 무작위 방식(random)을 제외한 모든 방식(ES, ES(Robust), Proposed, ERP, ERP(Robust))은 무시할만큼 작은 QoS 위반 확률을 보이나, 채널 오류를 고려하지 않는 ES 및 ERP 방식은 임계(

)가 증가함에 따라 높은 QoS 위반으로 인해 스펙트럼 효율(SE)의 속도 급저하 되는 것을 확인할 수 있다. also

When the channel state information (CSI) is accurate, all methods (ES, ES (Robust), Proposed, ERP, ERP (Robust)) except for the random method show a negligibly small QoS violation probability, but channel errors ES and ERP methods that are not considered are critical (

) increases, it can be seen that the speed of the spectral efficiency (SE) is rapidly degraded due to high QoS violations.

의 영향을 조사하기 위해 임계(

)가 -55dBm 일 때 각각의

에 대한 장치간 사용자 장비(DUE) 트랜스시버 쌍(TP)의 스펙트럼 효율(SE)과 QoS 위반 확률을 비교한 그래프이다. 5 is a comparison graph for spectral efficiency (SE), FIG. 6 is a comparison graph for QoS violation probability, and FIGS. 5 and 6 are graphs for performance

In order to investigate the effect of criticality (

) is -55 dBm, each

This is a graph comparing spectral efficiency (SE) and QoS violation probability of a user equipment (DUE) transceiver pair (TP) between devices for

의 증가함에 따라 증가하는 것을 알 수 있다. 특히, ES와 ERP 방식의 스펙트럼 효율(SE)은 채널 상태 정보(CSI)의 부정확성을 고려하지 않기 때문에 ES(Robust)와 ERP(Robust)에 비해 현저히 낮은 것을 확인할 수 있다. The spectral efficiency (SE) of all schemes decreases, and the QoS violation probability

It can be seen that increases with the increase of In particular, it can be seen that the spectral efficiency (SE) of the ES and ERP methods is significantly lower than that of ES (Robust) and ERP (Robust) because inaccuracy of channel state information (CSI) is not considered.

가 0에서 0.5로 변경됨에 따라 본 발명(도면에서 Proposed에 해당)의 스펙트럼 효율(SE)은 3 %만 감소되나, ES (Robust)의 스펙트럼 효율(SE)은 8.1% 감소되는 것을 확인할 수 있다. 따라서 본 발명은 QoS 위반 확률이 무시할 정도로 작은

= 0.5 일 때 가장 높은 스펙트럼 효율(SE)을 달성한다. According to one embodiment of the present invention,

As is changed from 0 to 0.5, it can be confirmed that the spectral efficiency (SE) of the present invention (corresponding to proposed in the figure) is reduced by only 3%, but the spectral efficiency (SE) of ES (Robust) is reduced by 8.1%. Therefore, the present invention has a negligibly small QoS violation probability.

= 0.5, the highest spectral efficiency (SE) is achieved.

가 0이며, 채널 오류가 없을 때는

가 0.5에 해당함)에 따라 장치간 사용자 장비(DUE) 트랜스시버 쌍(TP)의 수(

)를 변경하여 장치간 사용자 장비(DUE) 트랜스시버 쌍(TP)의 평균 스펙트럼 효율(SE)와 계산 시간에 대한 그래프이다.7 and 8 show the presence or absence of a channel error (when there is a channel error)

is 0 and there is no channel error

is equal to 0.5), the number of user equipment (DUE) transceiver pairs (TPs) between devices (

) is a graph of the average spectral efficiency (SE) and computation time of a user equipment (DUE) transceiver pair (TP) between devices by changing

)가 6.2일 경우 200시간 이상이 소요된다. 이처럼 수(

)가 크면 과도한 계산 시간을 발생하므로 ES 방식은 도 7의 그래프에서 생략하였다.The calculation time of the ES method is the number of user equipment (DUE) transceiver pairs (TP) between devices (

) is 6.2, it takes more than 200 hours. Like this number (

) is large, excessive calculation time occurs, so the ES method is omitted from the graph of FIG. 7 .

The ES and ERP schemes that do not consider channel state information (CSI) in an inaccurate environment are equivalent to the calculation time of ES (robust) and ERP (robust) schemes, so ES (robust) and ERP (robust) in the graph of FIG. method was omitted.

The present invention exhibits the highest spectral efficiency (SE) among the considered techniques while achieving the lowest computational time. This can prove feasibility and robustness.

As described above, according to the present invention, the transmit power control apparatus considering inaccurate channel state information uses deep learning to determine the optimal transmit power even in an environment in which channel information is inaccurate, thereby transmitting D2D user equipment (DUE) in each channel. The power is optimized to maximize the average spectral efficiency (SE) and compensate for the effects of inaccurate channel state information (CSI) using artificially distorted CSI to reduce channel error. Reliable transmission power control.

Although the above has been described with reference to the illustrated embodiments of the present invention, these are only examples, and those skilled in the art to which the present invention belongs can variously It will be apparent that other embodiments that are variations, modifications and equivalents are possible. Therefore, the true technical protection scope of the present invention should be determined by the technical spirit of the appended claims.

1000: Transmission power control device considering inaccurate channel state information
100: first information module
110: normalization unit
120: reconstruction unit
200: second information module
300: DNN control module
310: first DNN unit
311: sigmoid layer
320: second DNN unit
321: soft max layer
400: decision module

Claims (11)

a first information module for inputting first channel state information (CSI);
a second information module inputting randomly distorted second channel state information (CSI) by adding randomly generated channel state information to the first channel state information;
a DNN control module for controlling transmit power of transceiver pairs (TPs) of D2D user equipments (DUEs) between devices based on the first channel state information (CSI); and
and a determination module for determining a loss function using transmit power of the DNN control module and the second channel state information (CSI), and updating parameters of the DNN control module according to the determined loss function. Transmission power control device considering information.
According to claim 1,
The first channel state information,
Transmission power control device considering inaccurate channel state information, characterized in that the channel state information estimated in an inaccurate environment.
delete According to claim 1,
The DNN control module,
Transmission power control device considering inaccurate channel state information, characterized in that for acquiring the optimal transmission power of user equipment (DUE) transceiver pairs (TPs) between devices using the updated parameters of the DNN module.
According to claim 1,
The first information module,
a normalization unit normalizing transmit powers of the device-to-device user equipment (DUE) transceiver pairs (TPs) based on the first channel state information (CSI);
Transmit power control device considering inaccurate channel state information further comprising: a reconfiguration unit for reconstructing the transmit power of the user equipment (DUE) transceiver pairs (TPs) from three dimensions to one dimension.
According to claim 1,
The DNN control module,
a first DNN unit for approximating transmit power of user equipment (DUE) transceiver pairs (TPs) between devices;
A transmission power control device considering inaccurate channel state information including; a second DNN unit for approximating a transmission power ratio of a user equipment (DUE) transceiver pair (TP) between each device.
According to claim 6,
The first DNN unit,
Transmission power control device considering inaccurate channel state information, characterized in that the output layer is a sigmoid layer.
According to claim 6,
The second DNN unit,
Transmission power control device considering inaccurate channel state information, characterized in that the output layer is a softmax layer.
According to claim 1,
The decision module,
Transmission power control device considering inaccurate channel state information, characterized in that for learning to minimize the value of the loss function.
According to claim 6,
The decision module,
determining a set of transmit power by calculating a transmit power ratio of the first DNN unit, a ratio of transmit power of the second DNN unit, and a maximum transmit power of user equipment (DUE) transceiver pairs (TPs) between devices Transmission power control device considering one channel state information.
According to claim 10,
The decision module,
Transmit power control apparatus considering inaccurate channel state information, characterized in that for determining a loss function using the set of transmit powers and the second channel state information (CSI).

Images