CN112770398A - Far-end radio frequency end power control method based on convolutional neural network - Google Patents
Far-end radio frequency end power control method based on convolutional neural network Download PDFInfo
- Publication number
- CN112770398A CN112770398A CN202011506528.8A CN202011506528A CN112770398A CN 112770398 A CN112770398 A CN 112770398A CN 202011506528 A CN202011506528 A CN 202011506528A CN 112770398 A CN112770398 A CN 112770398A
- Authority
- CN
- China
- Prior art keywords
- neural network
- convolutional neural
- user
- radio frequency
- remote
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 103
- 238000000034 method Methods 0.000 title claims abstract description 38
- 210000004027 cell Anatomy 0.000 claims abstract description 36
- 238000012549 training Methods 0.000 claims abstract description 30
- 230000003044 adaptive effect Effects 0.000 claims abstract description 12
- 238000005562 fading Methods 0.000 claims abstract description 12
- 210000002569 neuron Anatomy 0.000 claims abstract description 12
- 230000005540 biological transmission Effects 0.000 claims description 27
- 238000011176 pooling Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 4
- 230000006835 compression Effects 0.000 claims description 3
- 238000007906 compression Methods 0.000 claims description 3
- 238000001228 spectrum Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 7
- 238000004364 calculation method Methods 0.000 abstract description 6
- 230000006870 function Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000010295 mobile communication Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W72/00—Local resource management
- H04W72/04—Wireless resource allocation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W72/00—Local resource management
- H04W72/12—Wireless traffic scheduling
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
- Y02E40/70—Smart grids as climate change mitigation technology in the energy generation sector
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The invention discloses a far-end radio frequency end power control method based on a convolutional neural network, which comprises the following steps: calculating channel gain of users in a cell according to the positions of the users in the cell in the cloud wireless access network and multipath fading; training a convolutional neural network by using the channel gain, and calculating the transmitting power of a radio frequency end to each user; optimizing the neuron weight and bias of the convolutional neural network based on a random gradient descent algorithm of the adaptive moment estimation, minimizing a loss function, and finishing the training of the convolutional neural network; and obtaining the optimal transmitting power distributed to each user by the radio frequency end according to the channel gain of the user and the convolutional neural network which completes training. The invention can realize the real-time processing of power distribution and effectively reduce the calculation complexity of power control.
Description
Technical Field
The invention relates to the technical field of power control of wireless communication, in particular to a far-end radio frequency end power control method based on a convolutional neural network.
Background
With the deep development of the next generation mobile communication system research, the demand of people on the communication service quality is increasing day by day; meanwhile, users in wireless communication systems are becoming more and more dense, so that the originally scarce radio resources are becoming more and more intense. Therefore, management and allocation of appropriate radio resources becomes increasingly important.
The cloud wireless access network adopts optical fiber Remote combined Remote Radio Head (RRH) cell deployment, can effectively improve the convenience of cell deployment and realize denser cell coverage; meanwhile, the baseband processing unit carries out centralized processing, so that when the number of users in each coverage area is changed, centralized shared resources can be directly managed and distributed in the baseband resource pool. The deep learning technology based on the deep neural network can solve the complex nonlinear problem through a simple back propagation algorithm without deriving a complex mathematical model, and is widely applied to various fields to solve the nonlinear problem.
In the existing power control research, a general transmitting power optimization problem is a non-convex optimization problem, a closed-form solution of transmitting power cannot be obtained by directly solving the optimization problem, and a suboptimal solution is found by adopting iterative algorithms such as weighted minimum mean square error and the like. When the number of users increases, the number of iterations increases explosively, resulting in too high computational complexity and thus affecting the real-time requirements.
Disclosure of Invention
The invention aims to provide a far-end radio-frequency end power control method based on a convolutional neural network, and the method is used for solving the problems that the power control of a far-end radio-frequency end is carried out through a weighted minimum mean square error algorithm, the calculation complexity is high, and the real-time performance is poor in the prior art.
To solve the above technical problem, an embodiment of the present invention provides the following solutions:
a far-end radio frequency end power control method based on a convolutional neural network comprises the following steps:
calculating channel gain of users in a cell according to the positions of the users in the cell in the cloud wireless access network and multipath fading;
training a convolutional neural network by using the channel gain of the users in the cell, and calculating the transmitting power of a radio frequency end to each user;
optimizing the weight and bias of the neurons of the convolutional neural network by utilizing a random gradient descent algorithm based on adaptive moment estimation according to the transmitting power of a radio frequency end in the convolutional neural network to each user, minimizing a loss function and finishing the training of the convolutional neural network;
and obtaining the optimal transmitting power distributed to each user by the radio frequency end by utilizing the trained convolutional neural network according to the channel gain of the users in the cell.
Preferably, in a downlink transmission scenario of a cloud wireless access network based on OFDMA, a remote radio frequency terminal is deployed in the center of each cell to implement cell coverage, all the remote radio frequency terminals use channels with the same bandwidth for transmission, spectrum resources are shared among the remote radio frequency terminals, and the total transmission power in the coverage range of the nth remote radio frequency terminal is:
wherein p isn,kRepresents the downlink transmission power allocated to the user k by the nth remote radio frequency end; suppose that the maximum transmission power limit of the nth remote RF end is pnmaxIf the total transmission power in the coverage area of the nth remote RF end in the system is required to satisfy 0 < pn≤pnmax。
Preferably, the channel gain of the users in the cell is related to the multipath fading and the path loss of the users in the cell, and the channel gain from the nth remote rf end to user k is:
hn,k＝|φn,k|2βdn.k -α
wherein phi isn,kIs the multipath fading from the nth far-end radio frequency end to the user k, and follows the cyclic symmetry with the mean value of 0 and the variance of 1A complex Gaussian distribution; beta dn.k -αIs the path loss from the nth remote RF end to user k, beta and alpha are path loss parameters, dn.kIs the distance from the nth remote rf end to user k.
Preferably, the convolutional neural network structure includes a convolutional part, a fully-connected part, and a Sigmoid part.
Preferably, the convolution portion consists of NcEach subblock comprises a convolution layer, a ReLU layer and a pooling layer; the convolutional layer is used for performing convolution on input data and extracting the characteristics of the input data; the ReLU layer is used for introducing nonlinearity into the convolutional neural network to relieve the overfitting phenomenon; the pooling layer is used for performing feature compression on input data, extracting important data features and reducing the operation amount; the full-connection part is used for combining the outputs of the convolution parts so as to distribute the power transmitted by the far-end radio frequency end to the user.
Preferably, the training process of the convolutional neural network comprises two parts of forward propagation and backward propagation; the forward propagation calculates the input data from the bottom layer to the high layer, and the input of each layer of data is the output of the upper layer; and calculating the gradient of the convolutional neural network by back propagation, and updating the neuron weight and the bias of the convolutional neural network by using a random gradient descent algorithm based on adaptive moment estimation according to the calculated gradient so as to minimize a loss function.
Preferably, the input data of the convolutional neural network is the normalized user channel gain in the coverage range of the remote radio frequency end, and the output data is the normalized transmission power distributed to the user by the corresponding remote radio frequency end;
nth far-end radio frequency end normalization input vector
Expressed as:
wherein,
in order to normalize the user channel gain,
the output data of the convolutional neural network is a normalized transmission power vector distributed to a user by the nth remote radio frequency end
The transmission power of the nth remote radio frequency terminal to the users in the coverage area is
Preferably, the optimal transmitting power of the remote radio frequency end to the user is solved through a weighted minimum mean square error algorithm
As a label for a convolutional neural network, based on the output data p of the convolutional neural networknCalculating a loss function:
in the process of training the convolutional neural network, a stochastic gradient descent algorithm based on adaptive moment estimation is used for training a convolutional neural network model, so that a loss function is minimized.
The technical scheme provided by the embodiment of the invention has the beneficial effects that at least:
in the embodiment of the invention, the channel gain of the users in the cell is calculated according to the positions of the users in the cell in the cloud wireless access network and multipath fading; training a convolutional neural network by using the channel gain, and calculating the transmitting power of a radio frequency end to each user; optimizing the neuron weight and bias of the convolutional neural network based on a random gradient descent algorithm of the adaptive moment estimation, minimizing a loss function, and finishing the training of the convolutional neural network; and obtaining the optimal transmitting power distributed to each user by the radio frequency end according to the channel gain of the user and the convolutional neural network which completes training. The invention can realize the real-time processing of power distribution and effectively reduce the calculation complexity of power control.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a flowchart of a far-end rf-end power control method based on a convolutional neural network according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a downlink transmission scenario of an OFDMA-based cloud radio access network according to an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of a convolutional neural network provided in an embodiment of the present invention;
FIG. 4 is a schematic diagram of a training process of a convolutional neural network according to an embodiment of the present invention;
FIG. 5 is a graph illustrating loss and accuracy curves for a training set and a test set provided by an embodiment of the present invention;
fig. 6 is a comparative schematic diagram of the run time curve provided by the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
The invention provides a far-end radio frequency end power control method based on a convolutional neural network, aiming at the problems of high computational complexity and poor real-time performance of far-end radio frequency end power control through a weighted minimum mean square error algorithm in the existing OFDMA-based cloud radio access network downlink transmission scene.
As shown in fig. 1, the remote rf terminal power control method based on convolutional neural network provided in the embodiment of the present invention includes the following steps:
s01, calculating the channel gain of the users in the cell according to the positions of the users in the cell in the cloud wireless access network and multipath fading;
s02, training a convolutional neural network by using the channel gain of the users in the cell, and calculating the transmitting power of the radio frequency end to each user;
s03, optimizing the neuron weight and bias of the convolutional neural network by using a random gradient descent algorithm based on adaptive moment estimation according to the transmitting power of a radio frequency end in the convolutional neural network to each user, minimizing a loss function and finishing the training of the convolutional neural network;
and S04, obtaining the optimal transmitting power distributed to each user by the radio frequency end by using the trained convolutional neural network according to the channel gain of the users in the cell.
According to the remote radio frequency end power control method based on the convolutional neural network, channel gains of users in a cell are calculated according to the positions of the users in the cell and multipath fading in a cloud wireless access network; training a convolutional neural network by using the channel gain, and calculating the transmitting power of a radio frequency end to each user; optimizing the neuron weight and bias of the convolutional neural network based on a random gradient descent algorithm of the adaptive moment estimation, minimizing a loss function, and finishing the training of the convolutional neural network; the optimal transmitting power of each user distributed by the radio frequency end is obtained according to the channel gain of the user and the convolutional neural network which completes training, the complexity of calculation can be effectively reduced, and the real-time processing of power distribution is realized.
For better understanding of the remote rf terminal power control method based on convolutional neural network according to the embodiment of the present invention, the following detailed description is provided for the implementation thereof:
firstly, in a downlink transmission scene of a cloud radio access network based on OFDMA, a plurality of Virtual Machines (VMs) are concentrated in a baseband resource pool, a remote radio frequency end is connected to the baseband resource pool through a large-bandwidth low-latency forward link, the forward link uses a standard ecpri (enhanced cpri) interface, and the remote radio frequency end provides services for users within a coverage area.
As shown in fig. 2, fig. 2 is a schematic diagram of a downlink transmission scenario of a cloud radio access network based on OFDMA, where a far-end radio frequency end is deployed in the center of each cell to implement cell coverage, each far-end radio frequency end is equipped with L transmitting antennas, the distance between the far-end radio frequency ends is R, the radius of the cell is greater than R/2, and all the far-end radio frequency ends transmit using channels with the same bandwidth, that is, the far-end radio frequency ends share spectrum resources.
The set of all remote radio-frequency terminals in the system is:
wherein, N represents the nth remote rf end.
In this embodiment, each user in a cell is served by at most one remote radio frequency end, and the user positions comply with poisson point distribution, when a user is in a cell coverage overlapping area, the user receives co-channel interference from an adjacent remote radio frequency end, and the set of all users is:
wherein K(n)And the number of the users served by the nth remote radio frequency end is represented. Let p ben,kThe downlink transmission power allocated to user k by the nth remote rf end is represented, and the total transmission power in the coverage area of the nth remote rf end is:
suppose that the maximum transmission power limit of the nth remote RF end is pnmaxWithin the coverage of the nth remote RF end in the systemThe total transmitting power needs to satisfy 0 < pn≤pnmax。
In this embodiment, the channel gain of the user in the cell is related to the multipath fading and the path loss, and the channel gain from the nth remote rf end to the user k is:
hn,k＝|φn,k|2βdn.k -α
wherein phi isn,kMultipath fading from the nth far-end radio frequency end to a user k obeys circularly symmetric complex Gaussian distribution with the mean value of 0 and the variance of 1; beta dn.k -αIs the path loss from the nth remote RF end to user k, beta and alpha are path loss parameters, dn.kIs the distance from the nth remote rf end to user k.
In this embodiment, a convolutional neural network structure of the far-end radio frequency end power control method based on a convolutional neural network is shown in fig. 3, and the convolutional neural network structure is composed of a convolutional portion, a Fully Connected (FC) portion, and a Sigmoid portion.
In this embodiment, the convolution portion is composed of NcThe sub-blocks are composed in series, and each sub-block comprises a convolutional Layer (Conv Layer), a ReLU (rectified linear Unit) Layer and a Pooling Layer (Max Pooling).
The convolutional layer performs convolution on input data, extracts the characteristics of the input data, and assumes that the size of the input data is L multiplied by 1; the convolution layer depth of the ith sub-block is Ci(ii) a The step size of the convolution filter is set to 1, and 0 padding is not used; convolution kernel size set to
The output data size of each convolutional layer is
Wherein:
ReLU layer introduces nonlinearity into convolutional neural networkThe overfitting phenomenon is relieved; the pooling layer performs feature compression on input data to extract important data features while reducing the amount of computation, and sets the size of convolution kernel to be
The step length is set to 1, and the output data size of the pooling layer is
Wherein:
in this embodiment, the fully connected portion combines the outputs of the convolved portions and converts them into L in sizeFCThe x 1 vector is used as its output to distribute the power transmitted by the remote RF end to the user, where LFCThe number of neurons set.
In this embodiment, the Sigmoid part takes the output of the fully connected part as its input, and the output of the ith Sigmoid part is
Wherein [. ]]iRepresenting the ith element of the output vector.
In this embodiment, the input data of the convolutional neural network is the normalized user channel gain within the coverage of the remote rf end, and the output data is the normalized transmit power allocated to the user by the corresponding remote rf end.
Nth far-end radio frequency end normalization input vector
Wherein,
in order to normalize the user channel gain,
then, the output data of the convolutional neural network is the normalized transmission power vector distributed to the user by the nth remote radio frequency terminal
So that the transmission power given to the users in the coverage area by the nth remote radio frequency end is
In this embodiment, the training process of the convolutional neural network is divided into two parts, namely, a forward propagation part and a backward propagation part, namely, two sub-processes, namely, a forward propagation part and a backward propagation part. The forward propagation is to calculate the input data from the bottom layer to the upper layer, and the input of each layer of data is the output of the upper layer. The back propagation is to calculate the gradient of the convolutional neural network, and according to the calculated gradient, update the neuron weight and bias of the convolutional neural network by using a random gradient descent algorithm based on adaptive moment estimation, so that the loss function is minimized.
The convolutional neural network training process is shown in fig. 4. Firstly, randomly initializing the neuron weight and the bias of the convolutional neural network; inputting the normalized input vector into a convolutional neural network, and obtaining output data of the convolutional neural network through forward propagation; calculating the gradient of the convolutional neural network; and updating the neuron weight and bias of the convolutional neural network by using a random gradient descent algorithm based on self-adaptive moment estimation, so that a loss function is minimized, and the optimal model of the convolutional neural network controlled by the power of the remote radio frequency end is obtained.
In this embodiment, the optimal transmit power from the remote rf end to the user is solved by a weighted minimum mean square error algorithm
As a label for a convolutional neural network, based on the output data p of the convolutional neural networknCalculating the lossFunction:
in the process of training the convolutional neural network, a stochastic gradient descent algorithm based on adaptive moment estimation is used for training a convolutional neural network model, so that a loss function is minimized.
In this embodiment, the optimal transmit power of the remote radio frequency end can be determined through the trained convolutional neural network according to the user channel gain after any normalization processing in the cell. Aiming at the far-end radio-frequency end power control method based on the convolutional neural network, the trained convolutional neural network is an approximator for solving the optimal transmitting power of the far-end radio-frequency end through a weighted minimum mean square error algorithm, so that the far-end radio-frequency end power control method based on the convolutional neural network can achieve the performance similar to that of the far-end radio-frequency end power control method adopting the weighted minimum mean square error algorithm. In addition, although a large amount of calculation is needed for training the convolutional neural network, the convolutional neural network can be trained offline, the problem of power control of a far-end radio frequency end can be solved with low cost for the trained convolutional neural network, the complexity of calculation is effectively reduced, and real-time processing of power distribution is realized.
Fig. 5 is a schematic diagram of loss and accuracy curves of a training set and a test set, and it can be known from fig. 5 that when the training frequency Epoch of the convolutional neural network is 90, the trained convolutional neural network model achieves an accuracy of 95.3% in the test set, and the convolutional neural network model has very good robustness, and can well predict the transmission power of the remote radio frequency end to the user according to the user channel gain.
Fig. 6 is a comparison diagram of operation time curves, and it can be seen from fig. 6 that, as the average number of users increases, both the operation time of the convolutional neural network-based remote rf power control method and the operation time of the weighted minimum mean square error algorithm-based remote rf power control increase, however, the operation time of the convolutional neural network-based remote rf power control method is much lower than the operation time of the weighted minimum mean square error algorithm-based remote rf power control when the average number of users increases, and the operation time difference between the two algorithms increases as the average number of users increases.
In summary, the far-end radio-frequency end power control method based on the convolutional neural network according to the embodiment applies the deep learning technology to the field of wireless communication, and in the downlink transmission scenario of the cloud radio access network based on the OFDMA, the far-end radio-frequency end power control is realized, so that the running time can be reduced, the real-time requirement can be met, and the computational complexity can be reduced. The far-end radio-frequency end power control method based on the convolutional neural network is applied to the next generation mobile communication technology, so that the real-time requirement can be met, the design cost can be saved, the design flow can be simplified, a new thought is provided for far-end radio-frequency end power control, and the method is suitable for scenes and has generality.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (8)
1. A far-end radio frequency end power control method based on a convolutional neural network is characterized by comprising the following steps:
calculating channel gain of users in a cell according to the positions of the users in the cell in the cloud wireless access network and multipath fading;
training a convolutional neural network by using the channel gain of the users in the cell, and calculating the transmitting power of a radio frequency end to each user;
optimizing the weight and bias of the neurons of the convolutional neural network by utilizing a random gradient descent algorithm based on adaptive moment estimation according to the transmitting power of a radio frequency end in the convolutional neural network to each user, minimizing a loss function and finishing the training of the convolutional neural network;
and obtaining the optimal transmitting power distributed to each user by the radio frequency end by utilizing the trained convolutional neural network according to the channel gain of the users in the cell.
2. The far-end radio-frequency terminal power control method based on the convolutional neural network as claimed in claim 1, wherein in an OFDMA-based cloud radio access network downlink transmission scenario, one far-end radio-frequency terminal is deployed in the center of each cell to achieve cell coverage, all the far-end radio-frequency terminals transmit using channels with the same bandwidth, spectrum resources are shared among the far-end radio-frequency terminals, and the total transmission power in the coverage range of the nth far-end radio-frequency terminal is:
wherein p isn,kRepresents the downlink transmission power allocated to the user k by the nth remote radio frequency end; suppose that the maximum transmission power limit of the nth remote RF end is pnmaxIf the total transmission power in the coverage area of the nth remote RF end in the system is required to satisfy 0 < pn≤pnmax。
3. The convolutional neural network-based remote rf terminal power control method as claimed in claim 1, wherein the channel gain of the users in the cell is related to the multipath fading and path loss of the users in the cell, and the channel gain from the nth remote rf terminal to user k is:
hn,k＝|φn,k|2βdn.k -α
wherein phi isn,kMultipath fading from the nth far-end radio frequency end to a user k obeys circularly symmetric complex Gaussian distribution with the mean value of 0 and the variance of 1; beta dn.k -αIs the path loss from the nth remote RF end to user k, beta and alpha are path loss parameters, dn.kIs the distance from the nth remote rf end to user k.
4. The far-end radio-frequency terminal power control method based on the convolutional neural network as claimed in claim 1, wherein the convolutional neural network structure comprises a convolutional part, a fully-connected part and a Sigmoid part.
5. The convolutional neural network-based remote RF end power control method as claimed in claim 4, wherein the convolutional part is composed of NcEach subblock comprises a convolution layer, a ReLU layer and a pooling layer; the convolutional layer is used for performing convolution on input data and extracting the characteristics of the input data; the ReLU layer is used for introducing nonlinearity into the convolutional neural network to relieve the overfitting phenomenon; the pooling layer is used for performing feature compression on input data, extracting important data features and reducing the operation amount; the full-connection part is used for combining the outputs of the convolution parts so as to distribute the power transmitted by the far-end radio frequency end to the user.
6. The convolutional neural network-based remote rf-terminal power control method as claimed in claim 1, wherein the training process of the convolutional neural network includes two parts of forward propagation and backward propagation; the forward propagation calculates the input data from the bottom layer to the high layer, and the input of each layer of data is the output of the upper layer; and calculating the gradient of the convolutional neural network by back propagation, and updating the neuron weight and the bias of the convolutional neural network by using a random gradient descent algorithm based on adaptive moment estimation according to the calculated gradient so as to minimize a loss function.
7. The convolutional neural network-based remote rf port power control method as claimed in claim 6, wherein the input data of the convolutional neural network is the normalized user channel gain in the coverage of the remote rf port, and the output data is the normalized transmit power allocated to the user by the corresponding remote rf port;
nth far-end radio frequency end normalization input vector
Expressed as:
wherein,
in order to normalize the user channel gain,
the output data of the convolutional neural network is a normalized transmission power vector distributed to a user by the nth remote radio frequency end
The transmission power of the nth remote radio frequency terminal to the users in the coverage area is
8. The convolutional neural network-based remote RF power control method as claimed in claim 6, wherein the optimal transmit power from the remote RF to the user is solved by weighted least mean square error algorithm
As a label for a convolutional neural network, based on the output data p of the convolutional neural networknCalculating a loss function:
in the process of training the convolutional neural network, a stochastic gradient descent algorithm based on adaptive moment estimation is used for training a convolutional neural network model, so that a loss function is minimized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011506528.8A CN112770398A (en) | 2020-12-18 | 2020-12-18 | Far-end radio frequency end power control method based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011506528.8A CN112770398A (en) | 2020-12-18 | 2020-12-18 | Far-end radio frequency end power control method based on convolutional neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112770398A true CN112770398A (en) | 2021-05-07 |
Family
ID=75694192
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011506528.8A Pending CN112770398A (en) | 2020-12-18 | 2020-12-18 | Far-end radio frequency end power control method based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112770398A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113708781A (en) * | 2021-08-13 | 2021-11-26 | Oppo广东移动通信有限公司 | Radio frequency gain control method, device and communication equipment |
| CN114268965A (en) * | 2021-12-21 | 2022-04-01 | 中国科学院数学与系统科学研究院 | Mobile communication network coverage rate calculation method and device based on deep learning |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109474980A (en) * | 2018-12-14 | 2019-03-15 | 北京科技大学 | A kind of wireless network resource distribution method based on depth enhancing study |
| WO2019190036A1 (en) * | 2018-03-27 | 2019-10-03 | 경상대학교산학협력단 | Method and apparatus for transmission power control based on neural network in wireless communication system |
| CN111328087A (en) * | 2018-12-17 | 2020-06-23 | 上海大学 | Deep learning-based high-energy-efficiency heterogeneous network sub-channel distribution and power distribution method |
-
2020
- 2020-12-18 CN CN202011506528.8A patent/CN112770398A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019190036A1 (en) * | 2018-03-27 | 2019-10-03 | 경상대학교산학협력단 | Method and apparatus for transmission power control based on neural network in wireless communication system |
| CN109474980A (en) * | 2018-12-14 | 2019-03-15 | 北京科技大学 | A kind of wireless network resource distribution method based on depth enhancing study |
| CN111328087A (en) * | 2018-12-17 | 2020-06-23 | 上海大学 | Deep learning-based high-energy-efficiency heterogeneous network sub-channel distribution and power distribution method |
Non-Patent Citations (1)
| Title |
|---|
| WOONGSUP LEE等: "Deep Power Control: Transmit Power Control Scheme Based on Convolutional Neural Network", 《IEEE COMMUNICATIONS LETTERS》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113708781A (en) * | 2021-08-13 | 2021-11-26 | Oppo广东移动通信有限公司 | Radio frequency gain control method, device and communication equipment |
| WO2023016231A1 (en) * | 2021-08-13 | 2023-02-16 | Oppo广东移动通信有限公司 | Radio frequency gain control method and apparatus, and communication device |
| CN113708781B (en) * | 2021-08-13 | 2023-02-17 | Oppo广东移动通信有限公司 | Radio frequency gain control method, device and communication equipment |
| CN114268965A (en) * | 2021-12-21 | 2022-04-01 | 中国科学院数学与系统科学研究院 | Mobile communication network coverage rate calculation method and device based on deep learning |
| CN114268965B (en) * | 2021-12-21 | 2023-09-08 | 中国科学院数学与系统科学研究院 | Mobile communication network coverage rate calculation method and device based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109474980B (en) | Wireless network resource allocation method based on deep reinforcement learning | |
| CN110798849A (en) | Computing resource allocation and task unloading method for ultra-dense network edge computing | |
| CN110418416A (en) | Resource allocation methods based on multiple agent intensified learning in mobile edge calculations system | |
| CN111314935B (en) | Method for minimizing downlink transmission delay based on NOMA-MEC system | |
| CN110191489B (en) | Resource allocation method and device based on reinforcement learning in ultra-dense network | |
| CN112770398A (en) | Far-end radio frequency end power control method based on convolutional neural network | |
| CN113038612B (en) | Cognitive radio power control method based on deep learning | |
| CN110166088A (en) | The power control algorithm without cell mimo system of customer-centric | |
| Yang et al. | Joint time allocation and power control in multicell networks with load coupling: Energy saving and rate improvement | |
| CN106792451A (en) | A kind of D2D communication resource optimization methods based on Multiple-population Genetic Algorithm | |
| CN113596785A (en) | D2D-NOMA communication system resource allocation method based on deep Q network | |
| CN111405596A (en) | Resource optimization method for large-scale antenna wireless energy-carrying communication system under Rice channel | |
| CN108965009A (en) | A kind of load known users correlating method based on gesture game | |
| CN113507712B (en) | Resource allocation and calculation task unloading method based on alternate direction multiplier | |
| CN113162662A (en) | User clustering and power distribution method under CF-mMIMO | |
| CN111542110A (en) | User scheduling and power allocation optimization method for multi-user physical layer secure communication | |
| CN111417125A (en) | Method and device for optimizing capacity of space-air-ground cooperative marine communication network | |
| CN111277308A (en) | Wave width control method based on machine learning | |
| CN108449737B (en) | D2D-based downlink energy-efficient power distribution method in distributed antenna system | |
| Liu et al. | Power allocation in ultra-dense networks through deep deterministic policy gradient | |
| CN110602718B (en) | Heterogeneous cellular network power distribution method and system based on alternative direction multiplier method | |
| CN112994762A (en) | MIMO-NOMA downlink self-adaptive wireless transmission method based on statistical CSI | |
| CN116887344A (en) | Cloud edge end collaborative system optimization method and device based on improved binary multi-universe algorithm | |
| CN111314938A (en) | Optimization method for time-frequency domain resource allocation of cellular network of single cell | |
| CN112243283B (en) | Cell-Free Massive MIMO network clustering calculation method based on successful transmission probability |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210507 |
|
| RJ01 | Rejection of invention patent application after publication |