CN112203345B - D2D communication energy efficiency maximization power distribution method based on deep neural network - Google Patents
D2D communication energy efficiency maximization power distribution method based on deep neural network Download PDFInfo
- Publication number
- CN112203345B CN112203345B CN202011051195.4A CN202011051195A CN112203345B CN 112203345 B CN112203345 B CN 112203345B CN 202011051195 A CN202011051195 A CN 202011051195A CN 112203345 B CN112203345 B CN 112203345B
- Authority
- CN
- China
- Prior art keywords
- link
- base station
- cellular
- user
- state information
- 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.)
- Active
Links
- 238000004891 communication Methods 0.000 title claims abstract description 82
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 64
- 238000000034 method Methods 0.000 title claims abstract description 38
- 230000001413 cellular effect Effects 0.000 claims abstract description 155
- 238000001228 spectrum Methods 0.000 claims abstract description 48
- 230000005540 biological transmission Effects 0.000 claims abstract description 44
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 13
- 230000006870 function Effects 0.000 claims description 27
- 238000010295 mobile communication Methods 0.000 claims description 26
- 238000012549 training Methods 0.000 claims description 25
- 238000010606 normalization Methods 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- IJJWOSAXNHWBPR-HUBLWGQQSA-N 5-[(3as,4s,6ar)-2-oxo-1,3,3a,4,6,6a-hexahydrothieno[3,4-d]imidazol-4-yl]-n-(6-hydrazinyl-6-oxohexyl)pentanamide Chemical compound N1C(=O)N[C@@H]2[C@H](CCCCC(=O)NCCCCCC(=O)NN)SC[C@@H]21 IJJWOSAXNHWBPR-HUBLWGQQSA-N 0.000 claims description 5
- 238000007726 management method Methods 0.000 claims description 5
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 230000002829 reductive effect Effects 0.000 claims description 3
- 230000008901 benefit Effects 0.000 abstract description 6
- 230000003595 spectral effect Effects 0.000 description 11
- 210000004027 cell Anatomy 0.000 description 9
- 238000013135 deep learning Methods 0.000 description 8
- 238000005457 optimization Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 230000007423 decrease Effects 0.000 description 5
- 238000010801 machine learning Methods 0.000 description 5
- 230000002452 interceptive effect Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005562 fading Methods 0.000 description 2
- 238000011478 gradient descent method Methods 0.000 description 2
- 238000003058 natural language processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 230000003213 activating effect Effects 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000010267 cellular communication Effects 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 230000001010 compromised effect Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000036961 partial effect Effects 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 238000013468 resource allocation Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W52/00—Power management, e.g. TPC [Transmission Power Control], power saving or power classes
- H04W52/04—TPC
- H04W52/18—TPC being performed according to specific parameters
-
- 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
-
- 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
- H04W16/14—Spectrum sharing arrangements between different networks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W52/00—Power management, e.g. TPC [Transmission Power Control], power saving or power classes
- H04W52/04—TPC
- H04W52/18—TPC being performed according to specific parameters
- H04W52/24—TPC being performed according to specific parameters using SIR [Signal to Interference Ratio] or other wireless path parameters
- H04W52/241—TPC being performed according to specific parameters using SIR [Signal to Interference Ratio] or other wireless path parameters taking into account channel quality metrics, e.g. SIR, SNR, CIR, Eb/lo
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W52/00—Power management, e.g. TPC [Transmission Power Control], power saving or power classes
- H04W52/04—TPC
- H04W52/18—TPC being performed according to specific parameters
- H04W52/24—TPC being performed according to specific parameters using SIR [Signal to Interference Ratio] or other wireless path parameters
- H04W52/243—TPC being performed according to specific parameters using SIR [Signal to Interference Ratio] or other wireless path parameters taking into account interferences
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W52/00—Power management, e.g. TPC [Transmission Power Control], power saving or power classes
- H04W52/04—TPC
- H04W52/38—TPC being performed in particular situations
- H04W52/42—TPC being performed in particular situations in systems with time, space, frequency or polarisation diversity
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W72/00—Local resource management
- H04W72/04—Wireless resource allocation
- H04W72/044—Wireless resource allocation based on the type of the allocated resource
- H04W72/0473—Wireless resource allocation based on the type of the allocated resource the resource being transmission power
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W72/00—Local resource management
- H04W72/50—Allocation or scheduling criteria for wireless resources
- H04W72/54—Allocation or scheduling criteria for wireless resources based on quality criteria
- H04W72/541—Allocation or scheduling criteria for wireless resources based on quality criteria using the level of interference
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W72/00—Local resource management
- H04W72/50—Allocation or scheduling criteria for wireless resources
- H04W72/54—Allocation or scheduling criteria for wireless resources based on quality criteria
- H04W72/542—Allocation or scheduling criteria for wireless resources based on quality criteria using measured or perceived quality
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W72/00—Local resource management
- H04W72/50—Allocation or scheduling criteria for wireless resources
- H04W72/54—Allocation or scheduling criteria for wireless resources based on quality criteria
- H04W72/543—Allocation or scheduling criteria for wireless resources based on quality criteria based on requested quality, e.g. QoS
-
- 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)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The invention discloses a D2D communication energy efficiency maximization power distribution method based on a deep neural network, which is suitable for the field of communication. Firstly, a base station selects a cellular user through a scheduling algorithm to share frequency resources to a plurality of selected D2D links for use; then the cell users which are scheduled and activated and a plurality of pairs of D2D users feed back channel state information to the base station; then, the base station leads the obtained channel state information of the cellular user and the D2D user into a trained deep neural network system to obtain a power distribution scheme with maximized D2D link energy efficiency; and finally, the D2D link transmitting end completes data transmission according to the power distribution scheme. The method can be effectively applied to an actual scene, the base station can quickly calculate the power distribution scheme after acquiring the real-time channel state information by using the trained deep neural network system, has the advantage of low time delay, and can improve the spectrum efficiency to a greater extent, improve the energy efficiency and realize green communication.
Description
Technical Field
The invention relates to a D2D communication energy efficiency maximization power distribution method, in particular to a D2D communication energy efficiency maximization power distribution method based on a deep neural network, which is suitable for the field of mobile communication equipment and equipment self-adaptive resource distribution.
Background
In recent years, deep Learning (DL) has been used with great success in the fields of Computer Vision (CV), natural Language Processing (NLP), and Automatic Speech Recognition (ASR), which are being attempted to solve some problems in mobile communications, and is currently being used in particular in cognitive radio, resource management, line adaptation, modulation recognition, decoding, and detection, among other aspects.
DL is a branch of Machine Learning (ML), and by means of a multi-layer non-linear processing unit, useful information can be hierarchically extracted from original data, so as to perform prediction or complete corresponding tasks according to a set target. Common DL frameworks are multi-layer perceptrons (MLPs), boltzmann machines (RBMs), automatic Encoders (AEs), convolutional Neural Networks (CNNs), recurrent Neural Networks (RNNs), generative countermeasure networks (GANs), and Deep Reinforcement Learning (DRLs).
Compared with the conventional ML algorithm, the DL has the advantages that: automatic feature extraction can be achieved from complex, inherently relevant data: this advantage is magnified in processing mobile data; large amounts of data can be processed: traditional ML algorithms, such as Support Vector Machines (SVMs), require space for data storage, which results in that efficient computation cannot be achieved in the context of large data; in the process of training the Neural Network (NN), only partial data is needed in each step, so that the expandability of deep learning in the aspect of big data processing is ensured, and overfitting can be avoided by training a large amount of data; being able to learn useful patterns in an unsupervised manner (using data without tags); one model can accomplish multiple tasks that other ML paradigms (e.g., linear regression, random forests) cannot do.
DL is also called Deep Neural Network (DNN), and its neural network layer can be roughly divided into three categories, i.e., input layer, hidden layer and output layer, and the complex objectives or functions are realized by preset weighted combination of some neurons, nonlinear activation function and loss function, and these objectives can be roughly divided into three categories, i.e., classification, regression and control. The Back Propagation (BP) of DNN follows a chain rule, and a gradient descent method (GD) is used for minimizing a loss function, so that the weight of the model is continuously optimized in the training process of the neural network, and the learning purpose is achieved.
As the depth of neural network models increases, some objective functions are generally non-convex and have many local minima, critical points, and saddle points. In this case, the conventional SGD may converge very slowly. In recent years, new optimization algorithms, such as SGD combined with Nesterov momentum variants, have emerged that converge very fast when optimizing convex functions; adam is a self-adaptive learning rate optimization algorithm, and the model has high convergence speed and strong robustness through the first moment joint momentum of the gradient. In addition, mature, open source deep learning platforms such as Tensorflow of Google, caffe of Facebook, etc. have emerged, which have accelerated the development and development of DNN products.
At present, some researchers successfully use DNN to solve the resource allocation problem in some specific communication scenarios and achieve better effect, but the present invention mainly considers the D2D communication scenario in the cellular mobile communication system, specifically, in one cell, a plurality of D2D links simultaneously reuse the uplink spectrum resource of one cellular user to realize direct communication without base station forwarding. Most of the existing D2D researches limit the spectrum resource of one cellular user to be used by only one D2D link, and if the spectrum resource of the cellular link is shared by a plurality of D2D links, the spectrum resource utilization rate can be improved to a greater extent. The introduction of a plurality of D2D links into a cellular mobile communication network inevitably brings co-channel interference to the communication of cellular users, and co-channel interference also exists among the plurality of D2D links. In the communication scenario considered by the invention, there are multiple optimization requirements, on the premise of ensuring the communication quality of the cellular user, the access of multiple D2D links improves the spectrum efficiency, and simultaneously optimizes the energy efficiency of the D2D links, and when the traditional optimization algorithm is used for solving such complex optimization problems, the optimal solution can be achieved by multiple iterations, and some optimal solutions even cannot achieve the global optimal solution. The iteration times are multiple, the corresponding calculation time is long, the requirement of low time delay in real-time communication cannot be met undoubtedly, DNN has the advantage in the aspect of calculation time efficiency, and the time efficiency of the trained neural network system in the output power distribution scheme can reach within millisecond level. Therefore, the DNN-based model is tried to be built to solve the power distribution problem of maximizing the D2D communication energy efficiency in the scene, and the method has great practical significance.
Disclosure of Invention
The purpose of the invention is as follows: aiming at the defects of the prior art, the D2D communication energy efficiency maximization power distribution method based on the deep neural network is low in communication time delay and power consumption, is beneficial to solving the problems of lack of frequency spectrum resources, overload of data traffic and the like at present, can greatly improve the utilization rate of the frequency spectrum resources by sharing the frequency spectrum resources of the cellular network to a plurality of D2D links, and ensures the communication service experience of cellular users.
The technical scheme is as follows: in order to achieve the technical purpose, the method for maximizing power distribution of D2D communication energy efficiency based on the deep neural network specifically comprises the following steps:
a D2D communication energy efficiency maximization power distribution method based on a deep neural network is characterized in that a used cellular mobile communication system comprises a base station arranged at the center of a mobile communication cell, cellular users arranged in the mobile communication cell and a plurality of devices forming a D2D link, the devices comprise a transmitting user and a receiving user, and form a D2D user pair according to needs, and the method is characterized by comprising the following steps of:
acquiring channel state information among a used base station, a used cellular user, a D2D link transmitting user and a used receiving user;
a base station schedules and selects a cellular user and a plurality of D2D user pairs in a mobile communication cell, and shares the frequency spectrum resource of the cellular user to a direct communication link formed by the plurality of D2D user pairs for performing device-to-device direct communication;
the frequency sharing of the link between the cellular user and the base station and the D2D link can effectively improve the spectrum efficiency, but co-channel interference can be generated at the same time, so that the base station is utilized to acquire the real-time channel state information of the transmission link and the interference link, the real-time channel state information is input into the trained deep neural network system, and the optimal power distribution scheme meeting the power constraint condition is calculated so as to ensure that the system performance is improved as much as possible within the tolerable range of the co-channel interference;
and configuring the power of the D2D user of the cellular user by using the optimal power allocation scheme, thereby completing the data transmission of a plurality of D2D links between the cellular user and the base station.
The selection of the cellular user and the D2D user is specifically as follows:
a1, in a cellular mobile communication system, a base station adopts a centralized control wireless resource management method, uses a common scheduling algorithm in the existing mobile communication system to schedule cellular users, and shares uplink spectrum resources of the cellular users to a plurality of proper D2D links to carry out direct communication between devices;
a2, the D2D link refers to a link for performing direct device-to-device communication between a transmitting user and a receiving user by using a shared spectrum, and the suitable multiple D2D links refer to that the distances from the scheduled multiple D2D links to a base station are required to be greater than the distances from cellular users to the base station, the distance between each D2D link and a cellular user needs to exceed a preset value, and the distance between each D2D link and the rest of the D2D links needs to exceed the preset value.
The method for acquiring the real-time channel state information of the transmission link and the interference link by the base station comprises the following steps: the base station selects a cellular user sharing spectrum resources and a plurality of D2D links sharing uplink spectrum resources of the cellular user, firstly, the cellular user carries out channel estimation to obtain channel state information of a transmission link between the cellular user and the base station, and simultaneously, channel state information of interference links between the cellular user and receiving ends of all the D2D links is obtained through the channel estimation, then, the channel state information is quantized according to a codebook which is known by the cellular user and the base station, and the quantized channel state information is fed back to the base station; meanwhile, each D2D link transmitting user first performs channel estimation to obtain channel state information of a transmission link between the transmitting user and a corresponding receiving end, and then performs channel estimation to obtain channel state information of interference links between the transmitting user and the other D2D link receiving users, and in addition, each D2D link transmitting user also needs to obtain channel state information of the interference links between the transmitting user and a base station through channel estimation, quantizes the channel state information according to a codebook commonly known by the transmitting user and the base station, and feeds back the quantized channel state information to the base station.
The method for setting the power constraint condition in communication specifically comprises the following steps:
in a cellular mobile communication system, a communication link between a cellular user and a base station is marked by 0, the total number of a plurality of D2D links multiplexing uplink spectrum resources of the cellular user is N, and the D2D links are marked by 1,2, \ 8230and N;
when the cellular mobile communication system supports D2D direct communicationIn time, if the cellular user has a higher priority, the interference to the D2D link sharing its resources is not reduced by technical means, i.e. the cellular user uses power p when communicating with the base station 0 ＝P 0 Transmitting a signal based on which:
when a plurality of D2D links share the uplink frequency spectrum resource with the cellular user, wherein the transmitting power of the nth D2D link satisfies that p is more than or equal to 0 n ≤p max The constraint of (2) is that,
because the signals transmitted by the N D2D links sharing the spectrum resource can interfere the communication of the cellular user, in order to ensure the communication quality of the cellular user, the received signal-to-interference-and-noise ratio (SINR) at the base station 0 To satisfy the constraint:
wherein: n represents the total number of D2D users simultaneously multiplexing the uplink frequency spectrum resources of the cellular users;
p＝[p 1 ,p 2 ,...,p n ,...p N ]，p n representing the transmitting power of the transmitting end of the nth D2D link;
p max representing the maximum transmitting power of the D2D links, and assuming that the maximum transmitting power of the N D2D links is the same;
r min a signal to interference plus noise ratio threshold value representing a transmission link of a cellular subscriber;
g 0,0 channel state information representing a transmission link of a cellular subscriber to a base station;
g n,0 representing channel state information of an interference link from the nth D2D link to the base station, wherein N belongs to {1, 2.. N };
σ 2 representing a gaussian white noise variance.
Received signal to interference and noise ratio (SINR) at a base station 0 The power allocation scheme meeting the power constraint condition takes the optimal energy efficiency eta of a plurality of D2D links sharing uplink spectrum resources as a target, and the optimized mathematical expression is as follows:
Subjectto：0≤p n ≤p max
in optimizing the target eta, p c Representing the line loss in the communication of a single D2D link, and assuming that the line losses of the N D2D links are the same; g n,m Represents channel state information from the n-th D2D link transmitting end to the m-th D2D link receiving end, and in particular, g when n = m n,m Representing the channel state information of a transmission link of the nth D2D link, wherein N, m belongs to {1, 2.. N }; g 0,n Channel state information representing an interfering link between a cellular user and an nth D2D link receiver, N ∈ {1, 2.. N }; sigma 2 And representing the Gaussian white noise variance of the D2D link, and setting the Gaussian white noise variances of all links to be the same.
The training process of the trained deep neural network system comprises the following steps:
the deep neural network system used at the base station consists of a normalization module, a hidden layer kernel and a multiplier, and can quickly obtain a transmitting power distribution scheme of a D2D link by training and perfecting the base station based on a large number of channel samples at a server end supporting high-speed and parallel computation before being applied to real-time D2D communication at the base station as long as channel state information obtained in real time is input into the deep neural network system
The configuration and the function of each module of the deep neural network system are as follows:
b1, in order to accelerate the training speed of the following neural network, firstly, a normalization module is used for normalizing the channel state information. Specifically, the channel state information is normalized by the following formula:
b2, the kernel of the deep neural network hidden layer is formed by connecting two full connection layers (FC), three convolution layers (CNN) and a Sigmoid function module in series, and also comprises three reshape operation layers;
the first full link layer (FC 1) fuses the normalized real-time channel state information, then three convolutional layers are used for extracting the characteristics of a channel, the second full link layer (FC 2) fuses the channel characteristics extracted by the convolutional layers, each convolutional layer has the same filter setting, specifically, each convolutional layer uses a filter with the size of 3 multiplied by 3, the convolutional depth is 16, the step size is 1, zero padding is used, and each convolutional layer is provided with a ReLU activation function in front, which aims to prevent negative values, on the one hand, the nonlinearity of the network can be enhanced, and the learning capability of the network can be improved. The Sigmoid function module may limit the output value of the network to be between 0 and 1, resulting in a normalized transmission power value of the D2D link. The first and third reshape layers are used for reshaping the input into a vector form suitable for being led into the full-connection layer, and the second reshape layer is used for reshaping the output of the FC1 into a matrix form, so that the feature extraction of the convolutional layer is facilitated;
b3, in order to obtain the actual transmitting power of the D2D link, the multiplier multiplies the normalized power value with the maximum transmitting power of the D2D link, so as to obtain the actual transmitting power distribution scheme of the D2D link meeting the power constraint condition
The deep neural network adopts an unsupervised learning training method:
firstly, collecting a large number of channel samples, explaining by using 100000 groups of channels as training samples, leading in data of one bach once for each 100 groups of channels, training, initializing the weight of a neural network by using truncated normal distribution (truncated normal distribution), carrying out backward propagation by using an Adam algorithm, and obtaining a loss function f loss The definition is as follows:
f loss ＝-ωη+(1-ω)[r min -SINR 0 ] + ,
wherein the balance factor omega epsilon (0, 1) is used for adjusting the compromise between the energy efficiency optimal target and the interference constraint condition, the data of training 1000 bachs is 1 epoch until the receiving signal-to-interference-and-noise ratio requirement of the cellular link is met, and the loss function f loss And stopping training when the training platform does not descend any more.
Optimally obtained specific power allocation scheme, cellular user per transmit power P 0 Completing data transmission with the base station, wherein each D2D link is according to the transmission power
And completing data transmission, wherein N belongs to {1,2,. And N }.
Has the advantages that: compared with the prior art, the technical scheme of the invention has the following beneficial technical effects:
1) Most of the existing researches related to D2D communication limit that the frequency spectrum resource of a cellular user can only be used by a pair of D2D users, the power distribution method with maximized D2D communication energy efficiency provided by the invention can ensure that the uplink frequency spectrum resource of the cellular user is shared by a plurality of D2D links on the premise of ensuring the communication quality of the cellular user, and under the condition of objectively having same frequency interference, the communication quality of each link is ensured through power control, the utilization rate of the frequency spectrum resource is improved to a greater extent, the energy efficiency is improved, and green communication is realized.
2) The D2D communication energy efficiency maximization power distribution method provided by the invention is different from the traditional power distribution method, and based on the power distribution method of the deep neural network system, if the base station is provided with the trained deep neural network system, the power distribution scheme can be quickly obtained after real-time channel state information is obtained, so that the method has the advantages of high time efficiency and low time delay, and the problem of high time delay caused by excessive iteration times of the traditional algorithm on the complex problems is effectively solved.
Drawings
Fig. 1 is a diagram of a model of a cellular mobile communication system supporting D2D communication according to the present invention;
FIG. 2 is a block diagram of a deep neural network based system of the present invention;
fig. 3 is a diagram illustrating the variation of the energy efficiency of D2D communication with different signal to interference plus noise ratio (sinr) constraints of cellular users.
Fig. 4 is a schematic diagram illustrating the variation of D2D communication energy efficiency with the total number of D2D links under different cellular user signal to interference plus noise ratio thresholds;
fig. 5 is a schematic diagram illustrating the changes of the spectral efficiency of cellular users and the total system spectral efficiency with the total number of D2D links under different cellular user signal to interference plus noise ratio threshold conditions.
Detailed Description
The technical scheme of the invention is further explained by combining the accompanying drawings as follows:
as shown in fig. 1, a cellular mobile communication system used in the method for maximizing power distribution based on D2D communication energy efficiency of a deep neural network of the present invention includes a base station disposed at a central position of a mobile communication cell, cellular users disposed in the mobile communication cell, and a plurality of devices forming a D2D link, where the devices include a transmitting user and a receiving user, and form a D2D user pair as required, and the D2D user pair communicates with each other to form a D2D link, and in the figure, a schematic diagram of a relevant effective communication link and a co-channel interference link is given by taking uplink frequency sharing as an example;
s1, scheduling and activating users, scheduling and selecting a cellular user and a plurality of D2D links by a base station, sharing uplink spectrum resources of the cellular user to the plurality of D2D links for device-to-device direct communication:
in the cell, a base station adopts a centralized control wireless resource management method, uses a common scheduling algorithm in the existing communication system to schedule cellular users, shares uplink frequency spectrum resources of the cellular users to a plurality of proper D2D links and simultaneously carries out device-to-device communication; scheduling a "suitable D2D link" to minimize interference between communications of a base station and a cellular user and communications of the D2D link, and to ensure QoS requirements of the base station and the D2D links on the premise of resource sharing, where, for example, a distance is taken as an example, a distance between each D2D link and the base station, a distance between each D2D link and the cellular user, and a distance between a plurality of D2D links are all suitable, and generally, a distance between a scheduled D2D link and the base station is required to be greater than a distance between the cellular user and the base station, and distances between a D2D link and the cellular user, and between a D2D link and the remaining D2D links are required to exceed a preset value;
s2, the base station acquires real-time channel state information of a transmission link and an interference link, and sets a power constraint condition in communication:
the energy efficiency optimization operation is completed under the condition that the requirements of the D2D link transmitting power and the cellular user signal-to-interference-and-noise ratio are limited, and the related transmission links comprise channels from the cellular user to the base station and channels from a plurality of D2D link transmitting ends to corresponding receiving ends; the interference link comprises channels from a plurality of D2D transmitting ends to a base station, channels from a cellular user to a plurality of D2D receiving ends and channels from the D2D transmitting ends to other D2D receiving ends; fig. 1 illustrates the associated transmission link and interfering link, using uplink frequency sharing as an example.
Based on the base station scheduling selected cellular user and the plurality of corresponding D2D links in the step S1, firstly, the cellular user performs channel estimation to obtain channel state information of a transmission link between the cellular user and the base station, and simultaneously obtains the channel state information of interference links between the cellular user and all D2D receiving ends through the channel estimation, and then quantizes the channel state information according to a codebook which is known by the cellular user and the base station, and feeds back the quantized channel state information to the base station; meanwhile, the sending end of each D2D link performs channel estimation to obtain channel state information of a transmission link between the sending end and a corresponding receiving end and channel state information of interference links between the sending end and the receiving ends of the other D2D links, and in addition, the sending end of each D2D link also needs to obtain the channel state information of the interference links between the sending end and the base station through channel estimation, quantizes the channel state information according to a codebook known by the sending end and the base station, and feeds back the quantized channel state information to the base station. The base station may obtain real-time channel state information for the transmission link and the interfering link.
In the system model illustrated in fig. 1, the communication link between the cellular subscriber and the base station is identified by 0, and the uplink spectrum resources of the cellular subscriber are multiplexedThe total number of D2D links of the source is N, and these D2D links are identified by 1,2, \ 8230; N. When a cellular mobile communication system supports direct communication, if a cellular user has an absolute priority, the interference to a D2D link sharing its resources is not reduced by technical means, and the cellular user uses power P when communicating with a base station 0 Transmitting signals, i.e.
p 0 ＝P 0
The plurality of D2D links share spectrum resources with cellular users, wherein the transmission power of the nth D2D link satisfies the constraint condition:
0≤p n ≤p max
the signals transmitted by N D2D links sharing spectrum resources can cause interference to the communication of cellular users, and in order to guarantee the communication quality of the cellular users, the received signal to interference plus noise ratio (SINR) at a base station 0 To satisfy the constraint:
wherein: n represents the total number of D2D users simultaneously multiplexing the uplink frequency spectrum resources of the cellular users; p = [ p ] 1 ,p 2 ,...,p n ,...p N ]，p n Representing the transmitting power of the transmitting end of the nth D2D link; p is a radical of formula max Representing the maximum transmitting power of the D2D links, and assuming that the maximum transmitting power of the N D2D links is the same; r is min A signal to interference and noise ratio threshold value representing a transmission link of a cellular subscriber; g is a radical of formula 0,0 Channel state information representing a transmission link of a cellular subscriber to a base station; g n,0 Representing channel state information of an interference link from the nth D2D link to the base station, wherein N belongs to {1, 2.. N }; sigma 2 Representing a gaussian white noise variance;
s3, the base station inputs the acquired real-time channel state information into the trained deep neural network system to obtain a power distribution scheme meeting the power constraint condition:
specifically, power allocation is performed by taking optimal energy efficiency η of multiple D2D links sharing uplink spectrum resources as a target, where an expression of the energy efficiency is as follows:
wherein p is c The line loss power of a single D2D link in communication is represented, and the line losses of the N D2D links are assumed to be the same; g is a radical of formula n,m Represents channel state information from the n-th D2D link transmitting end to the m-th D2D link receiving end, and in particular, g when n = m n,m Representing the channel state information of a transmission link of the nth D2D link, wherein N, m belongs to {1, 2.. N }; g is a radical of formula 0,n Channel state information representing an interfering link between a cellular user and an nth D2D receiver, N ∈ {1, 2.. N }; sigma 2 And representing the Gaussian white noise variance of the D2D link, and assuming that the Gaussian white noise variances of all links are the same.
As shown in fig. 2, a trained deep neural network system used by a base station is composed of a normalization module, a hidden layer kernel and a multiplier, the deep neural network system is trained and completed based on a large number of channel samples before being applied to real-time communication, the base station can quickly obtain a D2D transmission power allocation scheme as long as channel state information obtained in real time is input into the deep neural network system, and the specific steps of obtaining the trained deep neural network system are as follows:
in order to accelerate the training speed of the neural network system to carry out normalization processing on the channel state information, the following formula is utilized to carry out normalization on the channel state information:
the kernel of the deep neural network hidden layer is formed by connecting two full connection layers (FC), three convolution layers (CNN) and a Sigmoid function module in series, and also comprises three reshape operation layers. The first full link layer (FC 1) fuses the normalized real-time channel state information, then three convolutional layers are used for extracting the characteristics of channels, the second full link layer (FC 2) fuses the channel characteristics extracted by the convolutional layers, each convolutional layer has the same filter setting, specifically, each convolutional layer uses a filter with the size of 3 x 3, the convolutional depths are 16, the step lengths are 1, and zero padding is used. And each convolution layer is provided with a ReLu activation function in front, which aims to prevent negative values on one hand, and can also enhance the nonlinearity of the network and improve the learning capability of the network on the other hand. The Sigmoid function module may limit the output value of the front-end kernel network to between 0 and 1, resulting in a normalized D2D transmitter power value. The first and third reshape layers are both for reshaping the input into a vector form suitable for importing to the fully-connected layer, and the second reshape layer reshapes the output of FC1 into a matrix form, so that the feature extraction of the convolutional layer is facilitated. Specifically, the output and activation functions of each layer of the deep neural network hidden kernel are shown in the following table, including the output size, the dimension and the activation function of each layer of the deep neural network hidden kernel:
the multiplier is used for obtaining the actual transmitting power of the D2D link, and because the deep neural network obtains the normalized power value, the normalized power value is multiplied by the maximum transmitting power of the D2D link, and the actual power distribution scheme of the D2D link meeting the power constraint condition can be obtained
The deep neural network adopts an unsupervised learning training method: firstly, collecting a large number of channel samples, explaining by taking 100000 groups of channels as training samples, leading in data of one bach once for training every 100 groups of channel data, initializing the weight of a neural network by using truncated normal distribution (truncated normal distribution), carrying out backward propagation by using an Adam algorithm, and obtaining a loss function f loss The definition is as follows:
f loss ＝-ωη+(1-ω)[r min -SINR 0 ] +
wherein the balance factor ω ∈ (0, 1) is used to adjust the trade-off between the energy efficiency optimization objective and the interference constraint. The data of 1000 bachs is trained to be 1 epoch, and the training is stopped until the requirement of the received signal-to-interference-and-noise ratio of the cellular link is met and the loss function does not drop any more.
S4, completing data transmission of the D2D link between the cellular user and the base station according to the power distribution scheme:
specifically, according to the power distribution scheme obtained by the deep neural network-based system of step S3, the cellular users transmit power P 0 Completing data transmission with the base station, and transmitting ends of all D2D links respectively according to the transmitting power
And completing data transmission of the link, wherein N belongs to {1, 2.,. N }, and the data transmission performance of a plurality of D2D links between the cellular user and the base station is improved as much as possible within the range of tolerable interference.
A sending end user and a receiving end user of each D2D link utilize a shared frequency spectrum to carry out direct communication between devices; the above-mentioned power allocation strategy can be used in both uplink and downlink, but since the base station transmission power is usually much larger than that of the mobile terminal, when the D2D link shares the downlink spectrum resource, the receiving end of the D2D link may suffer from severe interference from the base station, and in view of this, the 3GPP proposes that the D2D link shares the uplink spectrum resource. Since the D2D links share the intra-cell uplink spectrum resources, the co-channel interference level between the D2D links and the cellular communication link is low. Therefore, uplink spectrum resource sharing is adopted when a system model and simulation are shown.
The first embodiment,
In a communication scenario of a cellular mobile communication system supporting D2D communication, a cell has a radius of 500m, and a base station is located at the center of the cell and adopts a centralized resource management strategy. Transmission power P of cellular user 0 For 46dbm, n D2D links share the uplink channel resources of the cellular user at the same time. Line loss p per D2D link c Are all 0.05W, the transmission power threshold p max Is 20dBm; the distance between the cellular user and the base station is 250m, the cellular user and eachThe distance between a receiving user in the D2D link is 500m, the distance between a transmitting user and a base station in each D2D link is 300m, the distance between the transmitting user in each D2D link and the receiving user is 15m, and the distance between the transmitting user in each D2D link and the receiving user in other D2D links is 500m. The path loss model is:
L＝KβξD -α
wherein K is a path loss constant, beta represents small-scale fading and obeys exponential distribution, ξ represents shadow fading obeying log-normal, the standard deviation is 8dB, D is the distance between a transmitting end and a receiving end, m is a unit, and alpha is a path loss index. In particular, K =10 -2 ,α＝4。
Using the deep neural network system and training method shown in fig. 2, 100000 sets of channel data are generated by using the channel model described above to train the neural network, and 1000 sets of channels are generated to test the trained network. The size of one bach is set bach _ size =100 and the learning rate LR =0.00005.
Fig. 3 shows the energy efficiency of all D2D links at different balance factor ω settings and the variation of the sir constraints with different cellular users for a D2D user pair of 2, i.e. N = 2. "DNN, ω =1", "DNN, ω =0.1", "DNN, ω =0.01" and "DNN, ω =0.001" represent the loss function f loss ＝-ωη+(1-ω)[r min -SINR 0 ] + The balance factor ω in (1) is set to 1,0.1,0.01,0.001, respectively, of the performance of the deep neural network system trained. SINR required at different SINR of cellular user 0 Under the requirement of (2), constructing a loss function corresponding to the signal to interference and noise ratio requirement of the network to train the network. In particular, when ω =1, the built deep neural network is trained with the goal of maximizing D2D energy efficiency without considering the communication quality of the cellular user, so that "DNN, ω =1" is a straight line parallel to the x-axis and can be regarded as an upper bound of D2D energy efficiency when considering the signal-to-interference-plus-noise ratio requirement of the cellular user. When ω =0.1,0.01,0.001, the D2D communication energy efficiency decreases as the SINR requirements of the cellular user increase, and the gap from the upper bound (i.e., "DNN, ω = 1") increases as the SINR requirements of the cellular user increase. When the neural network is trained, the balance factorThe larger the omega is, the greater the proportion of the energy efficiency target is given on the two targets of energy efficiency maximization and cellular user signal-to-interference-and-noise ratio threshold requirement, so that the larger the balance factor omega is, the larger the energy efficiency of the D2D communication is under the same cellular user signal-to-interference-and-noise ratio condition. "Fixed power transmit" means that all D2D transmitters use a Fixed power p max 2 the performance of the communication and therefore also a straight line parallel to the x-axis. As can be seen from the figure, the energy efficiency of the D2D communication link of the optimization method proposed by this patent is higher than at a fixed transmit power p max Energy efficiency of D2D users of/2 communication, and when the SINR requirement of cellular users is small, the method based on the deep neural network provided by the invention is obviously superior to a fixed transmission power scheme.
Fig. 4 shows the variation of the energy efficiency of all D2D links with the total number of D2D links under different cellular user sinr threshold conditions. It can be seen from the figure that when the signal to interference plus noise ratio threshold of the cellular user is 30dB, as the number of D2D links multiplexing the spectrum resources increases, the energy efficiency sum of the D2D links decreases, and the effects of N =1 and N =2 are not very different, but when N ≧ 3, the energy efficiency sum of the D2D links decreases sharply, and it can be seen that when the communication quality of the cellular user is high, simply increasing the number of D2D links only adversely affects the energy efficiency performance of the D2D links. The energy efficiency sum of the D2D link is maximum when the signal to interference and noise ratio threshold of the cellular user is low, for example, 10db,15db,20db,25db, n = 2. Specifically, when the signal-to-interference-and-noise ratio threshold of the cellular user is 10db and 15db, the energy efficiency of N =5 is slightly higher than that of N =4, because the signal-to-interference-and-noise ratio of the cellular user at this time is very low, and the performance of the added D2D link is slightly higher than the interference influence caused by the added D2D link in some communication environments. In general, the amount of D2D links multiplexing uplink spectrum resources of a cellular user needs to be properly compromised and selected according to the signal to interference and noise ratio threshold of the cellular user in practical application, so that the system performance is better.
Fig. 5 examines the spectral utilization efficiency of the system. Under the condition of different cellular user signal to interference plus noise ratio thresholds, the spectral efficiency and the total spectral efficiency of cellular users vary with the total number of D2D links. It can be seen from the figure that the spectral efficiency of the cellular user decreases with the increase of the number of D2D links, because the increase of the D2D links brings more interference to the cellular links, and the lower the sinr threshold of the cellular user is, the smaller the spectral efficiency is, this is because the base station only needs to ensure that the cellular user occupies more resources for data transmission when the sinr of the cellular user meets the requirement, which is also why the total spectral efficiency is higher when the sinr of the cellular user is lower. Although the spectral efficiency of the cellular user may decrease as the number of D2D links increases, the total spectral efficiency of the cellular user and the D2D links may increase as the number of D2D links increases, and the more D2D links, the more the spectral efficiency increases, the higher the spectrum resource utilization. Generally speaking, in practical applications, it is necessary to select a setting that makes the system performance better according to the performance index of the actual communication requirement by a proper compromise.
It should be understood that the above examples are only for illustrating the specific embodiments of the technical solutions of the present invention, and are not intended to limit the scope of the present invention. Various equivalent modifications and alterations of this invention will occur to those skilled in the art after reading this disclosure, and they fall within the scope of this application, which is defined by the claims.
Claims (6)
1. A D2D communication energy efficiency maximization power distribution method based on a deep neural network is characterized in that a used cellular mobile communication system comprises a base station arranged at the center of a mobile communication cell, cellular users arranged in the mobile communication cell and a plurality of devices forming a D2D link, the devices comprise a sending user and a receiving user, and form a D2D user pair according to needs, and the method is characterized by comprising the following steps:
acquiring channel state information among a used base station, a used cellular user, a D2D link transmitting user and a used receiving user;
a base station schedules and selects a cellular user and a plurality of D2D user pairs in a mobile communication cell, and shares the frequency spectrum resource of the cellular user to a direct communication link formed by the plurality of D2D user pairs for carrying out direct communication between devices;
the frequency sharing of the link between the cellular user and the base station and the D2D link can effectively improve the spectrum efficiency, but can generate co-channel interference, so that the base station is utilized to acquire the real-time channel state information of the transmission link and the interference link, the real-time channel state information is input into the trained deep neural network system, and the optimal power distribution scheme meeting the power constraint condition is calculated to ensure that the system performance is improved as much as possible within the tolerable range of the co-channel interference;
the power of the D2D users of the cellular users is configured by using the optimal power distribution scheme, so that data transmission of a plurality of D2D links between the cellular users and the base station is completed;
received signal to interference and noise ratio (SINR) at base station 0 The power allocation scheme meeting the power constraint condition aims at optimizing the energy efficiency eta of a plurality of D2D links sharing uplink spectrum resources, and the optimized mathematical expression is as follows:
Subject to：0≤p n ≤p max
in optimizing the target eta, p c Representing the line loss in the communication of a single D2D link, and assuming that the line losses of the N D2D links are all the same; g n,m Indicating channel state information from a sending end of an nth D2D link to a receiving end of an mth D2D link, and when n = m, g n,m Representing the channel state information of a transmission link of the nth D2D link, wherein N, m belongs to {1, 2.. N }; g is a radical of formula 0,n Representing channel state information of an interference link between a cellular user and an nth D2D link receiving end, wherein N belongs to {1, 2.. N }; sigma 2 Representing the Gaussian white noise variance of the D2D link, and setting the Gaussian white noise variances of all links to be the same; n represents the total number of D2D users simultaneously multiplexing the uplink frequency spectrum resources of the cellular users; p = [ p ] 1 ,p 2 ,...,p n ,...p N ]，p n Representing the transmitting power of the transmitting end of the nth D2D link;p max representing the maximum transmitting power of the D2D links, and assuming that the maximum transmitting power of the N D2D links is the same; r is min A signal to interference and noise ratio threshold value representing a transmission link of a cellular subscriber; g is a radical of formula 0,0 Channel state information representing a transmission link of a cellular subscriber to a base station; g is a radical of formula n,0 Representing channel state information of an interference link from the nth D2D link to the base station, wherein N belongs to {1, 2.. N }; sigma 2 Representing a gaussian white noise variance;
the training process of the trained deep neural network system comprises the following steps:
the deep neural network system used at the base station and trained well consists of a normalization module, a hidden layer kernel and a multiplier, the deep neural network system is trained and perfected based on a large number of channel samples at a server end supporting high-speed and parallel computation before being applied to real-time D2D communication at the base station, and the base station can quickly obtain a transmitting power distribution scheme of a D2D link as long as channel state information obtained in real time is input into the deep neural network system
The configuration and the function of each module of the deep neural network system are as follows:
b1, in order to accelerate the training speed of the following neural network, firstly, a normalization module is used for normalizing the channel state information; specifically, the channel state information is normalized by the following formula:
in the formula, g i,j Representing the channel state information from the ith D2D link sending end to the jth D2D link receiving end;
b2, the kernel of the deep neural network hidden layer is formed by connecting two full connection layers FC, three convolution layers CNN and a Sigmoid function module in series, and also comprises three reshape operation layers;
the first full-connection layer FC1 fuses the normalized real-time channel state information, then three convolutional layers are used for extracting the characteristics of a channel, the second full-connection layer FC2 fuses the characteristics of the channel extracted by the convolutional layers, each convolutional layer has the same filter setting, specifically, each convolutional layer uses a filter with the size of 3 x 3, the convolutional depth is 16, the step length is 1, zero padding is used, and each convolutional layer is provided with a ReLU activation function in front, so that on one hand, the occurrence of a negative value is prevented, on the other hand, the nonlinearity of a network can be enhanced, and the learning capability of the network is improved; the Sigmoid function module can limit the output value of the network between 0 and 1 to obtain the normalized transmitting power value of the D2D link; the first and third reshape layers are used for reshaping the input into a vector form suitable for being led into the full-connection layer, and the second reshape layer is used for reshaping the output of the FC1 into a matrix form, so that the feature extraction of the convolutional layer is facilitated;
b3, in order to obtain the actual transmitting power of the D2D link, the multiplier multiplies the normalized power value with the maximum transmitting power of the D2D link, so as to obtain the actual transmitting power distribution scheme of the D2D link meeting the power constraint condition
n∈{1,2,...,N}。
2. The method for maximizing energy efficiency of D2D communication based on the deep neural network according to claim 1, wherein the selection of the cellular user and the D2D user is specifically as follows:
a1, in a cellular mobile communication system, a base station adopts a centralized control wireless resource management method, uses a common scheduling algorithm in the existing mobile communication system to schedule cellular users, and shares uplink spectrum resources of the cellular users to a plurality of proper D2D links to carry out direct communication between devices;
a2, the D2D link refers to a link for performing direct device-to-device communication between a transmitting user and a receiving user by using a shared spectrum, and the suitable multiple D2D links refer to that the distances from the scheduled multiple D2D links to a base station are required to be greater than the distances from cellular users to the base station, the distance between each D2D link and a cellular user needs to exceed a preset value, and the distance between each D2D link and the rest of the D2D links needs to exceed the preset value.
3. The method for energy efficiency maximization power distribution of D2D communication based on the deep neural network according to claim 1 or 2, wherein the method for the base station to obtain the real-time channel state information of the transmission link and the interference link is as follows:
the base station selects a cellular user forming a link and a plurality of D2D links sharing uplink spectrum resources of the cellular user, firstly, the cellular user carries out channel estimation to obtain channel state information of a transmission link between the cellular user and the base station, and simultaneously, channel state information of interference links between the cellular user and receiving ends of all the D2D links is obtained through the channel estimation, then, the channel state information is quantized according to a codebook known by the cellular user and the base station, and the quantized channel state information is fed back to the base station; meanwhile, each D2D link transmitting user first performs channel estimation to obtain channel state information of a transmission link between the transmitting user and a corresponding receiving end, and then performs channel estimation to obtain channel state information of interference links between the transmitting user and the other D2D link receiving users, and in addition, each D2D link transmitting user also needs to obtain channel state information of the interference links between the transmitting user and a base station through channel estimation, quantizes the channel state information according to a codebook commonly known by the transmitting user and the base station, and feeds back the quantized channel state information to the base station.
4. The D2D communication energy efficiency maximization power distribution method based on the deep neural network according to claim 3, characterized in that the method for setting the power constraint condition in communication is as follows:
in a cellular mobile communication system, a communication link between a cellular user and a base station is marked by 0, the total number of a plurality of D2D links multiplexing uplink spectrum resources of the cellular user is N, and the D2D links are marked by 1,2, \ 8230and N;
when the cellular mobile communication system supports D2D direct communication, if the cellular user has higher priority, it is not reduced by technical meansInterference to D2D links sharing their resources, i.e. using power p when cellular users communicate with base stations 0 ＝P 0 Transmitting a signal based on which:
when a plurality of D2D links share the uplink frequency spectrum resource with the cellular user, wherein the transmitting power of the nth D2D link satisfies that p is more than or equal to 0 n ≤p max The constraint of (a) to (b),
because the signals transmitted by the N D2D links sharing the spectrum resource can interfere the communication of the cellular user, in order to ensure the communication quality of the cellular user, the received signal-to-interference-and-noise ratio (SINR) at the base station 0 The constraint is to be satisfied.
5. The D2D communication energy efficiency maximization power distribution method based on the deep neural network as claimed in claim 1, characterized in that the deep neural network adopts an unsupervised learning training method:
firstly, collecting a large number of channel samples, explaining by using 100000 groups of channels as training samples, leading in data of one bach once for each 100 groups of channels, training, initializing the weight of a neural network by adopting truncation normal distribution, carrying out backward propagation by adopting an Adam algorithm, and obtaining a loss function f loss The definition is as follows:
f loss ＝-ωη+(1-ω)[r min -SINR 0 ] + ,
wherein, a balance factor omega epsilon (0, 1) is used for adjusting the compromise between the energy efficiency optimal target and the interference constraint condition, the data of 1000 bachs is trained to be 1 epoch until the receiving signal-to-interference-and-noise ratio requirement of a cellular link is met, and a loss function f loss And stopping training when the training platform does not descend any more.
6. The D2D communication energy efficiency maximization power distribution method based on the deep neural network according to claim 5, characterized in that: after the obtained optimal specific power distribution scheme, the cellular user sends power P 0 Completing data transmission with the base station, and each D2D link according to the transmission power
And completing data transmission, wherein N belongs to {1,2,. And N }.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011051195.4A CN112203345B (en) | 2020-09-29 | 2020-09-29 | D2D communication energy efficiency maximization power distribution method based on deep neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011051195.4A CN112203345B (en) | 2020-09-29 | 2020-09-29 | D2D communication energy efficiency maximization power distribution method based on deep neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112203345A CN112203345A (en) | 2021-01-08 |
| CN112203345B true CN112203345B (en) | 2022-11-15 |
Family
ID=74008041
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011051195.4A Active CN112203345B (en) | 2020-09-29 | 2020-09-29 | D2D communication energy efficiency maximization power distribution method based on deep neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112203345B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113038612B (en) * | 2021-03-01 | 2023-02-28 | 南京工业大学 | Cognitive radio power control method based on deep learning |
| CN113242069B (en) * | 2021-05-10 | 2022-06-17 | 东南大学 | Codebook design method based on neural network |
| CN113382477B (en) * | 2021-05-14 | 2022-09-27 | 北京邮电大学 | Method for modeling uplink interference between wireless network users |
| CN113411785B (en) * | 2021-06-22 | 2024-09-13 | 超越科技股份有限公司 | Minimum energy consumption control method and device for Overlay D2D network system |
| CN114928878B (en) * | 2021-09-17 | 2024-03-19 | 齐鲁工业大学 | Control method, device and medium for service quality constraint power in CRNs |
| CN116054893B (en) * | 2022-12-28 | 2024-09-06 | 西安电子科技大学 | Symbiotic non-orthogonal transmission method based on intelligent super surface |
| CN116321390A (en) * | 2023-05-23 | 2023-06-23 | 北京星河亮点技术股份有限公司 | Power control method, device and equipment |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108093411A (en) * | 2018-01-10 | 2018-05-29 | 重庆邮电大学 | Scheduling of resource optimization method based on channel signature in D2D communication networks |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110278546B (en) * | 2019-05-27 | 2022-02-22 | 东南大学 | Average energy efficiency maximization power distribution method in delay insensitive D2D communication system |
-
2020
- 2020-09-29 CN CN202011051195.4A patent/CN112203345B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108093411A (en) * | 2018-01-10 | 2018-05-29 | 重庆邮电大学 | Scheduling of resource optimization method based on channel signature in D2D communication networks |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112203345A (en) | 2021-01-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112203345B (en) | D2D communication energy efficiency maximization power distribution method based on deep neural network | |
| CN111800828B (en) | Mobile edge computing resource allocation method for ultra-dense network | |
| Lee et al. | Deep power control: Transmit power control scheme based on convolutional neural network | |
| CN110839184B (en) | Method and device for adjusting bandwidth of mobile fronthaul optical network based on flow prediction | |
| CN111328133B (en) | V2X resource allocation method based on deep neural network | |
| CN109729528A (en) | A kind of D2D resource allocation methods based on the study of multiple agent deeply | |
| CN116456493A (en) | D2D user resource allocation method and storage medium based on deep reinforcement learning algorithm | |
| CN111787543B (en) | 5G communication system resource allocation method based on improved wolf optimization algorithm | |
| CN108924799A (en) | The resource allocation algorithm of D2D communication in a kind of cellular network | |
| Azizi et al. | MIX-MAB: Reinforcement learning-based resource allocation algorithm for LoRaWAN | |
| Lee et al. | Robust transmit power control with imperfect CSI using a deep neural network | |
| Ma et al. | On-demand resource management for 6G wireless networks using knowledge-assisted dynamic neural networks | |
| CN110139282B (en) | Energy acquisition D2D communication resource allocation method based on neural network | |
| Li et al. | Computation scheduling of multi-access edge networks based on the artificial fish swarm algorithm | |
| CN113038583B (en) | Inter-cell downlink interference control method, device and system | |
| Zhang et al. | Power allocation in multi-cell system using distributed deep neural network algorithm | |
| CN111930501B (en) | Wireless resource allocation method based on unsupervised learning and oriented to multi-cell network | |
| CN112770398A (en) | Far-end radio frequency end power control method based on convolutional neural network | |
| CN115811788A (en) | D2D network distributed resource allocation method combining deep reinforcement learning and unsupervised learning | |
| CN108601083B (en) | Resource management method based on non-cooperative game in D2D communication | |
| Zhang et al. | A convolutional neural network based resource management algorithm for NOMA enhanced D2D and cellular hybrid networks | |
| CN111132312B (en) | Resource allocation method and device | |
| Zhang et al. | Deep Learning Based Resource Allocation for Full-duplex Device-to-Device Communication | |
| Chen et al. | Efficient Task Scheduling and Resource Allocation for AI Training Services in Native AI Wireless Networks | |
| CN113518457A (en) | Power distribution strategy based on one-dimensional deep convolutional neural network |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |