CN114786258A - Wireless resource allocation optimization method and device based on graph neural network - Google Patents

Wireless resource allocation optimization method and device based on graph neural network Download PDF

Info

Publication number
CN114786258A
CN114786258A CN202210339926.8A CN202210339926A CN114786258A CN 114786258 A CN114786258 A CN 114786258A CN 202210339926 A CN202210339926 A CN 202210339926A CN 114786258 A CN114786258 A CN 114786258A
Authority
CN
China
Prior art keywords
channel
sub
wireless network
power
frequency band
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
Application number
CN202210339926.8A
Other languages
Chinese (zh)
Inventor
张海君
马旭
刘向南
皇甫伟
隆克平
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Science and Technology Beijing USTB
Original Assignee
University of Science and Technology Beijing USTB
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by University of Science and Technology Beijing USTB filed Critical University of Science and Technology Beijing USTB
Priority to CN202210339926.8A priority Critical patent/CN114786258A/en
Publication of CN114786258A publication Critical patent/CN114786258A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • 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
    • H04W72/0453—Resources in frequency domain, e.g. a carrier in FDMA
    • 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
    • G06N3/048—Activation functions
    • 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
    • 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
    • G06N3/088—Non-supervised learning, e.g. competitive learning
    • 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
    • H04W16/22—Traffic simulation tools or models
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W72/00—Local resource management
    • H04W72/50—Allocation or scheduling criteria for wireless resources
    • H04W72/53—Allocation or scheduling criteria for wireless resources based on regulatory allocation policies

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Signal Processing (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

The invention discloses a wireless resource allocation optimization method and a device based on a graph neural network, wherein the method comprises the following steps: modeling large-scale wireless network deployment under a terahertz frequency band, and considering the power and sub-channel distribution problems of a plurality of receiver and transmitter pairs under the large-scale wireless network under the terahertz frequency band; interference in the signal transmission process is corrected, and a physical channel model of a large-scale wireless network downlink system under the terahertz frequency band is established; describing the optimization problem of wireless network resource allocation; modeling a wireless network into a wireless channel graph, and expressing a wireless network resource allocation optimization problem as a graph optimization problem; and finding a strategy which can map the wireless channel map to the optimal power allocation vector and the optimal sub-channel allocation vector, and realizing the joint allocation optimization of the power and the sub-channel. The method can solve the problem of resource allocation optimization of the large-scale wireless network under the terahertz frequency band.

Description

Wireless resource allocation optimization method and device based on graph neural network
Technical Field
The present invention relates to the field of wireless communication technologies, and in particular, to a method and an apparatus for optimizing wireless resource allocation based on a graph neural network.
Background
With the rapid development of modern communication services such as holographic communication, high-quality video online conferences, augmented reality/virtual display, 3D games and the like, the network data flow is increased rapidly, the requirements on network KPI (key performance indicator) such as data rate, time delay, connection number and the like are increased by orders of magnitude, and the data rate of future wireless communication is expected to exceed 100 Gbps. However, the currently available spectrum resources are too scarce to support such high data rates. Terahertz waves refer to electromagnetic waves in the frequency range of 0.1Thz to 10Thz, are located between microwave and infrared wave frequency bands in the whole electromagnetic spectrum, and have penetrability and absorptivity of the microwave frequency band and spectral resolution characteristics due to the special position of the electromagnetic spectrum. Terahertz communication is a technology for realizing wireless communication by taking a terahertz frequency band as a carrier wave, and the terahertz communication is a basic technology of a 6G mobile communication network, has available frequency band resources with ultra-large bandwidth, supports ultra-high communication rate and is regarded as a basic technology of the 6G mobile communication network. Therefore, considering the problem of resource allocation optimization of a large-scale wireless network in the terahertz frequency band, the available bandwidth and power resources are utilized as best as possible, and it becomes very important to meet the rapidly increasing user demand and the requirement of the number of access devices on the service quality.
The basic structure of the communication network is a graph, graph data is typical non-European space data and has complex correlation and inter-object dependency, and a method of the traditional graph theory is difficult to adapt to complex graph problems in a future network. Therefore, finding an algorithm for solving complex graph data to guide resource allocation and management scheduling of a communication network becomes an important scientific problem in a future network.
As an emerging technology in the field of artificial intelligence in recent years, the graph neural network opens up a new space for processing complex graph structure data. By means of artificial intelligence technologies such as deep learning and reinforcement learning, the graph neural network can quickly mine topological information and complex features in a graph structure, and a plurality of important problems in the fields of computer vision, recommendation systems, knowledge maps and the like are solved. Therefore, the combination of the neural network of the graph and a future network is an important way for solving the problem of network optimization, enhancing the reliability of the network and improving the utilization rate of network resources.
Disclosure of Invention
The invention provides a wireless resource allocation optimization method and device based on a graph neural network, which are used for meeting the increase of modern communication application on service requirements such as network scale, data rate, time delay and the like, solving the problem of resource allocation optimization of a large-scale wireless network under a terahertz frequency band and providing ubiquitous, consistent and reliable wireless communication service for ultra-high-speed wireless mobile scenes such as holographic communication, intelligent traffic and the like in terahertz communication.
In order to solve the technical problems, the invention provides the following technical scheme:
in one aspect, the present invention provides a method for optimizing radio resource allocation based on a graph neural network, where the method for optimizing radio resource allocation based on a graph neural network includes:
modeling large-scale wireless network deployment under a terahertz frequency band, and considering the power and sub-channel distribution problems of a plurality of receiver and transmitter pairs under the large-scale wireless network under the terahertz frequency band; correcting interference in the signal transmission process, and establishing a physical channel model of the large-scale wireless network downlink system under the terahertz frequency band;
describing the wireless network resource allocation optimization problem by taking the maximum energy efficiency as a target; modeling a wireless network into a wireless channel graph, and expressing a wireless network resource allocation optimization problem as a graph optimization problem;
through iterative updating of the graph neural network, a strategy that the wireless channel graph can be mapped to the optimal power allocation vector and the optimal sub-channel allocation vector is found, and joint allocation optimization of power and sub-channels is achieved.
Further, modeling is carried out on large-scale wireless network deployment under the terahertz frequency band, and the power and sub-channel distribution problems of a plurality of receiver and transmitter pairs under the large-scale wireless network under the terahertz frequency band are considered, wherein the modeling comprises the following steps:
assuming that each transceiver pair in the wireless network is allocated to a subchannel, N represents the number of subchannels, N represents a subchannel index, BW represents a total bandwidth, each subchannel is allocated averagely, and the bandwidth is B ═ BW/N; definition of SCnFor the nth sub-channel, KnThe amount of interference, U, generated for the nth subchannel with other subchannelsnTo be at SCnNumber of transceiver pairs inDenotes an index, U, of an ith transceiver pair at an nth subchannelmFor at mth sub-channel SCmNumber of transceiver pairs, jmThe index of the jth transceiver pair in the mth sub-channel is then selected from the jthmTransmitter to ith ofnThe channel response of the receiver is
Figure BDA0003578790570000021
Definition PmaxIs the maximum transmit power, p, of the systemnFor on sub-channel SCnThe power to be distributed is added to the power,
Figure BDA0003578790570000022
for SC on sub-channelnThe assigned transmit power of the ith transmitter; the transmit power constraint is expressed as:
Figure BDA0003578790570000023
thus, the receiver inThe received signal is a superposition of multiple transmitter channels, represented as:
Figure BDA0003578790570000024
wherein, the first and the second end of the pipe are connected with each other,
Figure BDA0003578790570000025
represents a mean of 0 and a variance of
Figure BDA0003578790570000026
Additive white gaussian noise of (1);
thus, the ithnThe signal-to-noise ratio of each receiver is expressed as:
Figure BDA0003578790570000031
wherein G istFor gain of the transmitting antenna, GrIn order to achieve the gain of the receiving antenna,
Figure BDA0003578790570000032
is the ithnThe channel responses of the sub-channels,
Figure BDA0003578790570000033
is the jthmTransmitter to ithnThe channel response of the receiver of the mobile station,
Figure BDA0003578790570000034
for on sub-channel SCnThe assigned transmit power of the ith transmitter,
Figure BDA0003578790570000035
for SC on sub-channelmThe assigned transmit power of the jth transmitter; thus, at sub-channel SCnI th of (1)nThe achievable rate for each receiver is expressed as:
Figure BDA0003578790570000036
the total data rate of the system is represented as:
Figure BDA0003578790570000037
wherein, CnIndicating the data rate sum for the nth subchannel.
Further, the correcting interference in the signal transmission process and establishing a physical channel model of the large-scale wireless network downlink system in the terahertz frequency band includes:
based on the multipath propagation model, the channel response of the nth sub-channel is defined as:
Figure BDA0003578790570000038
gain G of transmitter antennatAnd receiver antenna gain GrAre all set to 20dBi, X of MP pathnIs 1, i.e. only the LOS path exists, so the channel response of the nth sub-channel can be restated as:
Figure BDA0003578790570000039
Figure BDA00035787905700000310
wherein t is time, d is transmission distance, ΩLOSE {0,1} is the index variable when ΩLOSWhen 1, it indicates that a LOS path is included in the terahertz channel, when ΩLOSWhen the value is 0, the LOS path is not included in the terahertz channel; in the case of the LOS path,
Figure BDA00035787905700000311
denotes the attenuation, tLOSRepresenting a communication delay; in the case of the q-th emitted ray,
Figure BDA00035787905700000312
it is meant that the attenuation is,
Figure BDA00035787905700000313
representing a communication delay; the total number of MP paths is represented as:
Figure BDA00035787905700000314
wherein, the first and the second end of the pipe are connected with each other,
Figure BDA00035787905700000315
representing the amount of reflected light; the delta (.) function represents if and only if t ═ tLOSIs, delta (t-t)LOS)＝1。
Further, the wireless network resource allocation optimization problem is described with the goal of maximizing energy efficiency, and comprises the following steps:
first, a weighting factor is proposed
Figure BDA0003578790570000041
As an indicator of terahertz sub-channel allocation
Figure BDA0003578790570000042
Indicating that the nth sub-channel is occupied by i transceiver pairs, and vice versa
Figure BDA0003578790570000043
The total data rate of the system is thus restated as:
Figure BDA0003578790570000044
the terahertz system energy efficiency EE is defined as total data rate and total power consumption ptoTherefore, the energy efficiency of the large-scale wireless network in the terahertz frequency band is defined as follows:
Figure BDA0003578790570000045
wherein p iscRepresents the circuit power consumption;
thus, the optimization problem is expressed as:
Figure BDA0003578790570000046
s.t.C1:
Figure BDA0003578790570000047
C2:
Figure BDA0003578790570000048
C3:
Figure BDA0003578790570000049
C4:
Figure BDA00035787905700000410
C5:
Figure BDA00035787905700000411
wherein C1 and C2 are power constraints of the wireless network under the terahertz frequency band, and C3 is the minimum data rate CminConstraints, C4 and C5, are terahertz sub-channel allocation constraints, D represents the maximum number of transceivers per sub-channel; s. theminLower bound, S, representing the number of sub-channels allocated to the ith transceiver pairmaxAn upper bound indicating the number of sub-channels allocated to the ith transceiver pair,
Figure BDA00035787905700000412
denotes the ithnThe data rate of each receiver.
Further, one node of the wireless channel diagram represents a transceiving pair, the node characteristics comprise direct channel state and weight, the link of the diagram is an interference channel, and the link characteristics are interference channel state.
Further, the modeling the wireless network into a wireless channel map includes:
modeling the multiple transceivers for the interfering channel as a complete graph G ═ of labels with vertices and edges (V, E, s, t); where V represents a set containing vertices, E represents a set containing edges,
Figure BDA00035787905700000413
representing the mapping of a node to a feature of the node,
Figure BDA0003578790570000051
representing a feature that maps an edge to an edge, node viE.v denotes a pair of receiver and transmitter, V ═ V1,v2,...,v|V|}, vertex
Figure BDA0003578790570000052
The ith vertex, representing the nth subchannel, where N ∈ {1,2,. cndot, N }, i ∈ {1,2,. cndot, U ∈ {1,2,. cndotnThe feature matrix of the node is expressed as
Figure BDA0003578790570000053
Wherein Z(i,:)＝s(vi) (ii) a Edge
Figure BDA0003578790570000054
Define the slave vertex
Figure BDA0003578790570000055
To the top
Figure BDA0003578790570000056
Of the directed relationship of (A), connecting the vertices
Figure BDA0003578790570000057
Is represented as
Figure BDA0003578790570000058
Vertex point
Figure BDA0003578790570000059
Is expressed as
Figure BDA00035787905700000510
If it is used
Figure BDA00035787905700000511
Exist, then
Figure BDA00035787905700000512
If not, [ A ]nm]i,j＝0；
In case of radio interference, willEach pair of transceivers acts as a vertex, the interference pattern from the transmitter to the receiver acts as an edge, and the channel properties of each vertex
Figure BDA00035787905700000513
Including weights
Figure BDA00035787905700000514
Variance of noise
Figure BDA00035787905700000515
Direct channel response
Figure BDA00035787905700000516
Each edge is characterized by a channel response from the interfering transmitter to the interfered receiver.
Further, through iterative update of the graph neural network, a strategy for mapping the wireless channel graph to the optimal power allocation vector and the optimal sub-channel allocation vector is found, and joint allocation optimization of power and sub-channels is realized, wherein the strategy comprises the following steps:
defining different edge update functions phi for nodes on different sub-channelseAnd a vertex update function phivThese functions are parameterized by a multilayer perceptron; assuming that the number of convolution layers of the graph neural network is L, iteration is carried out from 1 to L in a circulating mode, in each layer, the nodes send node information to adjacent edges, the node characteristics and the edge characteristics are updated through the following formula, the node characteristics of all the nodes are output until the L-th layer, and meanwhile, an activation function is added when the L-th layer of the graph neural network outputs
Figure BDA00035787905700000517
Wherein p represents a power allocation vector, w represents a subchannel allocation vector, and | represents a vector modulo length;
updating edge characteristics:
Figure BDA00035787905700000518
updating the node characteristics:
Figure BDA00035787905700000519
aggregation node characteristics:
Figure BDA00035787905700000520
the loss function L (theta) is a negative expectation of a utility function realized on different channels, is optimized by a random gradient descent method, is propagated reversely in the loss function, and updates the neural network model parameter theta in an unsupervised mode:
Figure BDA00035787905700000521
wherein E isHRepresenting an expectation function;
obtaining a strategy pi (-) through the characteristic information of each vertex, and outputting an optimal power distribution vector and an optimal sub-channel distribution vector
Figure BDA00035787905700000522
And realizing the joint allocation optimization of power and sub-channels.
On the other hand, the invention also provides a wireless resource allocation optimization device based on the graph neural network, which comprises the following steps:
the wireless network modeling module is used for modeling large-scale wireless network deployment under the terahertz frequency band, and considering the power and sub-channel distribution problems of a plurality of receiver and transmitter pairs under the large-scale wireless network under the terahertz frequency band; correcting interference in the signal transmission process, and establishing a physical channel model of the large-scale wireless network downlink system under the terahertz frequency band;
the wireless channel map construction module is used for describing the wireless network resource allocation optimization problem by taking the maximum energy efficiency as a target; modeling a wireless network into a wireless channel graph, and expressing a wireless network resource allocation optimization problem as a graph optimization problem;
and the optimization module is used for finding a strategy which can map the wireless channel map to the optimal power distribution vector and the optimal sub-channel distribution vector through iterative update of the map neural network, and realizing the joint distribution optimization of the power and the sub-channel.
In yet another aspect, the present invention also provides an electronic device comprising a processor and a memory; wherein the memory has stored therein at least one instruction that is loaded and executed by the processor to implement the above-described method.
In yet another aspect, the present invention also provides a computer-readable storage medium having at least one instruction stored therein, the instruction being loaded and executed by a processor to implement the above method.
The technical scheme provided by the invention has the beneficial effects that at least:
according to the technical scheme, a wireless network is modeled into a wireless channel diagram, a target task structure for optimizing wireless network resource allocation is combined into a neural network framework, the problem of optimizing wireless network resource allocation is expressed as a diagram optimization problem, the problem of optimizing large-scale wireless network resource allocation under a terahertz frequency band is solved by using expansibility of the diagram neural network, and the performance of the large-scale wireless network under the terahertz frequency band is enhanced. The problem of resource allocation optimization of a large-scale wireless network under a terahertz frequency band is solved, and ubiquitous, consistent and reliable wireless communication services are provided for ultra-high-speed wireless mobile scenes such as holographic communication and intelligent traffic in terahertz communication.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the description below are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic flowchart of an implementation of a method for optimizing radio resource allocation based on a graph neural network according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a directed interference graph construction method based on a graph neural network according to an 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.
First embodiment
In order to meet the increase of service requirements of modern communication application on network scale, data rate, time delay and the like, solve the problem of resource allocation optimization of a large-scale wireless network under a terahertz frequency band, and provide ubiquitous, consistent and reliable wireless communication services for ultra-high-speed wireless mobile scenes such as holographic communication, intelligent traffic and the like in terahertz communication, the embodiment provides a wireless resource allocation optimization method based on a graph neural network. The technical scheme of the method comprises a large-scale wireless network model under the terahertz frequency band, a wireless channel random edge diagram neural network and a joint optimization scheme of power control and sub-channel allocation.
Specifically, the execution flow of the radio resource allocation optimization method includes the following steps:
s1, modeling large-scale wireless network deployment under the terahertz frequency band, and considering the power and sub-channel distribution problem of a plurality of receiver and transmitter pairs under the large-scale wireless network under the terahertz frequency band; correcting interference in the signal transmission process, and establishing a physical channel model of a large-scale wireless network downlink system under a terahertz frequency band;
s2, describing the optimization problem of wireless network resource allocation with the aim of maximizing energy efficiency; modeling a wireless network into a wireless channel graph, and expressing a wireless network resource allocation optimization problem as a graph optimization problem;
and S3, through iterative updating of the graph neural network, finding a strategy which can map the wireless channel graph to the optimal power allocation vector and the optimal sub-channel allocation vector, and realizing the joint allocation optimization of the power and the sub-channel.
Specifically, in this embodiment, the wireless network deployment is modeled in the above S1 as follows:
considering a plurality of receiver and transmitter pairs under a large-scale wireless network under a terahertz frequency bandPower and subchannel allocation, assuming that each pair of transceivers in the wireless network will be allocated to a subchannel, N denotes the number of subchannels, N denotes the subchannel index, where N ∈ {1, 2.., N }, BW denotes the total bandwidth, each subchannel is allocated evenly, and the bandwidth is B ═ BW/N; definition of SCnFor the nth sub-channel, KnThe amount of interference, U, generated for the nth subchannel with other subchannelsnTo be at SCnNumber of transceiver pairs, inIs shown in the nth sub-channel (SC)n) Index of the ith transceiver pair of (1), in∈{1,2,...,Un}；UmFor at mth sub-channel SCmNumber of transceiver pairs, jmIs shown in the m-th sub-channel (SC)m) J of the jth transceiver pair ofm∈{1,2,...,Um}; then from the jthmTransmitter to ith ofnThe channel response of the receiver is
Figure BDA0003578790570000071
Definition PmaxIs the maximum transmit power, p, of the systemnFor SC on sub-channelnThe power that is distributed to the power transmission line,
Figure BDA0003578790570000072
for on sub-channel SCnThe assigned transmission power of the ith transmitter; the transmit power constraint is expressed as:
Figure BDA0003578790570000081
thus, the receiver inThe received signal is a superposition of multiple transmitter channels, which can be expressed as:
Figure BDA0003578790570000082
wherein, the first and the second end of the pipe are connected with each other,
Figure BDA0003578790570000083
represents a mean of 0 and a variance of
Figure BDA0003578790570000084
Additive White Gaussian Noise (AWGN); thus, the ithnThe signal-to-noise ratio of each receiver can be expressed as:
Figure BDA0003578790570000085
wherein, GtGain for the transmitting antenna, GrIn order to be the gain of the receiving antenna,
Figure BDA0003578790570000086
is the ithnThe channel response of the sub-channels,
Figure BDA0003578790570000087
is the jthmTransmitter to ith ofnThe channel response of the receiver of the mobile station,
Figure BDA0003578790570000088
for SC on sub-channelnThe assigned transmit power of the ith transmitter,
Figure BDA0003578790570000089
for SC on sub-channelmThe assigned transmit power of the jth transmitter; thus, at subchannel SCnIth of (2)nThe receiver achievable rates are expressed as:
Figure BDA00035787905700000810
the total data rate of the system can be expressed as:
Figure BDA00035787905700000811
wherein, CnRepresenting the sum of the data rates of the nth subchannel.
Further, the implementation process of correcting the interference in the signal transmission process in S1 and establishing the physical channel model of the large-scale wireless network downlink system in the terahertz frequency band is as follows:
multipath Propagation (MP) models include line-of-sight (LoS) links and non-line-of-sight (NLoS) links, which include three portions, reflection, scattering, and diffraction paths, which are negligible because they accept less power. Based on the multipath propagation model, the channel response of the nth sub-channel is defined as:
Figure BDA00035787905700000812
where t is time, d is transmission distance, ΩLOSE {0,1} is the index variable when ΩLOSWhen 1, it indicates that a LOS path is included in the terahertz channel, when ΩLOSWhen the value is 0, the terahertz channel does not include a LOS path; in the case of the LOS path,
Figure BDA00035787905700000813
denotes the attenuation, tLOSRepresenting a communication delay; in the case of the q-th emitted ray,
Figure BDA0003578790570000091
it is meant that the attenuation is,
Figure BDA0003578790570000092
representing a communication delay; the total number of MP paths is represented as:
Figure BDA0003578790570000093
wherein the content of the first and second substances,
Figure BDA0003578790570000094
representing the amount of reflected light, the function δ (. -) representing if and only if t ═ tLOSWhen, delta (t-t)LOS)＝1。
Under the effect of the high-gain antenna, the transmission beam width becomes smaller, and the number of MP rays is reduced sharply. To this end, inTerahertz frequency band, the present embodiment proposes a highly directional antenna to prevent high path loss of terahertz communication, terahertz transmission having high directivity of a high-gain antenna, and thus, a transmitter antenna gain GtAnd receiver antenna gain GrAre all set to 20dBi, X of MP pathnIs 1, i.e. only the LOS path exists, the channel response of the nth sub-channel can be restated as:
Figure BDA0003578790570000095
Figure BDA0003578790570000096
all path losses of LOS propagating in the terahertz waveband are composed of airborne propagation attenuation and atmospheric attenuation, and the propagation attenuation can be expressed as:
Figure BDA0003578790570000097
where f is the frequency and c is the speed of light in vacuum.
The molecular absorption decay cannot be eliminated and the atmospheric decay is expressed as:
Labs(f,d)[dB]＝10k(f)d lg e
where k (f) is the absorption coefficient of the frequency-dependent medium.
Therefore, the path loss of the terahertz frequency band can be expressed as:
PL(d)[dB]＝Lspread(f,d)[dB]+Labs(f,d)[dB]
the relationship between path loss and channel gain is defined as:
Figure BDA0003578790570000098
further, in consideration of performance requirements of signal transmission in a large-scale wireless network under the terahertz frequency band, the above S2 first describes a joint optimization problem of power and sub-channel allocation with a goal of maximizing Energy Efficiency (EE), which is specifically as follows:
first, a weight factor is proposed
Figure BDA0003578790570000099
As an indicator of terahertz subchannel allocation
Figure BDA00035787905700000910
Indicating that the nth sub-channel is occupied by i transceiver pairs, and vice versa
Figure BDA00035787905700000911
The total data rate of the system is thus restated as:
Figure BDA00035787905700000912
in addition to the transmission power, the total power consumption p during wireless communicationtoInvolving a circuit power consumption pcTherefore, total power consumption ptoExpressed as:
Figure BDA0003578790570000101
the terahertz system energy efficiency EE is defined as total data rate and total power consumption ptoTherefore, the energy efficiency of the large-scale wireless network in the terahertz frequency band is defined as follows:
Figure BDA0003578790570000102
thus, the optimization problem is expressed as:
Figure BDA0003578790570000103
s.t.C1:
Figure BDA0003578790570000104
C2:
Figure BDA0003578790570000105
C3:
Figure BDA0003578790570000106
C4:
Figure BDA0003578790570000107
C5:
Figure BDA0003578790570000108
wherein, C1 and C2 are power constraints of the wireless network under the terahertz frequency band, C1 gives transmission power constraints, and total transmission power P is setmaxTo ensure that the total power consumption of all transceiver pairs is less than or equal to the maximum value. C2 guarantees a non-negative power limit for the power allocated to each terminal by each terahertz sub-channel; c3 is minimum data rate CminIn order to ensure the performance of the terahertz system, C3 specifies that the achievable data rate of each terminal must be greater than or equal to a minimum data rate RminThe minimum data rate is determined by the QoS requirement; c4 and C5 are terahertz sub-channel allocation constraints, and C4 ensures that each SC is occupied by no more than D transceiver pairs, which prevents a large number of transceiver pairs on the terahertz sub-channel; c5 sets S as the upper bound of the number of sub-channels allocated to the ith transceiver pairmaxLower bound is set to SminFairness between transceiver pairs is described. It can efficiently utilize too large bandwidth resources. By defining the appropriate SmaxAnd SminAnd reconstructing the priority of the transceiver to the occupied terahertz sub-channel. D represents the maximum number of transceivers per sub-channel,
Figure BDA0003578790570000109
denotes the ithnThe data rate of each receiver.
Further, one node of the radio channel map in S2 represents one transmission/reception pair, the node characteristics include a direct channel state and a weight, the link of the map is an interference channel, and the link characteristics are an interference channel state.
In the above S2, the modeling of the wireless network as the wireless channel map includes:
modeling the multiple transceivers for the interfering channel as a complete graph G ═ of labels with vertices and edges (V, E, s, t); where V represents a set containing vertices, E represents a set containing edges,
Figure BDA0003578790570000111
representing the mapping of a node to a feature of the node,
Figure BDA0003578790570000112
representing a feature that maps an edge to an edge, node viE.v denotes a pair of receiver and transmitter, V ═ V1,v2,...,v|V|H, vertex vinDenotes the nth sub-channel (SC)n) Is used, where N is in {1,2,. eta., N }, i is in {1,2,. eta., UnDenoted by the feature matrix of the node
Figure BDA0003578790570000113
Wherein Z(i,:)＝s(vi) (ii) a Edge
Figure BDA0003578790570000114
Define the slave vertex
Figure BDA0003578790570000115
To the top
Figure BDA0003578790570000116
The directed relationship of (2) connecting the vertices
Figure BDA0003578790570000117
Is represented as
Figure BDA0003578790570000118
Vertex point
Figure BDA0003578790570000119
Is expressed as
Figure BDA00035787905700001110
If it is used
Figure BDA00035787905700001111
Exist, then
Figure BDA00035787905700001112
If not present, [ A ]nm]i,j0; in the case of radio interference, each pair of transceivers is taken as a vertex, the interference pattern from the transmitter to the receiver is taken as an edge, and the channel property of each vertex
Figure BDA00035787905700001113
Including weights
Figure BDA00035787905700001114
Variance of noise
Figure BDA00035787905700001115
Direct channel response
Figure BDA00035787905700001116
Each edge
Figure BDA00035787905700001117
Is characterized by the channel response from the interfering transmitter to the interfered receiver
Figure BDA00035787905700001118
Based on the above, the optimization method of the embodiment models the interference relationship into the channel coefficient as
Figure BDA00035787905700001119
The directed graph G of (a), through iterative updating of the graph neural network,final edge characteristic information and node characteristic information are output at the last layer of the graph neural network, and the vector capable of mapping the directed graph to the optimal power distribution is found
Figure BDA00035787905700001120
And an optimal subchannel allocation vector
Figure BDA00035787905700001121
And (3) realizing the joint allocation optimization of power and sub-channels. Policy function piθ(. cndot.) is represented as
Figure BDA00035787905700001122
Where θ represents a learnable parameter, λ1、λ2Representing the policy allocation scaling factor.
Further, the above S3 trains the transfer function and the output function through the convolutional neural network by using an unsupervised learning method, so as to obtain the optimal strategy for the transmission power and the sub-channel allocation of each transmitter.
In contrast, in the study of the existing wireless network "learning optimization" method, MLPs are mostly used as a neural network architecture, and although MLPs can approximate a function with good performance, the data efficiency, robustness and generalization performance are poor. However, in the wireless network optimization method based on the graph neural network, the structure of the target task is incorporated into a neural network system architecture, the neural network does not need to learn the structures from data, the training efficiency is higher, the generalization can be better from experience, meanwhile, the graph neural network has the displacement equal variance performance, the neural network architecture has expandability, and the method can be expanded to a large-scale wireless network to solve the power and sub-channel distribution optimization problem. The power and sub-channel allocation joint optimization method based on the graph neural network effectively utilizes the graph structure, applies the traditional convolutional neural network and the traditional cyclic neural network to the graph optimization problem, rapidly mines topological information and complex features in the graph structure, and trains a transfer function and an output function by adopting an unsupervised learning method, thereby obtaining the optimal strategy of the transmission power and the sub-channel allocation of each transmitter.
The basic calculation unit on the graph convolution neural network is a graph neural block, each graph neural block comprises an updating function phi and an aggregation function rho, the graph neural block needs to generate output on each vertex to serve as a transmission scheme of each link, and target vertex information is sampled and aggregated from different subchannel vertices to obtain final state information.
Defining different edge update functions phi for nodes on different sub-channelseAnd a vertex update function phivThese functions are parameterized by multilayer perceptrons (MLPs); assuming that the number of convolution layers of the graph neural network is L, iteration is carried out from 1 to L in a circulating mode, in each layer, the nodes send node information to adjacent edges, the node characteristics and the edge characteristics are updated through the following formula, the node characteristics of all the nodes are output until the L-th layer, and meanwhile, an activation function is added when the L-th layer of the graph neural network outputs
Figure BDA0003578790570000121
And applying constraints; where p represents the power allocation vector, w represents the subchannel allocation vector, and | represents the vector modulo length. The method comprises the following specific steps:
the update and aggregation definition for the l-th layer is expressed as:
updating the ith layer edge characteristic:
Figure BDA0003578790570000122
updating the characteristic of the vertex at the l layer:
Figure BDA0003578790570000123
polymerization of layer I:
Figure BDA0003578790570000124
where l represents the first update of the neural network of the graph, the edge update function φ is applied firsteInitializing all vertex features at each edge
Figure BDA0003578790570000125
And all edge characteristics
Figure BDA0003578790570000126
Then each vertex is put
Figure BDA0003578790570000127
Applying an aggregation function pe→vPolymerizing to obtain an edge
Figure BDA0003578790570000128
And (4) updating. Then, the updating of the aggregation edge is combined
Figure BDA0003578790570000129
And current attributes of the node
Figure BDA00035787905700001210
Through phivThe update obtains new vertex features. When each vertex feature information is embedded into the edge feature update and applied to the updates of its neighboring vertices, the information transfer process is complete. The updating function of the graph neural network selects a neural network module, the aggregation function selects an expectation function, the updating of each vertex is independent, and the minimum loss function is ensured after each updating. After L times of updating, outputting the final node characteristics
Figure BDA00035787905700001211
And constructing a plurality of activation function applying constraints at the output of the neural network of the graph:
Figure BDA00035787905700001212
and the constraint conditions in the optimization problem are met.
The loss function L (theta) is a negative expectation of a utility function realized on different channels, is optimized by a random gradient descent method, is propagated reversely in the loss function, and updates the neural network model parameter theta in an unsupervised mode:
Figure BDA00035787905700001213
wherein, EHRepresenting an expectation function;
finally, obtaining a strategy pi (-) through the characteristic information of each vertex, and outputting an optimal power distribution vector and an optimal sub-channel distribution vector
Figure BDA0003578790570000131
And realizing the joint allocation optimization of power and sub-channels.
In summary, the power and subchannel allocation joint optimization method based on the graph neural network provided by this embodiment effectively utilizes the graph structure, incorporates the structure of the target task into the neural network architecture, quickly mines the topology information and complex features in the graph structure, and the neural network does not need to learn these structures from data, so that the training efficiency is higher, and the method can be generalized better from experience, and can be extended to a large-scale wireless network to solve the power and subchannel allocation optimization problem. Meanwhile, an unsupervised learning method is adopted to train a transfer function and an output function, so that the optimal strategy of the transmission power and the sub-channel distribution of each transmitter is obtained. Therefore, the problem of resource allocation optimization of a large-scale wireless network under the terahertz frequency band is solved, and ubiquitous, consistent and reliable wireless communication services are provided for ultra-high-speed wireless mobile scenes such as holographic communication and intelligent traffic in terahertz communication.
Second embodiment
The embodiment provides a wireless resource allocation optimizing device based on a graph neural network, which comprises:
the wireless network modeling module is used for modeling large-scale wireless network deployment under the terahertz frequency band, and considering the power and sub-channel distribution problems of a plurality of receiver and transmitter pairs under the large-scale wireless network under the terahertz frequency band; interference in the signal transmission process is corrected, and a physical channel model of a large-scale wireless network downlink system under the terahertz frequency band is established;
the wireless channel map building module is used for describing the wireless network resource allocation optimization problem by taking the maximum energy efficiency as a target; modeling a wireless network into a wireless channel graph, and expressing a wireless network resource allocation optimization problem as a graph optimization problem;
and the optimization module is used for finding a strategy which can map the wireless channel map to the optimal power distribution vector and the optimal sub-channel distribution vector through iterative update of the map neural network, and realizing the joint distribution optimization of the power and the sub-channel.
The wireless resource allocation optimization device based on the graph neural network of the present embodiment corresponds to the wireless resource allocation optimization method based on the graph neural network of the first embodiment; the functions realized by the functional modules in the wireless resource allocation optimization device based on the graph neural network of the present embodiment correspond to the flow steps in the wireless resource allocation optimization method based on the graph neural network one by one; therefore, it will not be described herein.
Third embodiment
The embodiment provides an electronic device, which comprises a processor and a memory; wherein the memory has stored therein at least one instruction that is loaded and executed by the processor to implement the method of the first embodiment.
The electronic device may have a relatively large difference due to different configurations or performances, and may include one or more processors (CPUs) and one or more memories, where at least one instruction is stored in the memory, and the instruction is loaded by the processor and executes the method.
Fourth embodiment
The present embodiment provides a computer-readable storage medium, which stores at least one instruction, and the instruction is loaded and executed by a processor to implement the method of the first embodiment. The computer readable storage medium may be, among others, ROM, random access memory, CD-ROM, magnetic tape, floppy disk, optical data storage device, and the like. The instructions stored therein may be loaded by a processor in the terminal and perform the above-described method.
Furthermore, it should be noted that the present invention may be provided as a method, apparatus or computer program product. Accordingly, embodiments of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, embodiments of the present invention may take the form of a computer program product embodied on one or more computer-usable storage media having computer-usable program code embodied in the medium.
Embodiments of the present invention are described with reference to flowchart illustrations and/or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, embedded processor, or other programmable data processing terminal to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks. These computer program instructions may also be loaded onto a computer or other programmable data processing terminal to cause a series of operational steps to be performed on the computer or other programmable terminal to produce a computer implemented process such that the instructions which execute on the computer or other programmable terminal provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It should also be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or terminal. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or terminal that comprises the element.
Finally, it should be noted that while the above describes a preferred embodiment of the invention, it will be appreciated by those skilled in the art that, once having the benefit of the teaching of the present invention, numerous modifications and adaptations may be made without departing from the principles of the invention and are intended to be within the scope of the invention. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the embodiments of the invention.

Claims (8)

1. A wireless resource allocation optimization method based on a graph neural network is characterized by comprising the following steps:
modeling large-scale wireless network deployment under a terahertz frequency band, and considering the power and sub-channel distribution problems of a plurality of receiver and transmitter pairs under the large-scale wireless network under the terahertz frequency band; correcting interference in the signal transmission process, and establishing a physical channel model of the large-scale wireless network downlink system under the terahertz frequency band;
describing the wireless network resource allocation optimization problem by taking the maximum energy efficiency as a target; modeling a wireless network into a wireless channel graph, and expressing a wireless network resource allocation optimization problem as a graph optimization problem;
through iterative update of the graph neural network, a strategy that the wireless channel graph can be mapped to the optimal power allocation vector and the optimal sub-channel allocation vector is found, and joint allocation optimization of power and sub-channels is achieved.
2. The method for optimizing wireless resource allocation based on the graph neural network according to claim 1, wherein modeling large-scale wireless network deployment in the terahertz frequency band, considering power and sub-channel allocation problems of multiple receiver and transmitter pairs in the large-scale wireless network in the terahertz frequency band, comprises:
assuming that each transceiver pair in the wireless network is allocated to a subchannel, N represents the number of subchannels, N represents a subchannel index, BW represents a total bandwidth, each subchannel is allocated averagely, and the bandwidth is B ═ BW/N; definition of SCnFor the nth sub-channel, KnThe amount of interference, U, generated for the nth sub-channel with other sub-channelsnTo be at SCnNumber of transceiver pairs, inDenotes an index, U, of an ith transceiver pair in the nth sub-channelmFor at mth sub-channel SCmNumber of transceiver pairs, jmThe index of the jth transceiver pair in the mth sub-channel is then calculated from the jthmTransmitter to ith ofnThe channel response of the receiver is
Figure FDA0003578790560000011
Definition PmaxIs the maximum transmit power, p, of the systemnFor on sub-channel SCnThe power that is distributed to the power transmission line,
Figure FDA0003578790560000012
for SC on sub-channelnThe assigned transmission power of the ith transmitter; the transmit power constraint is expressed as:
Figure FDA0003578790560000013
thus, the receiver inThe received signal is a superposition of multiple transmitter channels, represented as:
Figure FDA0003578790560000014
wherein, the first and the second end of the pipe are connected with each other,
Figure FDA0003578790560000015
represents a mean of 0 and a variance of
Figure FDA0003578790560000016
Additive white gaussian noise of (1);
thus, the ithnThe signal-to-noise ratio of each receiver is expressed as:
Figure FDA0003578790560000017
wherein G istFor gain of the transmitting antenna, GrIn order to be the gain of the receiving antenna,
Figure FDA0003578790560000021
is the ithnThe channel response of the sub-channels,
Figure FDA0003578790560000022
is the jthmTransmitter to ith ofnThe channel response of the receiver of the station,
Figure FDA0003578790560000023
for SC on sub-channelnThe assigned transmit power of the ith transmitter,
Figure FDA0003578790560000024
for on sub-channel SCmThe assigned transmit power of the jth transmitter; thus, at sub-channel SCnI th of (1)nThe achievable rate for each receiver is expressed as:
Figure FDA0003578790560000025
the total data rate of the system is represented as:
Figure FDA0003578790560000026
wherein, CnRepresenting the sum of the data rates of the nth subchannel.
3. The method for optimizing wireless resource allocation based on the graph neural network according to claim 2, wherein the step of correcting the interference in the signal transmission process and establishing a physical channel model of a large-scale wireless network downlink system in the terahertz frequency band comprises the following steps:
based on the multipath propagation model, the channel response of the nth sub-channel is defined as:
Figure FDA0003578790560000027
gain G of transmitter antennatAnd receiver antenna gain GrAre all set to 20dBi, X of MP pathnIs 1, i.e. only the LOS path exists, so the channel response of the nth sub-channel is restated as:
Figure FDA0003578790560000028
Figure FDA0003578790560000029
where t is time, d is transmission distance, ΩLOSEpsilon {0,1} is an index variable when omegaLOSWhen 1, it indicates that a LOS path is included in the terahertz channel, when ΩLOSWhen the value is 0, the LOS path is not included in the terahertz channel; in the case of the LOS path,
Figure FDA00035787905600000210
denotes the attenuation, tLOSRepresenting a communication delay; in the case of the q-th emitted ray,
Figure FDA00035787905600000211
which is indicative of the attenuation of the light beam,
Figure FDA00035787905600000212
representing a communication delay; the total number of MP paths is represented as:
Figure FDA00035787905600000213
wherein, the first and the second end of the pipe are connected with each other,
Figure FDA00035787905600000214
indicating the amount of reflected light, the delta function indicates if and only if t-tLOSIs, delta (t-t)LOS)＝1。
4. The method of claim 3, wherein describing the optimization problem of wireless network resource allocation with the goal of maximizing energy efficiency comprises:
first, a weight factor is proposed
Figure FDA00035787905600000215
As an indicator of terahertz subchannel allocation
Figure FDA00035787905600000216
Figure FDA00035787905600000217
Indicating that the nth sub-channel is occupied by i transceiver pairs, and vice versa
Figure FDA00035787905600000218
The total data rate of the system is thus restated as:
Figure FDA0003578790560000031
the terahertz system energy efficiency EE is defined as total data rate and total power consumption ptoTherefore, the energy efficiency of the large-scale wireless network in the terahertz frequency band is defined as:
Figure FDA0003578790560000032
wherein p iscRepresents the circuit power consumption;
thus, the optimization problem is expressed as:
Figure FDA0003578790560000033
Figure FDA0003578790560000034
Figure FDA0003578790560000035
Figure FDA0003578790560000036
Figure FDA0003578790560000037
Figure FDA0003578790560000038
wherein C1 and C2 are power constraints of the wireless network under the terahertz frequency band, and C3 is the minimum data rate CminConstraints, namely C4 and C5, are used for allocating constraints for the terahertz sub-channels, and D represents the maximum number of transceivers of each sub-channel; sminIndicating assignment to the ith receiverLower bound of number of subchannels of a transmitter pair, SmaxAn upper bound indicating the number of sub-channels allocated to the ith transceiver pair,
Figure FDA0003578790560000039
denotes the ithnThe data rate of each receiver.
5. The method as claimed in claim 4, wherein a node of the radio channel map represents a transceiver pair, the node characteristics include direct channel status and weight, the link of the map is an interference channel, and the link characteristics are interference channel status.
6. The method of claim 5, wherein modeling the wireless network as a wireless channel map comprises:
modeling the multiple transceivers for the interfering channel as a complete graph G of labels with vertices and edges (V, E, s, t); where V represents a set containing vertices, E represents a set containing edges,
Figure FDA00035787905600000310
representing the mapping of a node to a feature of the node,
Figure FDA0003578790560000041
representing a feature that maps an edge to an edge, node viE.v denotes a pair of receiver and transmitter, V ═ V1,v2,...,v|V|}, vertex
Figure FDA0003578790560000042
The ith vertex, representing the nth subchannel, where N ∈ {1,2,. cndot, N }, i ∈ {1,2,. cndot, U ∈ {1,2,. cndotnDenoted by the feature matrix of the node
Figure FDA0003578790560000043
WhereinZ(i,:)＝s(vi) (ii) a Edge
Figure FDA0003578790560000044
Define the slave vertex
Figure FDA0003578790560000045
To the top
Figure FDA0003578790560000046
Of the directed relationship of (A), connecting the vertices
Figure FDA0003578790560000047
Is represented as
Figure FDA0003578790560000048
Vertex point
Figure FDA0003578790560000049
Is expressed as
Figure FDA00035787905600000410
If it is used
Figure FDA00035787905600000411
Exist, then
Figure FDA00035787905600000412
If not, [ A ]nm]i,j＝0；
In the case of radio interference, each pair of transceivers is taken as a vertex, the interference pattern from the transmitter to the receiver is taken as an edge, and the channel property of each vertex
Figure FDA00035787905600000413
Including weights
Figure FDA00035787905600000414
Variance of noise
Figure FDA00035787905600000415
Direct channel response
Figure FDA00035787905600000416
Each edge is characterized by a channel response from the interfering transmitter to the interfered receiver.
7. The method as claimed in claim 6, wherein the step of finding a strategy for mapping the radio channel map to the optimal power allocation vector and the optimal sub-channel allocation vector through iterative update of the neural network of the map, and implementing joint allocation optimization of power and sub-channels comprises:
defining different edge update functions phi for nodes on different sub-channelseAnd a vertex update function phivThese functions are parameterized by a multilayer perceptron; assuming that the number of convolution layers of the graph neural network is L, iteration is carried out from 1 to L in a circulating mode, in each layer, the nodes send node information to adjacent edges, the node characteristics and the edge characteristics are updated through the following formula, the node characteristics of all the nodes are output until the L-th layer, and meanwhile, an activation function is added when the L-th layer of the graph neural network outputs
Figure FDA00035787905600000417
Wherein p represents a power allocation vector, w represents a sub-channel allocation vector, | · | represents a vector modular length;
updating edge characteristics:
Figure FDA00035787905600000418
and (3) updating node characteristics:
Figure FDA00035787905600000419
aggregation node characteristics:
Figure FDA00035787905600000420
the loss function L (theta) is a negative expectation of a utility function realized on different channels, is optimized by a random gradient descent method, is propagated reversely in the loss function, and updates the neural network model parameter theta in an unsupervised mode:
Figure FDA00035787905600000421
wherein E isHRepresenting a desired function;
and obtaining a strategy pi (-) through the characteristic information of each vertex, outputting an optimal power distribution vector and an optimal sub-channel distribution vector, and realizing the joint distribution optimization of the power and the sub-channels.
8. An apparatus for optimizing wireless resource allocation based on a graph neural network, comprising:
the wireless network modeling module is used for modeling large-scale wireless network deployment under the terahertz frequency band, and considering the power and sub-channel distribution problems of a plurality of receiver and transmitter pairs under the large-scale wireless network under the terahertz frequency band; correcting interference in the signal transmission process, and establishing a physical channel model of the large-scale wireless network downlink system under the terahertz frequency band;
the wireless channel map construction module is used for describing the wireless network resource allocation optimization problem by taking the maximum energy efficiency as a target; modeling a wireless network into a wireless channel graph, and expressing a wireless network resource allocation optimization problem as a graph optimization problem;
and the optimization module is used for finding a strategy for mapping the wireless channel map to the optimal power distribution vector and the optimal sub-channel distribution vector through iterative update of the map neural network, so as to realize the joint distribution optimization of the power and the sub-channel.
CN202210339926.8A 2022-04-01 2022-04-01 Wireless resource allocation optimization method and device based on graph neural network Pending CN114786258A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210339926.8A CN114786258A (en) 2022-04-01 2022-04-01 Wireless resource allocation optimization method and device based on graph neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210339926.8A CN114786258A (en) 2022-04-01 2022-04-01 Wireless resource allocation optimization method and device based on graph neural network

Publications (1)

Publication Number Publication Date
CN114786258A true CN114786258A (en) 2022-07-22

Family

ID=82426993

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210339926.8A Pending CN114786258A (en) 2022-04-01 2022-04-01 Wireless resource allocation optimization method and device based on graph neural network

Country Status (1)

Country Link
CN (1) CN114786258A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117560043A (en) * 2024-01-11 2024-02-13 大连海事大学 Non-cellular network power control method based on graph neural network
WO2024065476A1 (en) * 2022-09-29 2024-04-04 华为技术有限公司 Wireless policy optimization method and apparatus
WO2024096775A1 (en) * 2022-11-02 2024-05-10 Telefonaktiebolaget Lm Ericsson (Publ) Methods and apparatuses for training and using a graph neural network

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2024065476A1 (en) * 2022-09-29 2024-04-04 华为技术有限公司 Wireless policy optimization method and apparatus
WO2024096775A1 (en) * 2022-11-02 2024-05-10 Telefonaktiebolaget Lm Ericsson (Publ) Methods and apparatuses for training and using a graph neural network
CN117560043A (en) * 2024-01-11 2024-02-13 大连海事大学 Non-cellular network power control method based on graph neural network
CN117560043B (en) * 2024-01-11 2024-03-19 大连海事大学 Non-cellular network power control method based on graph neural network

Similar Documents

Publication Publication Date Title
CN114786258A (en) Wireless resource allocation optimization method and device based on graph neural network
CN113162682B (en) PD-NOMA-based multi-beam LEO satellite system resource allocation method
CN114389678B (en) Multi-beam satellite resource allocation method based on decision performance evaluation
CN111800828B (en) Mobile edge computing resource allocation method for ultra-dense network
CN109756874B (en) Ultra-dense millimeter wave D2D communication interference management method
CN115441939B (en) MADDPG algorithm-based multi-beam satellite communication system resource allocation method
CN113258980B (en) Information transmission rate optimization method and device for wireless communication system
CN113377533A (en) Dynamic computation unloading and server deployment method in unmanned aerial vehicle assisted mobile edge computation
CN113490219B (en) Dynamic resource allocation method for ultra-dense networking
CN115278707A (en) NOMA terahertz network energy efficiency optimization method based on assistance of intelligent reflecting surface
CN113596785A (en) D2D-NOMA communication system resource allocation method based on deep Q network
CN116436512A (en) Multi-objective optimization method, system and equipment for RIS auxiliary communication
CN106231665A (en) Resource allocation methods based on the switching of RRH dynamic mode in number energy integrated network
CN112954806B (en) Chord graph coloring-based joint interference alignment and resource allocation method in heterogeneous network
CN117674958A (en) Network resource optimization method and device for air-space-earth integrated network
CN117412391A (en) Enhanced dual-depth Q network-based Internet of vehicles wireless resource allocation method
CN116390056B (en) STAR-RIS-assisted vehicle networking SR system link optimization method
CN116528250A (en) Unmanned aerial vehicle auxiliary MEC resource optimization method based on NOMA
CN116033461A (en) Symbiotic radio transmission method based on STAR-RIS assistance
CN115765826A (en) Unmanned aerial vehicle network topology reconstruction method for on-demand service
CN115811788A (en) D2D network distributed resource allocation method combining deep reinforcement learning and unsupervised learning
CN115065384A (en) Multi-beam satellite communication system resource allocation method considering user association, sub-channel allocation and beam association
CN114125744A (en) Data acquisition method based on block chain rights and interests certification and terminal system
Zhong et al. STAR-RISs assisted NOMA networks: A tile-based passive beamforming approach
YOU et al. RIS-Assisted UAV Assisted UAV-D2D Communications D Communications Exploiting Deep Reinforcement Learning

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