CN114828266B - Intelligent resource allocation method for optical and wireless fusion access - Google Patents
Intelligent resource allocation method for optical and wireless fusion access Download PDFInfo
- Publication number
- CN114828266B CN114828266B CN202210591121.2A CN202210591121A CN114828266B CN 114828266 B CN114828266 B CN 114828266B CN 202210591121 A CN202210591121 A CN 202210591121A CN 114828266 B CN114828266 B CN 114828266B
- Authority
- CN
- China
- Prior art keywords
- link
- line
- sight
- power angle
- convolutional neural
- 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
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
- H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
- H04B7/06—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station
- H04B7/0613—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station using simultaneous transmission
- H04B7/0615—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station using simultaneous transmission of weighted versions of same signal
- H04B7/0617—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station using simultaneous transmission of weighted versions of same signal for beam forming
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
- H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
- H04B7/08—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the receiving station
- H04B7/0891—Space-time diversity
- H04B7/0897—Space-time diversity using beamforming per multi-path, e.g. to cope with different directions of arrival [DOA] at different multi-paths
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/70—Services for machine-to-machine communication [M2M] or machine type communication [MTC]
-
- 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
Abstract
The invention discloses an intelligent resource allocation method for optical and wireless integrated access, and belongs to the technical field of communication resource allocation. After a direct link is established between equipment, a power angle spectrum formed by power received by a receiver and an arrival angle is input into a convolutional neural network, the convolutional neural network performs characteristic learning on power angle spectrum images of line-of-sight and non-line-of-sight links through a convolutional layer to realize the function of accurately classifying the links, and then, carriers of different frequency bands are used for signal transmission according to the transmission characteristics of the line-of-sight/non-line-of-sight links. The invention can improve the classification accuracy of the device-to-device link, improve the transmission signal quality of communication between devices and improve the cell signal coverage rate. The invention can be applied to 5G optical wireless fusion communication, can effectively improve the transmission quality of D2D communication and reduce the load pressure of a base station.
Description
Technical Field
The invention belongs to the technical field of communication resource allocation, and particularly relates to an intelligent resource allocation method for optical and wireless integrated access.
Background
The Radio over Fiber (RoF) refers to long-distance signal transmission between a central station and a base station through an optical Fiber link, and the base station receives millimeter wave signals or optical signals and then directly transmits the millimeter wave signals to users through simple photoelectric conversion. With the mass connection and the rapid increase of the flow of the base station, great pressure is applied to the existing RoF system in the aspects of data transmission rate, spectrum efficiency, network capacity and the like, and the load pressure of the server and the base station is increased sharply, so that the base station cannot respond to each user request in time.
One of the effective solutions is to directly perform Device-to-Device (D2D) transmission of communication data in some cells, so as to reduce the burden of data traffic of cell base stations. D2D technology as one of the 5G key technologies, D2D user equipment can realize direct connection by multiplexing channels of a certain cellular user equipment. The frequency resources of the cell can be reused through the D2D technology, and the cell throughput is improved.
At present, most of D2D communication schemes are combined with millimeter wave technology for data transmission. Millimeter wave technology solves the problem of spectrum scarcity, and has faster transmission rate and larger system capacity, thus being widely applied. When a non-blocking line-of-sight link exists between the D2D devices, the millimeter wave beam forming technology is adopted, and the characteristics of narrow millimeter wave beam, good directivity and high antenna gain can be utilized, so that the quality of signals received by a D2D receiving end is guaranteed. However, when a non-line-of-sight link occurs in a link between D2D devices, errors occur in communication positioning between the devices due to more serious path loss and penetration loss experienced in millimeter wave communication, and signals cannot be transmitted normally. How to guarantee the communication quality of the D2D device when the non-line-of-sight link occurs is a problem worthy of research.
The main difference between the line-of-sight and non-line-of-sight links is reflected in the propagation delay, the received signal power, and the distribution of the angle of arrival. The document "a Hybrid Communication Model of Millimeter Wave and Microwave in D2D Network,"2016 IEEE 83rd Vehicular Technology reference (VTC Spring), 2016, pp.1-5 "sets a threshold value τ of a received signal in advance, if the signal received by a receiver is greater than τ, the link is a line-of-sight link, otherwise, the link is a non-line-of-sight link, the value of the threshold value τ needs to be manually set and continuously adjusted to represent the discrimination of the link, and the discrimination precision is not ideal because the condition is too simple. A novel distributed spectrum, 2018 Wireless Telecommunications Symposium (WTS), 2018, pp.1-6, obtains an arrival angle spectrum by sending pilot signals for multiple times, and compares whether the peak value of the arrival angle spectrum is fixed. Therefore, it is necessary to further explore how to reduce the overhead as much as possible while ensuring the discrimination accuracy.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, provides a way for distinguishing line-of-sight links from non-line-of-sight links with higher precision and lower expense, and configures a matched transmission way based on the two classification results so as to improve the transmission quality of D2D communication and reduce the load pressure of a base station.
In order to achieve the above object, the present invention first provides a resource allocation scheme based on convolutional neural network, which includes the following steps:
step 1, constructing a convolutional neural network for link classification;
the input data of the convolutional neural network is a power angle spectrogram, and the convolutional neural network sequentially comprises: a first sub-network 1, a first sub-network 2, a second sub-network 1, a second sub-network 2, a first sub-network 3, a fully connected layer 1, a fully connected layer 2 and a classifier; wherein the first sub-network comprises convolution, an activation function and pooling in sequence; the second sub-network in turn comprises convolution and activation functions; the classifier is used for predicting the class probability of the line-of-sight link and the non-line-of-sight link;
collecting a certain amount of channel data to construct a training data set, and dividing a channel into a line-of-sight link and a non-line-of-sight link for each sample in the training data set according to the actual condition in the collection process to obtain a category label of each sample;
and constructing a power angle spectrum PAS (t) of each sample:
where t represents the sampling time, the direction vectorTheta and->Respectively representing a horizontal angle and a tilt angle in the arrival angle to the time t, and R represents a spatial covariance matrix;
spatial covariance matrixWherein H meas Representing the measured transport channel frequency response, E { } representing the mathematical expectation;
normalizing the power angle spectrum of the sample to be in [0,1], and converting the power angle spectrum of each sample into a square matrix form in an interpolation mode to obtain a power angle spectrum of each sample;
inputting the power angle spectrogram of the sample into a convolutional neural network for network training, and obtaining the trained convolutional neural network when a preset training end condition (for example, the training frequency reaches the maximum training number or the classification precision meets a set condition) is met;
step 2, collecting channel state information of the D2D link to be distributed:
acquiring a beam combination of receiver equipment of a D2D link to be allocated, and corresponding signal strength and arrival angle, and feeding back the beam combination to a base station to which the receiver equipment belongs, so that the base station can classify link types (distinguish link attributes) based on a trained convolutional neural network;
step 4, the base station configures the frequency band of the D2D link to be allocated:
if the link attribute is a line-of-sight link, transmitting and receiving transmission are carried out by adopting a millimeter wave band;
and if the link attribute is a non-line-of-sight link, transmitting and receiving transmission are carried out by adopting a microwave frequency band.
Further, in step 1, the power angle spectrum of each sample is converted into a square matrix form by an interpolation mode, and the obtained power angle spectrum of each sample specifically comprises:
defining the size of a power angle spectrum of each sample as w multiplied by h, and defining the size of the converted power angle spectrum as M multiplied by M;
for any point (x, y) in the power angle of the sample, the point obtained by bicubic interpolation is:
wherein X, Y represent the coordinates of the power angle spectrogram, B (X, Y) represents the value of the point (X, Y),indicating a location in the power angle spectrum at the coordinate pick>The value of (by interpolation).
Further, in step 3, if the link attribute is a line-of-sight link, a sector antenna gain model is used for transmitting and receiving, that is, directional beam forming is used between the millimeter wave band D2D devices during transmitting and receiving; if the link attribute is a non-line-of-sight link, an omnidirectional antenna is adopted for transmitting and receiving.
The technical scheme provided by the invention at least has the following beneficial effects:
compared with the prior scheme of distinguishing the links based on only one characteristic, the method combines two characteristics of power and arrival angle to construct a power angle spectrum, finishes the distinguishing of the links by means of the image characteristic extraction and classification function based on the convolutional neural network, and has the distinguishing precision of 98.13 percent. If the link is a line-of-sight link, a millimeter wave frequency band is used for subsequent signal transmission, otherwise, a microwave frequency band is used, so that the transmission quality of D2D communication is effectively improved, and the load pressure of a base station is reduced.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a schematic view of an application scenario of an intelligent resource allocation method for optical and wireless converged access according to an embodiment of the present invention;
fig. 2 is a schematic processing procedure diagram of a resource intelligent allocation method for optical and wireless converged access according to an embodiment of the present invention;
fig. 3 is a power angle spectrum diagram corresponding to a line-of-sight link and a non-line-of-sight link in the embodiment of the present invention, where 3-a is a power angle spectrum corresponding to a certain non-line-of-sight link, and 3-b is a scene corresponding to the line-of-sight link, and it can be seen from the diagram that the propagation angles of the line-of-sight link are more concentrated and the power loss is smaller;
FIG. 4 is a schematic diagram of a network structure of a convolutional neural network for link classification identification in an embodiment of the present invention;
fig. 5 is a graph comparing SINR coverage analysis using millimeter wave and microwave band co-transmission and single band transmission in an example of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
The embodiment of the invention provides an intelligent resource allocation method for optical and wireless integrated access, aiming at the technical problems that in the prior art, when equipment-to-equipment direct communication is carried out, a direct link is blocked, millimeter wave signals are easily absorbed by the blocked objects, and the transmission performance is not ideal. After a direct link is established between equipment and the equipment, a power angle spectrum formed by power received by a receiver and an arrival angle is input into a constructed convolutional neural network, the convolutional neural network performs characteristic learning on power angle spectrum images of line-of-sight and non-line-of-sight links through a convolutional layer to realize the function of accurately classifying the links, and then, according to the transmission characteristics of the line-of-sight/non-line-of-sight links, carriers of different frequency bands (millimeter waves/microwaves) are used for signal transmission. The embodiment of the invention can improve the classification accuracy of the device-to-device link, improve the quality of transmission signals of communication between devices and improve the coverage rate of cell signals.
As shown in fig. 1, the method for intelligently allocating resources for optical and wireless convergence access according to the embodiment of the present invention may be used in an optical and wireless convergence network scenario including a core server, a cell base station, and users, where the cell base station obtains network content from the core server through a (5G) forwarding network, and then provides services for all users in a whole cell range in a wireless communication manner. In FIG. 1, CU is called Centralized Unit for all purposes and DU is called Distributed Unit for all purposes.
Referring to fig. 2, the specific implementation steps of the intelligent resource allocation method for optical and wireless converged access provided by the embodiment of the present invention include:
and S1, collecting channel data.
Step S101, a certain amount of channel data is collected to construct a data set, and the channel is divided into a line-of-sight link and a non-line-of-sight link according to the actual situation in the collection process, so that a link type label of the data is obtained.
Constructing a power angle spectrum, which is specifically shown as the following formula:
wherein the content of the first and second substances,theta and->The horizontal angle and the tilt angle among the arrival angles are respectively indicated. R represents a spatial covariance matrix.
The spatial covariance matrix R is:
wherein H meas For the measured transport channel frequency response, E { } is to get the corresponding mathematical expectation.
Step S102, data preprocessing: because the unit of the power angle spectrum is dB, the value is negative, the data is normalized to [0,1] uniformly, for the extraction of the features in the subsequent convolution kernel, PAS (power angle spectrum) diagram a with the original size of 70 × 180 uses double cubic interpolation to convert the power angle spectrum sample into a square diagram B with the size of 227 × 227, and then the coordinates obtained by double cubic interpolation at a certain point (x, y) in the diagram a are:
step S103, data division: randomly dividing the collected data into a training set and a testing set according to a specified proportion (such as 7:3);
and S2, constructing a convolutional neural network.
Step S201, inputting the samples processed in the step S1 through an input layer, thereby completing the training of the constructed convolutional neural network;
step S202, extracting characteristics of a data sample through a five-layer convolutional layer network structure, wherein 1 layer, 2 layer and 5 layer are first sub-networks consisting of convolution, pooling and activation functions; 3. layer 4 is a layer comprising a second sub-network comprising convolution and activation functions;
the operations involving convolution, pooling, and activation functions are as follows:
taking the input layer as an example of a 227 × 227 matrix, x being the input samples and y being the outputA feature matrix having l convolutional layers, with a convolution kernel of k, representing a convolution operation, b l The corresponding offset for that convolution kernel.
And (3) convolution operation: after passing through a convolution layer, the corresponding output is a l :
a l =f(a l-1 *k l +b l )
Wherein, f (a) l-1 *k l +b l ) Representing a convolution function.
A pooling layer: in order to reduce the size of the feature matrix, maximum pooling is adopted in the pooling layer part, the maximum element value is selected to represent the feature of the window, a part of element values are discarded, a part of errors are necessarily brought, and the corresponding errors can be represented as
The Relu function is selected as the activation function, the calculation is simpler while the nonlinearity is introduced, and the convergence speed is faster in parameter iteration, wherein the Relu function can be expressed as follows:
f(x)=max(0,x)
step S203, merging the characteristics through two full connection layers, and inputting the merged characteristics into a SoftMax classifier to classify links (a line-of-sight link and a non-line-of-sight link); the output of the SoftMax classifier is the probability that a sample is assigned to a class.
Specifically, in the present embodiment, each layer structure is specifically as follows:
the first layer is an input layer, samples are subjected to bicubic interpolation, and the size is 227 multiplied by 227;
the second layer is a convolutional layer and sequentially comprises three operation operations of convolution, activation function and pooling, a characteristic matrix with the size of 55 x 55 is output through 96 convolution kernels with the size of 11 x 11, and then the characteristic matrix is subjected to activation function Relu and 3 x 3 pooling layer to obtain a characteristic matrix with the size of 27 x 27;
the third layer is a convolutional layer and sequentially comprises three operation operations of convolution, activation function and pooling, the characteristic matrix output by the second layer passes through 256 convolution kernels with the size of 5 x 5 to obtain 256 characteristic matrices with the size of 27 x 27, and similarly, the characteristic matrices with the size of 13 x 13 are obtained after the activation function and pooling;
the fourth layer is a convolution layer and sequentially comprises convolution and two operation operations of an activation function, and the feature matrix of the previous layer outputs a feature matrix with the size of 13 multiplied by 13 after passing through 384 convolution kernels with the size of 3 multiplied by 3;
the fifth layer is a convolution layer and sequentially comprises convolution and activation function operation, and the feature matrix of the previous layer outputs a feature matrix with the size of 13 multiplied by 13 after passing through 384 convolution kernels with the size of 3 multiplied by 3;
the sixth layer is a convolution layer and comprises three parts of convolution, activation function and pooling, and the feature matrix of the previous layer outputs a feature matrix with the size of 6 multiplied by 6 after passing through 256 convolution kernels with the size of 3 multiplied by 3;
the seventh and eighth layers are all-connected layers, each layer has 4096 neurons, and the features obtained from the second to sixth layers are connected;
the ninth layer is a SoftMax layer, which contains two neurons, and sends the output value to a SoftMax classifier.
And S3, testing the constructed convolutional neural network.
Inputting the channel data in the test set into the convolutional neural network trained in the step S2, and carrying out scene recognition;
obtaining a scene corresponding to the channel according to the maximum element corresponding index in the output category vector; and comparing with the labels of the test set to obtain the prediction accuracy of the convolutional neural network.
And S4, pairing the devices (realizing microwave bands).
Step S401, a terminal in a cell sends a communication request to a base station, and a gateway discovers that both source and destination devices belong to the same base station by processing a data packet, and has a D2D communication condition.
Step S402, the base station broadcasts the D2D communication link setting information to all users, and waits for the response of the receiver device.
Step S403, after the receiver device receives the broadcast information, agrees to communicate with the transmitter, and feeds back the information to the base station, and the base station sends a control signaling to enable the D2D device pair to perform channel detection.
Step S5, D2D link detection (millimeter wave band).
Step S501, it is assumed that all D2D devices are equipped with a millimeter wave antenna array and can switch between millimeter wave and microwave bands. The number of transmitting antennas of the D2D transmitter is N TX The number of receiving antennas of the receiver is N RX The D2D transmitter sends test signals to the receiver where each antenna receives N RX If the signal is received, the receiver has N TX ×N RX And (3) keeping the beam with the maximum signal strength for each antenna to form a beam combination A, and recording the corresponding beam strength and the arrival angle.
Step S502, the beam combination A of the receiver is fed back to the base station for the link attribute discrimination.
Step S6, link differentiation (at the base station) is performed using a convolutional neural network.
Step S601, the base station processes the signal intensity and the arrival angle information fed back by the receiver to construct a power angle spectrum;
step S602, importing the power angle spectrum into a convolutional neural network trained in advance, and inputting the power angle spectrum through an input layer of the convolutional neural network; and then extracting the characteristics through a five-layer convolutional layer network structure, and establishing a mapping relation between the power angle spectrum and the link characteristics. For the convolution layer part, 1, 2 and 5 layers are networks consisting of convolution, pooling and activation functions; 3. the 4 layers are networks comprising convolution and activation functions; combining the characteristics through two full connection layers, and inputting the characteristics into a SoftMax classifier to classify links; and obtaining the category corresponding to the link according to the index corresponding to the maximum element in the category vector output by SoftMax.
S7, distributing corresponding frequency bands according to scenes:
step S701, if the channel corresponds to a line-of-sight link, using a millimeter wave frequency band to perform corresponding transceiving transmission; directional beam forming is adopted among the millimeter wave band D2D devices, a sector antenna gain model is adopted, main beams among the D2D devices are aligned with each other, the gain of a beam antenna received by a receiver is maximum, interference is greatly reduced, and cross beams transmitted by other users can be received.
Step S702, if the channel is corresponding to non-line-of-sight, then corresponding transceiving transmission is carried out by adopting a microwave frequency band; the omni-directional antenna is adopted for transmission, and interference from a base station and other D2D equipment is mainly received.
The effects of the embodiments of the present invention are further illustrated by the following simulation and experimental verification:
under the conditions that the base station transmitting power is 37dBm, the d2D transmitter power is 0dBm, the main lobe gain is 10dBi, the width is 30 degrees and the side lobe is-10 dBi, the simulation of the embodiment is carried out, and the Signal quality of the transmission by using the mixed band transmission scheme is better than that by using only single-band transmission as shown in figure 5, and under the condition that the threshold value of SINR (Signal to Interference plus Noise Ratio) is 0, the performance of the transmission by using the mixed band is better than that by using only microwave transmission by 15 percent.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
What has been described above are merely some of the embodiments of the present invention. It will be apparent to those skilled in the art that various changes and modifications can be made therein without departing from the spirit and scope of the invention.
Claims (4)
1. An intelligent resource allocation method for optical and wireless converged access is characterized by comprising the following steps:
step 1, constructing a convolutional neural network for link classification;
the input data of the convolutional neural network is a power angle spectrogram, and the convolutional neural network sequentially comprises: a first sub-network 1, a first sub-network 2, a second sub-network 1, a second sub-network 2, a first sub-network 3, a fully connected layer 1, a fully connected layer 2 and a classifier; the network structures of the first sub-networks 1, 2 and 3 are the same, and the network structures sequentially comprise convolution, an activation function and pooling; the network structures of the second sub-networks 1 and 2 are the same and sequentially comprise convolution and activation functions; the classifier is used for predicting the class probability of the line-of-sight link and the non-line-of-sight link;
collecting a certain amount of channel data to construct a training data set, and dividing a channel into a line-of-sight link and a non-line-of-sight link for each sample in the training data set according to the actual condition in the collection process to obtain a category label of each sample;
and constructing a power angle spectrum PAS (t) for each sample:
where t represents the sampling time, the direction vectorTheta and->Respectively representing a horizontal angle and a tilt angle in arrival angles to a time t, and R represents a spatial covariance matrix;
spatial covariance matrixWherein H meas Representing the measured transport channel frequency response, E { } representing the mathematical expectation;
normalizing the power angle spectrum of the sample to be in [0,1], and converting the power angle spectrum of each sample into a square matrix form in an interpolation mode to obtain a power angle spectrum of each sample;
inputting the power angle spectrogram of the sample into a convolutional neural network for network training, and obtaining a trained convolutional neural network when a preset training end condition is met;
step 2, collecting channel state information of the D2D link to be distributed:
acquiring a beam combination of receiver equipment of a D2D link to be distributed, and corresponding signal strength and arrival angle, and feeding back the beam combination to a base station to which the receiver equipment belongs, so that the base station can classify the link types based on a trained convolutional neural network;
step 3, the base station constructs a power angle spectrogram based on the signal intensity and the arrival angle fed back by the receiver equipment, inputs the power angle spectrogram into the convolutional neural network trained in the step 2, and obtains link attributes of the D2D link to be distributed based on the output of the convolutional neural network: line-of-sight or non-line-of-sight links;
step 4, the base station configures the frequency band of the D2D link to be allocated:
if the link attribute is a line-of-sight link, transmitting and receiving transmission are carried out by adopting a millimeter wave band;
and if the link attribute is a non-line-of-sight link, transmitting and receiving transmission are carried out by adopting a microwave frequency band.
2. The method according to claim 1, wherein in step 1, the power angle spectrum of each sample is converted into a square matrix form by interpolation, and the obtaining of the power angle spectrum of each sample specifically comprises:
defining the size of a power angle spectrum of each sample as w multiplied by h, and defining the size of the converted power angle spectrum as M multiplied by M;
for any point (x, y) in the power angle of the sample, the point obtained by bicubic interpolation is:
3. The method of claim 1, wherein in step 2, the activation function of the convolutional neural network is a Relu activation function, and the classifier is a SoftMax classifier.
4. The method according to any one of claims 1 to 3, wherein in step 3, if the link attribute is line-of-sight link, a sector antenna gain model is adopted for transceiving transmission; and if the link attribute is a non-line-of-sight link, transmitting and receiving transmission are carried out by adopting the omnidirectional antenna.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210591121.2A CN114828266B (en) | 2022-05-27 | 2022-05-27 | Intelligent resource allocation method for optical and wireless fusion access |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210591121.2A CN114828266B (en) | 2022-05-27 | 2022-05-27 | Intelligent resource allocation method for optical and wireless fusion access |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114828266A CN114828266A (en) | 2022-07-29 |
CN114828266B true CN114828266B (en) | 2023-04-07 |
Family
ID=82520182
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210591121.2A Active CN114828266B (en) | 2022-05-27 | 2022-05-27 | Intelligent resource allocation method for optical and wireless fusion access |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114828266B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109922427A (en) * | 2019-03-06 | 2019-06-21 | 东南大学 | Utilize the intelligent radio positioning system and method for large scale array antenna |
CN111212379A (en) * | 2020-01-06 | 2020-05-29 | 天津工业大学 | Novel CSI indoor positioning method based on convolutional neural network |
CN111372195A (en) * | 2020-01-23 | 2020-07-03 | 鹏城实验室 | Method, apparatus and storage medium for tracking position of mobile terminal in mobile communication network |
WO2021101333A1 (en) * | 2019-11-22 | 2021-05-27 | Samsung Electronics Co., Ltd. | Method and system for intelligent 3d aerial wireless transmission |
CN113469110A (en) * | 2021-07-15 | 2021-10-01 | 北京邮电大学 | UWB non-line-of-sight signal identification method based on deep learning |
CN114189809A (en) * | 2021-11-15 | 2022-03-15 | 华东师范大学 | Indoor positioning method based on convolutional neural network and high-dimensional 5G observation characteristics |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9923657B2 (en) * | 2013-03-12 | 2018-03-20 | Rearden, Llc | Systems and methods for exploiting inter-cell multiplexing gain in wireless cellular systems via distributed input distributed output technology |
US10129778B2 (en) * | 2016-08-30 | 2018-11-13 | Nxgen Partners Ip, Llc | SDN-based channel estimation for multiplexing between LOS mmWaves NLOS sub-6 GHz and FSO |
-
2022
- 2022-05-27 CN CN202210591121.2A patent/CN114828266B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109922427A (en) * | 2019-03-06 | 2019-06-21 | 东南大学 | Utilize the intelligent radio positioning system and method for large scale array antenna |
WO2021101333A1 (en) * | 2019-11-22 | 2021-05-27 | Samsung Electronics Co., Ltd. | Method and system for intelligent 3d aerial wireless transmission |
CN111212379A (en) * | 2020-01-06 | 2020-05-29 | 天津工业大学 | Novel CSI indoor positioning method based on convolutional neural network |
CN111372195A (en) * | 2020-01-23 | 2020-07-03 | 鹏城实验室 | Method, apparatus and storage medium for tracking position of mobile terminal in mobile communication network |
CN113469110A (en) * | 2021-07-15 | 2021-10-01 | 北京邮电大学 | UWB non-line-of-sight signal identification method based on deep learning |
CN114189809A (en) * | 2021-11-15 | 2022-03-15 | 华东师范大学 | Indoor positioning method based on convolutional neural network and high-dimensional 5G observation characteristics |
Non-Patent Citations (3)
Title |
---|
Asha等."An improvised scheme for cross-layer optimization to support QoS in MANET".《 2017 2nd International Conference for Convergence in Technology (I2CT)》.2017,全文. * |
李君瑶等."大规模3D MIMO中基于信道相关的LOS/NLOS识别算法".《北京邮电大学学报》.2019,全文. * |
胡青松等."目标定位中的非视距传播研究综述".《工矿自动化》.2020,全文. * |
Also Published As
Publication number | Publication date |
---|---|
CN114828266A (en) | 2022-07-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Zhang et al. | A survey on 5G millimeter wave communications for UAV-assisted wireless networks | |
CN105992236A (en) | Electronic device and method of performing wireless communication using beamforming thereof | |
CN103199906B (en) | The method for spacial multiplex that millimeter wave high-speed communication system beam sector is intercepted | |
CN110430542B (en) | Rapid beam tracking method for unmanned aerial vehicle station group networking | |
US9942778B2 (en) | Virtual base station apparatus and communication method | |
CN110972312B (en) | Inter-cell interference coordination method of millimeter wave cellular network | |
CN109526042B (en) | Network access point selection method of LWA system based on OWMAD | |
CN110650525B (en) | Multi-beam distributed power MAC protocol communication method | |
Salehi et al. | Machine learning on camera images for fast mmWave beamforming | |
CN114828266B (en) | Intelligent resource allocation method for optical and wireless fusion access | |
Kaddoura et al. | Greenfield design in 5G FWA networks | |
CN110430584B (en) | Uplink multi-beam initial access method for non-active users in millimeter wave network | |
Shi et al. | Coverage probability of UAV-enabled millimeter wave communications in finite areas | |
CN109039410B (en) | Beam forming method of heterogeneous cloud wireless access network and communication network | |
CN115052296B (en) | Intelligent Rank downlink rate optimization method for 6G | |
CN107835514B (en) | Modeling method for wireless network resource matching in multi-host service scene | |
CN107249213B (en) | A kind of maximized power distribution method of D2D communication Intermediate Frequency spectrum efficiency | |
CN113301532B (en) | Channel allocation method for unmanned aerial vehicle-assisted millimeter wave emergency communication network | |
CN112788726B (en) | Power control method under network scene model of unmanned aerial vehicle auxiliary base station communication | |
CN109714093B (en) | Combined antenna selection method for heterogeneous cellular network | |
Zhang et al. | InferBeam: A fast beam alignment protocol for millimeter-wave networking | |
CN112512075B (en) | Pilot frequency collision detection method, device and system | |
CN112564761B (en) | Method for constructing wireless communication network between high-speed aircrafts based on millimeter wave communication | |
CN108307545B (en) | Data sending device and method after BBU and RRU function reconstruction | |
Chehri | Utility-Based Beam Selection Algorithms for Sub-6 GHz Backhauls |
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 |