CN113645700A - Deep learning-based resource allocation method and device for improving SCMA system performance - Google Patents
Deep learning-based resource allocation method and device for improving SCMA system performance Download PDFInfo
- Publication number
- CN113645700A CN113645700A CN202110800659.5A CN202110800659A CN113645700A CN 113645700 A CN113645700 A CN 113645700A CN 202110800659 A CN202110800659 A CN 202110800659A CN 113645700 A CN113645700 A CN 113645700A
- Authority
- CN
- China
- Prior art keywords
- time
- resource
- group
- resource blocks
- frequency resources
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000013468 resource allocation Methods 0.000 title claims abstract description 45
- 238000000034 method Methods 0.000 title claims abstract description 27
- 238000013135 deep learning Methods 0.000 title claims abstract description 23
- 238000003062 neural network model Methods 0.000 claims abstract description 51
- 238000012549 training Methods 0.000 claims abstract description 31
- 239000011159 matrix material Substances 0.000 claims description 18
- 230000005540 biological transmission Effects 0.000 claims description 14
- 238000012795 verification Methods 0.000 claims description 8
- 238000004064 recycling Methods 0.000 claims description 3
- 230000008054 signal transmission Effects 0.000 abstract description 3
- 238000004891 communication Methods 0.000 description 5
- 230000002452 interceptive effect Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 3
- 238000010295 mobile communication Methods 0.000 description 3
- 238000001228 spectrum Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 238000012369 In process control Methods 0.000 description 1
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000010965 in-process control Methods 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 230000015654 memory Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000003208 petroleum Substances 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W72/00—Local resource management
- H04W72/04—Wireless resource allocation
- H04W72/044—Wireless resource allocation based on the type of the allocated resource
- H04W72/0446—Resources in time domain, e.g. slots or frames
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
- G06N3/04—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
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W72/00—Local resource management
- H04W72/50—Allocation or scheduling criteria for wireless resources
- H04W72/54—Allocation or scheduling criteria for wireless resources based on quality criteria
- H04W72/541—Allocation or scheduling criteria for wireless resources based on quality criteria using the level of interference
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W72/00—Local resource management
- H04W72/50—Allocation or scheduling criteria for wireless resources
- H04W72/56—Allocation or scheduling criteria for wireless resources based on priority criteria
Abstract
The invention discloses a resource allocation method and a resource allocation device for improving SCMA system performance based on deep learning, wherein the resource allocation method comprises the following steps: grouping time-frequency resources of an SCMA system to obtain a plurality of groups of resource blocks, wherein each group of resource blocks comprises a plurality of time-frequency resources; acquiring the first N interference signals of each group of resource blocks in real time, and taking the first N interference signals as the input of an LSTM neural network model to obtain the interference power value of each time-frequency resource of each group of resource blocks under the current interference signal; the LSTM neural network model is obtained by training according to the historical interference signals of each group of the resource blocks; n is a positive integer; and matching different time frequency resources for users of different levels according to the interference power value of each time frequency resource. The technical scheme of the invention realizes the high-efficiency configuration of time-frequency resources and the high-quality transmission of signal data in an SCMA system.
Description
Technical Field
The invention relates to the technical field of mobile communication, in particular to a resource allocation method and a resource allocation device for improving SCMA system performance based on deep learning.
Background
Fifth generation mobile communication systems (5G) support up to 105/km5The connection number of the system can be applied to the vertical fields of electric power, petroleum, chemical industry, automobiles and the like, and an internet of things system with all things interconnected is created. For example, in the power industry, a large number of different sensors can be applied to various links of power generation, transmission, distribution, use and the like of a power system to acquire key information such as voltage, current, power factors, harmonic waves and the like, so that advance prevention, in-process control and after-process treatment are realized. Sparse Cryptographic Multiple Access (SCMA) is considered a potential solution in fifth generation mobile communication systems (5G) due to its ability to achieve high spectrum utilization and large-scale connectivity. SCMA allows multiple users to signal using different codewords in overlapping time and frequency resources. In the receiver, inter-cell interference between users can be cancelled using a less complex Message Passing Algorithm (MPA), which can approach the performance of a maximum likelihood detector, but inter-cell interference is still a limiting factor for SCMA system performance.
In order to guarantee end-to-end performance of users, the 5G network introduces the concept of network slice, wherein in a 5G air interface, dedicated radio resources can be reserved in the cell for some users to realize high-reliability transmission. However, the user may still suffer from co-channel interference from neighboring cells, resulting in poor transmission quality and even communication failure. Although the same wireless resources can be reserved in the adjacent cells to reduce the co-channel interference, the reuse of the frequency spectrum resources is reduced, and the utilization rate of the frequency spectrum of the system is greatly reduced.
Disclosure of Invention
The invention provides a resource allocation method and a resource allocation device for improving SCMA system performance based on deep learning, which realize high-efficiency configuration of time-frequency resources and high-quality transmission of signal data in an SCMA system.
An embodiment of the present invention provides a resource allocation method for improving SCMA system performance based on deep learning, including:
grouping time-frequency resources of an SCMA system to obtain a plurality of groups of resource blocks, wherein each group of resource blocks comprises a plurality of time-frequency resources;
acquiring the first N interference signals of each group of resource blocks in real time, and taking the first N interference signals as the input of an LSTM neural network model to obtain the interference power value of each time-frequency resource of each group of resource blocks under the current interference signal; the LSTM neural network model is obtained by training according to the historical interference signals of each group of the resource blocks; n is a positive integer;
and matching different time frequency resources for users of different levels according to the interference power value of each time frequency resource.
Further, grouping time-frequency resources of the system to obtain a plurality of groups of resource blocks, wherein each group of resource blocks comprises a plurality of time-frequency resources, and specifically comprises:
extracting one time frequency resource from the ith time frequency resource at intervals of K time frequency resources, and extracting M time frequency resources as an ith group of resource blocks, wherein the M time frequency resources comprise the ith time frequency resource; wherein K is more than or equal to i, and i, K and M are positive integers.
Further, the first N interference signals of each group of resource blocks are obtained in real time, and the first N interference signals are used as input of the LSTM neural network model, so as to obtain an interference power value of each time-frequency resource of each group of resource blocks under the current interference signal, specifically:
acquiring the first N interference signals of each group of resource blocks in real time, and representing the first N interference signals as a PxN matrix, wherein each interference signal corresponds to a vector of Px1 in the matrix;
taking the PxN matrix of each group of resource blocks as the input of the LSTM neural network model to obtain an Mx1 output vector of each group of resource blocks;
and obtaining the interference power value of each time-frequency resource of each group of resource blocks under the current interference signal according to the output vector of one Mx1 of each group of resource blocks.
Further, the users of different grades are primary users and secondary users which are divided according to the importance degree of the service.
Further, the matching different time frequency resources for users of different grades according to the interference power value of each time frequency resource specifically includes:
according to the interference power value of each time-frequency resource of each group of resource blocks under the current interference signal, dividing the time-frequency resources of each group of resource blocks into primary time-frequency resources and secondary time-frequency resources, allocating the primary time-frequency resources to the primary users for use, and allocating the secondary time-frequency resources to the secondary users for use.
Further, the allocating the primary time-frequency resource to the primary user for use specifically includes:
the primary users are prioritized according to the comprehensive indexes of the primary users;
and matching the corresponding primary time frequency resource for the primary user according to the priority of the primary user and the interference power value of the primary time frequency resource.
Further, the allocating the secondary time-frequency resource to the secondary user for use specifically includes:
and selecting the secondary time-frequency resource for recycling to form a resource pool, and allocating the time-frequency resource in the resource pool to the secondary user for use.
Further, the secondary time frequency resource adopts stronger channel coding and lower order modulation coding compared with the primary time frequency resource when data transmission is carried out.
Further, the LSTM neural network model is obtained by training the historical interference signals of each group of resource blocks, and specifically includes:
dividing the historical interference signals of each group of the resource blocks into a training data set and a verification data set, training the LSTM neural network model to be convergent by using the training data set, and verifying the LSTM neural network model by using the verification data set; the data in the training data set are arranged according to time sequence.
On the basis of the above method item embodiments of the present invention, there are correspondingly provided apparatus item embodiments;
another embodiment of the present invention provides a resource allocation apparatus for improving SCMA system performance based on deep learning, which includes a resource grouping module, an interference power prediction module and a resource matching module,
the resource grouping module is used for grouping time-frequency resources of a system to obtain a plurality of groups of resource blocks, and each group of resource blocks comprises a plurality of time-frequency resources;
the interference power prediction module is used for taking the first N interference signals of each group of resource blocks as the input of an LSTM neural network model to obtain the interference power value of each time-frequency resource of each group of resource blocks under the current interference signal; the LSTM neural network model is obtained by training according to the historical interference signals of each group of the resource blocks; n is a positive integer;
and the resource matching module is used for matching different time frequency resources for users of different grades according to the interference power value of each time frequency resource.
The embodiment of the invention has the following beneficial effects:
the invention provides a resource allocation method and a resource allocation device for improving SCMA system performance based on deep learning, wherein the method comprises the steps of collecting interference signals for a period of time, training a long-term and short-term memory artificial neural network model to be convergent by utilizing the collected interference signals, inputting N interference signals before data transmission into the neural network model to obtain interference power of each time-frequency resource of a resource block corresponding to the current interference signals, and matching users of corresponding grades for each time-frequency resource according to the interference power; therefore, the technical scheme realizes accurate prediction of the interference power of different time-frequency resources, can match the time-frequency resources with low interference power for high-level users and match the time-frequency resources with relatively high interference power for low-level users, and realizes efficient configuration of the time-frequency resources in the SCMA system; meanwhile, the co-frequency interference of high-grade users during signal data transmission can be reduced to the minimum by predicting the interference power of different time-frequency resources in real time, and further the high-quality transmission of signal data in the SCMA system is realized.
Drawings
FIG. 1 is a flowchart of a deep learning-based resource allocation method for improving SCMA system performance according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a deep learning-based resource allocation apparatus for improving the performance of an SCMA system according to an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of an LSTM neural network model provided in an embodiment of the present invention;
FIG. 4 is a schematic diagram of data processing of an LSTM neural network model provided by an embodiment of the present invention;
fig. 5 is a graph of bit error rate versus Eb/no (dB) for the Random time-frequency resource allocation scheme and the MA time-frequency resource allocation scheme according to an embodiment of the present invention when SIR is 15 dB;
fig. 6 is a graph of bit error rate versus Eb/no (dB) for the "Random" and "MA" resource allocation schemes according to the present invention when SIR is 10dB according to an embodiment of the present invention;
fig. 7 is a graph of bit error rate versus Eb/no (dB) for the "Random" and "MA" resource allocation schemes according to the present invention when SIR is 6dB according to an embodiment of the present invention;
Detailed Description
The technical solutions in the present invention will be described clearly and completely with reference to the accompanying drawings, and it is obvious that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, a resource allocation method for improving SCMA system performance based on deep learning according to an embodiment of the present invention includes:
step S11: and grouping the time-frequency resources of the SCMA system to obtain a plurality of groups of resource blocks, wherein each group of resource blocks comprises a plurality of time-frequency resources.
Preferably, one of the embodiments of step S11 is:
extracting one time frequency resource from the ith time frequency resource at intervals of K time frequency resources, and extracting M time frequency resources as an ith group of resource blocks, wherein the M time frequency resources comprise the ith time frequency resource; wherein K is more than or equal to i, and i, K and M are positive integers.
Step S12: acquiring the first N interference signals of each group of resource blocks in real time, and taking the first N interference signals as the input of an LSTM neural network model to obtain the interference power value of each time-frequency resource of each group of resource blocks under the current interference signal; the LSTM neural network model is obtained by training according to the historical interference signals of each group of the resource blocks; n is a positive integer.
Preferably, one of the embodiments of step S12 is:
as an example, step S12 may include the following sub-steps:
substep S121: training the LSTM neural network model to be convergent according to the historical interference signals of each group of the resource blocks;
preferably, one of the embodiments of step S121 is:
dividing the historical interference signals of each group of the resource blocks into a training data set and a verification data set, training the LSTM neural network model to be convergent by using the training data set, and verifying the LSTM neural network model by using the verification data set; the data in the training data set are arranged according to time sequence; the LSTM neural network model includes 20 hidden units.
Substep S122: acquiring the first N interference signals of each group of resource blocks in real time, and representing the first N interference signals as a PxN matrix, wherein each interference signal corresponds to a vector of Px1 in the matrix;
substep S123: taking the PxN matrix of each group of resource blocks as the input of the LSTM neural network model to obtain an Mx1 output vector of each group of resource blocks;
substep S124: and obtaining the interference power value of each time-frequency resource of each group of resource blocks under the current interference signal according to the output vector of one Mx1 of each group of resource blocks.
Step S13: and matching different time frequency resources for users of different levels according to the interference power value of each time frequency resource.
As an example, step S13 may include the following sub-steps:
substep S131: dividing the time frequency resources of each group of resource blocks into a primary time frequency resource and a secondary time frequency resource according to the interference power value of each time frequency resource of each group of resource blocks under the current interference signal;
substep S132: dividing users into a primary user and a secondary user according to the importance degree of the service, and prioritizing the primary user according to the comprehensive index of the primary user;
substep S133: matching the corresponding primary time frequency resource for the primary user according to the priority of the primary user and the interference power value of the primary time frequency resource;
substep S134: selecting the secondary time-frequency resource for recycling to form a resource pool, and allocating the time-frequency resource in the resource pool to the secondary user for use; and the secondary time frequency resource adopts stronger channel coding and lower-order modulation coding compared with the primary time frequency resource when data transmission is carried out.
More detailed examples are as follows:
in an SCMA system, 600 time-frequency resources and 900 users communicate with a base station in an SCMA multiple access manner.
Step A: grading the users;
dividing 900 users into primary users and secondary users according to the importance degree of the service, wherein 600 users are primary users, and 300 users are secondary users;
and B: grouping the time-frequency resources;
dividing 600 time frequency resources into 100 groups of resource blocks, wherein each group of resource blocks comprises 6 time frequency resources; wherein, the 1 st group of resource blocks selects one time frequency resource from the first time frequency resource at intervals of 100 time frequency resources, and selects 6 time frequency resources in total, for example, the 1 st group of resource blocks are 1, 101, 201, 301, 401, 501; the 2 nd group of resource blocks selects one time frequency resource from the second time frequency resource at intervals of 100 time frequency resources, and selects 6 time frequency resources in total, for example, the 2 nd group of resource blocks are 2, 102, 202, 302, 402 and 502; in this way, the 100 th group of resource blocks selects one time frequency resource from the 100 th time frequency resource every 100 time frequency resources, and selects 6 time frequency resources, such as 100, 200, 300, 400, 500, and 600; each group of resource blocks carries 6 users communicating with the base station.
And C: and collecting historical interference signals of each group of resource blocks to train the LSTM neural network model.
Wherein, as an example, step C comprises the following sub-steps:
substep C1: constructing an LSTM neural network model; as shown in fig. 3 and 4, the LSTM neural network model includes an input layer, an LSTM layer, a fully-connected layer, and a regression output layer, and includes 20 hidden units.
Substep C2: training the LSTM neural network model;
collecting historical interference signals of a cell where a user is located, dividing the historical interference signals of each group of resource blocks into a training data set and a verification data set, training the LSTM neural network model to be convergent by using the training data set, and verifying the LSTM neural network model by using the verification data set; and determining the weight parameters of the LSTM neural network model after the model training is finished.
Preferably, one of the embodiments of the sub-step C2 is:
collecting historical interference subframes for each group of resource blocks, e.g. collection 105A number of subframes, the historical interference subframes including real and imaginary parts; if the interference power of the current interfering sub-frame is predicted from the previous N interfering sub-frames collected each time, a 12x10 input matrix is generated for each historical interfering sub-frame when training and validating the LSTM neural network model, the 12x10 input matrix comprising the last 10 interfering sub-frames in the pastInterference values, where the interference value for each interfering sub-frame is a vector corresponding to 12x 1; inputting the input matrix into the LSTM neural network model for training, and obtaining an output vector of 6x1, such as g' ═ 0.1,1.2,1.3,0.8,0.9,0.3]The 6 values of the output vector respectively correspond to interference power values of the current interference subframe to 6 time-frequency resources in the group of resource blocks;
the 12x10 input matrix input in training the LSTM neural network model is sequential data with time dependence; namely, the interference suffered by 6 time-frequency resources of the 1 st, 2 nd,. 10 th interference sub-frame corresponds to an input matrix of 12x10, the input matrix is input to the LSTM neural network model, an output vector of 6x1 is obtained in the LSTM neural network model, and the output vector corresponds to and predicts the interference power value of the 6 time-frequency resources of the 11 th interference sub-frame. And according to the method, the input matrix of another 12x10 corresponding to the 6 time-frequency resources of the 2 nd, 3 rd,. 11 th interference subframe is used as the input of the LSTM neural network model, and the interference power value of the 6 time-frequency resources of the 12 th interference subframe is predicted.
Step D: acquiring the first N interference signals of each group of resource blocks in real time, inputting the first N interference signals into the LSTM neural network model to predict the interference power value of the current interference signal, and distributing corresponding time-frequency resources for users of different levels according to the prediction result;
wherein, as an example, step D comprises the following sub-steps:
substep D1: acquiring the first N interference sub-frames of each group of resource blocks in real time, inputting a 12x10 input matrix corresponding to the first N interference sub-frames of each group of resource blocks into the LSTM neural network model, and obtaining an output vector of 6x1 of each group of resource blocks, such as g' [0.1,1.2,1.3,0.8,0.9,0.3 ═]The output vector is a binary vector of 6x 1; obtaining the current interference power value of the time-frequency resource of each group of resource blocks according to the output vector of 6x1 of each group of resource blocks; i.e. when predicting the 10 th5When the interference level of 6 time frequency resources of +1 interference subframe is higher, the (10 th) th interference subframe is used5-9)、(105-8)...(105) One interference subframe corresponds to one 12The input matrix of x10 is used as the input of the LSTM neural network model, and the output vector of 6x1 of the output represents the (10 th) th5+1) interference experienced by 6 time-frequency resources of a subframe.
Substep D2: according to the current interference power value of the time frequency resource of each group of resource blocks, the time frequency resource of each group of resource blocks is divided into a first-level time frequency resource and a second-level time frequency resource, namely, two time frequency resources corresponding to the two largest interference power values (1.2 and 1.3) in an output vector g' ([ 0.1,1.2,1.3,0.8,0.9 and 0.3 ]) are used as the second-level time frequency resources, and the remaining 4 time frequency resources with smaller interference power are used as the first-level time frequency resources.
Substep D3: recovering the secondary time-frequency resources to form a resource pool, and distributing the time-frequency resources in the resource pool to the secondary users for use, namely recovering 2 secondary time-frequency resources in each group of resource blocks to form a resource pool comprising 200 secondary time-frequency resources; and the secondary time frequency resource adopts stronger channel coding and lower-order modulation coding compared with the primary time frequency resource when data transmission is carried out.
Substep D4: the primary users are prioritized according to the comprehensive indexes of the primary users;
and matching the corresponding primary time frequency resource for the primary user according to the priority of the primary user and the interference power value of the primary time frequency resource, namely allocating the time frequency resource with smaller interference power value in the primary time frequency resource to the primary user with higher priority.
The embodiment of the invention preferentially allocates the time-frequency resource with low interference to the user with higher level through a user grading mechanism, thereby ensuring the communication quality of the user with higher level; by grouping time-frequency resources, the frequency selectivity of a channel is fully utilized, and the channel condition of an individual resource group is prevented from being interfered too strongly; the LSTM neural network model is trained by utilizing the historical interference characteristics of the time-frequency resources, so that the interference characteristics of the current subframe can be predicted through the LSTM neural network model, and better wireless resources are selected in real time for data transmission, so that the same frequency interference of adjacent cells is reduced, and the transmission reliability of high-grade users is improved;
the communication quality of the low-level user is guaranteed by recovering the second-level time-frequency resources, adopting lower-order modulation coding and stronger channel coding to carry out data transmission so as to fully utilize the system resources. A user grading and resource grouping mechanism is adopted in the SCMA system, and according to the historical interference power of system resources, the interference power of a current subframe is predicted through an LSTM neural network, so that time-frequency resources with low interference power are selected for high-grade users to carry out data transmission.
The effect of this embodiment is compared with the "Random" and "MA" time-frequency resource allocation schemes as follows:
as shown in fig. 5, in the case that the signal to interference power ratio SIR is 15dB, the present embodiment compares the Bit Error Rate (BER) performance of different SNR (Eb/No) values for the "Random" time-frequency resource allocation and the "MA" time-frequency resource allocation. Wherein, the "Random" time frequency resource allocation scheme randomly selects 4 RNs from 6 RNs in the system, and the "MA" (Moving Average) time frequency resource allocation scheme predicts the interference power of the current interference subframe by calculating the Average power of 5 nearest interference subframes, and then selects 4 RNs with the minimum interference power. It can be seen that, under all considered signal-to-noise ratios, the performance of the present embodiment is superior to the "Random" time-frequency resource allocation scheme and the "MA" time-frequency resource allocation scheme, and the performance of the "MA" time-frequency resource allocation scheme is superior to the "Random" time-frequency resource allocation scheme. Error rate of 10-3In this embodiment, compared with the MA time-frequency resource allocation scheme, the gain of more than 4dB is realized, and when the bit error rate is 2 × 10-3And the time-frequency resource allocation scheme realizes gain of more than 8dB compared with the time-frequency resource allocation scheme of 'Random'.
As shown in fig. 6 and 7, in the case that SIR is equal to 10dB and 6dB, respectively, the performance of this embodiment is still better than the other two time-frequency resource allocation schemes.
On the basis of the embodiment of the invention, the invention correspondingly provides an embodiment of the device.
As shown in fig. 2, another embodiment of the present invention provides a deep learning-based resource allocation apparatus for improving the performance of an SCMA system, comprising a resource grouping module, an interference power prediction module and a resource matching module,
the resource grouping module is used for grouping time-frequency resources of a system to obtain a plurality of groups of resource blocks, and each group of resource blocks comprises a plurality of time-frequency resources;
the interference power prediction module is used for taking the first N interference signals of each group of resource blocks as the input of an LSTM neural network model to obtain the interference power value of each time-frequency resource of each group of resource blocks under the current interference signal; the LSTM neural network model is obtained by training according to the historical interference signals of each group of the resource blocks; n is a positive integer;
and the resource matching module is used for matching different time frequency resources for users of different grades according to the interference power value of each time frequency resource.
It should be noted that the above embodiments of the apparatus of the present invention correspond to the embodiments of the method of the present invention, and can implement a deep learning-based resource allocation method for improving the performance of an SCMA system according to any embodiment of the present invention.
It should be noted that the above-described device embodiments are merely illustrative, where the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. In addition, in the drawings of the embodiment of the apparatus provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, and may be specifically implemented as one or more communication buses or signal lines. One of ordinary skill in the art can understand and implement it without inventive effort.
While the foregoing is directed to the preferred embodiment of the present invention, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention.
It will be understood by those skilled in the art that all or part of the processes of the above embodiments may be implemented by hardware related to instructions of a computer program, and the computer program may be stored in a computer readable storage medium, and when executed, may include the processes of the above embodiments. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
Claims (10)
1. A resource allocation method for improving SCMA system performance based on deep learning is characterized by comprising the following steps:
grouping time-frequency resources of an SCMA system to obtain a plurality of groups of resource blocks, wherein each group of resource blocks comprises a plurality of time-frequency resources;
acquiring the first N interference signals of each group of resource blocks in real time, and taking the first N interference signals as the input of an LSTM neural network model to obtain the interference power value of each time-frequency resource of each group of resource blocks under the current interference signal; the LSTM neural network model is obtained by training according to the historical interference signals of each group of the resource blocks; n is a positive integer;
and matching different time frequency resources for users of different levels according to the interference power value of each time frequency resource.
2. The deep learning-based resource allocation method for improving the performance of the SCMA system according to claim 1, wherein the time-frequency resources of the system are grouped to obtain a plurality of groups of resource blocks, each group of resource blocks comprises a plurality of time-frequency resources, and specifically:
extracting one time frequency resource from the ith time frequency resource at intervals of K time frequency resources, and extracting M time frequency resources as an ith group of resource blocks, wherein the M time frequency resources comprise the ith time frequency resource; wherein K is more than or equal to i, and i, K and M are positive integers.
3. The deep learning-based resource allocation method for improving the performance of the SCMA system according to claim 2, wherein the first N interference signals of each group of resource blocks are obtained in real time, and the first N interference signals are used as input of an LSTM neural network model to obtain an interference power value of each time-frequency resource of each group of resource blocks under a current interference signal, specifically:
acquiring the first N interference signals of each group of resource blocks in real time, and representing the first N interference signals as a PxN matrix, wherein each interference signal corresponds to a vector of Px1 in the matrix;
taking the PxN matrix of each group of resource blocks as the input of the LSTM neural network model to obtain an Mx1 output vector of each group of resource blocks;
and obtaining the interference power value of each time-frequency resource of each group of resource blocks under the current interference signal according to the output vector of one Mx1 of each group of resource blocks.
4. The deep learning-based resource allocation method for improving SCMA system performance according to claim 3, wherein the users of different classes are primary users and secondary users divided according to the importance of the service.
5. The deep learning-based resource allocation method for improving the performance of the SCMA system according to claim 4, wherein the matching of different time-frequency resources for users of different classes according to the interference power values of the time-frequency resources comprises:
according to the interference power value of each time-frequency resource of each group of resource blocks under the current interference signal, dividing the time-frequency resources of each group of resource blocks into primary time-frequency resources and secondary time-frequency resources, allocating the primary time-frequency resources to the primary users for use, and allocating the secondary time-frequency resources to the secondary users for use.
6. The deep learning-based resource allocation method for improving the performance of the SCMA system according to claim 5, wherein the allocating the primary time-frequency resources to the primary users comprises:
the primary users are prioritized according to the comprehensive indexes of the primary users;
and matching the corresponding primary time frequency resource for the primary user according to the priority of the primary user and the interference power value of the primary time frequency resource.
7. The deep learning-based resource allocation method for improving the performance of the SCMA system according to claim 6, wherein the allocating the secondary time-frequency resources to the secondary users comprises:
and selecting the secondary time-frequency resource for recycling to form a resource pool, and allocating the time-frequency resource in the resource pool to the secondary user for use.
8. The deep learning-based resource allocation method for improving SCMA system performance as claimed in claim 7, wherein the channel coding of the secondary time-frequency resource is stronger than that of the primary time-frequency resource when data transmission is performed, and the order of the modulation coding of the secondary time-frequency resource when data transmission is performed is lower than that of the primary time-frequency resource.
9. The deep learning-based resource allocation method for improving the performance of an SCMA system according to any of claims 1 to 8, wherein the LSTM neural network model is obtained by training the LSTM neural network model according to the historical interference signals of each group of the resource blocks, specifically:
dividing the historical interference signals of each group of the resource blocks into a training data set and a verification data set, training the LSTM neural network model to be convergent by using the training data set, and verifying the LSTM neural network model by using the verification data set;
the data in the training data set are arranged according to time sequence.
10. A resource allocation device for improving SCMA system performance based on deep learning is characterized by comprising a resource grouping module, an interference power prediction module and a resource matching module,
the resource grouping module is used for grouping time-frequency resources of a system to obtain a plurality of groups of resource blocks, and each group of resource blocks comprises a plurality of time-frequency resources;
the interference power prediction module is used for taking the first N interference signals of each group of resource blocks as the input of an LSTM neural network model to obtain the interference power value of each time-frequency resource of each group of resource blocks under the current interference signal; the LSTM neural network model is obtained by training according to the historical interference signals of each group of the resource blocks; n is a positive integer;
and the resource matching module is used for matching different time frequency resources for users of different grades according to the interference power value of each time frequency resource.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110800659.5A CN113645700A (en) | 2021-07-15 | 2021-07-15 | Deep learning-based resource allocation method and device for improving SCMA system performance |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110800659.5A CN113645700A (en) | 2021-07-15 | 2021-07-15 | Deep learning-based resource allocation method and device for improving SCMA system performance |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113645700A true CN113645700A (en) | 2021-11-12 |
Family
ID=78417406
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110800659.5A Pending CN113645700A (en) | 2021-07-15 | 2021-07-15 | Deep learning-based resource allocation method and device for improving SCMA system performance |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113645700A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2024022042A1 (en) * | 2022-07-29 | 2024-02-01 | 中兴通讯股份有限公司 | Interference detection method and apparatus, and electronic device and storage medium |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100208608A1 (en) * | 2009-02-12 | 2010-08-19 | Nokia Siemens Networks Oy | Transmission power control for sounding signal for wireless networks |
US20130003680A1 (en) * | 2010-03-23 | 2013-01-03 | Sumitomo Electric Industries, Ltd. | Base station device and terminal device |
CN105323050A (en) * | 2014-07-11 | 2016-02-10 | 普天信息技术有限公司 | Method and device for relieving uplink interference |
CN110267338A (en) * | 2019-07-08 | 2019-09-20 | 西安电子科技大学 | Federated resource distribution and Poewr control method in a kind of D2D communication |
CN111083668A (en) * | 2020-01-02 | 2020-04-28 | 山东大学 | D2D resource allocation method based on alliance game algorithm in Internet of vehicles |
CN111638488A (en) * | 2020-04-10 | 2020-09-08 | 西安电子科技大学 | Radar interference signal identification method based on LSTM network |
CN112713966A (en) * | 2020-12-30 | 2021-04-27 | 大连大学 | Code modulation switching method based on likelihood estimation correction signal-to-noise ratio |
-
2021
- 2021-07-15 CN CN202110800659.5A patent/CN113645700A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100208608A1 (en) * | 2009-02-12 | 2010-08-19 | Nokia Siemens Networks Oy | Transmission power control for sounding signal for wireless networks |
US20130003680A1 (en) * | 2010-03-23 | 2013-01-03 | Sumitomo Electric Industries, Ltd. | Base station device and terminal device |
CN105323050A (en) * | 2014-07-11 | 2016-02-10 | 普天信息技术有限公司 | Method and device for relieving uplink interference |
CN110267338A (en) * | 2019-07-08 | 2019-09-20 | 西安电子科技大学 | Federated resource distribution and Poewr control method in a kind of D2D communication |
CN111083668A (en) * | 2020-01-02 | 2020-04-28 | 山东大学 | D2D resource allocation method based on alliance game algorithm in Internet of vehicles |
CN111638488A (en) * | 2020-04-10 | 2020-09-08 | 西安电子科技大学 | Radar interference signal identification method based on LSTM network |
CN112713966A (en) * | 2020-12-30 | 2021-04-27 | 大连大学 | Code modulation switching method based on likelihood estimation correction signal-to-noise ratio |
Non-Patent Citations (1)
Title |
---|
梁天棋: "\"基于深度学习的通信抗干扰接收算法研究\"", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2024022042A1 (en) * | 2022-07-29 | 2024-02-01 | 中兴通讯股份有限公司 | Interference detection method and apparatus, and electronic device and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110417496B (en) | Cognitive NOMA network stubborn resource allocation method based on energy efficiency | |
CN101409884B (en) | Method for optimizing network frequency based on measurement report | |
CN111372314A (en) | Task unloading method and task unloading device based on mobile edge computing scene | |
CN104205666A (en) | Method and system for dynamic, joint assignment of power and scheduling of users for wireless systems | |
CN103916355A (en) | Distribution method for sub carriers in cognitive OFDM network | |
CN106411469B (en) | Multicast resource allocation and transmission method based on scalable video in multi-base-station heterogeneous network | |
CN113382477A (en) | Method for modeling uplink interference between wireless network users | |
CN113239632A (en) | Wireless performance prediction method and device, electronic equipment and storage medium | |
CN113645700A (en) | Deep learning-based resource allocation method and device for improving SCMA system performance | |
CN110139282B (en) | Energy acquisition D2D communication resource allocation method based on neural network | |
CN111865474A (en) | Wireless communication anti-interference decision method and system based on edge calculation | |
CN111246485A (en) | Internet of vehicles resource allocation method under high-density vehicle-mounted communication environment | |
CN114521023A (en) | SWIPT-assisted NOMA-MEC system resource allocation modeling method | |
US9253740B2 (en) | Method and apparatus for coordinated power-zone-assignment in wireless backhaul networks | |
CN105530203B (en) | The connection control method and system of D2D communication links | |
CN110139318B (en) | NOMA cellular heterogeneous network resource allocation method and system | |
CN109041009B (en) | Internet of vehicles uplink power distribution method and device | |
CN115915454A (en) | SWIPT-assisted downlink resource allocation method and device | |
Pinagapany et al. | Solving channel allocation problem in cellular radio networks using genetic algorithm | |
Lei et al. | Power and discrete rate adaptation in BER constrained wireless powered communication networks | |
CN110381590B (en) | Wireless network sub-channel power distribution method based on wireless energy collection | |
CN110381589B (en) | Cooperative wireless network sub-channel power distribution method based on wireless energy collection | |
US8654704B2 (en) | System and method for distributed control of spectrum power in a multi-cell communication system | |
CN110225537B (en) | Cooperative wireless network power distribution method based on wireless energy collection | |
Kwon | Average data rate analysis for hierachical cell structure under Nakagami-m fading channel with a two-layer feed-forward neural network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20240402 |
|
AD01 | Patent right deemed abandoned |