Method, apparatus, and storage medium for allocating base station resources

Abstract

Embodiments of the present disclosure provide a base station resource allocation method and device, electronic equipment and storage medium, the method comprising: obtaining a load feature matrix vector corresponding to each base station according to a base station resource sharable graph of a preset area, a current load of each base station and a GCN network; obtaining a load feature matrix vector of any base station and a load feature matrix vector of any sharable base station of any base station from the load feature matrix vector corresponding to each base station according to the base station resource sharable graph; determining a sharing party, a shared party and a sharing ratio when any base station and any sharable base station share resources according to the load feature matrix vector of any base station and any sharable base station; determining a resource allocation strategy corresponding to any base station according to the sharing party, the shared party and the sharing ratio, a resource complementary value between any base station and any sharable base station, and allocating resources to any base station based on the resource allocation strategy.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, and in particular to a method, apparatus, electronic device, and storage medium for allocating base station resources. Background Technology

[0002] Traditional base station resource allocation methods primarily rely on real-time resource allocation based on current network load and user demand to ensure network efficiency and reliability. However, with the rapid development of communication technology and the significant increase in the number of base stations, the cost of real-time network monitoring has risen sharply, rendering traditional methods inadequate for meeting the requirements of efficient and intelligent base station resource allocation.

[0003] In related technologies, intelligent algorithms such as genetic algorithms and particle swarm optimization can be used to allocate base station resources in order to achieve intelligent resource allocation. However, in resource sharing scenarios between base stations, when facing large-scale base stations, these technologies cannot avoid the problems of latency in the resource allocation process and increased costs due to multiple sharing. Summary of the Invention

[0004] This disclosure provides a method, apparatus, electronic device, and storage medium for allocating base station resources to address the latency issues in resource allocation processes during large-scale base station scenarios and the increased costs resulting from multiple sharing among base stations in related technologies.

[0005] In a first aspect, embodiments of this disclosure provide a method for allocating base station resources, the method comprising:

[0006] Based on the base station resource sharing graph corresponding to the base stations in the preset area, the current load of each base station, and the pre-trained graph convolutional neural network (GCN), the load feature matrix vector corresponding to each base station is obtained; wherein, the base station resource sharing graph is determined according to the location of the base stations in the preset area and is used to identify base stations that are close to each other in the preset area; the load feature matrix vector is used to characterize the load information of the corresponding base station in the preset area of ​​its neighboring base stations.

[0007] Based on the base station resource shareability graph, the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station of any base station are obtained from the load feature matrix vectors corresponding to each base station.

[0008] Based on the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station, determine the sharing party, the sharing recipient, and the sharing ratio when any base station shares resources with any shareable base station.

[0009] Based on the sharing party, the sharing recipient, and the sharing ratio when any base station shares resources with any shareable base station, and the resource complementarity value between any base station and any shareable base station, a resource allocation strategy corresponding to any base station is determined, and resources are allocated to any base station based on the resource allocation strategy; wherein, the resource complementarity value between any base station and any shareable base station is determined based on the base station resource shareability map, the current load and optimal load of any base station, and the current load and optimal load of any shareable base station.

[0010] Secondly, embodiments of this disclosure provide a base station resource allocation apparatus, the apparatus comprising:

[0011] The first acquisition module is used to acquire a load feature matrix vector corresponding to each base station based on the base station resource shareable graph corresponding to the base stations in the acquired preset area, the current load of each base station, and a pre-trained graph convolutional neural network (GCN). The base station resource shareable graph is determined based on the location of the base stations in the preset area and is used to identify base stations that are close to each other in the preset area. The load feature matrix vector is used to characterize the load information of the neighboring base stations of the corresponding base station in the preset area.

[0012] The second acquisition module is used to acquire, based on the base station resource shareability graph, the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station of any base station from the acquired load feature matrix vectors corresponding to each base station.

[0013] The first determining module is used to determine the sharing party, the sharing recipient, and the sharing ratio when any base station shares resources with any shareable base station based on the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station.

[0014] The second determining module is used to determine the resource allocation strategy corresponding to any base station based on the sharing party, the sharing recipient, and the sharing ratio when any base station shares resources with any shareable base station, and the resource complementarity value between any base station and any shareable base station, so as to allocate resources to any base station based on the resource allocation strategy; wherein, the resource complementarity value between any base station and any shareable base station is determined based on the base station resource sharing map, the current load and optimal load of any base station, and the current load and optimal load of any shareable base station.

[0015] Thirdly, embodiments of this disclosure provide a base station resource allocation device, including: a processor; and a memory configured to store computer-executable instructions, which, when executed, cause the processor to perform the steps of the method described in the first aspect above.

[0016] Fourthly, embodiments of this disclosure provide a computer-readable storage medium for storing computer-executable instructions that, when executed by a processor, implement the steps of the method described in the first aspect.

[0017] The at least one technical solution provided by the embodiments of the present invention can achieve the following technical effects:

[0018] In this embodiment of the invention, the load feature matrix vectors corresponding to each base station can first be obtained based on the base station resource sharing map corresponding to the base stations in a preset area, the current load of each base station, and the pre-trained GCN network. Then, based on the base station resource sharing map, the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station of that base station are obtained from the obtained load feature matrix vectors corresponding to each base station. Then, based on the load feature matrix vectors of any base station and the load feature matrix vectors of any shareable base station, the sharing party, the sharing party, and the sharing ratio when the base station shares resources with the shareable base station are determined. Based on the sharing party, the sharing party, the sharing ratio, and the resource complementarity value between the base station and the shareable base station, the resource allocation strategy corresponding to the base station is determined, and resources are allocated to the base station based on the resource allocation strategy. The resource complementarity value between the base station and the shareable base station is determined based on the base station resource sharing map, the current load and optimal load of the base station, and the current load and optimal load of the shareable base station.

[0019] As can be seen from the above embodiments, when allocating base station resources, the embodiments of the present invention can first determine the base stations that can share resources, and further determine the sharing party, the sharing recipient, and the sharing ratio. Then, the resource allocation strategy for the base stations can be determined based on the determined sharing party, the sharing recipient, the sharing ratio, and the resource complementarity value between the base stations. Since the resource sharing between base stations can be comprehensively determined by combining the resource complementarity value between the base stations, the sharing direction, and the sharing ratio, a reasonable base station resource allocation strategy can be quickly determined, effectively reducing the number of sharing interactions between base stations, thereby effectively improving resource allocation efficiency. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in one or more embodiments of this disclosure or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is one of the flowcharts illustrating a method for allocating base station resources according to an embodiment of the present invention;

[0022] Figure 2 This is a second flowchart illustrating a method for allocating base station resources according to an embodiment of the present invention.

[0023] Figure 3 This is a schematic diagram of a base station resource allocation strategy prediction model provided in one embodiment of the present invention;

[0024] Figure 4 A schematic diagram of the module composition of a base station resource allocation device 400 provided in one embodiment of the present invention;

[0025] Figure 5 This is a schematic diagram of the hardware structure of a base station resource allocation device provided in one embodiment of the present invention. Detailed Implementation

[0026] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this disclosure, and to make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0027] The technical solutions provided by the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0028] Please see Figure 1 , Figure 1 This is one of the flowcharts illustrating a method for allocating base station resources according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0029] Step 102: Based on the base station resource sharing graph corresponding to the base stations in the preset area, the current load of each base station, and the pre-trained graph convolutional neural network (GCN), obtain the load feature matrix vector corresponding to each base station; wherein, the base station resource sharing graph is determined according to the location of the base stations in the preset area and is used to identify base stations that are close to each other in the preset area; the load feature matrix vector is used to characterize the load information of the corresponding base station in the preset area of ​​its neighboring base stations.

[0030] Step 104: Based on the base station resource sharing graph, obtain the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station from the load feature matrix vectors corresponding to each base station.

[0031] Step 106: Based on the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station, determine the sharing party, the sharing recipient, and the sharing ratio when any base station shares resources with any shareable base station.

[0032] Step 108: Based on the sharing party, the sharing recipient, and the sharing ratio when any base station shares resources with any shareable base station, and the resource complementarity value between any base station and any shareable base station, determine the resource allocation strategy corresponding to any base station, and allocate resources to any base station based on the resource allocation strategy; wherein, the resource complementarity value between any base station and any shareable base station is determined based on the base station resource sharing map, the current load and optimal load of any base station, and the current load and optimal load of any shareable base station.

[0033] In this embodiment of the invention, a base station resource sharing map corresponding to base stations within a preset area can be obtained first. The base station resource sharing map is determined based on the locations of base stations within the preset area and is used to identify base stations that are close to each other within the preset area.

[0034] In one embodiment of the invention, the spatial distance between any two base stations in a preset area can be obtained based on the location of base stations in the preset area, and a resource base station shareability map can be constructed based on the spatial distance between any two base stations and a preset resource shareability distance threshold.

[0035] In one example, the latitude and longitude coordinates of each base station within the preset area can be: L = {l1, l2, ..., l...} n}, Where n (n>0) can represent the number of base stations. It can represent a base station longitude coordinates Indicates base station latitude coordinates Base stations can be regarded as nodes. According to Equation (1), the latitude and longitude relationship between base stations can be converted into spatial distance d(i,j). Based on the distance threshold D1 (D1>0) for resource sharing between base stations, a base station resource sharing graph G can be constructed. R Where the distance threshold D1 is preset. It can be used... Figure G represents the shareable base station resources. R Given an adjacency matrix, if d(i,j)≤D1, then R(i,j)=1, meaning that the distance between base station i and base station j does not exceed D1, and they can share resources; if d(i,j)>D1, then R(i,j)=0, meaning that base station i and base station j cannot share resources.

[0036]

[0037] Where, a = lat i -lat j b represents the difference in latitude between base station i and base station j, where b = lng i -lng j This represents the difference in longitude between base station i and base station j, where r is the Earth's equatorial radius, measured in kilometers.

[0038] In this embodiment of the invention, after obtaining the base station resource shareable graph corresponding to the base stations within a preset area, the load feature matrix vector corresponding to each base station can be obtained based on the base station resource shareable graph, the current load of each base station within the preset area, and the pre-trained GCN (Graph Convolutional Network). The load feature matrix vector is used to characterize the load information of neighboring base stations within the preset area.

[0039] In one embodiment of the invention, when obtaining the load feature matrix vector corresponding to each base station, the current load of the shareable base stations displayed in the base station resource shareable graph can be obtained first based on the current load of each base station. That is, after obtaining the base station resource shareable graph through the above steps, the base stations that have edges with other base stations in the base station resource shareable graph, that is, the base stations that can share resources with other base stations in the preset area, can be identified as shareable base stations, and the current load of the shareable base stations can be obtained. Then, an initial load feature matrix can be constructed based on the current load of the shareable base stations. The initial load feature matrix is ​​transposed to obtain the transposed matrix, and the transposed matrix is ​​input into the pre-trained GCN network. The load feature matrix vector corresponding to each base station is obtained after the GCN network processes the transposed matrix, the preset weight matrix, the fusion matrix, and the degree matrix of the fusion matrix through the activation function. The fusion matrix can be obtained by fusing the transposed matrix with the preset identity matrix.

[0040] In one example, the base station resource sharing graph constructed above can be represented as G. R ,Will The load of shareable base stations in the time-based base station resource-sharing graph is regarded as a node feature. According to equation (2), the base station load feature matrix is ​​extracted using the GCN network. in For the extracted base station The load feature matrix vector is used to represent the base station. Load information of neighboring base stations within the preset area.

[0041]

[0042] in, for The transpose of the initial load characteristic matrix of the base station at that time. I N This is the preset identity matrix; for The degree matrix, W1 is the pre-trained weight matrix; f(·) is the activation function.

[0043] After obtaining the load feature matrix vector corresponding to the shareable base station in the base station resource shareable graph, the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station of that base station can be obtained from the obtained load feature matrix vector. Based on the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station, the sharing party, the sharing party and the sharing ratio when the base station shares resources with the shareable base station can be determined.

[0044] In one embodiment of the invention, when determining the sharing party and the shared party when any base station shares resources with any shareable base station of any base station, the corresponding load feature matrix vectors of the two can be concatenated in different orders, and the concatenated vector is input into a feedforward neural network to obtain the corresponding output value. Then, the sharing party and the shared party in this resource sharing are determined according to the magnitude of the output value corresponding to the different concatenation orders. This embodiment of the invention can constrain the sharing direction of resource sharing by clearly defining and fixing the sharing party and the shared party, limiting resource sharing to one-way, and effectively avoiding the problem of repeated interaction caused by base stations sharing resources with each other.

[0045] In this embodiment of the invention, specifically, the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station can be concatenated sequentially to obtain a first concatenated matrix vector corresponding to the concatenation of any base station with any shareable base station. The first concatenated matrix vector is then input into a pre-constructed feedforward neural network to obtain the first feedforward network value output by the feedforward neural network corresponding to the concatenation of any base station with any shareable base station. Then, the load feature matrix vector of any shareable base station and the load feature matrix vector of any base station can be concatenated sequentially to obtain a second concatenated matrix vector corresponding to the concatenation of any base station with any other base station. The second concatenated matrix vector is then input into the feedforward neural network to obtain the second feedforward network value output by the feedforward neural network corresponding to the concatenation of any base station with any other base station.

[0046] Of course, in this embodiment of the invention, the second concatenation matrix vector can be obtained first, and then the first concatenation matrix vector can be obtained. This embodiment of the invention does not limit this.

[0047] After obtaining the first feedforward network value and the second feedforward network value, when the first feedforward network value is not less than the second feedforward network value, any base station can be identified as the sharing party when sharing resources with any shareable base station, and any shareable base station can be identified as the sharing recipient when sharing resources with any base station. This indicates that the idle level of any base station is higher than that of any shareable base station. Therefore, when sharing resources with any base station, any base station can share its idle resources with any shareable base station. Conversely, when the first feedforward network value is less than the second feedforward network value, any shareable base station can be identified as the sharing party when sharing resources with any base station, and any base station can be identified as the sharing recipient when sharing resources with any base station. This indicates that the idle level of any shareable base station is higher than that of any base station. Therefore, when sharing resources with any base station, any shareable base station can share its idle resources with any base station.

[0048] In one example of an embodiment of the present invention, it can be The degree of sharing between time-sharing base stations is represented as in, It can be determined by formula (3), and then by comparison. and Judging by size The direction of resource allocation between base stations at any given time, specifically, when hour, Then base station i cannot transmit resources to base station j; when hour, Therefore, base station j cannot transmit resources to base station i.

[0049]

[0050] Here, concat(·) is a vector concatenation operation, and f1(·) is a feedforward neural network.

[0051] After determining the sharing party and the receiving party when any base station shares resources with any shareable base station, the sharing ratio during this resource sharing can be further determined to reduce the number of interactions between base stations to a certain extent.

[0052] In one embodiment of the invention, when determining the sharing ratio when any base station shares resources with any shareable base station, the load feature matrix vector of any base station and the load feature matrix vectors of each shareable base station can be concatenated sequentially to obtain a first concatenated matrix vector corresponding to the concatenation of each shareable base station by any base station. This first concatenated matrix vector is then input into the feedforward neural network to obtain the first feedforward network value output by the feedforward neural network corresponding to the concatenation of each shareable base station. Then, the sharing ratio when any base station shares resources with any shareable base station can be determined based on the product of the first feedforward network value corresponding to the concatenation of each shareable base station and the resource complementarity value between the base station and the shareable base station, and the product of the first feedforward network value corresponding to the concatenation of each shareable base station and the resource complementarity value between the base station and the shareable base station.

[0053] In order to achieve optimal base station resource allocation while reducing the number of resource sharing sessions between base stations, this embodiment of the invention prioritizes base stations with high resource complementarity values ​​for resource allocation and increases the resource allocation ratio among these base stations. This effectively reduces the number of interactions between base stations, improves computational efficiency, and saves on the cost of interactions between base stations.

[0054] Therefore, when determining the risk ratio of resource sharing between base stations, the embodiments of the present invention need to consider not only the sharing direction, but also the resource complementarity value between base stations.

[0055] In one embodiment of the invention, a resource complementarity value between any base station and any sharing base station can be determined. This resource complementarity value can be determined based on a resource-sharing map of the base stations, the current and optimal load of the any base station, and the current and optimal load of the any sharing base station.

[0056] In this embodiment of the invention, the current load and optimal load of any base station, as well as the current load and optimal load of any shareable base station, can be obtained first. Then, a base station resource shareable graph can be obtained, and the resource complementarity value between the two can be calculated according to a preset calculation formula.

[0057] In one example of an embodiment of the present invention, it is possible to obtain Bandwidth utilization rate of each base station in the preset area during the time period in, express The load of n base stations at time t, and the base station shareability graph, can be represented as G. R The optimal load of a base station can be expressed as V (V > 0). When resource sharing is possible between base stations, it can be calculated according to equation (4). Resource complementarity value between base station i and base station j at time point

[0058]

[0059] in, They represent The current load of base station i and base station j at any given time. Let be the unit transition function, defined as shown in equation (5). From equation (5), it can be seen that as the sum of the shareable base station loads approaches 2V, The larger the value, the greater the complementary value of idle resources and additional demand resources between base station i and base station j. In this case, resource sharing between the base stations can more effectively meet each other's needs. At this time, the number of resource sharing times with other base stations can be reduced, thereby quickly approaching the optimal load.

[0060]

[0061] Considering that excessive load leads to degraded base station communication quality, while excessively low load results in low base station resource utilization, and that the task of base station resource allocation is to transfer bandwidth resources from low-load base stations to high-load base stations, it is essential to ensure that each base station within a pre-defined area is close to its optimal load. This allows for full utilization of base station bandwidth resources to guarantee user communication quality. This invention prioritizes resource sharing between two base stations with the highest complementarity between idle resources and additional demand resources, thereby effectively reducing the number of interactions with other base stations and improving resource sharing costs and computational efficiency.

[0062] In one example, the shareable distance D1 between base stations can be set to 0.8. Calculate the distance between any two base stations according to Equation (1), and construct a base station resource shareable graph G based on this. R , and part of the adjacency matrix R is shown in Table 1. For example, the distance d(1, 2) between base station 1 and base station 2 = 0.386, d(1, 2) < D1, then R(1, 2) = R(2, 1) = 1.

[0063] Table 1. Base station shareable matrix R

[0064] base station 1 2 3 4 5 … 200 1 0 1 0 1 0 … 0 2 1 0 1 1 0 … 0 3 0 1 0 1 0 … 0 4 1 1 1 0 0 … 0 5 0 0 1 1 0 … 0 … … … … … … … … 200 0 0 0 0 0 … 0

[0065] Then, obtain the load data of the base stations within the preset area. Among them, the load data of some base stations within the preset area can be shown in Table 2.

[0066] Table 2. Load of each base station on a certain day

[0067] base station 1 2 3 4 5 … 200 load 20％ 56％ 65％ 60％ 24％ … 17％

[0068] The optimal load V of the base station can be set to 35%. When resource sharing is possible between base stations, the resource complementarity value between base stations can be calculated according to Equation (4). The resource complementarity values between some base stations can be shown in Table 3, which reflects the complementarity degree of idle resources and additional demands between base stations. For example, the resource complementarity value between base station 1 and base station 2 at a certain moment is 0.92, while the resource complementarity value between base station 1 and base station 4 is 0.88, then base station 1 and base station 2 are more suitable for resource interaction at this moment. By preferentially considering the two base stations with the largest complementarity degree of idle resources and additional demand resources for resource interaction, the number of interactions with other base stations can be reduced.

[0069] Table 3. Resource complementarity values between base stations at a certain moment

[0070]

[0071] After determining the resource complementarity value between base stations, the embodiment of the present invention can determine the sharing ratio when any base station shares resources with any shareable base station according to the first feedforward network value corresponding to when any base station splices any shareable base station, the resource complementarity value between any base station and any shareable base station, the first feedforward network value corresponding to when any base station splices each shareable base station, and the resource complementarity value between any base station and each shareable base station.

[0072] In one example of the embodiment of the present invention, it can be calculated by Equation (6) Resource sharing ratio between base stations at a certain moment

[0073]

[0074] In this embodiment of the invention, after calculating the resource sharing ratio, the resource allocation strategy corresponding to any base station can be determined based on the sharing party, the sharing party and the sharing ratio when any base station shares resources with any shareable base station, and the resource complementarity value between any base station and any shareable base station, so as to allocate resources to any base station based on the resource allocation strategy.

[0075] In one example of an embodiment of the present invention, the above It can also represent the ratio of the amount of resources transmitted from base station i to base station j to the amount of idle resources available at base station i. Then, it can be... Consider the weights of the edges and construct... Time-based base station resource allocation diagram use Represent its adjacency matrix, Describe the resource allocation strategy among base stations.

[0076] Specifically, to ensure that all shareable base stations can implement the corresponding resource allocation strategy, the system is traversed row by row. When the load of base station i is lower than the optimal load and the load of base station j is higher than the optimal load V, base station resources are allocated and base station loads are updated according to equation (7). Among them, base station load It can reflect the resource allocation between base stations.

[0077]

[0078] Then, the loss function of the base station resource allocation model can be constructed based on this. The definition can be shown in equation (8), the loss function It can represent the difference between the load of each base station and the optimal load after resource allocation. The smaller the value, the more it indicates that it is based on The better the resource allocation strategy, the better. By minimizing the loss function... Perform backpropagation to bring the load of each base station as close as possible to its optimal load, thereby obtaining... Optimal base station resource allocation diagram Let represent the optimal resource allocation strategy at that moment.

[0079]

[0080] The construction can be carried out according to the technical steps corresponding to equations (7) and (8) above. Dynamic diagram of optimal base station resource allocation during the time period And the dynamic graph of resource allocation G ′Represented as an adjacency matrix

[0081] In one example of this embodiment of the invention, based on the base station resource shareability graph, the load of the base station can be regarded as a node feature, and the load information of the neighboring base stations of each base station can be extracted using GCN according to equation (2). Next, the sharing degree between base stations is calculated according to equation (3). and By comparing the magnitudes, the direction of resource allocation between base stations is determined, avoiding redundant resource interactions. For example, Then set That is, base station 2 cannot transmit resources to base station 1.

[0082] Based on the determined base station resource sharing degree and sharing direction, the resource sharing ratio is calculated according to equation (6), and a base station resource allocation matrix is ​​constructed at time t accordingly. The resource allocation matrix of some base stations at a certain moment can be shown in Table 4.

[0083] Table 4. Base station resource allocation matrix at a certain time.

[0084] base station 1 2 3 4 5 … 200 1 0 0.25 0 0.17 0 … 0 2 0 0 0 0.19 0 … 0 3 0 0.22 0 0.24 0 … 0 4 0 0 0 0 0 … 0 5 0 0 0.20 0.22 0 … 0 … … … … … … … … 200 0 0 0 0 0 … 0

[0085] Iterate through A row by row. t Base station resources are allocated and base station load is updated according to equation (7). This reflects the resource allocation among base stations. For example, according to Table 2, base station 1 has a load of 20%, base station 2 has a load of 56%, and A t If (1,2) = 0.25, then base station 1 transmits 0.25 * (35% - 20%) = 3.75% of its resources to base station 2. Then, the load on base station 1 is updated to 20% + 3.75% = 23.75%, and the load on base station 2 is updated to 56% - 3.75% = 52.25%. The load on base station 4 is 60%. t (1,4) = 0.17. Base station 1 transmits 0.17 * (35% - 23.75%) = 1.91% of its resources to base station 4. Then, the load of base station 1 is updated to 23.75% + 1.91% = 25.66%, and the load of base station 4 is updated to 60% - 1.91% = 58.09%. The partial base station loads after resource allocation are shown in Table 5.

[0086] Table 5. Load of each base station after resource allocation

[0087] base station 1 2 3 4 5 … 200 load 35％ 48.66％ 60.28％ 51.72％ 35％ … 30％

[0088] Base station resource allocation process as follows Figure 2As shown, the learning rate is set to 0.01. According to formula (8), the stochastic gradient descent method is used to iteratively update the weights of the network model until the loss function is reached. Convergence is achieved by bringing the load of each base station as close as possible to its optimal load, thus obtaining the optimal base station resource allocation graph G at that moment. t This represents the optimal resource allocation strategy at that moment. The optimal base station resource allocation map A′={A1,A2,…,A′} is obtained by applying the above steps to six months of data. 183}

[0089] In this embodiment of the invention, after determining the resource allocation strategy corresponding to any base station, a resource allocation prediction strategy corresponding to that base station can also be obtained through a pre-trained GAN (Generative Adversarial Networks) network; wherein, the pre-trained GAN network includes a resource allocation generation network and a resource allocation discrimination network.

[0090] In this embodiment of the invention, considering that the spatial distance between base stations and unidirectional resource sharing may lead to a lack of sufficient information to predict the future relationship between base stations, resulting in low accuracy of base station resource allocation prediction results, this embodiment of the invention further proposes a base station resource allocation prediction scheme to solve the problem that the limited spatial distance and resource sharing direction between base stations may lead to a lack of sufficient information to predict the relationship between base stations in the future, resulting in low accuracy of base station resource allocation prediction results, and to achieve a more reasonable allocation of future base station resources.

[0091] Among them, the base station resource allocation strategy prediction task is to predict the base station resource allocation strategy by giving the base station resource allocation matrix A″={A ... T-l+1 A T-l+2 ,…,A r To predict the base station resource allocation matrix at the next time step. To train the base station resource allocation prediction model, this embodiment of the invention can divide the adjacency matrix sequence of the base station resource allocation strategy obtained through the previous embodiments into multiple matrix sequences with a time window of l, and divide them into training set and test set according to the prediction ratio, such as 8:2. The training set is used to update the parameters of the base station resource allocation prediction model, i.e., the GAN network model, and the test set is used to evaluate the accuracy of the GAN network model.

[0092] In one example of an embodiment of the present invention, A′ in the above embodiment can be divided into multiple time windows of 40, forming a matrix sequence A″={A T-39 A T-38 A TThe task of the base station resource allocation strategy prediction model is to predict the base station resource allocation matrix at the next time step based on the first 40 optimal base station resource allocation matrices. Then, it is divided into training and test sets according to the prediction ratio, such as an 8:2 ratio.

[0093] In this embodiment of the invention, the GAN network may include a resource allocation generation module and a resource allocation discrimination module. The resource allocation generation module can predict the future base station resource allocation matrix by effectively utilizing the temporal characteristics of base station load and resource allocation relationships between base stations; while the resource allocation discrimination module can be used to evaluate the base station resource allocation map predicted by the resource allocation generation module and guide the model to predict a base station resource allocation map closer to the optimal one. A schematic diagram of the base station resource allocation strategy prediction model is shown below. Figure 3 As shown.

[0094] In this embodiment of the invention, the resource allocation generation module may include a standard convolutional kernel and a dilated convolutional kernel, wherein the standard convolutional kernel can be used to extract short-term temporal features, and the dilated convolutional kernel can be used to extract periodic features.

[0095] In this embodiment of the invention, considering that the base station load and the resource allocation relationship between base stations have obvious short-term temporal and periodic patterns, this embodiment first uses the GCN network to extract the load and resource allocation information around each base station at any given time. Based on this, standard convolution and dilated convolution are used to extract the short-term temporal and periodic features of the base station load and the resource allocation between base stations, respectively, and then predict the base station resource allocation map at future times. Standard convolution focuses on information from adjacent time points and is therefore suitable for extracting short-term temporal features; dilated convolution has a larger receptive field than standard convolution and is suitable for extracting periodic features. The steps of the base station resource allocation generation module to predict the base station resource allocation map at future times can be as follows:

[0096] First, based on the dynamic graph of optimal base station resource allocation in the time period [T-l+1, T], {G} T-l+1 G T-l+2 , ..., G T}, taking the base station load as a node characteristic, and using Equation (9) to obtain the resource allocation relationship feature matrix (Y) between base stations during this period using the GCN network. T-l+1 Y T-l+2 , ..., Y T ),in express The feature matrix of resource allocation relationship between time-based base stations, followed by... Reconstructed into row vectors The base station resource allocation vector matrix Y = (y T-l+1 y T-l+2 , ..., y T ), This facilitates the extraction of temporal features in subsequent convolution operations.

[0097]

[0098] in, To uniformly distribute noise, adding noise to the generator input in a GAN network can increase the diversity of generated samples, thereby improving the generalization ability of the base station resource allocation prediction model. W2 is the weight matrix.

[0099] Then, based on the base station resource allocation relationship feature matrix Y obtained from the above steps, k one-dimensional dilated convolution kernels and standard convolution kernels of size c are used respectively to extract the periodic features of the dynamic relationship of the base stations. and short-term time series characteristics The dilation rate of the dilated convolution kernel is skipped, corresponding to the duration of one cycle. Then, average pooling is performed on P1 and P2 to obtain the features. and

[0100] The periodic feature q1 and short-term time series feature q2 extracted in the above steps are concatenated into vectors and then input into a feedforward neural network to obtain the base station resource allocation matrix for the next time step. Then, the diagonal elements are set to 0 to eliminate the effect of self-connected edges, meaning that the base station does not need to perform resource interactions with itself.

[0101] The base station resource allocation discrimination module D calculates the discrimination probability of the input base station resource allocation map according to equation (10). The larger the value, the more the base station resource allocation discrimination module believes that the input is closer to the optimal base station resource allocation map. The discrimination probability is used to construct the subsequent loss function to guide the base station resource allocation generation module to predict a better base station resource allocation matrix.

[0102] D(A′)=sigmoid(sigmoid(A′W3+b1)W4+b2) (10)

[0103] in, W3 and b1 are the trainable weight matrices and bias terms of the hidden layer of the resource allocation discrimination module, respectively. W4 and b2 are the trainable weight matrices and bias terms of the output layer of the resource allocation discrimination module, respectively. sigmoid(·) is the activation function.

[0104] After constructing a resource allocation prediction model that includes a resource allocation generation module and a resource allocation discrimination module, the resource allocation prediction model can be trained and evaluated.

[0105] In training the resource allocation prediction model, considering that traditional GANs use cross-entropy as the loss function, gradient vanishing and other problems may occur in generative adversarial training, while WGAN (Wasserstein Generative Adversarial Network) measures the distance between the generated sample distribution and the real sample distribution by the difference between the outputs of the discriminator, which can provide a more stable and controllable training process, this embodiment of the invention uses WGAN for generative adversarial training.

[0106] In each training iteration, the parameters of the resource allocation generation module are first fixed, and the loss function of equation (11) is minimized. To update the parameters of the resource allocation discrimination module, a weight pruning method is used to limit its parameter range to [-c, c] to avoid gradient vanishing; then, the parameters of the resource allocation discrimination module are fixed, and the loss function of equation (12) is minimized. This updates the parameters of the resource allocation generation module. Through continuous iteration, the predictive ability of the resource allocation generation module and the discriminative ability of the resource allocation discrimination module are improved, as the loss function... and After convergence, the parameters of the resource allocation generation module are saved for future base station resource allocation matrix prediction.

[0107]

[0108]

[0109] Here, mean(·) is used to calculate the mean, in order to reduce the variance between samples.

[0110] When evaluating the resource allocation prediction model, the test set corresponding to the resource allocation strategy obtained in the above steps can be input into the trained base station resource allocation strategy prediction model to generate the base station resource allocation matrix at future time. The accuracy of the base station resource allocation strategy prediction model is evaluated using RMSE (Root Mean Square Error) and MR (Mismatch Rate). RMSE reflects the distance between the predicted base station resource allocation matrix and the optimal base station resource allocation matrix, and MR reflects the proportion of incorrectly predicted base station interactions. The smaller the RMSE and MR, the better the prediction ability of the base station resource allocation strategy prediction model.

[0111] The RMSE is defined as shown in equation (13).

[0112]

[0113] The definition of MR is shown in equation (14).

[0114]

[0115] Where, N false To prevent incorrect predictions about the number of base stations that can interact, N exist The number of interactions in the optimal base station resource allocation strategy.

[0116] In one example of this invention, the constructed base station resource allocation strategy prediction model is as follows: Figure 3 As shown, it is divided into a base station resource allocation generation module and a base station resource allocation discrimination module.

[0117] The base station resource allocation generation module can use base station load as a node feature, and extract the resource allocation relationship feature matrix (Y) between base stations in the time period [T-39, T] according to equation (9) using GCN. T-39 Y T-38 , ..., Y T ),in Next, Y t Reorganized into row vectors, the base station resource allocation vector matrix Y = (y T-39 y T-39 , ..., y T ), We then performed convolution operations on Y using 64 one-dimensional dilated convolution kernels of size 3 and standard convolution kernels to extract the periodic features of the dynamic relationship between base stations. and short-term time series characteristics The base station load cycle is set to 7 days; therefore, the dilation rate of the dilated convolution is set to skip = 7. Then, average pooling is performed on P1 and P2 to obtain the features. and Next, q1 and q2 are concatenated into vectors and output via a feedforward neural network. Finally, the diagonal elements are set to 0 to obtain the base station resource allocation matrix for the next time step. Some of the results are shown in Table 5.

[0118] Table 6. Predicted Base Station Resource Allocation Matrix

[0119] base station 1 2 3 4 5 … 200 1 0 0.56 0 0 0 … 0 2 0 0 0.19 0 0 … 0 3 0 0 0.46 0.67 0 … 0 4 0.22 0 0 0 0 … 0 5 0 0 0.36 0 0 … 0 … … … … … … … … 200 0 0 0 0 0 … 0

[0120] The base station resource allocation discrimination module can calculate the discrimination probability of the input base station resource allocation map according to equation (10), which is used to construct the subsequent loss function to guide the base station resource allocation generation module to predict a better base station resource allocation matrix. For example, the objective of the base station resource allocation discrimination module is D(A T+1 ) as large as possible, and The goal of the base station resource allocation generation module is to minimize the resource allocation matrix to accurately determine the predicted and optimal resource allocation matrices. The larger the value, the better the predicted resource allocation matrix, making it difficult for the base station resource allocation discrimination module to distinguish between them.

[0121] When training the base station resource allocation strategy prediction model, the training set corresponding to the base station resource allocation strategy obtained in the above embodiment is input into the base station resource allocation strategy prediction model for generative adversarial training, with a learning rate of 0.0001. In each training iteration, the parameters of the resource allocation generation module are first fixed, and the loss function of equation (11) is minimized. To update the parameters of the resource allocation discrimination module, a weight pruning method is used to limit its parameters to the range [-0.01, 0.01] to avoid gradient vanishing; then the parameters of the resource allocation discrimination module are fixed, and the loss function of equation (12) is minimized. This updates the parameters of the resource allocation generation module. Through continuous iteration, until the loss function is reached... and The convergence process is completed, and the parameters of the resource allocation generation module are saved for future base station resource allocation matrix prediction.

[0122] Then, the training set corresponding to the base station resource allocation strategy obtained in the above embodiment is input into the trained base station resource allocation strategy prediction model to generate the base station resource allocation matrix at future time. RMSE = 0.0183 and MR = 0.151 are calculated according to equations (13) and (14), thus reflecting the good predictive ability of the base station resource allocation strategy prediction model.

[0123] In this embodiment of the invention, the load feature matrix vectors corresponding to each base station can first be obtained based on the base station resource sharing map corresponding to the base stations in a preset area, the current load of each base station, and the pre-trained GCN network. Then, based on the base station resource sharing map, the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station of that base station are obtained from the obtained load feature matrix vectors corresponding to each base station. Then, based on the load feature matrix vectors of any base station and the load feature matrix vectors of any shareable base station, the sharing party, the sharing party, and the sharing ratio when the base station shares resources with the shareable base station are determined. Based on the sharing party, the sharing party, the sharing ratio, and the resource complementarity value between the base station and the shareable base station, the resource allocation strategy corresponding to the base station is determined, and resources are allocated to the base station based on the resource allocation strategy. The resource complementarity value between the base station and the shareable base station is determined based on the base station resource sharing map, the current load and optimal load of the base station, and the current load and optimal load of the shareable base station.

[0124] As can be seen from the above embodiments, when allocating base station resources, the embodiments of the present invention can first determine the base stations that can share resources, and further determine the sharing party, the sharing recipient, and the sharing ratio. Then, the resource allocation strategy for the base stations can be determined based on the determined sharing party, the sharing recipient, the sharing ratio, and the resource complementarity value between the base stations. Since the resource sharing between base stations can be comprehensively determined by combining the resource complementarity value between the base stations, the sharing direction, and the sharing ratio, a reasonable base station resource allocation strategy can be quickly determined, effectively reducing the number of sharing interactions between base stations, thereby effectively improving resource allocation efficiency.

[0125] This invention further proposes a base station resource allocation prediction scheme to address the problem that the lack of sufficient information to predict the relationship between base stations in the future, due to the limited spatial distance and resource sharing direction between base stations, results in low accuracy of base station resource allocation prediction results, thereby achieving a more reasonable allocation of future base station resources.

[0126] Corresponding to the above-described method for allocating base station resources, this embodiment of the invention also provides a device for allocating base station resources. Figure 4 This is a schematic diagram of the module composition of the base station resource allocation device 400 provided in an embodiment of the present invention, as shown below. Figure 4 As shown, the base station resource allocation device 400 includes:

[0127] The first acquisition module 401 is used to acquire a load feature matrix vector corresponding to each base station based on the base station resource shareable graph corresponding to the base stations in the preset area, the current load of each base station, and a pre-trained graph convolutional neural network (GCN). The base station resource shareable graph is determined based on the location of the base stations in the preset area and is used to identify base stations that are close to each other in the preset area. The load feature matrix vector is used to characterize the load information of the neighboring base stations of the corresponding base station in the preset area.

[0128] The second acquisition module 402 is used to acquire, based on the base station resource shareable graph, the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station of any base station from the acquired load feature matrix vectors corresponding to each base station.

[0129] The first determining module 403 is used to determine the sharing party, the sharing recipient, and the sharing ratio when the any base station and the any shareable base station share resources, based on the load feature matrix vector of the any base station and the load feature matrix vector of the any shareable base station.

[0130] The second determining module 404 is used to determine the resource allocation strategy corresponding to any base station based on the sharing party, the sharing recipient, and the sharing ratio when any base station shares resources with any shareable base station, and the resource complementarity value between any base station and any shareable base station, so as to allocate resources to any base station based on the resource allocation strategy; wherein, the resource complementarity value between any base station and any shareable base station is determined based on the base station resource shareability map, the current load and optimal load of any base station, and the current load and optimal load of any shareable base station.

[0131] Optionally, the device further includes ( Figure 4 (Not shown in the image):

[0132] Before obtaining the load feature matrix corresponding to each base station based on the base station resource sharing graph corresponding to the base station in the preset area, the current load of each base station, and the pre-trained graph convolutional neural network (GCN), the spatial distance between each pair of base stations in the preset area is obtained based on the location of the base stations in the preset area.

[0133] Based on the obtained spatial distance between each pair of base stations and the preset resource-sharing distance threshold, the resource base station sharing map is constructed.

[0134] The first determining module 403 is used for:

[0135] The load feature matrix vector of any base station and the load feature matrix vector of any shareable base station are concatenated in sequence to obtain the first concatenated matrix vector corresponding to the concatenation of any base station with the shareable base station. The first concatenated matrix vector is then input into a pre-built feedforward neural network to obtain the first feedforward network value output by the feedforward neural network corresponding to the concatenation of any base station with the shareable base station.

[0136] The load feature matrix vector of any shareable base station and the load feature matrix vector of any base station are concatenated in sequence to obtain the second concatenated matrix vector corresponding to the concatenation of any base station with any other base station. The second concatenated matrix vector is then input into the feedforward neural network to obtain the second feedforward network value output by the feedforward neural network corresponding to the concatenation of any base station with any other base station.

[0137] When the first feedforward network value is not less than the second feedforward network value, any base station is determined as the sharing party when sharing resources with any shareable base station, and any shareable base station is determined as the sharing recipient when sharing resources with any shareable base station; when the first feedforward network value is less than the second feedforward network value, any shareable base station is determined as the sharing party when sharing resources with any shareable base station, and any base station is determined as the sharing recipient when sharing resources with any shareable base station.

[0138] Optionally, the first determining module 403 is further configured to:

[0139] The load feature matrix vector of any base station and the load feature matrix vector of each shareable base station of any base station are concatenated in a sequential manner to obtain the first concatenated matrix vector corresponding to the concatenation of each shareable base station by any base station. The first concatenated matrix vector corresponding to the concatenation of each shareable base station by any base station is then input into the feedforward neural network to obtain the first feedforward network value output by the feedforward neural network corresponding to the concatenation of each shareable base station.

[0140] The sharing ratio when any base station shares resources with any shareable base station is determined by multiplying the first feedforward network value corresponding to any base station when splicing any base station to any shareable base station with the resource complementarity value between any base station and any shareable base station, and by multiplying the first feedforward network value corresponding to any base station when splicing any shareable base station with each shareable base station with the resource complementarity value between any base station and each shareable base station.

[0141] Optionally, the first acquisition module 401 is used to:

[0142] Based on the current load of each base station, obtain the current load of the shareable base stations shown in the base station resource shareable diagram;

[0143] Based on the current load of the shareable base stations shown in the base station resource shareability diagram, construct an initial load feature matrix;

[0144] The initial load feature matrix is ​​transposed to obtain the transposed matrix;

[0145] The transpose matrix is ​​input into the pre-trained GCN network to obtain the load feature matrix vectors corresponding to each base station output by the GCN network after processing the transpose matrix, the pre-trained weight matrix, the fusion matrix, and the degree matrix of the fusion matrix through the activation function; wherein, the fusion matrix is ​​obtained by fusing the transpose matrix with a preset identity matrix.

[0146] Optionally, the device further includes ( Figure 4 (Not shown in the image):

[0147] After determining the resource allocation strategy corresponding to any base station, a resource allocation prediction strategy corresponding to any base station is obtained through a pre-trained Generative Adversarial Network (GAN); wherein, the pre-trained GAN includes a resource allocation generation module and a resource allocation discrimination module.

[0148] Optionally, the resource allocation generation module includes a standard convolutional kernel and a dilated convolutional kernel, wherein the standard convolutional kernel is used to extract short-term temporal features, and the dilated convolutional kernel is used to extract periodic features.

[0149] In this embodiment of the invention, the load feature matrix vectors corresponding to each base station can first be obtained based on the base station resource sharing map corresponding to the base stations in a preset area, the current load of each base station, and the pre-trained GCN network. Then, based on the base station resource sharing map, the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station of that base station are obtained from the obtained load feature matrix vectors corresponding to each base station. Then, based on the load feature matrix vectors of any base station and the load feature matrix vectors of any shareable base station, the sharing party, the sharing party, and the sharing ratio when the base station shares resources with the shareable base station are determined. Based on the sharing party, the sharing party, the sharing ratio, and the resource complementarity value between the base station and the shareable base station, the resource allocation strategy corresponding to the base station is determined, and resources are allocated to the base station based on the resource allocation strategy. The resource complementarity value between the base station and the shareable base station is determined based on the base station resource sharing map, the current load and optimal load of the base station, and the current load and optimal load of the shareable base station.

[0150] As can be seen from the above embodiments, when allocating base station resources, the embodiments of the present invention can first determine the base stations that can share resources, and further determine the sharing party, the sharing recipient, and the sharing ratio. Then, the resource allocation strategy for the base stations can be determined based on the determined sharing party, the sharing recipient, the sharing ratio, and the resource complementarity value between the base stations. Since the resource sharing between base stations can be comprehensively determined by combining the resource complementarity value between the base stations, the sharing direction, and the sharing ratio, a reasonable base station resource allocation strategy can be quickly determined, effectively reducing the number of sharing interactions between base stations, thereby effectively improving resource allocation efficiency.

[0151] This invention further proposes a base station resource allocation prediction scheme to address the problem that the lack of sufficient information to predict the relationship between base stations in the future, due to the limited spatial distance and resource sharing direction between base stations, results in low accuracy of base station resource allocation prediction results, thereby achieving a more reasonable allocation of future base station resources.

[0152] Corresponding to the above-described method for allocating base station resources, this embodiment of the invention also provides a device for allocating base station resources. Figure 5 This is a schematic diagram of the hardware structure of a base station resource allocation device provided in one embodiment of the present invention.

[0153] The base station resource allocation device can be the terminal device or server provided in the above embodiments for allocating base station resources.

[0154] The base station resource allocation device can vary significantly due to differences in configuration or performance. It may include one or more processors 501 and memory 502, with memory 502 storing one or more application programs or data. Memory 502 can be temporary or persistent storage. The application programs stored in memory 502 may include one or more modules (not shown in the figure), each module including a series of computer-executable instructions for the base station resource allocation device. Furthermore, processor 501 may be configured to communicate with memory 502 and execute the series of computer-executable instructions in memory 502 on the base station resource allocation device. The base station resource allocation device may also include one or more power supplies 503, one or more wired or wireless network interfaces 504, one or more input / output interfaces 505, and one or more keyboards 506.

[0155] Specifically, in this embodiment, the base station resource allocation device includes a memory and one or more programs, wherein one or more programs are stored in the memory, and one or more programs may include one or more modules, and each module may include a series of computer-executable instructions for the base station resource allocation device, and is configured to be executed by one or more processors as described above.

[0156] In this embodiment of the invention, the load feature matrix vectors corresponding to each base station can first be obtained based on the base station resource sharing map corresponding to the base stations in a preset area, the current load of each base station, and the pre-trained GCN network. Then, based on the base station resource sharing map, the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station of that base station are obtained from the obtained load feature matrix vectors corresponding to each base station. Then, based on the load feature matrix vectors of any base station and the load feature matrix vectors of any shareable base station, the sharing party, the sharing party, and the sharing ratio when the base station shares resources with the shareable base station are determined. Based on the sharing party, the sharing party, the sharing ratio, and the resource complementarity value between the base station and the shareable base station, the resource allocation strategy corresponding to the base station is determined, and resources are allocated to the base station based on the resource allocation strategy. The resource complementarity value between the base station and the shareable base station is determined based on the base station resource sharing map, the current load and optimal load of the base station, and the current load and optimal load of the shareable base station.

[0157] As can be seen from the above embodiments, when allocating base station resources, the embodiments of the present invention can first determine the base stations that can share resources, and further determine the sharing party, the sharing recipient, and the sharing ratio. Then, the resource allocation strategy for the base stations can be determined based on the determined sharing party, the sharing recipient, the sharing ratio, and the resource complementarity value between the base stations. Since the resource sharing between base stations can be comprehensively determined by combining the resource complementarity value between the base stations, the sharing direction, and the sharing ratio, a reasonable base station resource allocation strategy can be quickly determined, effectively reducing the number of sharing interactions between base stations, thereby effectively improving resource allocation efficiency.

[0158] This invention further proposes a base station resource allocation prediction scheme to address the problem that the lack of sufficient information to predict the relationship between base stations in the future, due to the limited spatial distance and resource sharing direction between base stations, results in low accuracy of base station resource allocation prediction results, thereby achieving a more reasonable allocation of future base station resources.

[0159] Another embodiment of this disclosure also provides a computer-readable storage medium for storing computer-executable instructions that, when executed by a processor, implement the above-described process.

[0160] The storage medium in this embodiment can implement the various processes of the above-described base station resource allocation method embodiment and achieve the same effect and function, which will not be repeated here.

[0161] In various embodiments of this disclosure, the computer-readable storage medium includes read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, etc.

[0162] In the 1990s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many methodological improvements today can be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that a methodological improvement cannot be implemented using a hardware physical module. For example, a Programmable Logic Device (PLD) (e.g., a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program a digital system themselves to "integrate" it onto a PLD, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Furthermore, nowadays, instead of manually manufacturing integrated circuit chips, this programming is mostly implemented using "logic compiler" software. Similar to the software compiler used in program development, the original code before compilation must be written in a specific programming language, called a Hardware Description Language (HDL). There are many HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, and RHDL (Ruby Hardware Description Language). Currently, the most commonly used are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should understand that by simply performing some logic programming on the method flow using one of these hardware description languages ​​and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.

[0163] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0164] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0165] For ease of description, the above apparatus is described by dividing it into various functional units. Of course, in implementing the embodiments of this disclosure, the functions of each unit can be implemented in one or more software and / or hardware.

[0166] Those skilled in the art will understand that one or more embodiments of this disclosure can be provided as a method, system, or computer program product. Therefore, one or more embodiments of this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, one or more embodiments of this disclosure can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0167] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0168] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0169] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0170] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0171] One or more embodiments of this disclosure can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a particular task or implement a particular abstract data type. One or more embodiments of this disclosure can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules can reside in local and remote computer storage media, including storage devices.

[0172] The various embodiments in this disclosure are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0173] The above description is merely an embodiment of this disclosure and is not intended to limit the scope of this disclosure. Various modifications and variations can be made to this disclosure by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of the claims of this disclosure.

Claims

1. A method for allocating base station resources, characterized in that, The method includes: Based on the base station resource sharing graph corresponding to the base stations in the preset area, the current load of each base station, and the pre-trained graph convolutional neural network (GCN), the load feature matrix vector corresponding to each base station is obtained; wherein, the base station resource sharing graph is determined according to the location of the base stations in the preset area and is used to identify base stations that are close to each other in the preset area; the load feature matrix vector is used to characterize the load information of the corresponding base station in the preset area of ​​its neighboring base stations. Based on the base station resource shareability graph, the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station of any base station are obtained from the load feature matrix vectors corresponding to each base station. Based on the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station, determine the sharing party, the sharing recipient, and the sharing ratio when any base station shares resources with any shareable base station. Based on the sharing party, the sharing recipient, and the sharing ratio when any base station shares resources with any shareable base station, and the resource complementarity value between any base station and any shareable base station, a resource allocation strategy corresponding to any base station is determined, and resources are allocated to any base station based on the resource allocation strategy; wherein, the resource complementarity value between any base station and any shareable base station is determined based on the base station resource shareability map, the current load and optimal load of any base station, and the current load and optimal load of any shareable base station; Based on the load characteristic matrix vector of any base station and the load characteristic matrix vector of any shareable base station, determine the sharing ratio when any base station shares resources with any shareable base station, including: The load feature matrix vector of any base station and the load feature matrix vector of each shareable base station of any base station are concatenated in sequence to obtain the first concatenated matrix vector corresponding to the concatenation of each shareable base station by any base station. The first concatenated matrix vector corresponding to the concatenation of each shareable base station by any base station is then input into a feedforward neural network to obtain the first feedforward network value output by the feedforward neural network corresponding to the concatenation of each shareable base station. The sharing ratio when any base station shares resources with any shareable base station is determined by multiplying the first feedforward network value corresponding to any base station when splicing any base station to any shareable base station with the resource complementarity value between any base station and any shareable base station, and by multiplying the first feedforward network value corresponding to any base station when splicing any shareable base station with each shareable base station with the resource complementarity value between any base station and each shareable base station.

2. The method according to claim 1, characterized in that, Before obtaining the load feature matrix corresponding to each base station based on the base station resource shareable graph corresponding to the base stations in the preset area, the current load of each base station, and the pre-trained graph convolutional neural network (GCN), the method further includes: Based on the location of the base stations within the preset area, obtain the spatial distance between any two base stations within the preset area; Based on the obtained spatial distance between each pair of base stations and the preset resource-sharing distance threshold, the base station resource-sharing graph is constructed.

3. The method according to claim 1, characterized in that, Based on the load characteristic matrix of any base station and the load characteristic matrix of any shareable base station, determine the sharing party and the sharing recipient when any base station shares resources with any shareable base station, including: The load feature matrix vector of any base station and the load feature matrix vector of any shareable base station are concatenated in sequence to obtain the first concatenated matrix vector corresponding to the concatenation of any base station with the shareable base station. The first concatenated matrix vector is then input into a pre-built feedforward neural network to obtain the first feedforward network value output by the feedforward neural network corresponding to the concatenation of any base station with the shareable base station. The load feature matrix vector of any shareable base station and the load feature matrix vector of any base station are concatenated in sequence to obtain the second concatenated matrix vector corresponding to the concatenation of any base station with any other base station. The second concatenated matrix vector is then input into the feedforward neural network to obtain the second feedforward network value output by the feedforward neural network corresponding to the concatenation of any base station with any other base station. When the first feedforward network value is not less than the second feedforward network value, any base station is determined as the sharing party when sharing resources with any shareable base station, and any shareable base station is determined as the sharing recipient when sharing resources with any shareable base station; when the first feedforward network value is less than the second feedforward network value, any shareable base station is determined as the sharing party when sharing resources with any shareable base station, and any base station is determined as the sharing recipient when sharing resources with any shareable base station.

4. The method according to claim 1, characterized in that, The step of obtaining the load feature matrix vector corresponding to each base station based on the base station resource shareable graph corresponding to the base stations in the preset area, the current load of each base station, and the pre-trained graph convolutional neural network (GCN) includes: Based on the current load of each base station, obtain the current load of the shareable base stations shown in the base station resource shareable diagram; Based on the current load of the shareable base stations shown in the base station resource shareability diagram, construct an initial load feature matrix; The initial load feature matrix is ​​transposed to obtain the transposed matrix; The transpose matrix is ​​input into the pre-trained GCN network to obtain the load feature matrix vectors corresponding to each base station output by the GCN network after processing the transpose matrix, the pre-trained weight matrix, the fusion matrix, and the degree matrix of the fusion matrix through the activation function; wherein, the fusion matrix is ​​obtained by fusing the transpose matrix with a preset identity matrix.

5. The method according to claim 1, characterized in that, After determining the resource allocation strategy corresponding to any of the base stations, the method further includes: The resource allocation prediction strategy corresponding to any base station is obtained through a pre-trained generative adversarial network (GAN); wherein the pre-trained GAN includes a resource allocation generation module and a resource allocation discrimination module.

6. The method according to claim 5, characterized in that, The resource allocation generation module includes a standard convolutional kernel and a dilated convolutional kernel. The standard convolutional kernel is used to extract short-term temporal features, and the dilated convolutional kernel is used to extract periodic features.

7. A base station resource allocation device, characterized in that, The device includes: The first acquisition module is used to acquire a load feature matrix vector corresponding to each base station based on the base station resource shareable graph corresponding to the base stations in the acquired preset area, the current load of each base station, and a pre-trained graph convolutional neural network (GCN). The base station resource shareable graph is determined based on the location of the base stations in the preset area and is used to identify base stations that are close to each other in the preset area. The load feature matrix vector is used to characterize the load information of the neighboring base stations of the corresponding base station in the preset area. The second acquisition module is used to acquire, based on the base station resource shareability graph, the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station of any base station from the acquired load feature matrix vectors corresponding to each base station. The first determining module is used to determine the sharing party, the sharing recipient, and the sharing ratio when any base station shares resources with any shareable base station based on the load feature matrix vector of any base station and the load feature matrix vector of any shareable base station. The second determining module is used to determine the resource allocation strategy corresponding to any base station based on the sharing party, the sharing recipient, and the sharing ratio when any base station shares resources with any shareable base station, and the resource complementarity value between any base station and any shareable base station, so as to allocate resources to any base station based on the resource allocation strategy; wherein, the resource complementarity value between any base station and any shareable base station is determined based on the base station resource shareability map, the current load and optimal load of any base station, and the current load and optimal load of any shareable base station; The first determining module is further configured to perform matrix-vector concatenation on the load feature matrix vector of any base station and the load feature matrix vector of each shareable base station of any base station in a sequential order, to obtain the first concatenation matrix vector corresponding to the concatenation of each shareable base station by any base station, and input the first concatenation matrix vector corresponding to the concatenation of each shareable base station by any base station into the feedforward neural network to obtain the first feedforward network value output by the feedforward neural network corresponding to the concatenation of each shareable base station; The sharing ratio when any base station shares resources with any shareable base station is determined by multiplying the first feedforward network value corresponding to any base station when splicing any base station to any shareable base station with the resource complementarity value between any base station and any shareable base station, and by multiplying the first feedforward network value corresponding to any base station when splicing any shareable base station with each shareable base station with the resource complementarity value between any base station and each shareable base station.

8. A base station resource allocation device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store computer-executable instructions that, when executed by a processor, implement the steps of the method described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the method described in any one of claims 1 to 6.

Images