Resource scheduling methods, devices, electronic equipment, products and storage media

Abstract

This invention provides a resource scheduling method, apparatus, electronic device, product, and storage medium, relating to the field of mobile communication technology. The method includes: determining the structural features of a user graph and the feature fusion result between the node feature set of the user graph; clustering a target node and multiple user nodes based on the feature fusion result to obtain a target node cluster, the target node cluster including the target node and a set of neighboring nodes, the neighboring node set including several user nodes similar to the target node; updating the initial CQI value corresponding to the target node in the target node cluster based on the initial CQI value corresponding to each neighboring node in the neighboring node set of the target node cluster, obtaining an updated CQI value; and performing resource scheduling operations on the target terminal based on the CQI value. This invention can solve the problem of the specific lag in the CQI value sent by the user in the prior art, accurately reflecting the real-time wireless channel quality, thereby improving the accuracy of downlink resource scheduling.

Description

Technical Field

[0001] This invention relates to the field of mobile communication technology, and in particular to a resource scheduling method, apparatus, electronic device, product, and storage medium. Background Technology

[0002] In 5G systems, base stations transmit system messages or user data through shared channels, meaning that time-frequency resources are dynamically shared among users. Therefore, base stations need a scheduler to allocate uplink and downlink time-frequency resources to ensure system throughput and resource fairness, and to improve network performance and system capacity. The basic working principle of the scheduler includes: first, determining the scheduling priority of the bearer and selecting the users to be scheduled based on the scheduling input information; then, selecting a suitable MCS (Modulation and Coding Scheme) for each user; and finally, allocating resources based on the user data volume and the MCS selected by the base station for each user.

[0003] The initial MCS selected by the base station for each user is determined based on the channel quality measured and reported by the terminal. Channel quality is indicated by the CQI (Channel Quality Indicator) value, which is measured and calculated based on CSI-RS (Channel State Information Reference Signal). However, existing resource allocation schemes have some drawbacks. When the subcarrier spacing in 5G is set to 30kHz, the base station's scheduling period is 0.5ms, while the CSI-RS period is typically configured to 80ms, resulting in a significant time difference between the two periods. This leads to a lag in the CQI value reported by the user based on CSI-RS measurements at the scheduling time, failing to accurately reflect the real-time wireless channel quality and thus affecting the accuracy of downlink resource scheduling. Summary of the Invention

[0004] This invention provides a resource scheduling method, apparatus, electronic device, product, and storage medium to address the shortcomings of existing technologies where the CQI values ​​reported by users based on CSI-RS measurements during scheduling are lagging and cannot accurately reflect real-time wireless channel quality, thus affecting the accuracy of downlink resource scheduling.

[0005] This invention provides a resource scheduling method, comprising:

[0006] In response to a resource scheduling request sent to a target terminal, a feature fusion result is determined between the structural features of the scheduling user graph and the node feature set of the scheduling user graph. The structural features include topological features between the target node and multiple user nodes, and the node feature set includes the node features of the target node and the node features of the multiple user nodes.

[0007] Based on the feature fusion results, the target node and multiple user nodes are clustered to obtain a target node cluster. The target node cluster includes the target node and a set of neighboring nodes. The set of neighboring nodes includes several user nodes that are similar to the target node.

[0008] Based on the initial CQI value corresponding to each neighboring node in the neighboring node set of the target node cluster, the initial CQI value corresponding to the target node in the target node cluster is updated to obtain the updated CQI value;

[0009] Based on the CQI value, resource scheduling operations are performed on the target terminal.

[0010] According to the present invention, a resource scheduling method is provided, wherein the topological features include the weights of the edges between the target node and each user node among a plurality of user nodes, and the method further includes, before determining the feature fusion result between the structural features of the scheduling user graph and the node feature set:

[0011] Based on the latitude difference between the target node and each of the multiple user nodes and the longitude difference between the target node and each of the multiple user nodes, the spatial distance between the target node and each of the multiple user nodes is determined.

[0012] The weights of the edges between the target node and each of the multiple user nodes are determined based on the reciprocal of the spatial distance between the target node and each of the multiple user nodes.

[0013] According to the present invention, a resource scheduling method is provided, wherein the node characteristics of the target node and the node characteristics of multiple user nodes are determined based on the following method:

[0014] The multi-source information parameters uploaded by the target terminal and multiple user terminals are obtained respectively. The multi-source information parameters include primary service area measurement parameters and neighboring cell measurement parameters. The primary service area measurement parameters include channel state information reference signal resource identifier, rank identifier and precoding matrix identifier. The neighboring cell measurement parameters include neighboring cell identifier.

[0015] The multi-source information parameters uploaded by the target terminal are determined as the node characteristics of the target node;

[0016] The multi-source information parameters uploaded by the multiple user terminals are determined as the node characteristics of the multiple user nodes.

[0017] According to the present invention, a resource scheduling method is provided, which constructs a scheduling user subgraph based on the target node and the set of neighboring nodes in the target node cluster. The step of updating the initial CQI value corresponding to the target node in the target node cluster based on the initial CQI value corresponding to each neighboring node in the set of neighboring nodes in the target node cluster to obtain the updated CQI value includes:

[0018] The initial CQI value corresponding to each neighboring node in the neighboring node set of the target node cluster is encoded to obtain the encoded node feature of each neighboring node in the neighboring node set. The initial CQI value corresponding to the target node in the target node cluster is encoded to obtain the encoded node feature of the target node.

[0019] A graph convolution operation is performed on the encoded node features of each neighboring node in the neighboring node set and the encoded node features of the target node to obtain a graph convolution fusion feature set. Each graph convolution fusion feature in the graph convolution fusion feature set corresponds one-to-one with each node in the scheduling user subgraph.

[0020] The graph convolutional fusion features of the target node in the target node cluster and the graph convolutional fusion features of each neighboring node in the neighboring node set are aggregated to obtain the aggregated node features of the target node.

[0021] The aggregated node features of the target node are encoded and converted to obtain the updated CQI value of the target node.

[0022] According to the present invention, a resource scheduling method is provided, wherein the aggregation of graph convolutional fusion features of the target node in the target node cluster and graph convolutional fusion features of each neighboring node in the neighboring node set to obtain aggregated node features of the target node includes:

[0023] Calculate the similarity between the graph convolutional fusion feature of the target node in the target node cluster and the graph convolutional fusion feature of each neighboring node in the neighboring node set, and determine the weight of the edge between the target node and each neighboring node in the neighboring node set based on the calculation result;

[0024] The weights of the edges between the target node and each neighboring node in the neighboring node set are sorted, and the sub-neighboring node set is determined based on the sorting results;

[0025] The graph convolutional fusion features of the target node and the graph convolutional fusion features of each neighboring node in the sub-neighboring node set are aggregated to determine the aggregated node features of the target node.

[0026] According to the present invention, a resource scheduling method is provided, wherein before performing resource scheduling operation on the target terminal based on the CQI value, the method further includes:

[0027] The CQI value is determined by comparing the spectral efficiency corresponding to the initial CQI value of the target node with the spectral efficiency of the updated CQI value of the target node.

[0028] The present invention also provides a resource scheduling device, comprising:

[0029] The feature fusion module is used to respond to a resource scheduling request sent to the target terminal and determine the feature fusion result between the structural features of the scheduling user graph and the node feature set of the scheduling user graph. The structural features include the topological features between the target node and multiple user nodes, and the node feature set includes the node features of the target node and the node features of the multiple user nodes.

[0030] The clustering module is used to cluster the target node and multiple user nodes based on the feature fusion result to obtain a target node cluster. The target node cluster includes the target node and a set of neighboring nodes. The set of neighboring nodes includes several user nodes similar to the target node.

[0031] The CQI value update module is used to update the initial CQI value corresponding to the target node in the target node cluster based on the initial CQI value corresponding to each neighboring node in the neighboring node set in the target node cluster, so as to obtain the updated CQI value.

[0032] The resource scheduling module is used to perform resource scheduling operations on the target terminal based on the CQI value.

[0033] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the resource scheduling methods described above.

[0034] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the resource scheduling method as described above.

[0035] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements any of the resource scheduling methods described above.

[0036] The resource scheduling method, apparatus, electronic device, product, and storage medium provided by this invention, in response to a resource scheduling request sent to a target terminal, determine the feature fusion result between the structural features of the scheduling user graph and the node feature set of the scheduling user graph. The structural features include topological features between the target node and multiple user nodes, and the node feature set includes the node features of the target node and the node features of the multiple user nodes. Based on the feature fusion result, the target node and the multiple user nodes are clustered to obtain a target node cluster. The target node cluster includes the target node and a set of neighboring nodes, and the set of neighboring nodes includes several user nodes similar to the target node. Based on the initial CQI value corresponding to each neighboring node in the set of neighboring nodes in the target node cluster, the initial CQI value corresponding to the target node in the target node cluster is updated to obtain an updated CQI value. Based on the CQI value, a resource scheduling operation is performed on the target terminal. This invention constructs a scheduling user graph and performs graph convolution operations to fully fuse the topological and node features among users, resulting in richer and more accurate features for each node. By clustering the fused scheduling user graph, neighboring nodes more similar to the target node are selected, facilitating the selection of more suitable neighboring nodes for CQI value updates and avoiding the uncertainty of information from a single node. Updating the target node's CQI value using the CQI values ​​of neighboring nodes significantly reduces the CQI value lag problem caused by excessively long CSI-RS measurement cycles. Based on the updated CQI value, the base station can more accurately select appropriate MCS and time-frequency resources for allocation, thereby improving the accuracy of downlink resource scheduling. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in this invention 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0038] Figure 1 This is a flowchart illustrating the resource scheduling method provided by the present invention.

[0039] Figure 2 This is a schematic diagram of the resource scheduling device provided by the present invention.

[0040] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0042] In existing 5G systems, base stations transmit system messages or user data through shared channels, and time-frequency domain resources are dynamically shared among users. Through long-term research and practical application, the patent applicant has discovered significant shortcomings in the resource scheduling of existing 5G systems. Although base stations can select appropriate MCS (Modulation and Coding Scheme) using CQI (Channel Quality Indicator) values ​​to achieve uplink and downlink time-frequency resource allocation and ensure system throughput and resource fairness, this method suffers from severe lag. Specifically, existing technologies rely on CSI-RS (Channel State Information Reference Signal) measurements to report CQI values. However, the measurement period of CSI-RS is typically long, while the scheduling period of base stations is relatively short, causing the CQI value to fail to reflect the current channel quality in real time, thus affecting the accuracy and efficiency of resource allocation.

[0043] After in-depth research into existing technologies, the patent applicant found that some existing solutions attempt to shorten the CQI update cycle by increasing the CSI-RS measurement frequency in an attempt to reduce lag. While this method can improve the real-time performance of CQI values ​​to some extent, it significantly increases the system's measurement overhead and channel resource usage, leading to a decrease in overall system efficiency. Other solutions use prediction algorithms to estimate the user's CQI value to compensate for the time difference in the measurement cycle. However, these prediction algorithms have low accuracy and large prediction errors in real-world complex wireless environments, failing to guarantee the accuracy of CQI values ​​and thus affecting scheduling decisions.

[0044] To address the above problems, the present invention proposes the following embodiments.

[0045] Figure 1 This is a flowchart illustrating the resource scheduling method provided by the present invention, as follows: Figure 1 As shown, the method includes the following:

[0046] Step 110: In response to the resource scheduling request sent to the target terminal, determine the feature fusion result between the structural features of the scheduling user graph and the node feature set of the scheduling user graph. The structural features include the topological features between the target node and multiple user nodes, and the node feature set includes the node features of the target node and the node features of the multiple user nodes.

[0047] Here, a resource scheduling request refers to a request issued by a user terminal, requesting the base station to allocate uplink and downlink time-frequency resources for data transmission; the scheduling user graph represents the graph structure of the users participating in the scheduling and their connection relationships. The target node is the terminal that issues the resource scheduling request, user nodes are other user terminals participating in the scheduling, and edges represent the connection relationships between nodes; structural features include the target node and multiple user nodes, as well as the topological features between nodes, which are used to represent the relative positions and relationship strength between users, such as edge weights; the node feature set contains attribute information describing the individual user, which in this case refers to channel state information, reference signal resource identifier, rank identifier, precoding matrix identifier, and neighbor cell identifier, etc.; the feature fusion result refers to fusing the structural features and node feature set through graph convolution operations to generate a feature matrix containing comprehensive information. Each node feature in this feature matrix contains richer feature information, which is obtained by fusing the structural features and node features.

[0048] In one embodiment, in response to a resource scheduling request sent by a target terminal, indicating that the terminal needs to transmit data and the base station needs to allocate appropriate time-frequency resources for it, the feature fusion result between the structural features and the node feature set of the scheduling user graph is determined. The feature fusion result is generated by fusing the structural features and node features of the scheduling user graph through graph convolution operations to generate a feature matrix. This process combines the connection relationships between nodes (structural features) and the attributes of the nodes themselves (node ​​features), so that the generated features contain richer information and can more comprehensively reflect the state and interrelationships of each node. The base station can more accurately understand the real-time status of users and thus make more reasonable resource scheduling decisions. The specific construction process of the scheduling user graph can be referred to in the following embodiments, and will not be repeated here.

[0049] Step 120: Based on the feature fusion result, cluster the target node and multiple user nodes to obtain a target node cluster. The target node cluster includes the target node and a set of neighboring nodes. The set of neighboring nodes includes several user nodes similar to the target node.

[0050] Here, clustering refers to the process of grouping nodes with similar characteristics. In this step, the clustering algorithm divides nodes into several clusters based on their feature similarity, minimizing the intra-cluster distance and maximizing the inter-cluster distance. The cluster of the target node represents the set of the target node and several user nodes with similar features; the neighboring node set refers to the set of several user nodes similar to the target node within the target node cluster, where these neighboring nodes are relatively close to the target node in the feature space.

[0051] It should be noted that in this embodiment of the invention, only the processing procedure of the target node is described in detail. The processing procedures of other nodes are the same as those of the target node and should be included in the protection scope of this invention.

[0052] In one embodiment, based on the feature fusion results, the target node and multiple user nodes are clustered. Clustering is the process of grouping nodes with similar features. Clustering algorithms typically group nodes with similar features into the same cluster based on the feature similarity between nodes. The target node cluster refers to the cluster containing the target node. Through feature fusion and clustering, the actual similarity between users can be better reflected. Similar users are grouped into the same cluster, which allows for more accurate resource scheduling. Clustering helps identify user nodes with similar features to the target node; these nodes are usually in the same or similar wireless environments. Based on this information, the CQI value of the target node can be adjusted, thereby optimizing resource allocation.

[0053] It should be noted that there are various clustering methods, including but not limited to K-Means clustering, spectral clustering, and hierarchical clustering. Other clustering methods different from the embodiments of this invention should be regarded as equivalent substitutions and do not depart from the protection scope of this invention.

[0054] It should be noted that in this embodiment of the invention, only the operation of the target node cluster containing the target node is described in detail. The processing of other node clusters is the same as that of the target node cluster and should be included in the protection scope of this invention.

[0055] For example, the structural features and node feature sets in the scheduling user graph are fused through graph convolution operations to obtain a feature fusion result; then, spectral clustering is used to cluster the nodes in the feature fusion result to achieve user clustering and determine the target node cluster containing the target node; specifically as follows:

[0056] Suppose we need to divide n users into m clusters, and the user scheduling graph is represented by G. Let V represent the set of nodes in G. Let X represent the node features of each node in G, i.e., X represents the node feature set of the scheduling user graph. Since the node features are obtained by combining multi-source measurement information such as channel state information, reference signal resource identifier, rank identifier, precoding matrix identifier, and neighbor cell identifier, the node features have multi-dimensional feature information, expressed by the formula: in Represents a node The s-dimensional eigenvectors of G. Let E represent the set of edges in G. Let the adjacency matrix A represent the structural features of G. , Representing an edge The weight of the edge between two nodes is expressed as the reciprocal of the spatial distance between them when there is a connection between them. When there is no connection between two nodes (i.e., no edge exists between them), the edge has no weight. The specific formula is as follows: Let C represent the set of cluster labels to which each user in V belongs, then Furthermore, the value range of the cluster label to which each user belongs is from 0 to m-1.

[0057] The purpose of spectral clustering is to cluster the scheduling user graph, minimizing intra-cluster distances and maximizing inter-cluster distances. Therefore, the iteration terminates only if one of the following two conditions is met: first, the intra-cluster distances calculated in this iteration no longer decrease compared to the previous iteration, and the inter-cluster distances no longer increase; second, the number of iterations in this iteration is greater than m. The specific algorithm flow for clustering includes input data, processed data, and output data; where the input data includes a node set V, an adjacency matrix A, and a node feature set X; the output data includes a node label set C; the specific steps for processing the data are as follows:

[0058] Step 1201: Let the current iteration number t=0, and define the degree matrix of A as D. From this, we obtain the Laplacian matrix of the symmetric normalized graph. , where I represents the identity matrix;

[0059] Step 1202, repeat the following operations until... and , or t>m;

[0060] (1) Let t = t + 1, and assume that the convolution order of the graph is equal to t;

[0061] (2) Perform k-order graph convolution on the scheduling user graph to fuse the structural features and node feature sets of the scheduling user graph, thereby obtaining the feature matrix, i.e., the feature fusion result. The feature matrix is ​​expressed by the following formula: ;

[0062] (3) Calculate the similarity matrix between nodes Then, symmetry is performed to obtain a symmetric and non-negative similarity matrix. ,in This means taking the absolute value of each element in K;

[0063] (4) Perform spectral clustering on W, that is, calculate the eigenvectors corresponding to the first m largest eigenvalues ​​of W, form an N×m matrix from these m eigenvectors, where each row is regarded as a vector in m-dimensional space, and cluster each row using the K-Means clustering algorithm to obtain C;

[0064] (5) Calculate intra-cluster distance and inter-cluster distance ;

[0065] in This represents the average of the intra-cluster node distances for each cluster, where... for The i-th row vector. The specific formula is:

[0066]

[0067] in The average distance between the centroids of each cluster is given by the following formula:

[0068]

[0069] Among them clusters center of mass The calculation formula is as follows:

[0070]

[0071] Step 1203, from the above calculation, we obtain k=t-1, ;

[0072] Here, the degree matrix D is a diagonal matrix representing the degree of each node; the symmetric normalized graph Laplacian matrix L describes the topological structure of the graph and is the core part of the graph convolutional network; graph convolution is a convolution operation performed through a graph convolutional network (GCN) that fuses the feature information of nodes with the structural information of the graph; the similarity matrix 𝑆 represents the similarity between nodes and is used for subsequent clustering analysis; spectral clustering is a method of clustering using the feature vectors of the graph, which can effectively group similar nodes into one class; intra-cluster distance refers to the average distance between data points within the same cluster, measuring the compactness of the cluster; inter-cluster distance refers to the average distance between the centroids of different clusters, measuring the separation of the clusters.

[0073] This step, by calculating intra-cluster and inter-cluster distances, can assess the quality of clustering and ensure the reliability of the results.

[0074] Step 130: Based on the initial CQI values ​​corresponding to each neighboring node in the neighboring node set of the target node cluster, update the initial CQI value corresponding to the target node in the target node cluster to obtain the updated CQI value.

[0075] Here, the initial CQI value refers to the channel quality indication value measured and reported by each user based on the terminal equipment, which is typically used as a reference by the base station during initial scheduling. The updated CQI value refers to the CQI value of the target node that has been corrected and updated based on the initial CQI values ​​of neighboring nodes, reflecting a more real-time and accurate channel quality.

[0076] In one embodiment, the initial CQI value corresponding to the target node in the target node cluster is updated by using the initial CQI values ​​corresponding to each neighboring node in the neighboring node set of the target node cluster, resulting in an updated CQI value. This operation can effectively reduce the lag of the target node's CQI value, improve the timeliness and accuracy of CQI, and the updated CQI value better reflects the current channel quality of the target node. Based on this, the base station can make more accurate resource allocation and improve the overall network performance.

[0077] In another embodiment, the neighboring nodes in the target node cluster are further filtered to identify several neighboring nodes that are more relevant to the target node. The initial CQI value of the target node in the target node cluster is then updated using the initial CQI values ​​of these neighboring nodes, resulting in an updated CQI value. This further filtering of neighboring nodes ensures that the nodes participating in the update have a higher correlation with the channel quality of the target node, reducing noise and errors. This makes the updated CQI value more accurate, enabling the base station to make more precise resource allocation decisions, improving system throughput and user experience. The specific operation of further filtering neighboring nodes in the neighboring node cluster can be found in the following embodiments, and will not be elaborated further here.

[0078] For details on updating the initial CQI value of the target node in the target node cluster using the initial CQI value of each neighboring node in the neighboring node set, please refer to the following embodiment, which will not be elaborated further here.

[0079] Step 140: Based on the CQI value, perform resource scheduling operations on the target terminal.

[0080] Here, the CQI (Channel Quality Indicator) value reflects the channel quality measured by the user terminal. The base station determines the amount of resources allocated to the user and the MCS (Modulation and Coding Scheme) based on the CQI value. The CQI value is an indicator used to reflect the channel quality measured by the terminal, typically ranging from 0 to 15; a higher value indicates better channel quality. The MCS is the modulation and coding scheme used during data transmission. Different MCSs are selected based on channel quality to balance transmission rate and reliability. Through this step, the base station can perform resource scheduling operations based on a more accurate CQI value, thereby improving the overall system performance and user experience.

[0081] The resource scheduling method provided in this invention constructs a scheduling user graph and performs graph convolution operations to fully fuse the topological features and node features among users, making the features of each node richer and more accurate. By clustering the scheduling user graph after feature fusion, neighboring nodes more similar to the target node are selected, which helps to select more reasonable neighboring nodes for CQI value updates and avoids the uncertainty of single node information. Updating the CQI value of the target node using the CQI values ​​of neighboring nodes significantly reduces the CQI value lag problem caused by excessively long CSI-RS measurement periods. Based on the updated CQI value, the base station can more accurately select appropriate MCS and time-frequency resources for allocation, thereby improving the accuracy of downlink resource scheduling.

[0082] Based on any of the above embodiments, in this method, the topological features include the weights of the edges between the target node and each user node among multiple user nodes. Before determining the feature fusion result between the structural features of the scheduling user graph and the node feature set, the method further includes:

[0083] Based on the latitude difference between the target node and each of the multiple user nodes and the longitude difference between the target node and each of the multiple user nodes, the spatial distance between the target node and each of the multiple user nodes is determined.

[0084] The weights of the edges between the target node and each of the multiple user nodes are determined based on the reciprocal of the spatial distance between the target node and each of the multiple user nodes.

[0085] Here, spatial distance refers to the geographical distance between the target node and other user nodes. It is typically calculated using the Euclidean distance formula. Edge weights represent the strength or importance of the connection between two nodes in the scheduling user graph; here, the weights are determined based on the reciprocal of the spatial distance, reflecting the spatial proximity between nodes. Latitude difference refers to the difference in latitude between the target node and other user nodes in the geographic coordinate system, and longitude difference refers to the difference in longitude between the target node and other user nodes in the geographic coordinate system, used to calculate spatial distance.

[0086] In one embodiment, the weight of the edge between the target node and each of the multiple user nodes is determined by the reciprocal of the spatial distance between the target node and each of the multiple user nodes.

[0087] It should be noted that the weights of the edges between the target node and each user node in the multiple user nodes can also be determined by other methods, including but not limited to methods that determine the weights based on the similarity between the target node and each user node in the multiple user nodes. Other methods for determining the weights of the edges between the target node and each user node in the multiple user nodes are equivalent substitutions for the embodiments of the present invention and do not depart from the protection scope of the present invention.

[0088] For example, the spatial distance between the target node and each of the multiple user nodes can be calculated using the Haversine formula. For instance, the spatial distance between target user node i and user j is expressed as... The unit is kilometers, and the specific calculation formula is as follows:

[0089]

[0090] Where Lng1 and Lat1 represent the longitude and latitude of user i, respectively, and Lng2 and Lat2 represent the longitude and latitude of user j, respectively; ,Right now This represents the difference in dimensions between user i and user j; , where b represents the difference in longitude between user i and user j; 6378.137 represents the Earth's radius in kilometers.

[0091] The resource scheduling method provided in this invention calculates the latitude and longitude differences between a target node and multiple user nodes to determine their spatial distances. Then, the reciprocal of the spatial distance is used to determine the weights of the edges between the target node and each user node. Therefore, edges between two far apart nodes have lower weights, indicating a weaker connection, while edges between two close nodes have higher weights, indicating a stronger connection. The edge weights, based on the reciprocal of the spatial distance, accurately reflect the spatial proximity between nodes, making the topology in the user graph more meaningful. By considering the influence of spatial distance, the actual relationships between nodes can be reflected more accurately during feature fusion, improving the effectiveness of node features.

[0092] Based on any of the above embodiments, in this method, the node characteristics of the target node and the node characteristics of multiple user nodes are determined in the following manner:

[0093] The multi-source information parameters uploaded by the target terminal and multiple user terminals are obtained respectively. The multi-source information parameters include primary service area measurement parameters and neighboring cell measurement parameters. The primary service area measurement parameters include channel state information reference signal resource identifier, rank identifier and precoding matrix identifier. The neighboring cell measurement parameters include neighboring cell identifier.

[0094] The multi-source information parameters uploaded by the target terminal are determined as the node characteristics of the target node;

[0095] The multi-source information parameters uploaded by the multiple user terminals are determined as the node characteristics of the multiple user nodes.

[0096] Here, a node is defined as any user within the same SSB (Synchronization Block) beam. Therefore, node characteristics should be user-related information. This application uses multi-source measurement information reported by each user as node characteristics. The Channel State Information Reference Signal Resource Identifier (CRI) indicates the CSI-RS (Channel State Information Reference Signal) beam with the best channel quality measured by the user during beam scanning; one CRI value corresponds to one resource pair. The Rank Identifier (RI) indicates the rank of the channel matrix, corresponding to the number of layers capable of transmitting Multiple-Input Multiple-Output (MIMO). The Precoding Matrix Identifier (PMI) indicates the codebook number for precoding, and the selection of the precoding matrix is ​​determined by combining the number of layers indicated by the Rank Identifier (RI). The Neighbor Cell Identifier (NCI) consists of the PCIs of the top three neighboring cells in terms of user signal strength and their strongest beam index.

[0097] In one embodiment, multi-source information parameters uploaded by the target terminal and multiple user terminals are acquired. These parameters include primary service area measurement parameters and neighboring cell measurement parameters. The primary service area measurement parameters include channel state information reference signal resource identifier, rank identifier, and precoding matrix identifier, while the neighboring cell measurement parameters include neighboring cell identifiers. By integrating detailed primary service area measurement parameters and neighboring cell measurement parameters into multi-source information parameters, and using these multi-source information parameters to determine the node characteristics of the target node and multiple user nodes respectively, the channel quality and network status of each node can be accurately reflected. The multi-source information parameters provide rich information, making the node characteristics more comprehensive and facilitating feature fusion and clustering operations.

[0098] For example, the measurement parameters CRI (Channel State Information Reference Signal Resource Identifier), RI (Rank Identifier), and PMI (Precoding Matrix Identifier) ​​of the main service area are derived from the CSI (Channel State Information) reported by the user. The basis for using these parameters as node characteristics is that if the CSI (Channel State Information) of two users are similar, they should be classified into the same cluster in the clustering operation.

[0099] The measurement parameter NCI (Neighbor Identifier) ​​for neighboring cells comes from the MR reported by the user. The basis for using it as a node feature is that if the neighboring cell measurement results of two users are similar, they should be classified into the same cluster in the clustering operation.

[0100] The resource scheduling method provided in this invention determines the multi-source information parameters uploaded by the target terminal and user terminal as node features of the target node and user node, respectively. These features include primary service area measurement parameters and neighboring cell measurement parameters. This method can accurately reflect the channel quality and network status of nodes, enrich node feature information, improve the accuracy of feature fusion and clustering, thereby improving the accuracy of CQI value updates, optimizing resource scheduling, and ultimately improving the performance and system capacity of 5G networks, enhancing the robustness and stability of the system in complex network environments.

[0101] Based on any of the above embodiments, in this method, constructing a scheduling user subgraph based on the target node and the set of neighboring nodes in the target node cluster, and updating the initial CQI value corresponding to the target node in the target node cluster based on the initial CQI value corresponding to each neighboring node in the set of neighboring nodes in the target node cluster to obtain the updated CQI value, includes:

[0102] The initial CQI value corresponding to each neighboring node in the neighboring node set of the target node cluster is encoded to obtain the encoded node feature of each neighboring node in the neighboring node set. The initial CQI value corresponding to the target node in the target node cluster is encoded to obtain the encoded node feature of the target node.

[0103] A graph convolution operation is performed on the encoded node features of each neighboring node in the neighboring node set and the encoded node features of the target node to obtain a graph convolution fusion feature set. Each graph convolution fusion feature in the graph convolution fusion feature set corresponds one-to-one with each node in the scheduling user subgraph.

[0104] The graph convolutional fusion features of the target node in the target node cluster and the graph convolutional fusion features of each neighboring node in the neighboring node set are aggregated to obtain the aggregated node features of the target node.

[0105] The aggregated node features of the target node are encoded and converted to obtain the updated CQI value of the target node.

[0106] Here, the scheduling user subgraph is constructed based on the relationships between nodes in the target node cluster. Encoded node features refer to the feature values ​​obtained by encoding and transforming the initial CQI value of each node in the target node cluster. Graph convolutional fusion features are the feature values ​​obtained after performing graph convolution operations on the encoded node features between nodes in the target node cluster. Aggregated node features refer to the features obtained by aggregating the graph convolutional fusion features of the target node and its neighboring nodes, representing the expression of the target node in the new feature space. The encoding transformation operation is the process of converting the aggregated node features back to CQI values, so that the updated CQI values ​​can be used for actual resource scheduling.

[0107] In one embodiment, the initial CQI value of each node in the target node cluster is encoded to obtain the encoded node features of each node. The encoding process transforms the original CQI values ​​into a standardized feature form, making it suitable for graph convolution operations. Further, the encoded features of adjacent nodes and the target node are input into a graph convolutional network for processing, resulting in the graph convolutional fusion features of the target node and the graph convolutional fusion features of adjacent nodes. The graph convolution operation performs feature extraction and fusion on the scheduling user subgraph structure, processing the relationships and features between nodes. Graph convolution can comprehensively consider node features and their relationships, extracting a more comprehensive set of features. Feature information; further, the graph convolutional fusion features of the target node are aggregated with the graph convolutional fusion features of adjacent nodes to obtain the aggregated node features of the target node. By combining the features of multiple nodes, the bias of the target node features can be reduced, and the stability and robustness of the features can be improved. Further, the aggregated node features of the target node are encoded and transformed to obtain the updated CQI value. Through encoding and transformation, the aggregated features can be transformed into a more accurate CQI value, reflecting the current channel quality. The updated CQI value can be used more accurately for resource scheduling decisions, thereby improving the system's resource utilization efficiency and user experience.

[0108] The aggregation operation of the graph convolution fusion features of the target node and the graph convolution fusion features of the adjacent nodes can be referred to in the following embodiments, and will not be repeated here.

[0109] The resource scheduling method provided in this invention encodes the set of neighboring nodes in the target node cluster and the initial CQI value of the target node, and then performs graph convolution operations on the encoded node features. This method can fully utilize the topological information between nodes and the initial CQI value to achieve deep feature fusion. Then, by aggregating the graph convolution fusion features of the target node and its neighboring nodes, the representation of the target node in the new feature space is obtained. Finally, the aggregated node features are encoded and transformed to update the CQI value of the target node. This method effectively improves the accuracy and stability of the CQI value, making resource scheduling based on the updated CQI value more accurate and efficient.

[0110] Based on any of the above embodiments, in this method, the step of aggregating the graph convolutional fusion features of the target node in the target node cluster and the graph convolutional fusion features of each neighboring node in the neighboring node set to obtain the aggregated node features of the target node includes:

[0111] Calculate the similarity between the graph convolutional fusion feature of the target node in the target node cluster and the graph convolutional fusion feature of each neighboring node in the neighboring node set, and determine the weight of the edge between the target node and each neighboring node in the neighboring node set based on the calculation result;

[0112] The weights of the edges between the target node and each neighboring node in the neighboring node set are sorted, and the sub-neighboring node set is determined based on the sorting results;

[0113] The graph convolutional fusion features of the target node and the graph convolutional fusion features of each neighboring node in the sub-neighboring node set are aggregated to determine the aggregated node features of the target node.

[0114] In one embodiment, the similarity between the graph convolutional fusion features of the target node in the target node cluster and the graph convolutional fusion features of each neighboring node in the neighboring node set is sorted, and a sub-neighboring node set is determined based on the sorting result; nodes with higher weights are selected as the sub-neighboring node set, and nodes with lower weights are filtered out to reduce noise. By selecting nodes with high weights, the accuracy and reliability of the aggregated node features are improved.

[0115] For example, suppose the target node cluster contains There are nodes, and the node set is... Construct a scheduling user subgraph consisting of the nodes in the target node cluster. According to the 3GPP TS 38.214 specification, the CQI (Channel Quality Indicator) index ranges from 0 to 15. Therefore, The CQI (Channel Quality Indicator) reported by each user is converted into a 4-bit one-hot encoding, and this encoding is used as the feature of the corresponding node's encoding node, thus obtaining... Node feature matrix ;by express The edge set, for the node feature matrix Perform graph convolution operation to obtain From the target node and user nodes For example, node and nodes The corresponding graph convolution fusion feature is and Calculated using Gaussian radial kernel function and The similarity between the two is defined as the similarity between nodes. and nodes edge weight The specific calculation formula is as follows:

[0116]

[0117] in, This is a hyperparameter that, for each node, uses a proportional thresholding method to retain the top 50% of its edges, sorted by weight from largest to smallest, while setting the weights of the remaining edges to 0. (Based on the node...) For example, let the set of its neighboring nodes be... Through neighborhood aggregation, Self-graph convolution fusion features and The graph convolution fusion features of each node within the node are aggregated to obtain... Corresponding aggregation node features The formula is as follows:

[0118]

[0119] in, The aggregation weight of the self-graph convolutional fusion features is set to 1. Neighboring nodes The aggregation weight is set as the node and nodes edge weight Then aggregate node features. Convert the 4-bit encoding to decimal and use that value as the target node. The correction value corresponding to the user's initial CQI value (Channel Quality Indicator), i.e., the updated CQI value.

[0120] For example, the process of neighborhood aggregation of a scheduling user graph containing four nodes, V1-V4, is described in detail in the table below.

[0121]

[0122] The resource scheduling method provided in this invention sorts the similarity between the graph convolutional fusion features of target nodes in the target node cluster and the graph convolutional fusion features of each neighboring node in the neighboring node set. Based on the sorting result, nodes with higher weights are selected as the sub-neighboring node set, and nodes with lower weights are filtered out to reduce noise. This method effectively improves the accuracy and reliability of aggregated node features. By selecting nodes with high weights, the accuracy of feature aggregation is enhanced, thereby improving the update accuracy of the target node's CQI value and the efficiency of resource scheduling. Ultimately, the entire process significantly improves the system's scheduling performance and the overall network performance.

[0123] Based on any of the above embodiments, before performing resource scheduling operations on the target terminal based on the CQI value, the method further includes:

[0124] The CQI value is determined by comparing the spectral efficiency corresponding to the initial CQI value of the target node with the spectral efficiency of the updated CQI value of the target node.

[0125] Here, spectral efficiency refers to the amount of information that can be transmitted per unit of spectrum resources (bandwidth), and it is an important indicator for measuring the performance of wireless communication systems.

[0126] In one embodiment, by comparing the spectral efficiency corresponding to the initial CQI value of the target node with the spectral efficiency of the updated CQI value, it can be determined whether the updated CQI value can improve data transmission efficiency. If the spectral efficiency corresponding to the updated CQI value is higher, it indicates that the updated CQI value is superior, and the updated CQI value is used to perform resource scheduling operations on the target terminal.

[0127] If the updated CQI value corresponds to a lower spectral efficiency, it indicates that the initial CQI value of the target node is better. Based on the initial CQI value of the target node, a suitable MCS is selected to perform resource scheduling operations on the target terminal.

[0128] For example, a metric is introduced: spectral efficiency. The specific calculation formula is as follows: Where R represents the effective information rate of this scheduled transmission, and B represents the channel bandwidth of this scheduled transmission, i.e., the number of bits per second that can be transmitted per unit bandwidth, measured in bits / s / Hz. The CQI (Channel Quality Indicator) can be mapped to the MCS (Modulation and Coding Scheme), thus determining the number of bits corresponding to the modulation scheme selected for this scheduling as Q. Let L be the number of MIMO (Multiple-Input Multiple-Output) layers, N be the number of REs (Resource Particles), E be the coding efficiency, and T be the spectral efficiency evaluation time for this scheduling. Then, the formula for calculating R is as follows:

[0129]

[0130] pass Determine whether to update the user's CQI (Channel Quality Indicator) value, which is calculated from the initial CQI value of the target node. The CQI value is calculated using the updated CQI value of the target node. ,like This indicates that the updated CQI value can improve spectral efficiency, so the updated CQI value is used for scheduling; if This indicates that the updated CQI value actually reduces the spectral efficiency, therefore the initial CQI value should be used for scheduling.

[0131] The resource scheduling method provided in this invention compares the spectral efficiency of the initial CQI value and the updated CQI value of the target node and selects the CQI value with higher spectral efficiency. This method not only improves data transmission efficiency and system stability, but also optimizes resource scheduling, enabling the system to allocate and use spectrum resources more efficiently, thereby improving overall performance and user experience.

[0132] The resource scheduling device provided by the present invention is described below. The resource scheduling device described below and the resource scheduling method described above can be referred to in correspondence.

[0133] Figure 2 Figure 2 shows a schematic diagram of the resource scheduling device provided by the present invention. The resource scheduling device includes:

[0134] The feature fusion module 210 is used to respond to a resource scheduling request sent to the target terminal, and determine the feature fusion result between the structural features of the scheduling user graph and the node feature set of the scheduling user graph. The structural features include the topological features between the target node and multiple user nodes, and the node feature set includes the node features of the target node and the node features of the multiple user nodes.

[0135] Clustering module 220 is used to cluster the target node and multiple user nodes based on the feature fusion result to obtain a target node cluster. The target node cluster includes the target node and a set of neighboring nodes. The set of neighboring nodes includes several user nodes similar to the target node.

[0136] CQI value update module 230 is used to update the initial CQI value corresponding to the target node in the target node cluster based on the initial CQI value corresponding to each neighboring node in the neighboring node set in the target node cluster, so as to obtain the updated CQI value.

[0137] The resource scheduling module 240 is used to perform resource scheduling operations on the target terminal based on the CQI value.

[0138] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3As shown, the electronic device may include: a processor 310, a communications interface 320, a memory 330, and a communications bus 340, wherein the processor 310, the communications interface 320, and the memory 330 communicate with each other through the communications bus 340. The processor 310 can invoke logical instructions in the memory 330 to execute a resource scheduling method, the method comprising: in response to a resource scheduling request sent to a target terminal, determining a feature fusion result between structural features of a scheduling user graph and a set of node features of the scheduling user graph, wherein the structural features include topological features between a target node and multiple user nodes, and the set of node features includes node features of the target node and node features of the multiple user nodes; clustering the target node and the multiple user nodes based on the feature fusion result to obtain a target node cluster, wherein the target node cluster includes the target node and a set of neighboring nodes, wherein the set of neighboring nodes includes several user nodes similar to the target node; updating the initial CQI value corresponding to the target node in the target node cluster based on the initial CQI value corresponding to each neighboring node in the set of neighboring nodes in the target node cluster to obtain an updated CQI value; and performing a resource scheduling operation on the target terminal based on the CQI value.

[0139] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0140] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the resource scheduling method provided by the above methods. The method includes: in response to a resource scheduling request sent to a target terminal, determining a feature fusion result between structural features of a scheduling user graph and a node feature set of the scheduling user graph, wherein the structural features include topological features between a target node and multiple user nodes, and the node feature set includes node features of the target node and node features of the multiple user nodes; clustering the target node and the multiple user nodes based on the feature fusion result to obtain a target node cluster, wherein the target node cluster includes the target node and a set of neighboring nodes, and the set of neighboring nodes includes several user nodes similar to the target node; updating the initial CQI value corresponding to the target node in the target node cluster based on the initial CQI value corresponding to each neighboring node in the set of neighboring nodes in the target node cluster to obtain an updated CQI value; and performing a resource scheduling operation on the target terminal based on the CQI value.

[0141] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, is implemented to perform the resource scheduling method provided by the above methods. The method includes: in response to a resource scheduling request sent to a target terminal, determining a feature fusion result between structural features of a scheduling user graph and a set of node features of the scheduling user graph, wherein the structural features include topological features between a target node and multiple user nodes, and the set of node features includes node features of the target node and node features of the multiple user nodes; clustering the target node and the multiple user nodes based on the feature fusion result to obtain a target node cluster, wherein the target node cluster includes the target node and a set of neighboring nodes, the set of neighboring nodes including several user nodes similar to the target node; updating the initial CQI value corresponding to the target node in the target node cluster based on the initial CQI value corresponding to each neighboring node in the set of neighboring nodes in the target node cluster to obtain an updated CQI value; and performing a resource scheduling operation on the target terminal based on the CQI value.

[0142] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0143] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0144] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A resource scheduling method, characterized in that, include: In response to a resource scheduling request sent to a target terminal, a feature fusion result is determined between the structural features of the scheduling user graph and the node feature set of the scheduling user graph. The structural features include topological features between the target node and multiple user nodes, and the node feature set includes the node features of the target node and the node features of the multiple user nodes. Based on the feature fusion results, the target node and multiple user nodes are clustered to obtain a target node cluster. The target node cluster includes the target node and a set of neighboring nodes. The set of neighboring nodes includes several user nodes that are similar to the target node. Based on the initial CQI value corresponding to each neighboring node in the neighboring node set of the target node cluster, the initial CQI value corresponding to the target node in the target node cluster is updated to obtain the updated CQI value; Based on the CQI value, resource scheduling operations are performed on the target terminal; Specifically, a scheduling user subgraph is constructed based on the target node and the set of neighboring nodes in the target node cluster. The process of updating the initial CQI value corresponding to the target node in the target node cluster based on the initial CQI value corresponding to each neighboring node in the set of neighboring nodes in the target node cluster to obtain the updated CQI value includes: The initial CQI value corresponding to each neighboring node in the neighboring node set of the target node cluster is encoded to obtain the encoded node feature of each neighboring node in the neighboring node set. The initial CQI value corresponding to the target node in the target node cluster is encoded to obtain the encoded node feature of the target node. A graph convolution operation is performed on the encoded node features of each neighboring node in the neighboring node set and the encoded node features of the target node to obtain a graph convolution fusion feature set. Each graph convolution fusion feature in the graph convolution fusion feature set corresponds one-to-one with each node in the scheduling user subgraph. The graph convolutional fusion features of the target node in the target node cluster and the graph convolutional fusion features of each neighboring node in the neighboring node set are aggregated to obtain the aggregated node features of the target node. The aggregated node features of the target node are encoded and converted to obtain the updated CQI value of the target node.

2. The resource scheduling method according to claim 1, characterized in that, The topological features include the weights of the edges between the target node and each user node in the plurality of user nodes. Before determining the feature fusion result between the structural features of the scheduling user graph and the node feature set, the method further includes: Based on the latitude difference between the target node and each of the multiple user nodes and the longitude difference between the target node and each of the multiple user nodes, the spatial distance between the target node and each of the multiple user nodes is determined. The weights of the edges between the target node and each of the multiple user nodes are determined based on the reciprocal of the spatial distance between the target node and each of the multiple user nodes.

3. The resource scheduling method according to claim 1, characterized in that, The node characteristics of the target node and the node characteristics of the multiple user nodes are determined based on the following method: The multi-source information parameters uploaded by the target terminal and multiple user terminals are obtained respectively. The multi-source information parameters include primary service area measurement parameters and neighboring cell measurement parameters. The primary service area measurement parameters include channel state information reference signal resource identifier, rank identifier and precoding matrix identifier. The neighboring cell measurement parameters include neighboring cell identifier. The multi-source information parameters uploaded by the target terminal are determined as the node characteristics of the target node; The multi-source information parameters uploaded by the multiple user terminals are determined as the node characteristics of the multiple user nodes.

4. The resource scheduling method according to claim 1, characterized in that, The aggregation of graph convolutional fusion features of the target node in the target node cluster and graph convolutional fusion features of each neighboring node in the neighboring node set yields the aggregated node features of the target node, including: Calculate the similarity between the graph convolutional fusion feature of the target node in the target node cluster and the graph convolutional fusion feature of each neighboring node in the neighboring node set, and determine the weight of the edge between the target node and each neighboring node in the neighboring node set based on the calculation result; The weights of the edges between the target node and each neighboring node in the neighboring node set are sorted, and the sub-neighboring node set is determined based on the sorting results; The graph convolutional fusion features of the target node and the graph convolutional fusion features of each neighboring node in the sub-neighboring node set are aggregated to determine the aggregated node features of the target node.

5. The resource scheduling method according to claim 1, characterized in that, Before performing resource scheduling operations on the target terminal based on the CQI value, the method further includes: The CQI value is determined by comparing the spectral efficiency corresponding to the initial CQI value of the target node with the spectral efficiency of the updated CQI value of the target node.

6. A resource scheduling device, characterized in that, include: The feature fusion module is used to respond to a resource scheduling request sent to the target terminal and determine the feature fusion result between the structural features of the scheduling user graph and the node feature set of the scheduling user graph. The structural features include the topological features between the target node and multiple user nodes, and the node feature set includes the node features of the target node and the node features of the multiple user nodes. The clustering module is used to cluster the target node and multiple user nodes based on the feature fusion result to obtain a target node cluster. The target node cluster includes the target node and a set of neighboring nodes. The set of neighboring nodes includes several user nodes similar to the target node. The CQI value update module is used to update the initial CQI value corresponding to the target node in the target node cluster based on the initial CQI value corresponding to each neighboring node in the neighboring node set in the target node cluster, so as to obtain the updated CQI value. The resource scheduling module is used to perform resource scheduling operations on the target terminal based on the CQI value; Specifically, a scheduling user subgraph is constructed based on the target node and the set of neighboring nodes in the target node cluster. The process of updating the initial CQI value corresponding to the target node in the target node cluster based on the initial CQI value corresponding to each neighboring node in the set of neighboring nodes in the target node cluster to obtain the updated CQI value includes: The initial CQI value corresponding to each neighboring node in the neighboring node set of the target node cluster is encoded to obtain the encoded node feature of each neighboring node in the neighboring node set. The initial CQI value corresponding to the target node in the target node cluster is encoded to obtain the encoded node feature of the target node. A graph convolution operation is performed on the encoded node features of each neighboring node in the neighboring node set and the encoded node features of the target node to obtain a graph convolution fusion feature set. Each graph convolution fusion feature in the graph convolution fusion feature set corresponds one-to-one with each node in the scheduling user subgraph. The graph convolutional fusion features of the target node in the target node cluster and the graph convolutional fusion features of each neighboring node in the neighboring node set are aggregated to obtain the aggregated node features of the target node. The aggregated node features of the target node are encoded and converted to obtain the updated CQI value of the target node.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the resource scheduling method as described in any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the resource scheduling method as described in any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the resource scheduling method as described in any one of claims 1 to 5.

Images