Cell switching method and device oriented to load balancing, equipment and storage medium

Abstract

The invention provides a load balancing-oriented cell switching method, device and equipment and a storage medium, and belongs to the technical field of communication, and the method comprises the steps: constructing a load distribution diagram by using neighbor cell nodes determined based on candidate neighbor cells of a current cell and user nodes determined based on transfer users connected with the current cell; each edge of the load distribution diagram is used for associating a neighbor cell node and a user node; utilizing an edge weight prediction model to predict the edge weight of each edge in the load distribution diagram to obtain a load distribution diagram after the edge weight is updated; and determining a target cell switched by the transfer user based on the load distribution diagram after the edge weight is updated. According to the load balancing-oriented cell switching method, device and equipment and the storage medium provided by the invention, the load balancing problem is converted into the edge prediction problem of the heterogeneous bipartite graph, the topological structure information of a communication network can be fully utilized, the target cell most suitable for switching can be accurately matched for a user, and the user experience is improved. Therefore, the optimal load balancing among the cells is realized.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a cell switching method, device, equipment and storage medium oriented to load balancing. Background Art

[0002] Switching users in a high-load cell to a low-load neighboring cell of the high-load cell is an important way to achieve load balancing.

[0003] At present, the method for switching cells is mainly the threshold judgment method, that is, based on the comparison result of various data such as load interaction information and measurement report of a candidate neighboring cell with the preset threshold value, it is determined whether the candidate neighboring cell can be used as the target cell for user switching cells.

[0004] However, the threshold judgment method cannot fully utilize the topological structure information of the communication network, cannot efficiently extract useful features from complex and massive data, is difficult to accurately match users with the most suitable target cell for switching among many candidate neighboring cells, and is difficult to achieve optimal load balancing between cells. Summary of the invention

[0005] The present invention provides a cell switching method, device, equipment and storage medium for load balancing, which are used to solve the defect that the threshold judgment method in the prior art is difficult to accurately match the most suitable cell, and realize the optimal load balancing between cells.

[0006] The present invention provides a cell switching method oriented to load balancing, comprising: A load distribution graph is constructed using neighboring cell nodes and user nodes; the neighboring cell nodes are determined based on candidate neighboring cells of a current cell; the user nodes are determined based on transferred users connected to the current cell; each edge of the load distribution graph is used to associate a neighboring cell node with a user node; The edge weight prediction model is used to predict the edge weight of each edge in the load distribution graph to obtain a load distribution graph after the edge weight is updated; Based on the load distribution diagram after the edge weight is updated, a target cell for handover of the transferred user is determined.

[0007] According to a cell switching method for load balancing provided by the present invention, before constructing a load distribution map using neighboring cell nodes and user nodes, determining the candidate neighboring cell based on the neighboring cell of the current cell includes: Determining a comprehensive load index based on the load indexes of each dimension of the neighboring area; Based on the comprehensive load index, the candidate neighboring area is determined.

[0008] According to a cell switching method for load balancing provided by the present invention, before constructing a load distribution diagram using neighboring cell nodes and user nodes, determining the transfer user based on the user connected to the current cell includes: Determine the transfer priority based on the priorities of each dimension of the user; Based on the transfer priority, the transfer user is determined.

[0009] According to a cell switching method for load balancing provided by the present invention, the use of neighboring cell nodes and user nodes to construct a load distribution graph includes: Acquire measurement report information; the measurement report information is measured by the transferring user after receiving the event information of the candidate neighboring area; the event information of the candidate neighboring area is sent by the current cell to the transferring user; Determining, based on the measurement report information, an initial edge weight of each edge used to associate a neighboring node with a user node; The load distribution graph is constructed using the node characteristics of the neighboring nodes, the node characteristics of the user nodes, and the initial edge weight of each edge.

[0010] According to a cell switching method for load balancing provided by the present invention, the measurement report information includes a reference signal received power and a signal to interference plus noise ratio; and the edge weight of each edge for associating a neighboring cell node and a user node is determined based on the measurement report information, including: Determine a first quotient value based on the reference signal received power of the candidate neighboring cell associated with the one neighboring cell node and the reference signal received power of the current cell; Determine a second quotient value based on a signal to interference plus noise ratio of a candidate neighboring cell associated with the one neighboring cell node and a signal to interference plus noise ratio of the current cell; The edge weight is determined based on a product of the first quotient value and the second quotient value.

[0011] According to a cell switching method for load balancing provided by the present invention, the edge weight prediction model is used to predict the edge weight of each edge in the load distribution graph to obtain the load distribution graph after the edge weight is updated, including: Determine an updated node feature of the neighboring node based on a node feature of the neighboring node and a node feature of a user node adjacent to the neighboring node that is fused together; Determine an updated node feature of the user node based on a fusion of the node feature of the user node and the node features of neighboring nodes adjacent to the user node; Determining a connection possibility score based on the updated node characteristics of the neighboring node and the updated node characteristics of the user node; Based on the connection possibility score, the edge weight of each edge in the load distribution graph is updated to obtain the load distribution graph after the edge weight is updated.

[0012] The present invention also provides a cell switching device for load balancing, comprising: A load distribution graph construction module, used to construct a load distribution graph using neighboring cell nodes and user nodes; the neighboring cell nodes are determined based on candidate neighboring cells of the current cell; the user nodes are determined based on transferred users connected to the current cell; each edge of the load distribution graph is used to associate a neighboring cell node with a user node; An edge weight updating module, used to predict the edge weight of each edge in the load distribution graph by using an edge weight prediction model, and obtain a load distribution graph after the edge weight is updated; The target cell switching module is used to determine the target cell for switching the transferred user based on the load distribution diagram after the edge weight is updated.

[0013] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the load-balancing oriented cell switching method described above is implemented.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the load-balancing oriented cell switching method described above is implemented.

[0015] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned load balancing-oriented cell switching methods.

[0016] The cell switching method, device, equipment and storage medium for load balancing provided by the present invention use candidate neighboring cells and transferred users as nodes of a load distribution graph, and each edge of the load distribution graph is associated with a neighboring cell node and a user node, so as to construct a load distribution graph as a heterogeneous bipartite graph, thereby converting the load balancing problem into an edge prediction problem of the heterogeneous bipartite graph, and further using a neural network model to predict the edge weights of the load distribution graph, so as to determine the candidate neighboring cell corresponding to the neighboring cell node associated with the edge with the highest edge weight as the target cell for the transferred user, thereby making full use of the topological structure information of the communication network, efficiently mining and extracting useful features from the data, accurately matching the user to the most suitable target cell for switching, and thus achieving optimal load balancing between cells. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 It is a flow chart of a cell switching method for load balancing provided by the present invention.

[0019] Figure 2 It is an example diagram of the load distribution diagram provided by the present invention.

[0020] Figure 3 It is a structural schematic diagram of a cell switching device oriented to load balancing provided by the present invention.

[0021] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] It should be noted that, in the description of the present invention, the term "include" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0024] The terms "first", "second", etc. in the present invention are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present invention can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally of the same type, and the number of objects is not limited. For example, the first object can be one or more.

[0025] Combine the following Figure 1-Figure 4The present invention describes a load balancing-oriented cell switching method, device, equipment and storage medium.

[0026] In the 5G mobile communication system, the time and frequency resources of each cell are limited, so the number of users and traffic that can be accommodated are also limited.

[0027] On the one hand, in the spatial dimension, end users move randomly throughout the network, often showing an uneven distribution. For example, in scenarios such as large-scale sports events and concerts, people gather in the same area with high density, which will concentrate the cell resources in the area.

[0028] On the other hand, in terms of time, even for users with relatively stable locations, business needs may vary at different times of the day. For example, near a construction site, the data traffic during the lunch and evening breaks is much higher than at other times.

[0029] Therefore, cells in hot spots and hot time periods are prone to high load problems. Resource congestion will lead to degradation of cell performance and reduced user experience rate. At the same time, other relatively idle low-load cells with different frequencies and the same coverage have the problem of resource waste. This phenomenon of load imbalance among multiple cells will lead to a decrease in the throughput of the entire network, seriously affecting user perception. In order to make full use of the resources of each cell efficiently, it is necessary to share traffic among each cell through load balancing technology, transfer users in high-load cells to low-load cells, that is, switch the cell where the user is located to achieve load balancing.

[0030] It should be noted that the execution subject of the load balancing-oriented cell switching method provided by the present invention can be a server, a computer device, such as a mobile phone, a tablet computer, a laptop computer, a PDA, a vehicle-mounted electronic device, a wearable device, an ultra-mobile personal computer (Ultra-Mobile Personal Computer, UMPC), a netbook or a personal digital assistant (Personal Digital Assistant, PDA), etc.

[0031] Figure 1 is a flow chart of a cell switching method for load balancing provided by the present invention, such as Figure 1 As shown, the load balancing-oriented cell switching method includes but is not limited to steps 101 to 103.

[0032] Step 101: Use neighboring nodes and user nodes to construct a load distribution graph.

[0033] The neighboring cell node is determined based on the candidate neighboring cells of the current cell; the user node is determined based on the transferred user connected to the current cell; each edge of the load distribution graph is used to associate a neighboring cell node and a user node.

[0034] The current cell, also called the current primary service cell, is a cell whose load exceeds the set threshold and is determined by the base station to be in a high load state. Cells within the same base station can directly exchange load information, and cells between different base stations can send resource status request messages to other base stations through the interface between base stations, and then receive resource status information replied by other base stations.

[0035] A cell that has a neighboring cell relationship with the current cell, is in an activated state, has a different frequency from the current cell, and allows load balancing interaction with the current cell is also called a neighboring cell (referred to as a neighboring cell). The neighboring cell and the current cell can be in the same base station or in different base stations.

[0036] The candidate neighboring cells are part or all of the neighboring cells that can be used for load redistribution and for users to switch cells.

[0037] The transferred users are part or all of the users determined from all users connected to the current cell and who need to switch cells.

[0038] Specifically, candidate neighboring cells are determined in advance from all neighboring cells of the current cell, and transfer users are determined from all users connected to the current cell. Each candidate neighboring cell is used as each neighboring cell node in the load distribution graph, and each transfer user is used as each user node in the load distribution graph to determine all nodes in the load distribution graph. Further, an edge is used to connect any neighboring cell node and any user node, so that each edge in the load distribution graph is used to associate a neighboring cell node with a user node.

[0039] It can be understood that each neighboring cell node in the load distribution diagram corresponds one-to-one to a candidate neighboring cell, and each user node corresponds one-to-one to a transferred user.

[0040] Figure 2 is an example diagram of the load distribution diagram provided by the present invention, such as Figure 2 As shown in the figure, the dotted box on the left is a node set consisting of neighboring nodes corresponding to the three candidate neighboring areas, and the dotted box on the right is a node set consisting of user nodes corresponding to the five transferred users. Each candidate neighboring area and each transferred user are associated with each other using a black undirected edge. Finally, all neighboring nodes, all user nodes and all edges together constitute the load distribution graph.

[0041] The load distribution graph is a heterogeneous bipartite graph. On the one hand, a heterogeneous graph refers to a graph in which at least one of the nodes and edges has multiple types. In the load distribution graph, there are two types of nodes: neighbor nodes and user nodes. Therefore, the load distribution graph is a heterogeneous graph. On the other hand, the set of neighbor nodes and the set of user nodes do not intersect each other, and the two nodes associated with each edge in the load distribution graph belong to the cell node set and the user node set respectively. Therefore, the load distribution graph is a bipartite graph.

[0042] Step 102: Use the edge weight prediction model to predict the edge weight of each edge in the load distribution graph to obtain a load distribution graph with updated edge weights.

[0043] The edge weight prediction model is obtained by training using a training sample set, wherein the training sample set includes a plurality of training samples, and each training sample includes a load distribution graph sample and its corresponding edge weight label.

[0044] Specifically, after constructing a load distribution map using neighboring nodes and user nodes, the load distribution map is input into a pre-trained edge weight prediction model, and the pre-trained edge weight prediction model predicts the edge weight of each edge in the load distribution map to obtain the load distribution map after the edge weight is updated.

[0045] Optionally, the edge weight prediction model is constructed based on a graph neural network model.

[0046] It should be noted that the edge weight prediction model can be specifically constructed based on any one of the graph neural network models such as graph neural network (GNN), relation-aware graph convolutional neural network (RAGCN), etc. The loss function of the edge weight prediction model can be defined as a commonly used loss function such as cross entropy, and the present invention does not impose any limitation on this.

[0047] In one embodiment, each training sample is obtained based on the following method: In the same manner as the construction of the load distribution map, a sample of the constructed load distribution map at the first historical moment is constructed by using the candidate neighboring area data and the transfer user data at the first historical moment; Using the candidate neighboring area data and the transferred user data at the second historical moment, a load distribution graph at the second historical moment is constructed; the edge weight of each edge in the load distribution graph at the second historical moment is used as an edge weight label corresponding to the load distribution graph sample; Among them, the first historical moment precedes the second historical moment.

[0048] The edge weight prediction model is trained in batches using the training sample set. Once the model accuracy reaches the preset requirements, it can be used for edge prediction of cell switching.

[0049] Step 103: Determine the target cell for handover of the transferred user based on the load distribution diagram after the edge weight is updated.

[0050] Specifically, after obtaining the load distribution diagram after edge weight update, for each transferred user, the candidate neighboring cell corresponding to the neighboring cell node associated with the edge with the largest edge weight among all edges corresponding to the user node is determined as the target cell for the transferred user to switch from the current cell to the target cell. Similarly, after determining the target cell for all transferred users to switch, all transferred users are switched from the current cell with high load to the neighboring cell with low load, and finally load balancing is achieved.

[0051] For example, in the unupdated load distribution graph G, if the user node corresponding to the transferred user 1 is associated with the neighboring node a, neighboring node b, and neighboring node c corresponding to the candidate neighboring area A, candidate neighboring area B, and candidate neighboring area C through the edge α, edge β, and edge γ, respectively. The edge weights of edge α, edge β, and edge γ are 0.2, 0.4, and 0.4, respectively.

[0052] After the load distribution graph G is input into the edge weight prediction model, the edge weight prediction model updates the edge weight of each edge in the load distribution graph. In the load distribution graph G' after the edge weight update, the edge weights of edge α, edge β and edge γ are 0.8, 0.1 and 0.1 respectively.

[0053] Therefore, the candidate neighboring cell A corresponding to the neighboring cell node a is used as the target cell for the transfer user 1 to switch from the current cell to the candidate neighboring cell A. Similarly, the target cells for all the transfer users to switch from the current cell to the target cell are determined, so that all the transfer users can switch from the current cell to the target cell, thereby switching the user from the high-load current cell to the low-load neighboring cell, and finally achieving load balancing.

[0054] The cell switching method for load balancing provided by the present invention uses candidate neighboring cells and transferred users as nodes of a load distribution graph, and each edge of the load distribution graph is associated with a neighboring cell node and a user node, so as to construct a load distribution graph as a heterogeneous bipartite graph, thereby converting the load balancing problem into an edge prediction problem of the heterogeneous bipartite graph, and further using a neural network model to predict the edge weights of the load distribution graph, so as to determine the candidate neighboring cell corresponding to the neighboring cell node associated with the edge with the highest edge weight as the target cell for the transferred user, thereby making full use of the topological structure information of the communication network, efficiently mining and extracting useful features from the data, accurately matching the user to the most suitable target cell for switching, and thus achieving optimal load balancing between cells.

[0055] Based on the above embodiment, as an optional embodiment, before using the neighboring cell nodes and the user nodes to construct the load distribution map, determining the candidate neighboring cell based on the neighboring cell of the current cell includes: Determining a comprehensive load index based on the load indexes of each dimension of the neighboring area; Based on the comprehensive load index, the candidate neighboring area is determined.

[0056] The load indexes of each dimension include but are not limited to at least two of the load indexes such as the user number load index, the physical resource block (PRB) utilization load index and the control channel element (CCE) utilization load index.

[0057] The PRB utilization load index is an indicator that compares and measures the service channel resource conditions of the neighboring cell and the current cell.

[0058] The CCE utilization load index is an indicator that compares and measures the control channel resource conditions of the neighboring cell and the current cell.

[0059] Specifically, before constructing a load distribution diagram using neighboring cell nodes and user nodes, it is necessary to determine candidate neighboring cells from neighboring cells of the current cell and to determine transfer users from users connected to the current cell.

[0060] For determining candidate neighboring cells from neighboring cells of the current cell, first ensure that the neighboring cells are not in a high load state or an interference state. Then, according to the load index of each dimension of each neighboring cell, the comprehensive load index of each neighboring cell is calculated. Further, in the order of the comprehensive load index from small to large, the first number of neighboring cells are determined as candidate neighboring cells. Correspondingly, in the load distribution diagram, there are also a first number of neighboring cell nodes that correspond one-to-one to the candidate neighboring cells.

[0061] It can be understood that the data required to determine whether the neighboring cell is in a high load state or an interference state and to calculate the neighboring cell user number load index, PRB utilization load index and CCE utilization load index can all be obtained from the load interaction information between the current cell and the neighboring cell.

[0062] Optionally, the first quantity is determined according to a preset quantity threshold.

[0063] Optionally, the first number is determined based on the product of a preset ratio value and the total number of neighboring areas.

[0064] In one embodiment, the user load index is determined based on the number of downlink radio resource control (RRC) connected users in the neighboring cell, the number of downlink RRC connected users in the current cell, the proportion of downlink RRC connected users in the neighboring cell, and the proportion of downlink RRC connected users in the current cell.

[0065] Optionally, an expression for determining a user number load index of each neighboring cell based on the number of downlink RRC connected users in the neighboring cell, the number of downlink RRC connected users in the current cell, the proportion of downlink RRC connected users in the neighboring cell, and the proportion of downlink RRC connected users in the current cell is as follows: .

[0066] In one embodiment, the PRB utilization load index is determined based on the number of PRBs of downlink data channels actually occupied by neighboring cells, the number of PRBs of downlink data channels available in neighboring cells, the number of PRBs of downlink data channels actually occupied by the current cell, and the number of PRBs of downlink data channels available in the current cell.

[0067] Optionally, the expression for determining the PRB utilization load index of each neighboring cell based on the number of PRBs of the downlink data channel actually occupied by the neighboring cell, the number of PRBs of the downlink data channel available in the neighboring cell, the number of PRBs of the downlink data channel actually occupied by the current cell, and the number of PRBs of the downlink data channel available in the current cell is as follows: .

[0068] In one embodiment, the CCE utilization load index is determined based on the number of CCEs equivalent to the reference power used in the neighboring cell downlink, the number of CCEs equivalent to the reference power available in the neighboring cell downlink, the number of CCEs equivalent to the reference power used in the current cell downlink, and the number of CCEs equivalent to the reference power available in the current cell downlink.

[0069] Optionally, the expression for determining the CCE utilization load index based on the number of CCEs equivalent to the reference power used in the neighboring cell downlink, the number of CCEs equivalent to the reference power available in the neighboring cell downlink, the number of CCEs equivalent to the reference power used in the current cell downlink, and the number of CCEs equivalent to the reference power available in the current cell downlink is as follows: .

[0070] In one embodiment, a comprehensive load index is determined based on the load indexes of each dimension of the neighboring cell, including: determining the comprehensive load index based on the product of the user number load index, the PRB utilization load index and the CCE utilization load index.

[0071] It can be seen that the user number load index, PRB utilization load index and CCE utilization load index are all positively correlated with the comprehensive load index. The lower the comprehensive load index, the greater the difference between the number of users, PRB utilization and CCE utilization of this neighboring area and the current cell, and the more suitable it is as a candidate neighboring area for migrating users. Therefore, according to the calculated comprehensive load index, the neighboring areas of the current cell can be arranged in ascending order from small to large, and the first M neighboring areas can be selected as candidate neighboring areas.

[0072] The cell switching method for load balancing provided by the present invention calculates a comprehensive load index from multi-dimensional load indexes such as a user number load index, a PRB utilization load index, and a CCE utilization load index according to load interaction information between a current cell and a neighboring cell, taking into account both the user number load index to measure the network usage of each neighboring cell itself and the PRB utilization load index and the CCE utilization load index to measure the mutual influence between the current cell and the neighboring cell; further, candidate neighboring cells are determined according to the size arrangement of the comprehensive load indexes of all neighboring cells, and under the condition of better comprehensive multi-dimensional indicators, the selection of candidate neighboring cells is separated from the decision threshold of a single indicator set by experience, so that it can have better generalization performance when used in different scenarios.

[0073] Based on the above embodiment, as an optional embodiment, before using neighboring nodes and user nodes to construct a load distribution map, determining the transferred user based on the user connected to the current cell includes: Determine the transfer priority based on the priorities of each dimension of the user; Based on the transfer priority, the transfer user is determined.

[0074] Among them, the priorities of each dimension include but are not limited to at least two of the priorities such as allocation and reservation (Allocation and Retention Priority, ARP) priority, quality of service (Quality of Service, QoS) class identifier (QoS Class Identifier, QCI) priority, PRB utilization priority and far point priority.

[0075] ARP priority is based on the ARP threshold user selection policy priority. The user with higher ARP priority should be more likely to be selected as a transfer user; QCI priority is based on the QCI threshold user selection policy priority. The user with higher QCI priority should be more likely to be selected as a transfer user; PRB utilization priority is based on the PRB utilization threshold user selection policy priority. The user with higher PRB utilization priority should be more likely to be selected as a transfer user; far point priority is based on the far point threshold user selection policy priority. The user with higher far point priority should be more likely to be selected as a transfer user.

[0076] Specifically, for determining the transfer users from the users connected to the current cell, the transfer priority of each user is calculated according to the priorities of each dimension of each user. The priorities of each dimension include but are not limited to at least two of the priorities such as ARP priority, QCI priority, PRB utilization priority and far point priority. Further, in descending order of the transfer priority, a second number of users are determined as transfer users, such as selecting the first N users as transfer users. Correspondingly, in the load distribution diagram, there are a second number of user nodes corresponding to the transfer users one by one.

[0077] Optionally, the second quantity is determined according to a preset quantity threshold.

[0078] Optionally, the second number is determined based on the product of a preset ratio value and the total number of users.

[0079] In one embodiment, the transfer priority is determined based on the priorities of each dimension of the user, including: determining the transfer priority based on the weighted sum of the user's ARP priority, QCI priority, PRB utilization priority and far point priority.

[0080] Optionally, in one embodiment, the expression for determining the transfer priority based on the priorities of each dimension of the user is as follows: ; in, is the weight coefficient of ARP priority; is the weight coefficient of QCI priority; is the weight coefficient of the PRB utilization priority; is the weight coefficient of the far point priority.

[0081] Optionally, .

[0082] The cell switching method for load balancing provided by the present invention calculates the transfer priority from the user's ARP priority, QCI priority, PRB utilization priority and far point priority by multi-dimensional superposition. Under the condition of better comprehensive multi-dimensional indicators, the selection of the transferred user is freed from the decision threshold of a single indicator set by experience, and thus can have better generalization performance when used in different scenarios.

[0083] In another embodiment, before constructing the load distribution diagram using neighboring cell nodes and user nodes, the candidate neighboring cells are first determined based on the neighboring cells of the current cell; and the transfer users are then determined based on the users connected to the current cell.

[0084] By optimizing the method of determining candidate neighboring cells and transferred users, it is helpful to match more suitable candidate neighboring cells as target cells for transferred users, which plays a key role in effectively reducing the load of the current cell, avoiding causing high load in neighboring cells, and minimizing the impact on transferred users.

[0085] Based on the above embodiment, as an optional embodiment, the use of neighboring nodes and user nodes to construct a load distribution graph includes: Acquire measurement report information; the measurement report information is measured by the transferring user after receiving the event information of the candidate neighboring area; the event information of the candidate neighboring area is sent by the current cell to the transferring user; Determining, based on the measurement report information, an initial edge weight of each edge used to associate a neighboring node with a user node; The load distribution graph is constructed using the node characteristics of the neighboring nodes, the node characteristics of the user nodes, and the initial edge weight of each edge.

[0086] Node features are information related to a specific node.

[0087] Specifically, the current cell sends a measurement command to the determined N transferred users, and sends the event information of the M candidate neighboring cells to the transferred users. After receiving the measurement command and event information, the transferred users perform measurements and upload the measurement report information to the current cell if the event judgment conditions are met, so that the current cell obtains the measurement report information. The measurement report information uploaded by the transferred user includes but is not limited to the reference signal receiving power (RSRP) and signal to interference plus noise ratio (SINR) of the candidate neighboring cells and the current cell.

[0088] According to the measurement report information uploaded by the transferred user, the initial edge weight of each edge used to associate the user node corresponding to the transferred user with each neighboring node can be determined by a certain calculation method, thereby determining the initial edge weight of each edge used to associate a neighboring node and a user node in the load distribution graph. The load distribution graph is further constructed using the node characteristics of the neighboring nodes, the user characteristics of the user nodes, and the initial edge weight of each edge.

[0089] Optionally, the event information includes but is not limited to at least one of the event information such as relevant information of frequency point A4 or A5 events of the candidate neighboring area.

[0090] Optionally, the node characteristics of each neighboring node and the node characteristics of each user node are determined based on the following method: Determine the node set of all candidate neighboring areas as A , the node set of all transferred users is B .

[0091] Define the node characteristics of each neighboring node as a d A The vector of dimension 1 is obtained by combining the values ​​of the number of RRC connected users in the candidate neighboring cell corresponding to the neighboring cell node, the proportion of RRC connected users in the candidate neighboring cell, the PRB utilization rate of the cell, the interference of the cell, the uplink bandwidth of the cell, the downlink bandwidth of the cell, etc., and by coding conversion.

[0092] Define the node characteristics of each user node as a d B The vector of dimension is obtained by combining the values ​​of fields such as ARP priority, QCI priority, PRB utilization priority and far-point priority of the transferred user corresponding to the user node and through coding conversion.

[0093] The cell switching method for load balancing provided by the present invention, after determining the candidate neighboring cells and the transferred users, obtains the measurement report information of each transferred user relative to each candidate neighboring cell by sending a measurement command and event information of the candidate neighboring cells to the transferred users, and determines the initial edge weight of each edge in the load distribution diagram for associating a neighboring cell node and a user node according to the measurement report information, thereby improving the accuracy of determining the edge weight of each edge in the load distribution diagram, and helping to improve the accuracy of matching the most suitable target cell for switching for the transferred users.

[0094] Based on the above embodiment, as an optional embodiment, the measurement report information includes RSRP and SINR; and determining the edge weight of each edge for associating a neighboring node and a user node based on the measurement report information includes: Determine a first quotient value based on the RSRP of the candidate neighboring cell associated with the one neighboring cell node and the RSRP of the current cell; Determine a second quotient value based on the SINR of the candidate neighboring cell associated with the neighboring cell node and the SINR of the current cell; The edge weight is determined based on a product of the first quotient value and the second quotient value.

[0095] Specifically, the measurement report information uploaded by the transferred user includes the RSRP of several candidate neighboring areas covering the transferred user and the RSRP of the current cell of the transferred user. When determining the edge weight of each edge, a first quotient value is determined based on the RSRP of the candidate neighboring area associated with a neighboring area node and the RSRP of the current cell, and a second quotient value is determined based on the SINR of the candidate neighboring area associated with the same neighboring area node and the SINR of the current cell. Finally, the product of the first quotient value and the second quotient value is determined as the edge weight.

[0096] Optionally, the expression for determining the edge weight of each edge for associating a neighboring node and a user node based on the measurement report information is as follows: .

[0097] For example, a neighboring node of a candidate neighboring area A i User node with a transferred user B i The edge weight between them is equal to the product of the first quotient and the second quotient, where the first quotient is equal to the value measured by the transfer user. A i The second quotient value is equal to the RSRP of the transferred user measured by the transferred user. A i The SINR of the transferred user is compared with the SINR of the current cell measured by the transferred user.

[0098] The cell switching method for load balancing provided by the present invention sets the edge weight between the user node corresponding to the transferred user and the neighboring cell node corresponding to the candidate neighboring cell to be larger as the RSRP and SINR measured by the transferred user are closer to the candidate neighboring cell of the current cell. Thus, when determining the target cell for the transferred user to switch, the candidate neighboring cell similar to the current cell can be tended to be determined as the target cell, thereby reducing the impact on the connected user when switching cells.

[0099] Based on the above embodiment, as an optional embodiment, the edge weight prediction model is used to predict the edge weight of each edge in the load distribution graph to obtain the load distribution graph after the edge weight is updated, including: Determine an updated node feature of the neighboring node based on a node feature of the neighboring node and a node feature of a user node adjacent to the neighboring node that is fused together; Determine an updated node feature of the user node based on a fusion of the node feature of the user node and the node features of neighboring nodes adjacent to the user node; Determining a connection possibility score based on the updated node characteristics of the neighboring node and the updated node characteristics of the user node; Based on the connection possibility score, the edge weight of each edge in the load distribution graph is updated to obtain the load distribution graph after the edge weight is updated.

[0100] Specifically, after constructing a load distribution graph using the node features of all neighboring nodes, the node features of all user nodes, and the initial edge weights of each edge, the load distribution graph is input into the trained edge weight prediction model, and each layer of the network in the edge weight prediction model updates the node features of the neighboring nodes and the user nodes, and calculates the connection possibility scores based on the updated node features of the neighboring nodes and the node features of the user nodes. The calculated connection possibility scores are used as new edge weights for the edges associating the neighboring nodes and the user nodes, to obtain the load distribution graph after the edge weights are updated.

[0101] In one embodiment, the expression for determining the updated node feature of the neighboring node based on the node feature of the neighboring node and the node feature of the user node adjacent to the neighboring node is as follows: ; in, l is the edge weight prediction model l layer; Neighboring node A i Updated node features; Neighboring node A i own d A Dimension node features; N A,i Neighboring node A i The set of all adjacent user nodes; Is a collection N A,i The absolute value of N A,i The number of elements in the neighboring nodes A i The number of all adjacent user nodes; Is a collection N A,i Middle j User Node B j Node characteristics of yes The corresponding weight matrix is ​​the parameter obtained by training the edge weight prediction model using the training sample set; yes The corresponding weight matrix is ​​the parameter obtained by training the edge weight prediction model using the training sample set; is the activation function.

[0102] In one embodiment, the expression for determining the updated node feature of the user node based on the node feature of the user node and the node feature of the neighboring nodes adjacent to the user node is as follows: ; in, l is the edge weight prediction model l layer; User Node B i Updated node features; Is a user node B i own d B Dimension node features; N B,i Is a user node B i The set of all adjacent neighboring nodes; Is a collection N B,i The absolute value of N B,i The number of elements in the user node B i The number of all adjacent neighbor nodes; Is a collection N B,i Middle j Neighborhood Nodes A j Node characteristics of yes The corresponding weight matrix is ​​the parameter obtained by training the edge weight prediction model using the training sample set; yes The corresponding weight matrix is ​​the parameter obtained by training the edge weight prediction model using the training sample set; is the activation function.

[0103] In one embodiment, the expression for determining the connection possibility score based on the updated node features of the neighboring node and the updated node features of the user node is as follows: ; in, l is the edge weight prediction model l layer; Neighboring node A i and user nodes B i The connection probability score of Neighboring node Ai Updated node features; Is a user node B i Updated node features; The dimension is d A × d B The weight matrix is ​​the parameter obtained by training the edge weight prediction model using the training sample set.

[0104] It can be understood that, based on the heterogeneous bipartite graph property of the load distribution graph, the nodes adjacent to any neighboring node are all user nodes, and the nodes adjacent to any user node are all neighboring nodes.

[0105] The cell switching method for load balancing provided by the present invention updates itself by using the node characteristics of all user nodes adjacent to the neighboring cell node, updates itself by using the node characteristics of all neighboring cell nodes adjacent to the user node, and uses the updated node characteristics to calculate the connection possibility score to update the edge weight of each edge, so as to determine the candidate neighboring cell corresponding to the neighboring cell node associated with the edge with the highest updated edge weight as the target cell for transferring the user, thereby making full use of the topological structure information of the communication network, efficiently mining and extracting useful features from the data, accurately matching the user to the most suitable target cell for switching, and thus achieving optimal load balancing between cells.

[0106] On the whole, the cell switching method for load balancing provided by the present invention aims at the actual load balancing problem in the existing 5G network, abstractly models the candidate neighboring cells, transferred users and the relationship between them to design a load distribution diagram, and uses historical data such as network load status and load balancing strategy to construct a load distribution diagram dataset, trains the edge weight prediction model through the definition of input and output, and converts the technical problem of load balancing into the edge prediction problem of heterogeneous bipartite graph; compares the resource gap between the current cell and the neighboring cell based on multiple comprehensive indicators such as the number of users, PRB utilization and CCE utilization, and determines the candidate neighboring cell according to the indicator ranking, avoiding setting dependent The empirical candidate neighboring cell threshold is used to improve the generalization ability for different scenarios; active feature learning based on the graph neural network model can fully integrate multi-dimensional information such as the topological structure, load status and measurement results of the 5G network, and realize efficient extraction of useful features from complex network data, providing a basis for the accurate matching of candidate neighboring cells and transferred users from the perspective of data structure design; the prediction ability of the trained model is used to realize one-to-many matching between candidate neighboring cells and transferred users, and the most suitable neighboring cell is selected for the transferred user to switch according to the predicted edge weights, which can improve the decision-making efficiency of load balancing and provide an efficient auxiliary tool for future intelligent and automated decision-making of load balancing.

[0107] Figure 3 is a schematic diagram of the structure of a cell switching device for load balancing provided by the present invention, such as Figure 3 As shown, the cell switching device for load balancing includes but is not limited to a load distribution diagram construction module 301 , an edge weight updating module 302 and a target cell switching module 303 .

[0108] The load distribution diagram construction module 301 is used to construct a load distribution diagram using neighboring cell nodes and user nodes; the neighboring cell nodes are determined based on candidate neighboring cells of the current cell; the user nodes are determined based on transfer users connected to the current cell; each edge of the load distribution diagram is used to associate a neighboring cell node and a user node.

[0109] The edge weight updating module 302 is used to predict the edge weight of each edge in the load distribution graph by using the edge weight prediction model to obtain the load distribution graph after the edge weight is updated.

[0110] The target cell switching module 303 is used to determine the target cell for switching the transferred user based on the load distribution diagram after the edge weight is updated.

[0111] It should be noted that the cell switching device for load balancing provided by the present invention can execute the cell switching method for load balancing described in any of the above embodiments during specific operation, which will not be described in detail in this embodiment.

[0112] The cell switching device for load balancing provided by the present invention uses candidate neighboring cells and transferred users as nodes of a load distribution graph, and each edge of the load distribution graph is associated with a neighboring cell node and a user node, so as to construct a load distribution graph as a heterogeneous bipartite graph, thereby converting the load balancing problem into an edge prediction problem of the heterogeneous bipartite graph, and further using a neural network model to predict the edge weights of the load distribution graph, so as to determine the candidate neighboring cell corresponding to the neighboring cell node associated with the edge with the highest edge weight as the target cell for the transferred user, thereby making full use of the topological structure information of the communication network, efficiently mining and extracting useful features from the data, accurately matching the user to the most suitable target cell for switching, and thus achieving optimal load balancing between cells.

[0113] Figure 4 is a schematic diagram of the structure of the electronic device provided by the present invention, such as Figure 4As shown, the electronic device may include: a processor (Processor) 410, a communication interface (Communications Interface) 420, a memory (Memory) 430 and a communication bus 440, wherein the processor 410, the communication interface 420, and the memory 430 complete mutual communication through the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute the load-balancing-oriented cell switching method provided in any of the above embodiments, and the load-balancing-oriented cell switching method includes but is not limited to the following steps: constructing a load distribution graph using neighboring cell nodes and user nodes; the neighboring cell nodes are determined based on candidate neighboring cells of the current cell; the user nodes are determined based on the transferred users connected to the current cell; each edge of the load distribution graph is used to associate a neighboring cell node and a user node; the edge weight prediction model is used to predict the edge weight of each edge in the load distribution graph to obtain a load distribution graph after the edge weight is updated; based on the load distribution graph after the edge weight is updated, determining the target cell for the transferred user to switch.

[0114] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0115] On the other hand, the present invention also provides a computer program product, which includes a computer program, and the computer program 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 load-balancing oriented cell switching method provided in any of the above embodiments, and the load-balancing oriented cell switching method includes but is not limited to the following steps: using neighboring cell nodes and user nodes to construct a load distribution map; the neighboring cell nodes are determined based on candidate neighboring cells of the current cell; the user nodes are determined based on transfer users connected to the current cell; each edge of the load distribution map is used to associate a neighboring cell node and a user node; the edge weight prediction model is used to predict the edge weight of each edge in the load distribution map to obtain a load distribution map after the edge weight is updated; based on the load distribution map after the edge weight is updated, the target cell for switching the transfer user is determined.

[0116] On the other hand, 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 load-balancing oriented cell switching method provided in any of the above-mentioned embodiments, wherein the load-balancing oriented cell switching method includes but is not limited to the following steps: constructing a load distribution graph using neighboring cell nodes and user nodes; the neighboring cell nodes are determined based on candidate neighboring cells of the current cell; the user nodes are determined based on transfer users connected to the current cell; each edge of the load distribution graph is used to associate a neighboring cell node and a user node; the edge weight of each edge in the load distribution graph is predicted using an edge weight prediction model to obtain a load distribution graph after the edge weight is updated; and determining the target cell for switching the transfer user based on the load distribution graph after the edge weight is updated.

[0117] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0118] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0119] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A cell switching method for load balancing, characterized in that: include: Use neighboring nodes and user nodes to build a load distribution map; The neighboring cell node is determined based on the candidate neighboring cells of the current cell; The user node is determined based on the transferred user connected to the current cell; each edge of the load distribution graph is used to associate a neighboring cell node with a user node; Using an edge weight prediction model to predict the edge weight of each edge in the load distribution graph, to obtain a load distribution graph after the edge weight is updated; Based on the load distribution diagram after the edge weight is updated, a target cell for handover of the transferred user is determined.

2. The cell switching method for load balancing according to claim 1, characterized in that: Before constructing the load distribution diagram using the neighboring cell nodes and the user nodes, determining the candidate neighboring cells based on the neighboring cells of the current cell includes: Determining a comprehensive load index based on the load indexes of each dimension of the neighboring area; Based on the comprehensive load index, the candidate neighboring area is determined.

3. The cell switching method for load balancing according to claim 1, characterized in that: Before constructing the load distribution diagram using neighboring cell nodes and user nodes, determining the transferred user based on the user connected to the current cell includes: Determine the transfer priority based on the priorities of each dimension of the user; Based on the transfer priority, the transfer user is determined.

4. The cell switching method for load balancing according to claim 1, characterized in that: The method of constructing a load distribution diagram by using neighboring nodes and user nodes includes: Acquire measurement report information; the measurement report information is measured by the transferring user after receiving the event information of the candidate neighboring area; the event information of the candidate neighboring area is sent by the current cell to the transferring user; Determining, based on the measurement report information, an initial edge weight of each edge used to associate a neighboring node with a user node; The load distribution graph is constructed using the node characteristics of the neighboring nodes, the node characteristics of the user nodes, and the initial edge weight of each edge.

5. The cell switching method for load balancing according to claim 4, characterized in that: The measurement report information includes reference signal received power and signal to interference plus noise ratio; The determining, based on the measurement report information, an edge weight of each edge used to associate a neighboring node with a user node comprises: Determine a first quotient value based on the reference signal received power of the candidate neighboring cell associated with the one neighboring cell node and the reference signal received power of the current cell; Determine a second quotient value based on a signal to interference plus noise ratio of a candidate neighboring cell associated with the one neighboring cell node and a signal to interference plus noise ratio of the current cell; The edge weight is determined based on a product of the first quotient value and the second quotient value.

6. The cell switching method for load balancing according to claim 1, characterized in that: The method of using the edge weight prediction model to predict the edge weight of each edge in the load distribution graph to obtain the load distribution graph after the edge weight is updated includes: Determine an updated node feature of the neighboring node based on a node feature of the neighboring node and a node feature of a user node adjacent to the neighboring node that is fused together; Determine an updated node feature of the user node based on a fusion of the node feature of the user node and the node features of neighboring nodes adjacent to the user node; Determining a connection possibility score based on the updated node characteristics of the neighboring node and the updated node characteristics of the user node; Based on the connection possibility score, the edge weight of each edge in the load distribution graph is updated to obtain the load distribution graph after the edge weight is updated.

7. A cell switching device for load balancing, characterized in that: include: A load distribution map building module is used to build a load distribution map using neighboring nodes and user nodes; The neighboring cell node is determined based on the candidate neighboring cells of the current cell; The user node is determined based on the transferred user connected to the current cell; each edge of the load distribution graph is used to associate a neighboring cell node with a user node; An edge weight updating module, used to predict the edge weight of each edge in the load distribution graph by using an edge weight prediction model, and obtain a load distribution graph after the edge weight is updated; The target cell switching module is used to determine the target cell for switching the transferred user based on the load distribution diagram after the edge weight is updated.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the load-balancing oriented cell switching method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the load-balancing oriented cell switching method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the load-balancing oriented cell switching method according to any one of claims 1 to 6 is implemented.

Images