Cell load balancing method, apparatus, device and computer program product

Abstract

The application discloses a cell load balancing method, device, equipment and computer program product, and belongs to the technical field of communication. The method comprises the following steps: acquiring a pre-constructed cell distribution topology structure diagram, wherein the cell distribution topology structure diagram comprises a plurality of cells and the adjacency relationship between the cells; determining load prediction data of each cell according to historical load data of each cell in the cell distribution topology structure diagram; determining a target adjustment strategy of each cell according to the load prediction data of each cell and a target strategy network, so as to realize cell load balancing; wherein the target strategy network is obtained by training based on the historical load data of the plurality of cells and the adjacency relationship between each cell in the cell distribution topology structure diagram. In this way, the load condition of each cell at a future time can be dynamically predicted, and the target adjustment strategy can be determined in time, so that the resource utilization rate of the whole network is maximized.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of communication, and in particular, to a cell load balancing method, device, equipment and computer program product. BACKGROUND

[0002] In a mobile communication network, the loads of different cells can be different, some cells can be heavily loaded, while other cells are relatively lightly loaded. If the loads are not balanced, some cells can be over-utilized, while other cells are idle, which can affect the overall performance of the network and the experience of users. Therefore, by implementing cell load balancing, the utilization of network resources can be effectively optimized, the capacity and coverage of the network can be improved, and the communication quality and experience of users can be improved.

[0003] The related methods for implementing cell load balancing can be divided into the following types:

[0004] (1) Obtain the load information of adjacent cells, when the load of a certain cell is too heavy, perform load balancing by moving user equipment to an adjacent cell with relatively light load, for example, determine the mobility parameter that needs to be modified according to the load information of the cell and the load information of the adjacent cell, send a parameter modification request to the adjacent cell; perform corresponding operations according to the information fed back by the adjacent cell in response to the parameter modification request;

[0005] (2) According to a pre-established prediction model, monitor the predicted load rate of the cell, and switch the users of the cell with a predicted load rate greater than a first threshold to a cell with a predicted load rate less than a second threshold.

[0006] However, in the related art, when the cell is in a high load state, user switching is performed, which can increase the switching delay, and the risk of switching failure is high, and the use of pre-set threshold values for load adjustment can result in a situation where the load rate of some cells is close to the threshold value, while the load rate of its adjacent cells is very low, and the utilization rate of network resources is not high. SUMMARY

[0007] Embodiments of the present application provide a cell load balancing method, device, equipment and computer program product to at least solve the problem of low utilization of network resources in related solutions.

[0008] To solve the above technical problems, the present application is implemented as follows:

[0009] In a first aspect, an embodiment of the present application provides a cell load balancing method, comprising: obtaining a pre-constructed cell distribution topology graph, the cell distribution topology graph comprising a plurality of cells and adjacency relationships between the cells, the adjacency relationships being used to represent that user equipment switching can occur between the cells; determining load prediction data of each of the cells according to historical load data of each of the cells in the cell distribution topology graph; determining a target adjustment strategy of each of the cells according to the load prediction data of each of the cells and a target strategy network, so as to achieve cell load balancing; wherein the target strategy network is obtained based on historical load data of the plurality of cells and the adjacency relationships between each of the cells in the cell distribution topology graph.

[0010] In a second aspect, an embodiment of the present application provides a cell load balancing apparatus, comprising: an obtaining module configured to obtain a pre-constructed cell distribution topology graph, the cell distribution topology graph comprising a plurality of cells and adjacency relationships between the cells, the adjacency relationships being used to represent that user equipment switching can occur between the cells; a determining module configured to determine load prediction data of each of the cells according to historical load data of each of the cells in the cell distribution topology graph; and a balancing module configured to determine a target adjustment strategy of each of the cells according to the load prediction data of each of the cells and a target strategy network, so as to achieve cell load balancing; wherein the target strategy network is obtained based on historical load data of the plurality of cells and the adjacency relationships between each of the cells in the cell distribution topology graph.

[0011] In a third aspect, an embodiment of the present application provides an electronic device, comprising a processor and a memory, the memory storing programs or instructions executable on the processor, and the programs or instructions, when executed by the processor, implement steps of the method according to the first aspect.

[0012] In a fourth aspect, an embodiment of the present application provides a readable storage medium, the readable storage medium storing programs or instructions, and the programs or instructions, when executed by a processor, implement steps of the method according to the first aspect.

[0013] In a fifth aspect, an embodiment of the present application provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, and the program instructions, when executed by a computer, cause the computer to perform steps of the method according to the first aspect.

[0014] In the embodiment of the present application, a pre-constructed cell distribution topology structure diagram is acquired, the cell distribution topology structure diagram including a plurality of cells and adjacency relationships between the cells, the adjacency relationships being used to represent that user equipment switching can occur between the cells; historical load data of each cell in the cell distribution topology structure diagram is acquired, and load prediction data of each cell is determined according to the historical load data; a target adjustment strategy of each cell is determined according to the load prediction data of each cell and a target strategy network, so as to realize cell load balancing; wherein the target strategy network is obtained by training based on historical load data of a plurality of cells and the adjacency relationships between each cell in the cell distribution topology structure diagram. In this way, the target adjustment strategy of each cell is determined according to the load prediction data of each cell and the target strategy network, which can dynamically predict the load of each cell at a future time and determine the target adjustment strategy in time, so as to avoid cell overload. Moreover, the target strategy network is obtained by training based on historical load data of a plurality of cells and the adjacency relationships between each cell in the cell distribution topology structure diagram, which can enable the target strategy network to learn the implicit relationship between a cell and an adjacent cell from the cell distribution topology structure diagram, and determine the target adjustment strategy of each cell according to the load prediction data and the cell topology structure, so as to maximize the resource utilization of the entire network.

[0015] It should be understood that the general description above and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0016] The accompanying drawings, which are incorporated into and form part of the specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the application.

[0017] Figure 1 Fig. 1 shows a flow diagram of a cell load balancing method according to an embodiment of the present application;

[0018] Figure 2 Fig. 3 shows a structure diagram of a deep learning network model according to an embodiment of the present application;

[0019] Figure 3 Fig. 5 shows a flow diagram of a load prediction data determination method according to an embodiment of the present application;

[0020] Figure 4 Fig. 7 shows another flow diagram of a cell load balancing method according to an embodiment of the present application;

[0021] Figure 5 Fig. 9 shows a structure diagram of a cell load balancing device according to an embodiment of the present application;

[0022] Figure 6 Fig. 11 shows a structure diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] The exemplary embodiments will be described in detail herein with reference to the attached drawings. The description herein relates to the drawings, in which the same numbers represent the same or similar elements throughout. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of apparatuses and methods consistent with some aspects of the present application as detailed in the appended claims.

[0024] In a mobile communication network, when a certain cell is overloaded, load balancing can be performed by moving user equipment to a neighboring cell with relatively light load. However, when the cell is in a high load state, performing user equipment switching again can increase the switching delay and increase the risk of switching failure. In addition, some user equipment in the high load cell can already have the problem of increased delay before switching.

[0025] In related solutions, a cell load rate prediction model is established to monitor and adjust the load. Although this method can predict the load of each cell in advance, the established cell load rate prediction model is usually a simple linear weighting of historical load rate and current load rate to predict the current load rate, and cannot depict the increasingly complex changes in cell load rate. In addition, whether user equipment switching is needed is determined according to a preset threshold, which can result in a scenario in which the load rate of a part of cells is close to the threshold, while the relative load rate of its neighboring cell is very low. Such a scenario is also a waste of network resources in practice.

[0026] To solve the above problems in the cell load balancing process, the embodiments of the present application provide a cell load balancing method. The method determines a target adjustment strategy for each cell according to load prediction data of each cell and a target strategy network, to dynamically predict the load of each cell at a future time and timely determine the target adjustment strategy, thereby avoiding cell overload.

[0027] Figure 1This illustration shows a flowchart of a cell load balancing method provided in an embodiment of this application. The execution subject of this method can be a terminal device or a server. The terminal device can be a personal computer, a mobile terminal device such as a mobile phone or tablet, or a user-used terminal device. The server can be a standalone server or a server cluster composed of multiple servers. Furthermore, the server can be a backend server for a specific service, or a backend server for a platform or application (such as an intelligent scheduling platform, a cell load balancing system, etc.). This embodiment uses a server as the execution subject for illustration. For the case of a terminal device, the following related content can be used, and will not be elaborated further here. As shown in the figure, method 100 may include the following steps:

[0028] S101: Obtain a pre-constructed cell distribution topology map, which includes multiple cells and the adjacency relationships between the cells.

[0029] The adjacency relationship is used to characterize the ability of user equipment handover to occur between cells.

[0030] In practical implementation, the first step is to construct a cell distribution topology diagram. Specifically, each cell in the defined area is treated as a graph node v. i If user equipment handover can occur between cell i and cell j, then graph node v is considered to be... i And graph node v j There exists an undirected edge e. ij Based on this criterion, a topological undirected graph can be obtained. Then, obtain the distribution topology map of the community. This undirected graph topology includes multiple cells and the adjacency relationships between them.

[0031] Here, it can be determined whether a user equipment handover can occur between cell i and cell j by obtaining the neighbor cell list of the current cell i, or by using a pre-configured load balancing strategy.

[0032] S102: Determine the load forecast data for each of the cells based on the historical load data of each cell in the cell distribution topology diagram.

[0033] Historical load data can include historical graph data from multiple time points, with each historical graph data point representing a node v in the cell distribution topology map. iThe eigenvector can specifically include a PRB (Physical Resource Block) usage rate of a GBR (Guaranteed Bit Rate) service, a PRB usage rate of a non-GBR service, a PRB usage rate of a signaling bearer, a PRB usage rate of a common channel, and the like.

[0034] In a specific implementation, eigenvectors of all cells in the target area at the first t time points (with an interval of 15 minutes) can be acquired, that is, t historical graph data. The load conditions of each cell at the next time point are predicted according to the t historical graph data.

[0035] In a possible implementation, in the S102, the load prediction data of each cell is determined according to the historical load data of each cell in the cell distribution topology graph, and the method comprises the following steps of:

[0036] The historical load data of each cell in the cell distribution topology graph is acquired, and the historical load data comprises load data of each cell at multiple time points. The load data of each cell at the multiple time points is input into a deep learning network, the dependence relationship between the load data of each cell at the multiple time points is acquired by the deep learning network, and the load prediction data of each cell at the next time point is predicted.

[0037] In a specific implementation, a deep learning framework (such as TensorFlow, PyTorch, and the like) can be used to construct a deep learning network (Temporal Convolutional Network, TCN) model, as shown in Figure 2 The TCN network model can be composed of multiple residual blocks of different scales. By inputting the historical graph data at multiple time points into the TCN network model, the dependence relationship between the load data of each cell at the multiple time points is acquired by the TCN network model, and the load prediction data of each cell at the next time point is predicted.

[0038] In a possible implementation, in order to improve the prediction accuracy of the TCN network model, the real load data of each cell at the next time point in the cell distribution topology graph can be acquired; and the historical load data is updated according to the real load data.

[0039] In an exemplary embodiment, as shown in Figure 3 The load prediction data determination method can specifically comprise the following steps:

[0040] Step 301: acquiring a cell distribution topology graph;

[0041] Step 302: Obtain the previous t historical graph data, wherein the historical graph data includes the load data of each cell in the cell distribution topology graph at t time points;

[0042] Step 303: Input the previous t historical graph data into the TCN network model;

[0043] Step 304: Obtain the load prediction data of each cell at t+1 time point through the TCN network model;

[0044] Step 305: Obtain the real load data of each cell at t+1 time point, and return to step 302 to update the historical graph data and input the TCN network to predict the load data at the next time again, and so on.

[0045] S103: Determine the target adjustment strategy of each cell according to the load prediction data of each cell and the target strategy network, so as to realize the cell load balancing.

[0046] The target strategy network is trained based on the historical load data of the plurality of cells and the adjacency relationship between each cell in the cell distribution topology graph; the target adjustment strategy can include adjusting the cell's own center frequency channel number, adjusting the cell's antenna hanging height, adjusting the cell's antenna direction angle, adjusting the cell's electronic down angle, adjusting the cell's mechanical down angle, adjusting the cell's built-in down angle, and adjusting the key parameters for wireless network optimization.

[0047] In a specific implementation, the target strategy network is trained based on the historical load data of the plurality of cells and the adjacency relationship between each cell in the cell distribution topology graph, so that the target strategy network can learn the implicit relationship between the cell and the adjacent cell from the cell distribution topology graph, and determine the target adjustment strategy of each cell according to the load prediction data of each cell and the cell distribution topology, so as to realize the cell load balancing. In an exemplary embodiment, as shown in Figure 4 After step 304, the method further includes step 306: the reinforcement learning model adjusts the cell parameters according to the load prediction data and the target strategy network, and further obtains the real load data of each cell at t+1 time point.

[0048] In a possible implementation, the target strategy network can be trained in the following manner:

[0049] obtain first load data of a target cell in a target time point from the plurality of cells; determine neighboring cells of the target cell according to the cell distribution topology graph; pre-train a pre-constructed strategy network according to the first load data of the target cell and second load data of the neighboring cells in the target time point, obtain a reward value of the strategy network, and update network parameters of the strategy network to maximize the reward value, to obtain the target strategy network.

[0050] In a specific implementation, each cell is regarded as an agent, and a relationship between the agents is represented by an undirected graph If there is an edge between nodes, it means that the agents can perceive the state of each other and make decisions. In an example embodiment, the cell i first perceives a predicted load state s i,t of itself and the neighboring cells in a target time point t, the predicted load state s i,t is composed of first load data of the cell i in the target time point t and second load data of all neighboring cells in the target time point t; the agent takes an action a i,t according to the load rate of itself and the neighboring agents and the strategy π, that is, adjusts cell parameters; after the environment perceives the action of all agents, the state changes, and the environment feeds back a reward value r t to the selected action; the agent adjusts network parameters of the strategy network according to the reward value, to obtain a target strategy network. The predicted load state s i,t may be expressed as:

[0051] s i,t = [x i , x j1 , …, x jn ];

[0052] wherein j1, …, j n are indexes of all neighboring agents of the agent i, and x i is a feature vector of a graph node; the action a i,t taken by the agent includes adjustment of six parameters, that is, a self-center frequency channel number, an antenna hanging height, an antenna direction angle, an electronic downtilt angle, a mechanical downtilt angle, and a built-in downtilt angle.

[0053] In a possible implementation, the pre-training of the pre-constructed strategy network according to the first load data of the target cell and the second load data of the neighboring cells in the target time point, to obtain the reward value of the strategy network, includes:

[0054] The first load data and the second load data are input into a pre-constructed strategy network to obtain a parameter adjustment strategy of the target cell; and a reward value of the strategy network is determined according to the parameter adjustment strategy of the target cell and a preset reward function.

[0055] In a specific implementation, the first load data and the second load data are input into a pre-constructed strategy network, the strategy network outputs a parameter adjustment strategy (for example, adjusting a self-center carrier frequency channel number, an antenna hanging height, an antenna direction angle, an electronic down-tilt angle, a mechanical down-tilt angle, a built-in down-tilt angle, etc.) of the target cell, a reward value is obtained by evaluating the strategy adopted by the target cell through a reward function, and the calculated reward value is fed back to the strategy network, and the strategy network can adjust network parameters thereof according to the reward value.

[0056] The reward value of the strategy network is determined according to the parameter adjustment strategy of the target cell and a preset reward function, and includes:

[0057] Load data of each cell and a number of times of user equipment switching after the parameters of the target cell are adjusted based on the parameter adjustment strategy are obtained; a load balance degree is determined according to the load data of each cell, and the load balance degree is used to represent a balance degree of the load of each cell; and the reward value of the strategy network is determined according to the load balance degree and a reciprocal of the number of times of user equipment switching.

[0058] In a specific implementation, a setting rule of the reward function is as follows:

[0059] a) The selection of the action should make the load of each cell tend to be balanced;

[0060] b) The selection of the action cannot make the number of times of switching too much, otherwise the consumption of energy will be increased.

[0061] The load balance degree can be a variance, a standard deviation, an average value, etc. of the load of each cell, and an embodiment of the present application determines a maximum difference amount of an average load index of each cell according to the load data of each cell, and determines the load balance degree according to the maximum difference amount. In an exemplary embodiment, a calculation formula of the reward value r t is as follows:

[0062]

[0063] wherein ρ i is an average load index of the agent i, and specifically, a PRB usage rate in a feature vector can be added to obtain the average load index, S is a number of times of user switching of the entire area cell after the parameter adjustment. The strategy modeling is a random function π: s1×s2×…×s n→ A, parameterize the strategy through a deep neural network to improve the generalization ability of the model.

[0064] The embodiment of the application provides a cell load balancing method, a cell distribution topology structure graph is acquired, the cell distribution topology structure graph comprises a plurality of cells and adjacency relationships between the cells, the adjacency relationships are used for representing that user equipment switching can occur between the cells; load prediction data of each cell is determined according to historical load data of each cell in the cell distribution topology structure graph; a target adjustment strategy of each cell is determined according to the load prediction data of each cell and a target strategy network, so that cell load balancing is realized; wherein the target strategy network is obtained by training based on historical load data of the plurality of cells and the adjacency relationships between each cell in the cell distribution topology structure graph. In this way, the target adjustment strategy of each cell is determined according to the load prediction data of each cell and the target strategy network, the load conditions of each cell at a future time can be dynamically predicted and the target adjustment strategy can be determined in time, so that cell overload is avoided. Moreover, the target strategy network is obtained by training based on the historical load data of the plurality of cells and the adjacency relationships between each cell in the cell distribution topology structure graph, the target strategy network can learn the implicit relationships between the cells and the adjacent cells from the cell distribution topology structure graph, and the target adjustment strategy of each cell can be determined according to the load prediction data and the cell topology structure, so that the resource utilization rate of the whole network is maximized.

[0065] Figure 5 The structure schematic diagram of the cell load balancing device provided by the embodiment of the application is shown, and the cell load balancing device can realize all or part of the contents in the embodiments shown in Figure 1 or Figure 4 The cell load balancing device 500 comprises:

[0066] The acquisition module 510 is used for acquiring a cell distribution topology structure graph constructed in advance, the cell distribution topology structure graph comprises a plurality of cells and adjacency relationships between the cells, and the adjacency relationships are used for representing that user equipment switching can occur between the cells.

[0067] The determination module 520 is used for determining load prediction data of each cell according to historical load data of each cell in the cell distribution topology structure graph.

[0068] The balancing module 530 is used for determining a target adjustment strategy of each cell according to the load prediction data of each cell and a target strategy network, so that cell load balancing is realized; wherein the target strategy network is obtained by training based on historical load data of the plurality of cells and the adjacency relationships between each cell in the cell distribution topology structure graph.

[0069] In a possible implementation, the determining module 520 comprises:

[0070] The data obtaining unit is configured to obtain historical load data of each cell in the cell distribution topology graph, the historical load data comprising load data of each cell at multiple time points;

[0071] The load predicting unit is configured to input the load data of each cell at the multiple time points into a deep learning network, obtain a dependency relationship between the load data of each cell at the multiple time points through the deep learning network, and predict load prediction data of each cell at a next time point.

[0072] The determining module 520 further comprises:

[0073] The data updating unit is configured to obtain real load data of each cell at the next time point in the cell distribution topology graph, and update the historical load data according to the real load data.

[0074] In a possible implementation, the balancing module 530 is configured to train the target policy network in the following manner:

[0075] Obtain first load data of a target cell in the multiple cells at a target time point;

[0076] Determine adjacent cells of the target cell according to the cell distribution topology graph;

[0077] Pre-train a pre-constructed policy network according to the first load data of the target cell and second load data of the adjacent cells at the target time point, obtain a reward value of the policy network, and update network parameters of the policy network to maximize the reward value, to obtain the target policy network.

[0078] In a possible implementation, the balancing module 530, when pre-training the pre-constructed policy network according to the first load data of the target cell and the second load data of the adjacent cells at the target time point to obtain the reward value of the policy network, is specifically configured to:

[0079] Input the first load data and the second load data into the pre-constructed policy network, to obtain a parameter adjustment strategy of the target cell;

[0080] Determine the reward value of the policy network according to the parameter adjustment strategy of the target cell and a preset reward function.

[0081] The balancing module 530 is specifically configured to:

[0082] obtain load data of each cell and the number of user equipment handovers after adjusting the parameters of the target cell based on the parameter adjustment strategy;

[0083] determine a load balancing degree according to the load data of each cell, the load balancing degree being used to represent the balancing degree of the load of each cell;

[0084] determine the reward value of the strategy network according to the load balancing degree and the reciprocal of the number of user equipment handovers.

[0085] The embodiment of the present application provides a cell load balancing device, which comprises an obtaining module, a determining module and a balancing module. The obtaining module obtains a pre-constructed cell distribution topology structure diagram, the cell distribution topology structure diagram comprising a plurality of cells and adjacency relationships between the cells, the adjacency relationships being used to represent that user equipment handovers can occur between the cells. The determining module determines load prediction data of each cell according to historical load data of each cell in the cell distribution topology structure diagram. The balancing module determines a target adjustment strategy of each cell according to the load prediction data of each cell and a target strategy network, so as to realize cell load balancing. The target strategy network is obtained based on historical load data of a plurality of cells and the adjacency relationships between the cells in the cell distribution topology structure diagram. In this way, the target adjustment strategy of each cell is determined according to the load prediction data of each cell and the target strategy network, the load conditions of each cell at a future time can be dynamically predicted, and the target adjustment strategy can be determined in time, so that cell overload is avoided. Moreover, the target strategy network is obtained based on the historical load data of a plurality of cells and the adjacency relationships between the cells in the cell distribution topology structure diagram, the target strategy network can learn the implicit relationships between the cells and the adjacent cells from the cell distribution topology structure diagram, and the target adjustment strategy of each cell can be determined according to the load prediction data and the cell topology structure, so that the resource utilization rate of the whole network is maximized.

[0086] Figure 6A hardware structure schematic diagram of an electronic device is shown, and reference is made to the diagram. At the hardware level, the electronic device 600 includes a processor 610, and optionally, an internal bus 620, a network interface 630, and a memory 640. The memory 640 can include a memory 641, such as a random-access memory (RAM), and can also include a non-volatile memory 642, such as at least one disk memory. Of course, the electronic device can also include other hardware required by a business.

[0087] The processor 610, the network interface 630, and the memory can be connected to each other through the internal bus 620, which can be an industry standard architecture (ISA) bus, a peripheral component interconnect (PCI) bus, or an extended industry standard architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one bidirectional arrow is shown in the diagram, but it does not mean that there is only one bus or only one type of bus.

[0088] The memory 640 stores programs. Specifically, the programs can include program code, which includes computer operation instructions. The memory 640 can include the memory 641 and the non-volatile memory 642, and provide instructions and data to the processor 610.

[0089] The processor 610 reads the corresponding computer program from the non-volatile memory 642 into the memory and then runs, and forms a device for positioning a target user at the logical level. The processor 610 executes the programs stored in the memory, and specifically executes: Figure 1 Or Figure 4 The method disclosed in the illustrated embodiment achieves the functions and beneficial effects of each method described in the foregoing method embodiment, and thus will not be described here.

[0090] The above as described in the present application Figure 1 Or Figure 4The method disclosed by the embodiment shown can be applied to the processor 610 or implemented by the processor 610. The processor 610 can be an integrated circuit chip with processing capability. In the implementation process, each step of the above method can be completed by integrated logic circuits in hardware or instructions in software form in the processor 610. The processor 610 described above can be a general processor, including a central processing unit (CPU), a network processor (NP), etc.; can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component. Each method, step and logic block diagram disclosed in the embodiment of the present application can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in combination with the embodiment of the present application can be directly embodied as a hardware code processor for execution, or a combination of hardware and software modules in the code processor for execution. The software module can be located in a random memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium in the storage, the processor reads the information in the storage, and combines the hardware to complete the steps of the above method.

[0091] The computer device can also execute the methods described in the foregoing method embodiments, and achieve the functions and beneficial effects of the methods described in the foregoing method embodiments, which will not be repeated here.

[0092] Of course, in addition to the software implementation, the electronic device of the present application does not exclude other implementation manners, such as logic devices or a combination of software and hardware, etc. That is, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or a logic device.

[0093] The embodiment of the present application also proposes a computer readable storage medium, the computer readable medium stores one or more programs, the one or more programs when being executed by the electronic device including a plurality of application programs, make the electronic device execute Figure 1 Or Figure 4 The method disclosed by the embodiment shown and the functions and beneficial effects of the methods described in the foregoing method embodiments are not repeated here.

[0094] The computer readable storage medium includes a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0095] Further, the embodiment of the present application further provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions which, when executed by a computer, implement the following processes: Figure 1 or Figure 4 The method disclosed in the embodiments shown and the functions and beneficial effects of the methods described in the foregoing method embodiments are not described here again.

[0096] The embodiments of the present application can be applied to various electronic device cooperation or interconnection scenarios, including: mobile phone and notebook computer / tablet computer cooperation and interconnection; mobile terminal and smart television / display cooperation and interconnection; mobile phone, or tablet computer and vehicle entertainment system cooperation and interconnection; mobile terminal and smart conference system cooperation and interconnection, etc. Thus, the needs of users in smart home, smart office, smart travel and other diversified scenarios are met.

[0097] In summary, the above only describes the preferred embodiments of the present application, and does not limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0098] The system, device, module or unit described in the above embodiments can be specifically implemented by a computer chip or entity, or by a product with certain functions. A typical implementation device is a computer. Specifically, the computer may, for example, be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0099] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can implement information storage by any method or technology. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can store information accessible by a computing device. According to the definition herein, computer-readable media does not include transitory media such as modulated data signals and carriers.

[0100] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusions, so that a process, method, article or apparatus that includes a list of elements does not only include those elements, but also includes other elements not explicitly listed, or inherent to such a process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0101] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments. In particular, for the system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the part of the method embodiments.

Claims

1. A method for cell load balancing, characterized in that, The method comprises the following steps: obtaining a pre-constructed cell distribution topology graph, wherein the cell distribution topology graph comprises a plurality of cells and the adjacency relationship between each of the cells, and the adjacency relationship is used to represent that user equipment switching can occur between cells; determining load prediction data of each of the cells according to historical load data of each of the cells in the cell distribution topology graph; determining a target adjustment strategy of each of the cells according to the load prediction data of each of the cells and a target strategy network, so as to achieve cell load balancing; wherein the target strategy network is obtained by training based on historical load data of a plurality of the cells and the adjacency relationship between each of the cells in the cell distribution topology graph; wherein the target strategy network is trained in the following manner: obtaining first load data of a target cell in a target time point in the plurality of cells; determining an adjacent cell of the target cell according to the cell distribution topology graph; pre-training a pre-constructed strategy network according to the first load data of the target cell and second load data of the adjacent cell in the target time point, obtaining a reward value of the strategy network, and updating network parameters of the strategy network to maximize the reward value to obtain the target strategy network.

2. The method of claim 1, wherein, The determination of the load prediction data of each of the cells according to the historical load data of each of the cells in the cell distribution topology graph comprises: obtaining historical load data of each of the cells in the cell distribution topology graph, wherein the historical load data comprises load data of each of the cells at a plurality of time points; inputting the load data of each of the cells at a plurality of time points into a deep learning network, obtaining a dependency relationship between the load data of each of the cells at a plurality of time points through the deep learning network, and predicting load prediction data of each of the cells at a next time point.

3. The method of claim 2, wherein, After determining the parameter adjustment strategy of each target cell according to the load prediction data of each of the cells and the target strategy network, the method further comprises: obtaining real load data of each of the cells at the next time point in the cell distribution topology graph; updating the historical load data according to the real load data.

4. The method of claim 1, wherein, The pre-training of the pre-constructed strategy network according to the first load data of the target cell and the second load data of the adjacent cell in the target time point, and the obtaining of the reward value of the strategy network, comprise: inputting the first load data and the second load data into the pre-constructed strategy network to obtain the parameter adjustment strategy of the target cell; determining the reward value of the strategy network according to the parameter adjustment strategy of the target cell and a preset reward function.

5. The method of claim 4, wherein, The determination of the reward value of the strategy network according to the parameter adjustment strategy of the target cell and the preset reward function comprises: obtaining load data of each of the cells and the number of times of user equipment switching after adjusting the parameters of the target cell based on the parameter adjustment strategy; determining a load balancing degree according to the load data of each of the cells, wherein the load balancing degree is used to represent the balancing degree of the load of each of the cells; According to the load balancing degree and the reciprocal of the number of times of user equipment switching, a reward value of the policy network is determined.

6. A cell load balancing apparatus, characterized by comprising: The method comprises the following steps: An acquisition module is configured to acquire a pre-constructed cell distribution topology graph, wherein the cell distribution topology graph comprises a plurality of cells and adjacency relationships between the cells, and the adjacency relationships are used to represent that user equipment switching can occur between the cells. A determination module is configured to determine load prediction data of each cell according to historical load data of each cell in the cell distribution topology graph. An equalization module is configured to determine a target adjustment policy for each cell according to the load prediction data of each cell and a target policy network, so as to achieve cell load balancing, wherein the target policy network is obtained by training based on historical load data of a plurality of cells and the adjacency relationships between each cell in the cell distribution topology graph. The target policy network is trained in the following manner: First load data of a target cell at a target time point is acquired. Adjacent cells of the target cell are determined according to the cell distribution topology graph. The pre-constructed policy network is pre-trained according to the first load data of the target cell and second load data of the adjacent cells at the target time point, a reward value of the policy network is obtained, and network parameters of the policy network are updated to maximize the reward value, thereby obtaining the target policy network.

7. An electronic device, comprising: The electronic device comprises a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and the programs or instructions are executed by the processor to implement the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores programs or instructions, and the programs or instructions are executed by the processor to implement the steps of the method according to any one of claims 1 to 5.

9. A computer program product, comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions that, when executed by a computer, cause the computer to perform the steps of the method according to any one of claims 1 to 5.

Images