Network slice management method, apparatus and device, and storage medium and product

By setting the on/off status and time window of small base stations in network slices, and combining Markov decision process and federated game algorithm, the slice configuration and resource allocation are optimized, which solves the contradiction between energy consumption and service quality in large-scale network slice deployment, and achieves energy consumption reduction and spectrum efficiency improvement.

WO2026232644A1PCT designated stage Publication Date: 2026-11-12PENG CHENG LAB
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/CN2025/096426
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-05-07
Filing Date
2025-05-22
Publication Date
2026-11-12

AI Technical Summary

Technical Problem

Existing technologies have failed to effectively reduce energy consumption while ensuring service quality in large-scale network slicing deployments, and have neglected network decision optimization with mixed variables.

Method used

By setting the on/off status of small base stations, dividing time windows and time slots, and combining Markov decision processes and coalition game algorithms, slice configuration and resource allocation strategies are optimized to achieve slice energy consumption management.

Benefits of technology

While reducing network slicing energy consumption, we can avoid service interruption and ensure slicing spectrum efficiency and long-term performance requirements.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2025096426_12112026_PF_FP_ABST
    Figure CN2025096426_12112026_PF_FP_ABST
Patent Text Reader

Abstract

Disclosed in the present application are a network slice management method, apparatus and device, and a storage medium and a product. The method comprises: on the basis of slice energy consumption of a network slice, setting on / off states corresponding to a preset number of small base stations; dividing the network slice into a plurality of time windows, and on the basis of the on / off states, determining a slice configuration strategy for each small base station within each time window; and dividing each time window into a plurality of time slots, and determining a resource allocation strategy corresponding to the network slice in each time slot, so as to perform slice management.
Need to check novelty before this filing date? Find Prior Art

Description

Network slicing management methods, devices, equipment, storage media and products

[0001] Related applications

[0002] This application claims priority to Chinese patent application No. 202510583355.6, filed on May 7, 2025, the entire contents of which are incorporated herein by reference. Technical Field

[0003] This application relates to the field of network resource management technology, and in particular to a network slice management method, apparatus, device, storage medium and product. Background Technology

[0004] 6G wireless access networks support a variety of applications, including immersive communication, ultra-reliable low-latency communication, and massive data acquisition. Applications with varying requirements for data rate, latency, reliability, and connection density demand different levels of service quality and user experience. To achieve this, network slicing provides flexible network management and supports diverse services in large-scale heterogeneous wireless networks, demonstrating significant application potential in 6G networks.

[0005] Existing solutions, by combining resource management and reinforcement learning methods, can adaptively allocate resources based on current network conditions and user needs, effectively achieving real-time monitoring and predictive analysis of Radio Access Network Slicing (RAN Slicing). This promotes the development of intelligent resource management for network slices and improves network performance and resource utilization efficiency to some extent. In practical applications, energy consumption, as a key performance indicator for network operation, receives widespread attention from operators and slice customers. However, most existing methods optimize energy consumption from the perspective of the system or equipment, rarely focusing on the operational energy consumption of the slice itself, and ignoring the additional energy consumption caused by deploying a large number of slices in large-scale network scenarios to ensure service isolation and quality of service. Furthermore, real-world network environments are often highly complex, requiring hybrid network decisions involving both discrete and continuous variables. Existing solutions are mostly designed for network decisions with only a single variable type. These factors mean that existing technologies still have room for optimization in terms of slice energy consumption and hybrid action space algorithms. How to reduce slice energy consumption while ensuring quality of service requirements remains a critical challenge that urgently needs research. Summary of the Invention

[0006] The main objective of this application is to provide a network slice management method, apparatus, device, storage medium, and product, which aims to solve the technical problem of how to reduce energy consumption while ensuring slice performance when deploying a large number of network slices.

[0007] To achieve the above objectives, this application proposes a network slice management method, the method comprising:

[0008] The on / off states of a preset number of small base stations are set according to the slice energy consumption of the network slice. The slice energy consumption is the total energy consumption generated by all devices in the network slice. The small base station is the base station covered by the network slice whose power is less than the power threshold and whose coverage area is less than the range threshold.

[0009] The network slice is divided into multiple time windows, and the slice configuration strategy for each small base station in each time window is determined according to the switch status.

[0010] The time window corresponding to each network slice is divided into multiple time slots, and the resource allocation strategy for each network slice in each time slot is determined.

[0011] Slice management is performed according to the slice configuration strategy and the resource allocation strategy.

[0012] In one embodiment, the step of setting the on / off states of a preset number of small base stations based on the slice energy consumption of the network slice includes:

[0013] Within the network slicing time window, the energy consumption of the virtual device and the energy consumption of the remote wireless head are summed to obtain the network virtual energy consumption, wherein the energy consumption of the virtual device refers to the energy consumption generated by the optical devices in the virtualization server and the baseband unit pool.

[0014] When the network slice switches from the on / off state to the on state in the current time window, determine the slice restart energy consumption corresponding to the virtual machine that restarts the network slice;

[0015] The network virtual energy consumption and the slice restart energy consumption are summed to obtain the slice energy consumption of the network slice in each time window;

[0016] When the energy consumption of the network slice in the time window is lower than a preset energy consumption threshold, the switch status of a preset number of small base stations is set to the off state.

[0017] In one embodiment, the step of dividing the network slice into multiple time windows and determining the slice configuration strategy for each small base station within each time window based on the on / off state includes:

[0018] The network slice is divided into multiple time windows, and the bandwidth reservation ratio is determined based on the total bandwidth of each small base station and the number of network slices in each small base station whose switch status is on.

[0019] The slicing configuration strategy for each small base station within each time window is determined based on the bandwidth reservation ratio.

[0020] In one embodiment, the step of determining the slice configuration strategy for each small base station within each time window based on the bandwidth reservation ratio includes:

[0021] The bandwidth reservation decision for each small base station within each time window is determined based on the bandwidth reservation ratio.

[0022] A Markov decision process is constructed based on at least one of the following: the number of user devices within the time window corresponding to the network slice, the switch status, the bandwidth reservation decision, the preset slice cost, and the discount factor. The discount factor is used to adjust the reward of the Markov decision process.

[0023] The Markov decision process is trained by storing historical data generated during the training process in the experience replay area, and random sample data is randomly drawn from the experience replay area to iteratively train the Markov decision process to obtain the optimal strategy.

[0024] The total number of selectable actions in the optimal strategy is used as input and mapped to an output header with a preset number of dimensions, where the preset number of dimensions is the number of action dimensions of the optimal strategy.

[0025] The sub-actions of the overall action in each dimension are determined based on each of the output heads;

[0026] The sub-actions are combined into an optimal action, which is then used as the optimal slicing configuration strategy for each small base station within each time window.

[0027] In one embodiment, the step of dividing the time window corresponding to each network slice into multiple time slots and determining the resource allocation strategy corresponding to each network slice in each time slot includes:

[0028] The time window corresponding to each network slice is divided into multiple time slots;

[0029] The associated user is determined based on the on / off state of each network slice within the time window, and the associated user represents the user equipment connected to the small base station.

[0030] The bandwidth allocation capacity for the associated users is determined based on the reserved bandwidth capacity.

[0031] The quality of service for each time slot of each network slice is determined based on the minimum transmission rate.

[0032] The resource allocation strategy for each network slice in each time slot is determined based on the bandwidth allocation capacity and the quality of service.

[0033] In one embodiment, the step of determining the resource allocation strategy for each network slice in each time slot based on the bandwidth allocation capacity and the quality of service includes:

[0034] Based on the user equipment corresponding to the associated user, the small base station covered by the network slice, the bandwidth allocation capacity, and the quality of service, construct a coalition game model for each network slice in each time slot;

[0035] User relationships are determined based on the priority order of user devices in the coalition game model.

[0036] The user association relationship is iteratively solved using convex optimization until Nash equilibrium is reached, thereby obtaining the optimal resource allocation strategy for each network slice in each time slot.

[0037] Furthermore, to achieve the above objectives, this application also proposes a network slice management device, which includes:

[0038] The slice switch status setting module is used to set the switch status of a preset number of small base stations according to the slice energy consumption of the network slice. The slice energy consumption is the total energy consumption generated by all devices in the network slice. The small base station is the base station covered by the network slice whose power is less than the power threshold and whose coverage area is less than the range threshold.

[0039] The slice configuration strategy determination module is used to divide the network slice into multiple time windows and determine the slice configuration strategy of each small base station in each time window according to the switch status.

[0040] The resource allocation strategy determination module is used to divide the time window corresponding to each network slice into multiple time slots and determine the resource allocation strategy corresponding to each network slice in each time slot.

[0041] The network slice management module is used to manage slices according to the slice configuration policy and the resource allocation policy.

[0042] In addition, to achieve the above objectives, this application also proposes a network slice management device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the network slice management method as described above.

[0043] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the network slice management method described above.

[0044] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the network slice management method described above.

[0045] This application provides a network slicing management method. It sets the on / off states of a preset number of small base stations (SNSs) based on the energy consumption of the network slice. SNSs are base stations covered by the network slice that have small power and coverage area. The network slice is divided into multiple time windows, and the slice configuration strategy for each SNS within each time window is determined based on the on / off state. The time windows are further divided into multiple time slots, and a resource allocation strategy is determined for each time slot. Slicing management is then performed based on the slice configuration strategy and the resource allocation strategy. This application designs a switch based on slice energy consumption load. When the slice load is low, slices on some SNSs on some SNSs are turned off to reduce energy consumption, while slices on large base stations remain on, thus avoiding service interruptions caused by turning off SNSs on SNSs. The slice configuration strategy and resource allocation strategy are determined based on the long-term scale of the time window and the short-term scale of the divided time slots, respectively. Slicing management is performed using both time scales, reducing network slice energy consumption while ensuring slice spectral efficiency and meeting long-term slice performance requirements. Attached Figure Description

[0046] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0047] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 is a flowchart of the network slice management method according to Embodiment 1 of this application;

[0049] Figure 2 is a schematic diagram of the physical network scenario in the network slice management method of this application;

[0050] Figure 3 is a diagram of the dual-time-scale management framework in the network slice management method of this application;

[0051] Figure 4 is a flowchart provided in Embodiment 2 of the network slice management method of this application;

[0052] Figure 5 is a diagram of the parameterized depth Q network structure in the network slice management method of this application;

[0053] Figure 6 is a flowchart provided in Embodiment 3 of the network slice management method of this application;

[0054] Figure 7 is a flowchart of the resource allocation method based on alliance game theory in the network slice management method of this application;

[0055] Figure 8 is a schematic diagram of the overall process of slice configuration and resource allocation in the network slice management method of this application;

[0056] Figure 9 is a schematic diagram of the module structure of the network slice management device according to an embodiment of this application;

[0057] Figure 10 is a schematic diagram of the hardware operating environment involved in the network slicing management method in this application embodiment.

[0058] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0059] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0060] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0061] The main solution of this application embodiment is as follows: A preset number of small base stations are set to switch on / off states based on the slice energy consumption of the network slice. The slice energy consumption is the total energy consumption generated by all devices in the network slice. The small base stations are base stations covered by the network slice whose power is less than a power threshold and whose coverage area is less than a range threshold. The network slice is divided into multiple time windows, and the slice configuration strategy for each small base station in each time window is determined based on the switch on / off states. The time window corresponding to each network slice is divided into multiple time slots, and the resource allocation strategy corresponding to each network slice in each time slot is determined. Slice management is performed based on the slice configuration strategy and the resource allocation strategy.

[0062] Existing solutions, by combining resource management and reinforcement learning methods, can adaptively allocate resources based on current network conditions and user needs, effectively achieving real-time monitoring and predictive analysis of Radio Access Network Slicing (RAN Slicing). This promotes the development of intelligent resource management for network slices and improves network performance and resource utilization efficiency to some extent. In practical applications, energy consumption, as a key performance indicator for network operation, receives widespread attention from operators and slice customers. However, most existing methods optimize energy consumption from the perspective of the system or equipment, rarely focusing on the operational energy consumption of the slice itself, and ignoring the additional energy consumption caused by deploying a large number of slices in large-scale network scenarios to ensure service isolation and quality of service. Furthermore, real-world network environments are often highly complex, requiring hybrid network decisions involving both discrete and continuous variables. Existing solutions are mostly designed for network decisions with only a single variable type. These factors mean that existing technologies still have room for optimization in terms of slice energy consumption and hybrid action space algorithms. How to reduce slice energy consumption while ensuring quality of service requirements remains a critical challenge that urgently needs research.

[0063] This application provides a solution that sets the on / off status of a preset number of small base stations (SNSs) based on the energy consumption of network slices. SNSs are base stations covered by the network slice that have small power and coverage area. The network slice is divided into multiple time windows, and the slice configuration strategy for each SNS within each time window is determined based on the on / off status. The time window is further divided into multiple time slots, and a resource allocation strategy is determined for each time slot. Slice management is performed based on the slice configuration strategy and the resource allocation strategy. This application designs a switch based on slice energy consumption load. When the slice load is low, slices on some SNSs on some SNSs are turned off to reduce energy consumption, while slices on large base stations remain on, thus avoiding service interruption caused by turning off SNSs on SNSs. The slice configuration strategy and resource allocation strategy are determined based on the long scale of the time window and the short scale of the divided time slots, respectively. Slice management is performed jointly on both time scales, reducing network slice energy consumption while ensuring slice spectral efficiency and meeting long-term slice performance requirements.

[0064] It should be noted that the executing entity of the method in this embodiment can be a network slice management device with network slice management, network communication, and program execution functions, wherein the network slice management device is a device connected to a 6G wireless access network. This embodiment and the following embodiments will be described using a network slice management device as an example.

[0065] Based on this, this application provides a network slice management method. Referring to Figure 1, which is a flowchart of the first embodiment of the network slice management method of this application.

[0066] In this embodiment, the network slice management method includes steps S10 to S40:

[0067] Step S10: Set the on / off state of a preset number of small base stations according to the slice energy consumption of the network slice. The slice energy consumption is the total energy consumption generated by all devices in the network slice. The small base station is the base station covered by the network slice whose power is less than the power threshold and whose coverage area is less than the range threshold.

[0068] It should be noted that 6G networks rely on large base stations to provide wide-area coverage, and use small base stations to achieve local capacity optimization and signal gap filling, forming a multi-layered heterogeneous network of "macro-micro-pico-fly". Large base stations (also known as macro base stations) are traditional base station types, with high power (over 10W per carrier transmit power) and wide coverage (hundreds of meters to several kilometers), suitable for large-scale user access in open outdoor areas (such as suburbs and urban main roads). Small base stations (also known as micro base stations / pico base stations / fly base stations) are base station types with lower power and narrower coverage. Each network slice can simultaneously cover multiple small base stations or large base stations; a physical network scenario diagram can be found in Figure 2.

[0069] Understandably, this embodiment proposes a slice switching scheme based on real-time traffic, which can solve the problem of excessive energy consumption caused by large-scale slice deployment. The basic idea of ​​the slice switching scheme is to adaptively make slice switching decisions in each slice window to adapt to the dynamic changes in service load and reduce slice energy consumption. Specifically, under low traffic load conditions, it may cause unnecessary energy consumption if a slice is always kept on on the covered base station without any user equipment access. Therefore, two switching states are defined for each slice on the small base stations it covers: on and off. The specific decision is determined based on the slice traffic load (i.e., the slice energy consumption generated by all devices in the network slice). The switching state of a preset number of network slices on the covered small base stations is switched according to the slice energy consumption of each slice.

[0070] In one feasible implementation, step S10 may include steps S101 to S103:

[0071] Step S101: Within the network slicing time window, sum the virtual device energy consumption and the remote wireless head energy consumption to obtain the network virtual energy consumption, wherein the virtual device energy consumption is the energy consumption generated by the optical devices in the virtualization server and the baseband unit pool.

[0072] Understandably, in large-scale slice networks, the deployment of numerous slices to support diverse emerging services leads to significant energy consumption issues. From an energy perspective, network slice energy consumption performance is a key indicator reflecting its overall operational status. System-level energy consumption comprises multiple components, making it impossible to assess the impact of individual slices. By establishing a network slice energy consumption model, operators can monitor the performance of slices in real time when providing them to tenants. If expected demands are not met, operators can directly adjust specific slices, thus avoiding additional overhead. Therefore, this implementation establishes a network slice energy consumption model to characterize the energy required for slice operation in each time window, including network virtual energy consumption and slice restart energy consumption.

[0073] In this context, network virtual energy consumption is achieved through network virtualization. Software-Defined Networking (SDN) controllers can utilize virtual resource blocks and separate network functions to achieve flexible network management. The energy consumption of network virtualization is modeled as the sum of the energy consumption of virtualized cloud baseband unit (BBU) servers and remote radio heads (RRHs).

[0074] It should be noted that the power consumption of the virtualized cloud baseband unit combined server (i.e., the energy consumption of the virtual device) is related to the power consumption of the sliced ​​virtualization server, the power consumption of the optical devices in the baseband unit pool, and the loss factors that are similar to AC-DC, DC-DC, and cooling losses. Meanwhile, the power consumption of the remote wireless head is closely related to the power consumption of the optical devices, the DC loss factor and MS loss factor of the remote wireless head, and the power consumption of the RF unit, etc.

[0075] It's worth noting that the power consumption of a sliced ​​virtualization server is primarily related to its base power consumption, the number of running virtual servers, and the amount of processing bandwidth resources. The power consumption of a network virtual machine mainly consists of four modules: Random Access Memory (RAM), Central Processing Unit (CPU), Network Interface Card (NIC), and Hard Drive (HDD). Therefore, the power consumption of a sliced ​​virtualization server can be expressed as the sum of the operating power consumption of the server's RAM, CPU, NIC, and HDD, the server's base power consumption, and the power consumption of processing bandwidth resources.

[0076] Step S102: When the network slice switches from the on / off state to the on state in the current time window, determine the slice restart energy consumption corresponding to the virtual machine that restarts the network slice.

[0077] It should be noted that at the start of each slice window, the system adjusts slice switching decisions based on dynamic changes in traffic status. If a slice was closed on the base station in the previous time window and is opened in the current time window, the relevant virtual machine will be restarted, resulting in additional energy consumption, known as slice restart energy consumption. Here, it is assumed that the time required for the base station to restart a slice is controlled within one time slot.

[0078] Step S103: Summate the network virtual energy consumption and the slice restart energy consumption to obtain the slice energy consumption of the network slice in each time window.

[0079] It is understandable that the slice energy consumption in the current time window is obtained by summing the network virtual energy consumption and the slice restart energy consumption.

[0080] Step S104: When the energy consumption of the network slice in the time window is lower than a preset energy consumption threshold, the switch status of a preset number of small base stations is set to the off state.

[0081] It should be understood that within a slice window, if the total traffic volume of a slice is low, the slice will be shut down on some small base stations to reduce energy consumption; conversely, if the total traffic volume of a slice is high, it will be turned on on small base stations to meet service quality requirements. Meanwhile, to avoid service interruptions that may result from shutting down slices on small base stations, the proposed slice switching scheme only applies to the small base stations covered by the slice, while all slices deployed on large base stations are always kept on in each slice window.

[0082] In this embodiment, the slice energy consumption of a network slice within each time window is determined by summing the network virtual energy consumption and slice restart energy consumption. Within a slice window, if the slice energy consumption is lower than a preset energy consumption threshold, the slice is shut down on some small base stations to reduce energy consumption. Slices are switched on and off in units of slice time windows. When traffic load is high, to avoid affecting service quality, the slice is turned on on the base stations it covers, ensuring user equipment can transmit tasks. When traffic load is low, to reduce slice operation energy consumption, intelligent decision-making is used to select and shut down slices on some base stations, achieving energy saving while maintaining service quality. This intelligent slice switching scheme controls slice operation based on the actual network conditions, significantly reducing slice energy consumption and operating costs.

[0083] Step S20: Divide the network slice into multiple time windows, and determine the slice configuration strategy for each small base station in each time window according to the switch status.

[0084] Understandably, existing solutions are designed only for network decisions involving a single variable type, leading to room for optimization in slice energy consumption and hybrid action space algorithms. Therefore, this embodiment proposes a dual-timescale management scheme. Since slice configuration and resource allocation are interdependent, the dual-timescale management framework is designed for efficient slice management. The dual-timescale management framework diagram can be found in Figure 3. Utilizing the characteristics of the proposed management scheme, the optimization problem can be naturally decoupled into two sub-problems at different timescales: a long-timescale slice configuration sub-problem and a short-timescale resource allocation sub-problem, which are then solved iteratively.

[0085] Specifically, the network slice is divided into multiple time windows as long-term time scales. At these long-term time scales, the slice configuration strategy is determined by the on / off status of each small base station and the bandwidth resources reserved for it by the corresponding base station. The long-term time scale can be set as the length of each slice's time window, with slice on / off decisions and bandwidth reservation decisions made only once at the beginning of each slice window. The slice configuration strategy remains unchanged within a given time window.

[0086] Step S30: Divide the time window corresponding to each network slice into multiple time slots, and determine the resource allocation strategy corresponding to each network slice in each time slot.

[0087] It should be understood that the time window corresponding to each network slice is divided into multiple time slots as a short time scale. On the short time scale, the time window corresponding to each network slice can be divided or equally divided into a preset number of time slots T. The long-term decision remains unchanged, while the resource allocation strategy for each network slice in each time slot is determined based on user association and bandwidth allocation issues.

[0088] Step S40: Perform slice management according to the slice configuration strategy and the resource allocation strategy.

[0089] This embodiment provides a network slice management method. It sets the on / off status of a preset number of small base stations (SNSs) based on the energy consumption of the network slice. SNSs are base stations covered by the network slice that have small power and coverage area. The network slice is divided into multiple time windows, and the slice configuration strategy for each SNS within each time window is determined based on the on / off status. The time window is further divided into multiple time slots, and a resource allocation strategy is determined for each time slot. Slice management is performed based on the slice configuration strategy and the resource allocation strategy. This embodiment designs a switch based on slice energy consumption load. When the slice load is low, slices on some SNSs on some SNSs are turned off to reduce energy consumption, while slices on large base stations remain on, thus avoiding service interruption caused by turning off SNSs on small base stations. The slice configuration strategy and resource allocation strategy are determined based on the long scale of the time window and the short scale of the divided time slots, respectively. Slice management is performed on both time scales, reducing network slice energy consumption while ensuring slice spectral efficiency and meeting long-term slice performance requirements.

[0090] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description and will not be repeated hereafter. Based on this, please refer to Figure 4, step S20, the network slice management method includes steps S201 to S202:

[0091] Step S201: Divide the network slice into multiple time windows, and determine the bandwidth reservation ratio based on the total bandwidth of each small base station and the number of network slices in each small base station whose switch state is on.

[0092] Understandably, dividing network slices into multiple time windows, on a long-term scale (i.e., within the time window corresponding to a network slice), the slice configuration strategy is accomplished by determining the on / off state of the slice on each small base station and reserving bandwidth resources for each slice. This aims to optimize slice costs within the slice window, minimizing energy consumption at the slice level while ensuring spectral efficiency. However, frequently switching slices on and off based on immediate service demands may cause unnecessary service interruptions and additional restart overhead.

[0093] Understandably, to adapt to long-term network traffic changes and avoid additional energy consumption from frequent operations, slice switching and bandwidth reservation decisions are only performed once at the beginning of each slice window. Slice configuration remains unchanged within a given time window. A binary variable represents whether a slice is enabled on the base station, and the bandwidth reservation ratio is represented as a continuous variable between 0 and 1. Considering that the bandwidth reservation decision is affected by the slice switching decision, the base station will only reserve bandwidth resources for a slice when it is enabled; otherwise, no reservations are made. The slice configuration subproblem is formulated as a long-term cost function weighted by slice energy consumption and spectral efficiency. During optimization, the following constraints must be met: First, the slice switching variable is a binary integer variable of 0 and 1, where a value of 1 indicates that the slice is enabled on the base station, and a value of 0 indicates that the slice is disabled on the base station. Furthermore, the bandwidth reservation ratio must satisfy relevant constraints. Specifically, the bandwidth reservation ratio cannot exceed the total bandwidth of the base station, and bandwidth resources are reserved only for slices enabled on the base station.

[0094] Step S202: Determine the slicing configuration strategy for each small base station within each time window based on the bandwidth reservation ratio.

[0095] It is understandable that after determining the bandwidth reservation ratio, the amount of reserved bandwidth for each small base station in each time window can be determined based on this ratio, thus obtaining the slicing configuration strategy.

[0096] In one possible implementation, step S202 may include steps S2021 to S2026:

[0097] Step S2021: Determine the bandwidth reservation decision for each small base station within each time window based on the bandwidth reservation ratio.

[0098] Understandably, this implementation proposes a management scheme based on a parameterized reinforcement learning architecture and optimization method. By optimizing slice configuration and resource allocation across two time scales, the overall cost of slice operation is reduced. The proposed slice configuration and resource allocation algorithm iteratively optimizes the intelligent slice management strategy. Based on the characteristics of the two different time scales, the slice configuration and resource management scheme can be divided into two parts: a PDQN (Parametrized Deep Q-Network) algorithm and a federated game theory algorithm. The former is used to learn the optimal action as the slice configuration decision over a longer time scale; the structure of the parameterized deep Q-Network can be seen in Figure 5. The latter, based on federated game theory and convex optimization theory, designs a resource allocation strategy for a shorter time scale, achieving a two-layer closed-loop solution.

[0099] It should be understood that, firstly, the bandwidth reservation decision for each small base station within each time window is determined based on the bandwidth reservation ratio, that is, the bandwidth reserved for the small base stations covered by each slice.

[0100] Step S2022: Construct a Markov decision process based on at least one of the following: the number of user devices in the time window corresponding to the network slice, the switch status, the bandwidth reservation decision, the preset slice cost, and the discount factor. The discount factor is used to adjust the reward of the Markov decision process.

[0101] Understandably, to minimize long-term network slicing costs, the problem can be reformulated as a Markov decision process. The agent collects current state information and inputs it into a neural network to determine slice switching and bandwidth reservation actions. After executing these actions in the environment, it receives a reward and moves to the next state. Here, the state, action, and reward of the Markov decision process are defined as follows: State: State information includes the number of user devices for each slice within the current time window, and the slice switching decision in the previous time window. Action: In the current time window, the action corresponds to the slice switching decision and the bandwidth reservation decision, i.e., the slice switching on the base stations covered by each slice and the bandwidth reserved for it by the base stations. Reward: The reward is defined in accordance with the optimization objective, represented by the preset slice cost for the current time window.

[0102] It should be noted that at the beginning of each time window, a slice configuration strategy guides the agent to determine slice switching and bandwidth reservation based on the current state information. Let Π represent the set of slice configuration strategies; the optimization objective can be transformed into determining the optimal slice configuration strategy. Maximize the cumulative reward. The problem can be reformulated as: max π∈∏ [∑ w∈W β w R w (S w A w )|π],

[0103] Here, β∈(0,1) represents the discount factor. When the discount factor is close to 1, the weight of future rewards increases. When it is close to 0, more attention is paid to immediate rewards. When the discount factor is close to 1, this problem can effectively approximate the original problem.

[0104] Step S2023: Train the Markov decision process by storing the historical data generated during the training process in the experience replay area, and randomly selecting sample data from the experience replay area to iteratively train the Markov decision process to obtain the optimal strategy.

[0105] It should be noted that, to overcome the instability of network learning during training, an experience replay strategy is employed. The Markov decision process is trained first, and then historical data from the training process is stored in the experience replay area, instead of training only with the current data. By randomly sampling a small batch of data from the experience replay area each time, the correlation between samples is reduced, ensuring the global optimality of the policy. Simultaneously, a greedy strategy is used during the exploration process to ensure that the agent fully explores the environment during learning and obtains the optimal policy.

[0106] Step S2024: The total number of selectable actions in the optimal strategy is used as input and mapped to an output head with a preset number of dimensions, where the preset number of dimensions is the number of action dimensions of the optimal strategy.

[0107] It is worth noting that, considering the slice switch involves multi-dimensional discrete action decisions, with an action dimension of I×J, the total number of selectable actions is 2. I×J As the number of slices and base stations increases, the total number of available actions grows exponentially. This huge action space negatively impacts the network's training performance. To overcome this problem, a multi-discrete-participant network approach is used to design the network structure. Specifically, the total number of available actions in the optimal policy is mapped as input to I×J output heads through a shared intermediate layer in the neural network, the number of which equals the action dimension of the optimal policy.

[0108] It should be noted that the parametric reinforcement learning algorithm employs a fully connected neural network architecture. Its computational complexity is primarily related to the dimensions of the input and output layers, the number of hidden layers, the number of neurons within these layers, and the size of the training dataset. Therefore, for a fully connected neural network with a fixed structure for hidden layers and neuron counts, the computational complexity is mainly related to the dimensions of the input and output layers. Based on our previous examination of the state and action spaces, the input and output dimensions of the parametric network are I×J+1 and 2×I×J, respectively. Here, the input corresponds to the number of user devices in the previous time window and the slice switching decision, while the output corresponds to the continuous action parameters, i.e., the bandwidth reservation decision for each discrete switching action taken in the slice. Meanwhile, the input and output dimensions of the action network are 3×I×J+1 and 2×I×J, respectively. In this case, the input contains the state and continuous action parameters corresponding to each discrete action, and the output is the Q-value for each discrete action.

[0109] Step S2025: Determine the sub-actions of the overall action in each dimension based on each output head.

[0110] Understandably, since the switching decisions of each slice on a single base station are independent of each other, each output head can independently determine the sub-actions of the overall action in each dimension, that is, the switching decisions of a single slice on a single base station.

[0111] Step S2026: Combine the sub-actions into an optimal action, and use the optimal action as the best slicing configuration strategy for each small base station within each time window.

[0112] It should be understood that by combining all sub-actions into the optimal overall action, the overall slice switching action is executed within the environment. Through this design, the action dimension of each output head becomes 1, and the number of actions is reduced to 2, effectively reducing the difficulty of action selection and network training. This optimal action is used as the best slice configuration strategy for network slices on each small base station, and the slice configuration strategy is not updated until the next slice window.

[0113] In this embodiment, due to constraints, the bandwidth reservation ratio cannot exceed the total bandwidth of the base station, and it only reserves bandwidth resources for slices that are enabled on the base station. Then, the PDQN algorithm is used to solve the long-term slice configuration sub-problem to minimize the long-term network slice cost. The problem is reformulated as a Markov decision process, and a discount factor is introduced for training that maximizes the cumulative reward. An experience replay strategy is employed to overcome the instability of network learning during training, and the input is mapped to multiple output heads, effectively reducing the difficulty of action selection and network training.

[0114] Based on the first embodiment of this application, in the third embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description and will not be repeated hereafter. Based on this, please refer to Figure 6, step S30, the network slice management method includes steps S301 to S305:

[0115] Step S301: Divide the time window corresponding to each network slice into multiple time slots.

[0116] Understandably, on a short timescale, each slice window can be divided into T hourly slots. Considering that service requests and channel conditions change more rapidly in reality, bandwidth resources are allocated to user equipment within a short timescale. According to the established slice energy consumption model, the energy consumption of a slice is determined based on the slice configuration at the beginning of the time window; user association and bandwidth allocation decisions on a short timescale do not affect the calculation of slice energy consumption. Therefore, the overall goal of this sub-problem is to make optimal resource allocation decisions for the transmission tasks of user equipment within each scheduling hourly slot, thereby maximizing the slice spectral efficiency.

[0117] Step S302: Determine the associated user based on the on / off state of each network slice within the time window. The associated user represents the user equipment connected to the small base station.

[0118] Understandably, the following constraints need to be met during the optimization process. First, the user association variable is a binary integer variable of 0 and 1. A value of 1 indicates that the user equipment is associated with the base station, and a value of 0 indicates that the user equipment is not associated with the base station. A user equipment can only connect to one base station to access the corresponding slice within the same time slot. Considering that slices undergo on / off decisions over a long time scale, a user equipment can only access a slice through the associated base station when the slice is enabled on the base station.

[0119] Step S303: Determine the bandwidth allocation capacity for the associated user based on the reserved bandwidth capacity.

[0120] Step S304: Determine the quality of service for each time slot of each network slice based on the minimum transmission rate.

[0121] Step S305: Determine the resource allocation strategy for each network slice in each time slot based on the bandwidth allocation capacity and the quality of service.

[0122] It should be understood that bandwidth allocation needs to meet resource capacity constraints and service quality requirements. The bandwidth allocated to all associated devices cannot exceed the bandwidth capacity reserved by the base station for the slice. The service quality of each time slot of each network slice is determined based on the minimum transmission rate.

[0123] In one feasible implementation, step S305 may include steps S3051 to S3053:

[0124] Step S3051: Construct a coalition game model for each network slice in each time slot based on the user equipment corresponding to the associated user, the small base station covered by the network slice, the bandwidth allocation capacity, and the quality of service.

[0125] It should be noted that for the resource allocation subproblem on a short timescale, a convex optimization method is first used to solve bandwidth allocation given user relationships, obtaining the corresponding revenue for each confederation partition. Then, based on the idea of ​​confederation game theory, the process is iterated until a Nash equilibrium is reached, yielding the final user relationships and resource allocation decisions. The flowchart of the resource allocation method based on confederation game theory can be found in Figure 7.

[0126] It is understandable that the user association problem in slicing can be represented as a coalition game: G i =(U i C i D i ),

[0127] U i Let C be the set of user devices in this game. i For the league division, D iThis represents a set of utilities. In the user association problem, each user equipment (UE) accesses its corresponding slice by associating with a base station. After the association is determined, the base station allocates reserved bandwidth resources to the UE. The base stations covered by the slice are defined as the smallest unit of the confederation, and the confederation is denoted as: C i ={C i,1 C i,2 ,···,C i,j},

[0128] Where C i,j This represents an alliance of user equipment. An alliance must meet the following requirements: And C i,j′ ∪C i,j =U i The utility of a consortium partition in a specific time slot within a time window can be expressed as:

[0129] Step S3052: Determine user association relationships based on the priority order of user devices in the alliance game model.

[0130] Next, we define the priority order of user devices to determine user associations, represented as follows: For any user equipment, given two federation partitions C i and C′ i hour, This means that user equipment is more likely to be associated with a specific base station to form a zone C. i Instead of being associated with another base station to form partition C′ i The mathematical expression is:

[0131] Given partition C′ i ={C i,1 C i,2 ,···,C i,j′ When, if and only if User equipment chooses to leave its associated base station alliance C i,j′ and with another base station C i,j (j≠j′) Association forms partition C i The mathematical representation of user equipment handover operations: {C i,j C i,j′}→{C i,j′ {k},C i,j ∪{k}}.

[0132] Step S3053: Iteratively solve the user association relationship using convex optimization until Nash equilibrium is reached, and obtain the optimal resource allocation strategy for each network slice in each time slot.

[0133] It should be understood that after determining user association relationships, the bandwidth resources reserved by the base station are allocated to the associated user equipment to maximize utility. Given the slice configuration decision and user association decision, since the objective function is convex and the constraints are all linear, the simplified resource allocation subproblem is also a convex problem. The optimal bandwidth allocation decision can be directly obtained using a convex optimization solver. The overall process of slice configuration and resource allocation methods can be seen in Figure 8.

[0134] Understandably, resource allocation algorithms based on federation game theory can guarantee convergence to a stable partition after a finite number of iterations, regardless of the initial partition. Since the number of user equipment and base stations is finite, the federations that user equipment can form are also finite. For any given federation partition, if a user equipment chooses a handover operation to change its federation, the total utility of the partition will increase. Therefore, after a finite number of iterations, the algorithm will reach a Nash equilibrium and eventually converge to a stable partition.

[0135] It should be noted that the computational complexity of the resource allocation algorithm based on coalition game theory primarily depends on the number of iterations. In each iteration, a user device and its current coalition are randomly selected. The spectral efficiency of the slice is calculated as the coalition utility, and the user device's preference for the current and potential coalitions is evaluated. If a user device prefers to associate with a potential coalition, it is removed from the current coalition and joins a new one. Each iteration randomly selects a user device to complete the above process, and then the next iteration begins. Assuming the total number of iterations is M, the computational complexity of the resource allocation algorithm can be expressed as O(M).

[0136] In this embodiment, associated users are determined based on their on / off states, and bandwidth allocation that satisfies resource capacity constraints and quality of service requirements is determined. The overall goal of this resource allocation problem is to make optimal resource allocation decisions for user equipment transmission tasks within each scheduling time slot, thereby maximizing the efficiency of the sliced ​​spectrum. For the resource allocation sub-problem on a short timescale, a convex optimization method is used to solve the bandwidth allocation given user association relationships, obtaining the corresponding confederation partition's revenue; then, based on the idea of ​​confederation game theory, it is iterated until a Nash equilibrium is reached, obtaining the final user association and resource allocation decisions.

[0137] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the network slice management method of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0138] This application also provides a network slice management device, as shown in Figure 9, the network slice management device includes:

[0139] The slice switch status setting module 10 is used to set the switch status of a preset number of small base stations according to the slice energy consumption of the network slice. The slice energy consumption is the total energy consumption generated by all devices in the network slice. The small base station is the base station covered by the network slice whose power is less than the power threshold and whose coverage area is less than the range threshold.

[0140] The slice configuration strategy determination module 20 is used to divide the network slice into multiple time windows and determine the slice configuration strategy of each small base station in each time window according to the switch status.

[0141] The resource allocation strategy determination module 30 is used to divide the time window corresponding to each network slice into multiple time slots and determine the resource allocation strategy corresponding to each network slice in each time slot.

[0142] The network slice management module 40 is used to manage slices according to the slice configuration policy and the resource allocation policy.

[0143] The network slice management device provided in this application, employing the network slice management method described in the above embodiments, can solve the technical problem. Compared with the prior art, the beneficial effects of the network slice management device provided in this application are the same as those of the network slice management method described in the above embodiments, and other technical features in the network slice management device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0144] This application provides a network slice management device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the network slice management method in Embodiment 1 above.

[0145] Referring now to Figure 10, a schematic diagram of a network slice management device suitable for implementing embodiments of this application is shown. The network slice management device in embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. The network slice management device shown in Figure 10 is merely an example and should not impose any limitations on the functionality and scope of use of embodiments of this application.

[0146] As shown in Figure 10, the network slicing management device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory 1002 or a program loaded from a storage device 1003 into a random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the network slicing management device. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems can be connected to the input / output interface 1006: input devices 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. The communication device 1009 allows the network slice management device to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows network slice management devices with various systems, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0147] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0148] The network slice management device provided in this application, employing the network slice management method described in the above embodiments, can solve the technical problems of network slice management. Compared with the prior art, the beneficial effects of the network slice management device provided in this application are the same as those of the network slice management method described in the above embodiments, and other technical features of this network slice management device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0149] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0150] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0151] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the network slice management method described in the above embodiments.

[0152] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0153] The aforementioned computer-readable storage medium may be included in the network slice management device; or it may exist independently and not be assembled into the network slice management device.

[0154] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by a network slice management device, the network slice management device: sets the on / off states of a preset number of small base stations based on the slice energy consumption of the network slice, where the slice energy consumption is the total energy consumption generated by all devices in the network slice, and the small base stations are base stations covered by the network slice whose power is less than a power threshold and whose coverage area is less than a range threshold; divides the network slice into multiple time windows, and determines the slice configuration strategy for each small base station in each time window based on the on / off states; divides the time window corresponding to each network slice into multiple time slots, and determines the resource allocation strategy corresponding to each network slice in each time slot; and performs slice management based on the slice configuration strategy and the resource allocation strategy.

[0155] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0156] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0157] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0158] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described network slice management method, and is capable of solving the technical problem. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the network slice management method provided in the above embodiments, and will not be repeated here.

[0159] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the network slice management method described above.

[0160] The computer program product provided in this application can solve the technical problem. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the network slicing management method provided in the above embodiments, and will not be repeated here.

[0161] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A network slice management method, wherein, The method includes: The on / off states of a preset number of small base stations are set according to the slice energy consumption of the network slice. The slice energy consumption is the total energy consumption generated by all devices in the network slice. The small base station is the base station covered by the network slice whose power is less than the power threshold and whose coverage area is less than the range threshold. The network slice is divided into multiple time windows, and the slice configuration strategy for each small base station in each time window is determined according to the switch status. The time window corresponding to each network slice is divided into multiple time slots, and the resource allocation strategy for each network slice in each time slot is determined. Slice management is performed according to the slice configuration strategy and the resource allocation strategy.

2. The method as described in claim 1, wherein, The step of setting the on / off states of a preset number of small base stations based on the slice energy consumption of the network slice includes: Within the network slicing time window, the energy consumption of the virtual device and the energy consumption of the remote wireless head are summed to obtain the network virtual energy consumption, wherein the energy consumption of the virtual device refers to the energy consumption generated by the optical devices in the virtualization server and the baseband unit pool. When the network slice switches from the on / off state to the on state in the current time window, determine the slice restart energy consumption corresponding to the virtual machine that restarts the network slice; The network virtual energy consumption and the slice restart energy consumption are summed to obtain the slice energy consumption of the network slice in each time window; When the energy consumption of the network slice in the time window is lower than a preset energy consumption threshold, the switch status of a preset number of small base stations is set to the off state.

3. The method as described in claim 1, wherein, The step of dividing the network slice into multiple time windows and determining the slice configuration strategy for each small base station within each time window based on the on / off state includes: The network slice is divided into multiple time windows, and the bandwidth reservation ratio is determined based on the total bandwidth of each small base station and the number of network slices in each small base station whose switch status is on. The slicing configuration strategy for each small base station within each time window is determined based on the bandwidth reservation ratio.

4. The method of claim 3, wherein, The step of determining the slice configuration strategy for each small base station within each time window based on the bandwidth reservation ratio includes: The bandwidth reservation decision for each small base station within each time window is determined based on the bandwidth reservation ratio. A Markov decision process is constructed based on at least one of the following: the number of user devices within the time window corresponding to the network slice, the switch status, the bandwidth reservation decision, the preset slice cost, and the discount factor. The discount factor is used to adjust the reward of the Markov decision process. The Markov decision process is trained by storing historical data generated during the training process in the experience replay area, and random sample data is randomly drawn from the experience replay area to iteratively train the Markov decision process to obtain the optimal strategy. The total number of selectable actions in the optimal strategy is used as input and mapped to an output header with a preset number of dimensions, where the preset number of dimensions is the number of action dimensions of the optimal strategy. The sub-actions of the overall action in each dimension are determined based on each of the output heads; The sub-actions are combined into an optimal action, which is then used as the optimal slicing configuration strategy for each small base station within each time window.

5. The method of claim 1, wherein, The step of dividing the time window corresponding to each network slice into multiple time slots and determining the resource allocation strategy corresponding to each network slice in each time slot includes: The time window corresponding to each network slice is divided into multiple time slots; The associated user is determined based on the on / off state of each network slice within the time window, and the associated user represents the user equipment connected to the small base station. The bandwidth allocation capacity for the associated users is determined based on the reserved bandwidth capacity. The quality of service for each time slot of each network slice is determined based on the minimum transmission rate. The resource allocation strategy for each network slice in each time slot is determined based on the bandwidth allocation capacity and the quality of service.

6. The method of claim 5, wherein, The step of determining the resource allocation strategy for each network slice in each time slot based on the bandwidth allocation capacity and the quality of service includes: Based on the user equipment corresponding to the associated user, the small base station covered by the network slice, the bandwidth allocation capacity, and the quality of service, construct a coalition game model for each network slice in each time slot; User relationships are determined based on the priority order of user devices in the coalition game model. The user association relationship is iteratively solved using convex optimization until Nash equilibrium is reached, thereby obtaining the optimal resource allocation strategy for each network slice in each time slot.

7. A network slice management device, wherein, The network slice management device includes: The slice switch status setting module is used to set the switch status of a preset number of small base stations according to the slice energy consumption of the network slice. The slice energy consumption is the total energy consumption generated by all devices in the network slice. The small base station is the base station covered by the network slice whose power is less than the power threshold and whose coverage area is less than the range threshold. The slice configuration strategy determination module is used to divide the network slice into multiple time windows and determine the slice configuration strategy of each small base station in each time window according to the switch status. The resource allocation strategy determination module is used to divide the time window corresponding to each network slice into multiple time slots and determine the resource allocation strategy corresponding to each network slice in each time slot. The network slice management module is used to manage slices according to the slice configuration policy and the resource allocation policy.

8. A network slicing management device, wherein, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the network slice management method as described in any one of claims 1 to 6.

9. A storage medium, wherein, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the network slice management method as described in any one of claims 1 to 6.

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