Heterogeneous network intelligent slicing resource scheduling method, device, electronic device and medium

Abstract

The present application relates to the technical field of network resource scheduling, and provides a method, device, electronic device and medium for intelligent slicing resource scheduling of heterogeneous networks. A dynamic resource state tensor is obtained by performing three-dimensional joint modeling of heterogeneous network physical resources, a dynamic resource allocation decision is made through a deep reinforcement learning model combined with a business demand matrix to obtain an initial slice resource allocation matrix, a mixed integer linear programming is performed through a preemptive decision model combined with a URLLC business request queue to obtain a resource allocation sequence, the occupancy rate of the dynamic resource state tensor is updated according to the resource allocation sequence to obtain ground network resource data, and resource fusion is performed in combination with satellite beam capacity data to obtain a space-ground integrated resource pool tensor, thereby performing traffic prediction through a long-short memory network combined with historical traffic data to obtain a bandwidth pre-adjustment strategy. The present application improves the overall network resource utilization by organically integrating technologies such as mixed integer linear programming and traffic prediction.

Description

Technical Field

[0001] The present application relates to the technical field of network resource scheduling, and in particular to a method, device, electronic device and medium for intelligent slicing resource scheduling in heterogeneous networks. Background Art

[0002] Meeting the ever-growing and diverse business needs within limited physical resources and ensuring the stability and efficiency of network operations have become important functions of network resource scheduling in modern communication networks.

[0003] Traditional heterogeneous network resource scheduling methods typically rely on pre-set rules or static allocation strategies, often implementing resource scheduling through fixed time-frequency domain partitioning or simple priority sorting. These methods have certain limitations in network resource allocation. Their scheduling decision-making process lacks sufficient intelligence and flexibility, and they are unable to fully integrate ground and satellite resources. This can lead to problems such as uneven resource allocation, response delays, and resource waste when network conditions change rapidly and service demand surges. Summary of the Invention

[0004] In view of this, the present application provides a heterogeneous network intelligent slicing resource scheduling method, device, electronic device and medium to solve the problem of difficulty in capturing dynamic changes in network resource status in real time and effectively responding to sudden business needs.

[0005] In a first aspect, the present application provides a method for intelligent slicing resource scheduling in a heterogeneous network, the method comprising:

[0006] Perform three-dimensional joint modeling of the preset heterogeneous network physical resources in the time domain, frequency domain, and spatial domain to obtain a dynamic resource state tensor;

[0007] Performing dynamic resource allocation decision processing on the dynamic resource state tensor and the received business demand matrix through a preset deep reinforcement learning model to obtain an initial slice resource allocation matrix;

[0008] Performing mixed integer linear programming on the initial slice resource allocation matrix and the received URLLC service request queue through a preset preemption decision model to obtain a resource allocation sequence;

[0009] performing occupancy update processing on the dynamic resource state tensor according to the resource allocation sequence to obtain ground network resource data, and performing resource fusion processing on the ground network resource data and the received satellite beam capacity data to obtain a ground-ground integrated resource pool tensor;

[0010] Traffic prediction processing is performed based on preset historical traffic data and the space-ground integrated resource pool tensor through a preset long-short memory network to obtain a bandwidth pre-adjustment strategy.

[0011] In an optional embodiment, performing three-dimensional joint modeling processing of the preset heterogeneous network physical resources in the time domain, frequency domain, and space domain to obtain a dynamic resource state tensor includes:

[0012] Performing time slot division processing on the heterogeneous network physical resources according to a preset time window to obtain a time slot set;

[0013] Performing subcarrier division and priority weighting processing on the heterogeneous network physical resources to obtain a weighted frequency band set;

[0014] Performing gridding processing on the coverage area of the heterogeneous network physical resources according to preset base station and satellite coverage areas to obtain a set of spatial units;

[0015] A three-dimensional joint construction process is performed on the time slot set, the weighted frequency band set, and the spatial unit set to obtain the dynamic resource state tensor.

[0016] In an optional embodiment, performing dynamic resource allocation decision processing on the dynamic resource state tensor and the received business demand matrix through a preset deep reinforcement learning model to obtain an initial slice resource allocation matrix includes:

[0017] Performing joint coding processing on the service demand matrix to obtain a state vector;

[0018] The state vector and the dynamic resource state tensor are subjected to time-domain-frequency-space three-dimensional resource dynamic allocation processing through the deep reinforcement learning model to obtain the initial slice resource allocation matrix.

[0019] In an optional embodiment, performing mixed integer linear programming processing on the initial slice resource allocation matrix and the received URLLC service request queue using a preset preemption decision model to obtain a resource allocation sequence includes:

[0020] Performing urgency sorting on the URLLC service request queue to obtain a priority sequence;

[0021] Performing mixed integer linear function solving processing on the priority sequence through the preemption decision model to obtain a resource preemption flag vector and a path reconstruction instruction set;

[0022] The initial slice resource allocation matrix is updated according to the resource preemption flag vector and the path reconstruction instruction set to obtain the resource allocation sequence.

[0023] In an optional embodiment, performing occupancy updating processing on the dynamic resource state tensor according to the resource allocation sequence to obtain ground network resource data, and performing resource fusion processing on the ground network resource data and the received satellite beam capacity data to obtain a ground-ground integrated resource pool tensor includes:

[0024] Performing resource occupancy extraction processing on the resource allocation sequence to obtain a resource occupancy parameter;

[0025] updating the dynamic resource state tensor according to the resource occupancy rate parameter to obtain the ground network resource data;

[0026] Performing virtualization mapping processing on the satellite beam capacity data to obtain a virtual resource block set;

[0027] The ground network resource data is combined with the virtual resource block set to obtain the ground-ground integrated resource pool tensor.

[0028] In an optional embodiment, the performing traffic prediction processing based on preset historical traffic data and the space-ground integrated resource pool tensor using a preset long-short-term memory network to obtain a bandwidth pre-adjustment strategy includes:

[0029] Normalizing the historical traffic data to obtain a historical traffic input sequence;

[0030] Performing flow prediction processing on the historical flow input sequence and the space-ground integrated resource pool tensor through the long short-term memory network to obtain flow prediction data;

[0031] The traffic prediction data is dynamically adjusted for slice bandwidth according to the average traffic in the historical traffic data to obtain the bandwidth pre-adjustment strategy.

[0032] In an optional embodiment, the method further comprises:

[0033] Performing local resource gap calculation processing according to the bandwidth pre-adjustment strategy and the space-ground integrated resource pool tensor to obtain a resource gap amount;

[0034] When the resource gap is greater than zero, constructing transaction request data for the resource gap according to a preset transaction priority coefficient to obtain a cross-operator resource transaction request;

[0035] Performing operator quotation matching processing on the cross-operator resource transaction request according to a preset blockchain smart contract to obtain a resource allocation encrypted certificate;

[0036] The space-ground integrated resource pool tensor is updated according to the resource allocation encryption certificate and the resource gap amount.

[0037] A second aspect of the present application provides a heterogeneous network intelligent slice resource scheduling device, the device comprising:

[0038] A three-dimensional joint module is used to perform three-dimensional joint modeling of the preset heterogeneous network physical resources in the time domain, frequency domain, and spatial domain to obtain a dynamic resource state tensor;

[0039] A resource allocation module is configured to perform dynamic resource allocation decision processing on the dynamic resource state tensor and the received business demand matrix through a preset deep reinforcement learning model to obtain an initial slice resource allocation matrix;

[0040] A linear programming module is configured to perform mixed integer linear programming on the initial slice resource allocation matrix and the received URLLC service request queue using a preset preemption decision model to obtain a resource allocation sequence;

[0041] a resource fusion module, configured to perform occupancy update processing on the dynamic resource state tensor according to the resource allocation sequence to obtain ground network resource data, and perform resource fusion processing on the ground network resource data and the received satellite beam capacity data to obtain a ground-ground integrated resource pool tensor;

[0042] The traffic prediction module is used to perform traffic prediction processing based on preset historical traffic data and the space-ground integrated resource pool tensor through a preset long-short term memory network to obtain a bandwidth pre-adjustment strategy.

[0043] The third aspect of the present application provides an electronic device, which includes a memory, a processor, and a computer program stored on the memory and runnable on the processor. When the processor executes the computer program, the steps of the heterogeneous network intelligent slicing resource scheduling method as described above are implemented.

[0044] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the heterogeneous network intelligent slicing resource scheduling method as described above are implemented.

[0045] In summary, this application has at least the following beneficial technical effects:

[0046] 1. By leveraging the advantages of combining deep learning with mathematical programming, it can quickly respond to dynamically changing network environments, especially when facing the scheduling needs of urgent services, showing high response speed and resource allocation flexibility.

[0047] 2. Through the construction of a space-ground integrated resource pool tensor and refined traffic prediction, various resources can be allocated more reasonably, reducing resource waste, while ensuring the timely allocation of key resources when emergency business occurs. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0049] Figure 1 This is a flow chart of a method for scheduling intelligent slicing resources in a heterogeneous network provided by an embodiment of the present application;

[0050] Figure 2 This is a functional module diagram of a heterogeneous network intelligent slice resource scheduling device provided in an embodiment of the present application;

[0051] Figure 3 It is a structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0053] like Figure 1 FIG2 is a flowchart of a method for scheduling resources in a heterogeneous network intelligent slice according to an embodiment of the present application. The method for scheduling resources in a heterogeneous network intelligent slice according to an embodiment of the present application includes the following steps.

[0054] Step S1: Perform three-dimensional joint modeling processing in the time domain, frequency domain, and space domain on the preset heterogeneous network physical resources to obtain a dynamic resource state tensor.

[0055] It should be understood that network resource status exhibits both periodic and bursty variations over time. Time slot partitioning can quantitatively describe the distribution of resources in the time domain, facilitating precise control of dynamic resource scheduling and allocation. In this embodiment, all physical resources in a heterogeneous network (i.e., heterogeneous network physical resources) are partitioned according to preset time windows, thereby forming a clear sequence of time periods. These partitioned time periods are referred to as "time slots," and each time slot represents the availability of network resources within that time period. Specifically, real-time status data of the heterogeneous network physical resources is first acquired through a network management platform or monitoring system. This real-time status data includes operating parameters of 5G base stations, signaling information from satellite communication systems, and status information of IoT terminals. Operators also predetermine time windows based on service requirements and network characteristics. For example, these can be set to 1 millisecond, 10 milliseconds, or longer. The selection of time windows must take into account network quality of service requirements and the dynamic nature of resource changes. After partitioning the preset time window, each resulting time slot has a fixed duration and predefined start and end times for network resource scheduling. Subsequently, the entire time period is evenly divided according to the preset windows. For example, set the total time period to T and divide it into n time slots, forming a time slot set T={t1,……,t n}. In which, each time slot t i Indicates that the network resources are in a relatively stable state during the corresponding time period. The resource capacity corresponding to each time slot is recorded as C t , which reflects the available amount of resources in that time period. For example, if the transmission resources, computing power and spectrum resources of the base station in a time slot all reach a certain percentage, then the C t The value can be calculated from these data. It should be understood that time slot division is achieved with the help of precise clock synchronization technology (for example, the Precision Time Protocol) to ensure time consistency between devices. After the division is completed, the resulting time slot set provides discretized information in the time domain, which helps to predict and optimize the resource conditions within each time slot. In an optional embodiment, time slot division also takes into account the peaks and troughs of network traffic, and associates the divided time slots with the network load conditions to ensure that the statistical information of the resources in each time slot accurately reflects the current network status.

[0056] Since different services have different demands for spectrum resources, the embodiment of the present application adopts weighted division of the spectrum part in the heterogeneous network physical resources in the frequency domain to make resource scheduling more in line with actual business needs and improve the effectiveness of the scheduling strategy. For example, due to the strict requirements on latency and reliability, URLLC services can be given a higher weight when dividing the frequency band, while eMBB services may obtain a relatively low weight. Specifically, spectrum information is first extracted from the physical resources of the heterogeneous network. The information usually includes parameters such as frequency band range, frequency resolution, and actual available bandwidth. According to the preset spectrum planning scheme, the entire frequency band is divided into several subcarriers. The division method can be uniform division or non-uniform division based on actual needs. The specific method depends on the spectrum utilization of the network and business needs. After the division is completed, a subcarrier set F={F1,……,F m}, each subcarrier F i Represents an independent, smaller frequency band unit in the spectrum. During the division process, an initial weight coefficient w is assigned to each subcarrier. f The initial weight coefficient reflects the importance of the frequency band to a specific service. The key to priority weighting is to assign different weight coefficients to different frequency bands based on preset service requirements (for example, URLLC, eMBB, and mMTC). For example, to ensure that URLLC services have sufficient resources, the weight of the frequency band related to this service can be set to 1, while other services can be set to a weight value less than 1. The initial weight coefficient setting can be determined based on historical service data, real-time service requirements, and network resource utilization. The weighting process takes into account both the physical characteristics of the spectrum and the service quality requirements.

[0057] At the same time, because in actual networks, there are large differences in user density, traffic demand and resource availability in different areas. Discretizing the spatial distribution information of physical resources in heterogeneous networks and dividing the geographical area into multiple grid units to form a clear spatial distribution model can provide a spatial reference basis for resource scheduling. By discretizing the region into multiple spatial units, the resource status of each region can be evaluated separately, and refined management can be achieved in the scheduling process. Specifically, the coverage range of base stations and satellites is first obtained through geographic information systems and network coverage planning data. For base stations, coverage data usually comes from base station planning maps and signal strength distribution maps; for satellites, it is obtained based on onboard data and coverage information provided by ground stations. The preset coverage area is usually determined by geographical boundaries, administrative divisions or signal strength thresholds. Based on this information, the entire service area is divided into multiple smaller grid units, each of which represents a specific geographical area. Next, each grid unit in the division process is defined as a spatial unit to obtain a set of spatial units S={S1,……,S kIn each spatial unit, the resource demand density ρ is calculated S This reflects the intensity of resource demand in the region. Resource demand density is typically calculated based on factors such as the number of users, service volume, and geographic area. By utilizing GIS data and real-time user statistics to calculate user density and service data within each grid cell, the boundaries of each spatial cell can be standardized to ensure uniform area or a predetermined distribution ratio.

[0058] Finally, since the distribution of network resources in the time domain, frequency domain, and spatial domain has a coupling effect, the three sets of discrete data obtained are jointly constructed to form a three-dimensional tensor (i.e., dynamic resource state tensor) that describes the state of heterogeneous network resources, which can more comprehensively capture the dynamic changes in resource distribution. Among them, each element of the dynamic resource state tensor reflects the available state of resources at a specific time, a specific frequency band, and a specific spatial unit, providing a global view for subsequent scheduling algorithms. Specifically, by performing a Cartesian product operation on the time slot set, the weighted frequency band set, and the spatial unit set, that is, in each spatial unit in each frequency band in each time slot, there is a corresponding resource state value. Among them, the resource state value is recorded as r t,f,s Represents the remaining available amount of resources in time slot t, frequency band f and spatial unit s. The constructed dynamic resource state tensor is recorded as .

[0059] During data processing, the r of each combination t,f,s The values need to be calculated based on the actual measurement data and preset parameters. First, the C of each time slot is obtained from the time slot division step. t Value; secondly, the weight w of each subcarrier is obtained from the weighted frequency band set f ; Then, the required density ρ of each grid is obtained from the spatial gridding process S ; Finally, based on the current resource allocation, calculate the occupancy rate of each location δ i(t,f,s) , to further calculate the remaining allocatable resource ratio based on the occupancy rate. t Value, weight w f , demand density ρ S and the proportion of allocable resources obtained are multiplied to obtain r for each combination t,f,s The dynamic resource state tensor R is constructed by combining the values of the dynamic resource state tensor and the resource state tensor. Each element of the dynamic resource state tensor R reflects the current available status of network resources within a specific time slot, frequency band, and spatial unit. It can not only be used for real-time resource scheduling, but also as historical data storage for subsequent prediction model training and scheduling strategy optimization.

[0060] Step S2: Perform dynamic resource allocation decision processing on the dynamic resource state tensor and the received business demand matrix through a preset deep reinforcement learning model to obtain an initial slice resource allocation matrix.

[0061] Since the received business demand matrix itself is a multi-dimensional data structure, in order to fuse it with the dynamic resource state tensor, it must first be feature extracted and encoded to form a low-dimensional and descriptive state vector, so that the subsequent model can take into account the specific needs of each business slice while jointly considering the time domain, frequency domain and spatial domain resource information. It should be understood that the obtained business demand matrix Each row corresponds to a network slice, and each slice contains demand data in three dimensions: bandwidth demand B i,req , delay requirement L i,req and reliability requirements R i,req This data is obtained through network operator configuration or application layer negotiation and is transmitted in a structured format (such as JSON or XML) during transmission.

[0062] First, the demand data in the business demand matrix will be preprocessed before entering the coding processing module, such as normalization, denoising and other operations to ensure that the data range is consistent and there are no outliers. Next, the preprocessed business demand matrix is jointly encoded with other auxiliary information. The auxiliary information here includes various indicators in the business matrix and global features that may be obtained through statistical analysis, such as the average bandwidth requirements of all slices, the statistics of delay distribution and the reliability distribution. The joint coding process can adopt methods such as feature splicing, linear transformation and nonlinear mapping. For example, each slice demand data vector in the original matrix is spliced with the global statistical vector to generate an extended vector. Afterwards, the coding state vectors of all slices are integrated to form an overall state vector S, which is used to represent the distribution of business demand at the entire network slice level. By stacking the state vectors of all slices, S=[s1;s2;……;s N The ";" denotes the vertical concatenation of vectors, generating a matrix of dimension N x d, where d is the dimension of the state vector for each slice after joint encoding. As one of the inputs to the deep reinforcement learning model, the state vector S reflects both the individual needs and overall trends of each slice, providing data support for the model's resource allocation decisions.

[0063] During the encoding process, activation functions (such as ReLU or sigmoid) are used to perform nonlinear mapping on the concatenated data, improving the expressive power of the state vector and enabling each component in the vector to better represent the nonlinear characteristics of the slice service requirements. The encoding module can be designed as a feedforward neural network, whose input layer receives the original concatenated vector and, through a series of fully connected layers and nonlinear activations, outputs the final state vector. The weight parameters in this process can be pre-trained offline using historical data and supervisory signals, ensuring that the encoding results are robust and discriminative across different scenarios. Furthermore, during the joint encoding process, attention must be paid to data scale matching and dimensionality conversion. Because the original demand data and statistical data may have different dimensions, normalization techniques (such as min-max normalization or z-score normalization) are often used to convert the data to a uniform scale to prevent any single data feature from dominating the encoding due to its excessively large value. The normalized data is input into the encoding network and then nonlinearly mapped to form the state vector. This ensures that each input feature contributes evenly during joint encoding, facilitating subsequent dynamic resource allocation.

[0064] Finally, the encoded results are verified through data validation and visualization to ensure that the state vector fully captures the key information in the business requirements and that its distribution characteristics reflect the differences between different slices. The encoded state vector not only transforms the original business requirements into a format that can be directly processed by the deep reinforcement learning algorithm, but also enhances the robustness of the data to a certain extent, providing a solid foundation for subsequent dynamic resource scheduling based on the state vector.

[0065] Furthermore, network resources change dynamically in time, frequency, and space, and the business demand state vector provides demand information for each slice. The deep reinforcement learning model can be used to learn the optimal resource allocation strategy and achieve optimization of global resource scheduling. The deep reinforcement learning model jointly processes the state vector obtained based on joint coding and the previously constructed dynamic resource state tensor to generate an initial slice resource allocation matrix. The initial slice resource allocation matrix describes the proportion of resources that each slice should obtain in each time slot, each frequency band, and each spatial unit. It should be understood that the dynamic resource state tensor R describes the real-time distribution of network resources in the time slot set T, the weighted frequency band set F, and the spatial unit set S; the state vector S∈R N×dThis reflects the business needs of each slice. In order for the deep reinforcement learning model to simultaneously consider both types of information, the two pieces of data must first be appropriately fused. This is typically accomplished by scaling the state vector to match its dimensions with the dynamic resource state tensor, or by employing an attention mechanism to organically combine the two. Specifically, the deep reinforcement learning model employs a joint feature fusion approach, mapping the state vector through a fully connected neural network layer to the same feature space as the dynamic resource state tensor. This mapping results in a new state vector S'.

[0066] Furthermore, a deep reinforcement learning model (such as DQN, DDPG, or Actor-Critic architecture) receives the mapped state vector S' and the dynamic resource state tensor R as input. The model is designed with a specific network structure that extracts local features in the time, frequency, and spatial domains through convolutional layers or 3D convolutional layers, while capturing business demand features in the state vector through fully connected layers. Finally, the deep reinforcement learning model generates an initial slice resource allocation matrix. . Among them, each element a in the initial slice resource allocation matrix t,f,s It represents the resource allocation ratio obtained in time slot t, frequency band f and spatial unit s. The initial slice resource allocation matrix reflects the preliminary decision of the deep reinforcement learning model on global resource scheduling.

[0067] It should be understood that the reward function is used as the optimization target during the deep reinforcement learning model training process. The reward function design needs to consider both the business needs of the slice and the resource utilization. The reward function used in the embodiment of the present application is as follows:

[0068]

[0069] Among them, R is the reward value, which represents the overall performance of the deep reinforcement learning model under the current resource allocation. i , L i , R i Represent the bandwidth, delay and reliability indicators actually allocated to slice i. i,req , L i,req , R i,req where α, β, and γ represent the service demand values corresponding to slice i. α, β, and γ are weight coefficients, and they satisfy α + β + γ = 1. λ is the coefficient of the resource fragmentation penalty term.

[0070] During the deep reinforcement learning decision-making process, the model continuously interacts with the environment, evaluates rewards, and uses backpropagation to adjust internal parameters to achieve the optimal allocation of the predicted initial slice resource allocation matrix. During this decision-making process, the model must effectively integrate information from the state vector and the dynamic resource state tensor. This allows it to consider both the individual needs of each slice and the distribution of resources across time, frequency, and space. After repeated training and optimization, the learned policy automatically outputs the resource allocation ratio for each slice in each time slot, frequency band, and spatial unit.

[0071] Step S3: Perform mixed integer linear programming on the initial slice resource allocation matrix and the received URLLC service request queue through a preset preemption decision model to obtain a resource allocation sequence.

[0072] Since there are significant differences in the urgency of each request in the network, in order to prevent service interruptions due to insufficient resource allocation for critical tasks, all requests must be clearly sorted so that high-urgency requests can be given priority when making preemption decisions. The embodiment of the present application implements relative urgency sorting of requests by quantifying the service level and latency deadline of each request. The sorting results will directly affect subsequent resource preemption decisions, ensuring that emergency services are scheduled first when resources are limited, thereby meeting the ultra-low latency and high reliability requirements of Ultra Reliable Low Latency Communication (URLLC). Specifically, the URLLC request queue Q is obtained from the edge node or real-time monitoring system. URLLC ={q1,……,q m}. Each request q j Contains at least two key attributes: Service Level SL j and deadline j The service level is usually expressed as a numerical value, with a higher value indicating a higher priority; the latency deadline indicates how long it takes for a request to be responded to. To take both factors into account, the embodiment of the present application uses the following priority calculation formula to score each request:

[0073]

[0074] Among them, p j represents the comprehensive priority score of the jth request. j Indicates the service level of the jth request. A higher value indicates a more important request. jrepresents the latency deadline for the jth request. A smaller value indicates a faster response time, so its reciprocal is larger. μ and ν are weight coefficients that balance the importance of service level and deadline. Both coefficients are positive and are preset based on actual needs.

[0075] After obtaining the priority score of each request, queue Q is sorted from high to low according to the score. URLLC Sort. The priority sequence P generated after sorting URLLC ={p1,……,p m The order in the} determines which requests are prioritized when preempting resources. The sorting process uses a standard sorting algorithm (such as quick sort or merge sort) to ensure accuracy and efficiency. The sorted priority sequence provides a clear goal for the subsequent mixed-integer linear programming solution: maximizing the total preemption benefit.

[0076] Furthermore, the preemption decision model set in the system constructs a mixed integer linear programming mathematical model based on the obtained priority sequence, and obtains a resource preemption flag vector and the corresponding path reconstruction instruction by solving the model. That is, it determines which requests should be preempted under given resource constraints, and how to reallocate resources after preemption to achieve priority response to emergency services. Mixed integer linear programming solvers can achieve global optimal or suboptimal decisions under limited resources using mathematical optimization methods. Specifically, assuming that the initial slice resource allocation matrix is For each URLLC request, a mixed integer linear programming mathematical model is constructed based on its slice and priority score. j Indicates whether the jth request is successfully preempted, and the value is 0 or 1. The objective function of the constructed mixed integer linear programming mathematical model is designed to maximize the total priority benefit of all emergency requests, by ensuring that under limited resource conditions, those high-priority requests are selected for resource preemption, thereby optimizing the overall service quality. At the same time, the mixed integer linear programming mathematical model needs to introduce constraints to ensure that the resources preempted in each time slot-frequency band-space unit do not exceed the resource state value capacity of the location (that is, satisfying By solving the mixed integer linear programming mathematical model, we can obtain a set of binary decision variables X=[x1,……,x m ], which is the resource preemption flag vector, indicating whether each request is selected for preemption. The optimal solution generated during the solution process also provides a path reconstruction instruction set corresponding to each request. This instruction set defines how to adjust the transmission path in real time after resource preemption, such as updating the SDN flow table and modifying the satellite beam pointing, to ensure that the preempted service flow can be smoothly migrated to the newly configured path.

[0077] The preemption decision model uses existing mathematical optimization tools (such as CPLEX or Gurobi) to solve the mixed integer linear programming mathematical model. The solution process is strictly based on the objective function and constraints to ensure the global optimal or near-optimal solution. After the solution is completed, the obtained resource preemption flag vector and path reconstruction instruction set are saved as system parameters and passed to subsequent modules for further processing. Mixed integer linear programming can accurately handle binary decision problems and their combined constraints to ensure that the total priority of emergency requests is maximized under limited resource conditions, making network scheduling decisions more targeted and reasonable. In actual data processing, the solution of the mixed integer linear programming mathematical model needs to take into account the dynamic changes of each resource unit and update the resource status value r in real time. t,f,s During the solution process, the values of each variable are affected by historical data and real-time monitoring data, and are standardized to ensure the comparability of each data in the model.

[0078] After the preemption decision, some URLLC service requests have received priority resource scheduling. These preemption results are then fed back into the initial allocation matrix to update the resource allocation within each time slot, frequency band, and spatial unit. A new set of resource allocation decisions is generated, ensuring that the actual allocation meets the latency and reliability requirements of all critical services. Specifically, each element in the initial slice resource allocation matrix is adjusted. This adjustment is based on the decision results of each request in the resource preemption flag vector. If the resource preemption flag vector corresponding to a request is 1, it indicates that the request has successfully preempted resources. At this point, the original allocation value for the time slot, frequency band, and spatial unit corresponding to the slice will be increased by the proportion of resources that need to be reallocated from the original resources, while the original resource allocation will be reduced accordingly. The path reconstruction instruction set provides operational instructions for network path updates, such as updating switch flow tables and reconfiguring satellite beams. These instructions are directly issued to the corresponding devices to complete the actual switching of physical resources.

[0079] After the update operation is completed, the resource allocation sequence obtained is in the form of a matrix and is recorded as Each element represents the actual resource proportion allocated to each slice in the corresponding time slot, frequency band, and spatial unit in the final resource allocation decision. Simultaneously, the various instructions in the path reconstruction instruction set are issued to network devices via the control platform, adjusting network transmission paths in real time to ensure that service data can be transmitted along the new path after preemption, thus meeting strict latency requirements.

[0080] The update process combines batch processing with real-time feedback. After each solution of the mixed-integer linear programming model, the initial slice resource allocation matrix is updated and stored in a database for subsequent query by the scheduling and monitoring modules. The update process also requires recalculating the remaining resources of each resource unit to ensure that the new matrix is consistent with the actual physical resource status.

[0081] Step S4: performing occupancy update processing on the dynamic resource status tensor according to the resource allocation sequence to obtain ground network resource data, and performing resource fusion processing on the ground network resource data and the received satellite beam capacity data to obtain a ground-ground integrated resource pool tensor.

[0082] While the initial slice resource allocation matrix reflects the preliminary decision results for resource allocation in various dimensions, subsequent resource fusion requires a more refined occupancy rate that represents the actual resource utilization at each location. Therefore, occupancy extraction processing is performed on the resource allocation sequence to quantify the proportion of resources allocated at each location and, in turn, calculate the remaining available resources. Specifically, the resource allocation sequence first extracts the resource proportion allocated to each slice i in time slot t, frequency band f, and spatial unit s. The extraction process calculates the occupancy of each slice in each resource unit according to predetermined rules and sums the resource proportions occupied by all slices in that resource unit to calculate a comprehensive resource occupancy parameter δ(t, f, s). The calculated resource occupancy parameter δ(t, f, s) provides quantified occupancy information for each resource unit and serves as a key parameter for updating the dynamic resource status tensor, helping to determine the current amount of resources available in the terrestrial network.

[0083] Furthermore, the original dynamic resource state tensor R is updated according to the resource occupancy parameter δ(t,f,s), thereby obtaining the ground network resource data R ground The dynamic resource state tensor R originally describes the theoretical total amount of resources in each time slot, frequency band, and spatial unit, but in practice, some resources have been allocated to each slice. Therefore, it is necessary to deduct the occupied resources and update them to reflect the actual available ground resource status (i.e., ground network resource data R). ground Specifically, a simple proportional deduction method is used to reduce each resource state value r in the initial dynamic resource state tensor. t,f,s The occupied resource ratio is deducted to obtain the unoccupied resource amount in each resource unit, thereby obtaining ground network resource data that reflects the actual available resources.

[0084] At the same time, the satellite beam capacity data obtained from the satellite communication system is processed and virtualized into resource blocks corresponding to the ground network resources. Satellite beam capacity data usually exists in the form of the remaining resources of the beam, which directly reflects the available communication resources on the satellite side. In order to achieve integrated space-ground resource scheduling, satellite resources must be integrated with ground network resources so that the entire network forms a unified resource pool between the air and the ground. Through virtualization mapping processing, the satellite beam data is converted into a set of virtual resource blocks, which is convenient for subsequent merging with ground network resource data to form a unified resource pool. Specifically, first obtain the beam capacity data C from the satellite communication system. sat =[c1,……,c p ], where each c j The virtualization mapping process is to convert the capacity data of each beam c into j Mapped to the time slots, frequency bands and spatial units of the ground network, thus forming a virtual resource block. The mapping method relies on preset rules, such as allocating satellite beam data to several equivalent ground resource units in a certain proportion. The capacity of each satellite beam is c j After mapping, it is allocated to the corresponding virtual resource block v j The virtual resource block not only contains the resource quantity information, but also retains the time-frequency-space coordinates corresponding to the ground network resources. j Collected into a virtual resource block set V sat ={v1,……,v p}.

[0085] Individual ground resources or satellite resources each have their limitations, but through reasonable merging, the two can give full play to their respective advantages, achieve resource complementarity, and thus improve overall resource utilization and service coverage. The acquired ground network resource data is merged with the virtual resource block set to construct a unified resource pool covering the ground and satellite. The construction of the space-ground integrated resource pool tensor enables unified scheduling of resources in different domains, solving the scheduling difficulties caused by resource heterogeneity. Specifically, the data format of the ground network resource data and the virtual resource block set is first standardized to ensure that the two have consistent representations in the time, frequency, and space dimensions. Then, the standardized ground network resource data and the standardized virtual resource block set are summed element by element to obtain the space-ground integrated resource pool tensor.

[0086] During data processing, the merging operation requires first converting the set of virtual resource blocks into a format consistent with the dimensionality of the ground resource data. During this conversion, mapping rules are used to determine the position of each virtual resource block in the tensor, and the corresponding resource amount is filled in at that position to form a virtual resource tensor. Subsequently, mathematical operations are used to add the values at corresponding positions in the two tensors to obtain the fused ground-to-ground integrated resource pool tensor. The merged result not only reflects the real-time available resources of the ground network, but also includes the additional resources provided by satellite beams. This resource pool provides a unified and comprehensive resource view for subsequent resource scheduling and dynamic optimization, effectively addressing the issue of uneven resource distribution within the network.

[0087] Step S5: Perform traffic prediction processing based on preset historical traffic data and the space-ground integrated resource pool tensor through a preset long short-term memory network to obtain a bandwidth pre-adjustment strategy.

[0088] It should be understood that historical traffic data records the network traffic conditions within each slice and time slot over a period of time. These data typically have different dimensions and data ranges. Directly using the raw data may lead to excessive numerical fluctuations during subsequent deep learning model training due to inconsistent numerical scales, affecting the model's convergence and accuracy. Therefore, normalization is an essential preprocessing step. Its main function is to scale the data to the same range, thereby making the features comparable and improving the model's training performance. Among them, normalization generally adopts two common methods: one is minimum-maximum normalization, and the other is standardization. The embodiment of the present application adopts a normalization method, converting each data point in the historical traffic data according to its mean and standard deviation, so that all data follows a distribution with zero mean and unit variance. The main reason for this process is that the standardized data not only facilitates numerical calculations of the network model, but also improves the model's robustness to outliers. Specifically, the historical traffic data is in vector form and reflects the traffic statistics of each slice. Calculating the mean and standard deviation of this data is the first step in normalization. After obtaining the mean and standard deviation, each data point in the historical traffic data is normalized to obtain a normalized historical traffic input sequence. After normalization, the historical traffic input sequence has been uniformly scaled, laying the foundation for the subsequent long-short-term memory network input.

[0089] Network traffic is characterized by temporal and dynamic changes. Long-Short Term Memory (LSTM) networks (LSTMs) can effectively capture long-term dependencies and nonlinear relationships within time series, predict future traffic trends, and provide a reference for subsequent bandwidth adjustment strategies. The LSTM network combines the normalized historical traffic input sequence with data from the integrated space-ground resource pool to predict traffic data for a specific period of time. The integrated space-ground resource pool tensor reflects the integration of ground and satellite resources and describes the resource distribution within each time slot, frequency band, and spatial unit. While analyzing historical traffic changes, the LSTM network also considers the current state of network resources to provide more accurate predictions of future traffic. Specifically, the normalized historical traffic input sequence and the integrated space-ground resource pool tensor are first appropriately preprocessed. The integrated space-ground resource pool tensor needs to be flattened or reconstructed according to the model input requirements to be combined with the time series data. Data preprocessing involves dimensionality conversion of the integrated space-ground resource pool tensor to align it with the historical traffic data in the temporal dimension, forming the joint input feature X. The LSTM network architecture consists of an input layer, multiple LSTM layers, and an output layer. The input layer receives the combined input feature X, which combines historical traffic information with current resource status. The LSTM network uses a gating mechanism to control information forgetting and remembering, thereby capturing long-term dependencies in the time series. The output layer generates predicted future traffic data. The combined input feature X first passes through the embedding layer and the convolution layer to extract local features before being input into the multi-layer LSTM network. The network is trained using a backpropagation algorithm, using the mean squared error between historical data and actual traffic as the loss function. The network weights are continuously updated to ensure that the predicted results are as close as possible to the actual future traffic.

[0090] Traffic forecast data is normalized and typically requires denormalization based on the normalization parameters to restore it to its original scale. This forecast not only provides future traffic trends but also reflects the relationship between current network resource status and traffic fluctuations, providing a scientific basis for subsequent bandwidth adjustment strategies. The processing of joint input features during the forecasting process ensures that traffic forecasts not only rely on historical trends but also take into account current resource conditions, improving forecast accuracy and robustness.

[0091] Furthermore, using historical average traffic as a benchmark, the degree of deviation from predicted traffic can be quantified, and the adjustment ratio can be determined to ensure that resource allocation is more aligned with future traffic demand, thereby improving service quality and resource utilization. By comparing traffic forecast data with the average traffic volume in historical traffic data, the bandwidth of each slice is dynamically adjusted based on the ratio between the two, forming a bandwidth pre-adjustment strategy. This bandwidth pre-adjustment strategy ensures that slice bandwidth allocation is adjusted in advance of traffic peaks to avoid network congestion and releases redundant bandwidth during traffic troughs, achieving flexible resource scheduling and efficient utilization. Specifically, a calculation method similar to that used for normalized statistical indicators is first used to average historical traffic data to obtain the average traffic volume. The ratio of the predicted traffic volume to the average traffic volume in each time slot is then calculated to determine the bandwidth adjustment factor. This factor is used to modify the original bandwidth, so that when the traffic forecast is high, the allocated bandwidth of the slice is proportionally increased. When the traffic forecast is low, the bandwidth adjustment factor is less than 1, and the allocated bandwidth is automatically reduced, thereby optimizing resource utilization and service quality. The adjustment process strictly adheres to a preset upper limit to ensure that the adjustment range does not exceed the system's tolerance. The bandwidth obtained after adjustment constitutes a bandwidth pre-adjustment strategy. The data structure of the bandwidth pre-adjustment strategy is a vector or a matrix, which reflects the bandwidth allocation ratio of each slice that should be adjusted in the future period.

[0092] It is difficult for any resource scheduling strategy to foresee all emergencies, especially when facing critical services such as ultra-reliable low-latency communications, where local resource shortages are inevitable. In this case, relying solely on internal resource scheduling and forecasting can no longer meet business needs. External resources must be introduced through cross-domain resource transactions to achieve rapid resource replenishment and dynamic balance. In an optional implementation, through the blockchain smart contract automated transaction mechanism, after detecting local resource gaps, cross-operator resource transactions are triggered in a timely manner, thereby ensuring that the network scheduling system has sufficient resource guarantees at all times. The specific operations are as follows:

[0093] After obtaining the bandwidth pre-adjustment policy, the difference between actual local available resources and expected demand is quantified using the bandwidth pre-adjustment policy and the integrated space-ground resource pool tensor, thereby calculating the resource gap. This combination of the two determines whether local resources are currently sufficient and to what extent. When predicted demand exceeds existing resources, a resource gap forms. Calculating this gap is crucial for subsequent cross-operator resource transactions. Only by determining the gap size can resource replenishment be initiated to ensure sufficient resources during peak service hours, meeting ultra-low latency and high reliability requirements. The resource gap is first determined by comparing the total bandwidth demand adjusted according to the bandwidth pre-adjustment policy with the total resources provided by the integrated space-ground resource pool tensor. It should be understood that before calculating the resource gap, the resources across all frequency bands and spatial units in time slot t in the integrated space-ground resource pool tensor must be summed to determine the total available resources in that time slot. The calculated resource gap is the difference between predicted demand and available resources. When predicted demand exceeds actual available resources, the resource gap is positive; otherwise, it is zero.

[0094] When the resource shortfall is greater than zero, cross-operator resource transaction request data is constructed. The construction of a cross-operator resource transaction request requires consideration of the transaction priority coefficient, a predefined coefficient used to differentiate the urgency of different service types (such as URLLC, eMBB, and mMTC). This priority coefficient ensures that urgent services are given higher priority when resources are insufficient, allowing them to secure necessary resources first during cross-domain transactions. Specifically, when the resource shortfall is greater than zero, the transaction request data is constructed based on the predefined transaction priority coefficient α for each service type (for example, URLLC services are assigned a highest priority coefficient of 3, eMBB services are assigned a priority coefficient of 2, and mMTC services are assigned a priority coefficient of 1). Assuming that each slice or service module has a corresponding priority coefficient based on service requirements, the cross-operator resource transaction request data structure is constructed based on the shortfall and priority coefficient. Key fields in the transaction request data typically include the resource shortfall, the transaction priority coefficient, and the resource delivery deadline. The deadline is determined based on service requirements and traffic forecasts. The constructed cross-operator resource transaction request is then transmitted to the blockchain smart contract module via a data interface, awaiting subsequent matching with operator quotations.

[0095] Cross-operator resource transactions require consensus among multiple operators. Only through automated quote matching via smart contracts can efficient, low-latency, and secure resource transactions be achieved. Leveraging a pre-configured blockchain smart contract, cross-operator resource transaction requests are matched against operator quotes to generate encrypted credentials for resource allocation. Blockchain smart contracts are used to automatically enforce transaction rules, verify quote consistency with budgets, and generate secure, tamper-proof encrypted credentials. Blockchain technology ensures transparent and tamper-proof transaction information, and cryptographically verifies the legitimacy of both parties.

[0096] Specifically, when a blockchain smart contract is triggered (i.e., when the resource shortfall is greater than zero), the budget cap is first calculated according to the blockchain smart contract's built-in transaction rules (e.g., budget cap calculation formula and bid matching conditions). The budget cap is typically calculated by multiplying the transaction priority coefficient, the resource shortfall, and the preset benchmark resource unit price to obtain the budget cap (i.e., the maximum allowable fee) for time slot t. Next, the list of resource bids submitted by participating operators in the blockchain network is traversed according to the bid matching conditions in the blockchain smart contract, and each bid is verified. When a bid combination that meets the conditions is found (i.e., an operator's bid is less than or equal to the budget cap), the blockchain smart contract generates a resource allocation encrypted certificate. This certificate is generated using an asymmetric cryptographic signature to ensure the security and immutability of transaction information. During the bid matching and certificate generation process, the blockchain smart contract records the entire transaction information and writes it to the blockchain ledger, ensuring data transparency and immutability.

[0097] After obtaining the resource allocation encryption certificate and the amount of resource gap, the resources obtained from the outside after the resource transaction is completed need to be promptly merged into the internal resource pool to achieve dynamic balance and optimal utilization of the entire network resources. The resource allocation encryption certificate and the amount of resource gap are used to update the space-ground integrated resource pool tensor, so that the new resource data after the supplementary transaction can be reflected in the overall resource pool, thereby dynamically adjusting the distribution of resources in the network to ensure that after the cross-operator resource transaction is completed, the actual available resources can meet future business needs, and at the same time provide the latest data support for subsequent resource scheduling decisions. Specifically, first, based on the obtained resource gap amount and resource allocation encryption certificate, the amount of resources that have been successfully supplemented is regarded as newly added resources. Among them, the newly added resources are the resource gap amount or part of it, depending on the actual matching results. The update operation requires merging the newly added resources into the original space-ground integrated resource pool tensor to obtain an updated space-ground integrated resource pool tensor.

[0098] In actual implementation, the newly added resources must undergo data verification and format standardization before the update to ensure consistency in data format, dimensions, and dimensionality with the pre-update ground-based resource pool tensor. The encrypted resource allocation certificate verifies the legitimacy and integrity of the transaction, ensuring that the update is based on authentic and valid transaction results. The data processing platform first reads the encrypted transaction data from the blockchain ledger and extracts the newly added resource quantities. The newly added resource data is then added item by item to the original resource pool data, completing the update. The updated ground-based resource pool tensor will reflect the latest network resource status after the cross-operator resource transaction, including both the original ground and satellite resources and the supplementary resources acquired through the transaction.

[0099] The present application is applied to the field of network resource scheduling technology. It obtains a dynamic resource state tensor by performing three-dimensional joint modeling of heterogeneous network physical resources, makes dynamic resource allocation decisions on the dynamic resource state tensor and the service demand matrix through a deep reinforcement learning model to obtain an initial slice resource allocation matrix, performs mixed integer linear programming on the initial slice resource allocation matrix and the URLLC service request queue through a preemption decision model to obtain a resource allocation sequence, updates the occupancy rate of the dynamic resource state tensor according to the resource allocation sequence to obtain ground network resource data, and performs resource fusion on the ground network resource data and the satellite beam capacity data to obtain a space-ground integrated resource pool tensor. Traffic prediction is performed based on historical traffic data and the space-ground integrated resource pool tensor through a long short-term memory network to obtain a bandwidth pre-adjustment strategy. The present application organically integrates multi-dimensional modeling, deep reinforcement learning, mixed integer linear programming, traffic prediction and blockchain technology, which not only improves the intelligence level and real-time response capability of resource scheduling, but also realizes the efficient integration of ground and satellite resources, ensures low-latency and high-reliability transmission of key services, and ultimately achieves the goal of improving overall network resource utilization and service quality.

[0100] like Figure 2 The figure shows a functional module diagram of a heterogeneous network intelligent slice resource scheduling device provided in an embodiment of the present application.

[0101] In some embodiments, the heterogeneous network intelligent slice resource scheduling device 2 may include multiple functional modules composed of computer program segments. The computer program of each program segment in the heterogeneous network intelligent slice resource scheduling device 2 may be stored in the memory of the server and executed by at least one processor to perform (see Figure 1 (Describe) the functions of the intelligent slicing resource scheduling method for heterogeneous networks.

[0102] In this embodiment, the heterogeneous network intelligent slicing resource scheduling device 2 can be divided into multiple functional modules based on the functions it performs. These functional modules may include: a three-dimensional joint module 21, a resource allocation module 22, a linear programming module 23, a resource fusion module 24, a traffic prediction module 25, and a resource trading module 26. As used herein, a module refers to a series of computer program segments that can be executed by at least one processor and can perform fixed functions, and are stored in a memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments.

[0103] The three-dimensional joint module 21 is used to perform three-dimensional joint modeling processing on the preset heterogeneous network physical resources in the time domain, frequency domain and space domain to obtain a dynamic resource state tensor.

[0104] In an optional embodiment, the three-dimensional joint module 21 is specifically configured to:

[0105] Performing time slot division processing on the heterogeneous network physical resources according to a preset time window to obtain a time slot set;

[0106] Performing subcarrier division and priority weighting processing on the heterogeneous network physical resources to obtain a weighted frequency band set;

[0107] Performing gridding processing on the coverage area of the heterogeneous network physical resources according to preset base station and satellite coverage areas to obtain a set of spatial units;

[0108] A three-dimensional joint construction process is performed on the time slot set, the weighted frequency band set, and the spatial unit set to obtain the dynamic resource state tensor.

[0109] The resource allocation module 22 is used to perform dynamic resource allocation decision processing on the dynamic resource state tensor and the received business demand matrix through a preset deep reinforcement learning model to obtain an initial slice resource allocation matrix.

[0110] In an optional implementation, the resource allocation module 22 is specifically configured to:

[0111] Performing joint coding processing on the service demand matrix to obtain a state vector;

[0112] The state vector and the dynamic resource state tensor are subjected to time-domain-frequency-space three-dimensional resource dynamic allocation processing through the deep reinforcement learning model to obtain the initial slice resource allocation matrix.

[0113] The linear programming module 23 is used to perform mixed integer linear programming processing on the initial slice resource allocation matrix and the received URLLC service request queue through a preset preemption decision model to obtain a resource allocation sequence.

[0114] In an optional embodiment, the linear programming module 23 is specifically configured to:

[0115] Performing urgency sorting on the URLLC service request queue to obtain a priority sequence;

[0116] Performing mixed integer linear function solving processing on the priority sequence through the preemption decision model to obtain a resource preemption flag vector and a path reconstruction instruction set;

[0117] The initial slice resource allocation matrix is updated according to the resource preemption flag vector and the path reconstruction instruction set to obtain the resource allocation sequence.

[0118] The resource fusion module 24 is used to perform occupancy update processing on the dynamic resource status tensor according to the resource allocation sequence to obtain ground network resource data, and to perform resource fusion processing on the ground network resource data and the received satellite beam capacity data to obtain a ground-ground integrated resource pool tensor.

[0119] In an optional implementation, the resource fusion module 24 is specifically configured to:

[0120] Performing resource occupancy extraction processing on the resource allocation sequence to obtain a resource occupancy parameter;

[0121] updating the dynamic resource state tensor according to the resource occupancy rate parameter to obtain the ground network resource data;

[0122] Performing virtualization mapping processing on the satellite beam capacity data to obtain a virtual resource block set;

[0123] The ground network resource data is combined with the virtual resource block set to obtain the ground-ground integrated resource pool tensor.

[0124] The traffic prediction module 25 is used to perform traffic prediction processing based on preset historical traffic data and the space-ground integrated resource pool tensor through a preset long short-term memory network to obtain a bandwidth pre-adjustment strategy.

[0125] In an optional embodiment, the traffic prediction module 25 is specifically configured to:

[0126] Normalizing the historical traffic data to obtain a historical traffic input sequence;

[0127] Performing flow prediction processing on the historical flow input sequence and the space-ground integrated resource pool tensor through the long short-term memory network to obtain flow prediction data;

[0128] The traffic prediction data is dynamically adjusted for slice bandwidth according to the average traffic in the historical traffic data to obtain the bandwidth pre-adjustment strategy.

[0129] In an optional embodiment, the heterogeneous network intelligent slice resource scheduling device 2 further includes a resource trading module 26, and the resource trading module 26 is specifically used to:

[0130] Performing local resource gap calculation processing according to the bandwidth pre-adjustment strategy and the space-ground integrated resource pool tensor to obtain a resource gap amount;

[0131] When the resource gap is greater than zero, constructing transaction request data for the resource gap according to a preset transaction priority coefficient to obtain a cross-operator resource transaction request;

[0132] Performing operator quotation matching processing on the cross-operator resource transaction request according to a preset blockchain smart contract to obtain a resource allocation encrypted certificate;

[0133] The space-ground integrated resource pool tensor is updated according to the resource allocation encryption certificate and the resource gap amount.

[0134] It should be understood that the various variations and specific embodiments of the methods provided in the above embodiments are also applicable to the heterogeneous network intelligent slice resource scheduling device of this embodiment. Through the above detailed description of the heterogeneous network intelligent slice resource scheduling method, those skilled in the art can clearly understand the implementation method of the heterogeneous network intelligent slice resource scheduling device in this embodiment. For the sake of brevity of the specification, it will not be described in detail here.

[0135] like Figure 3 , which is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0136] In a preferred embodiment of the present invention, the electronic device 3 may include, but is not limited to: a memory 31 , at least one processor 32 and at least one communication bus 33 .

[0137] Those skilled in the art should understand that Figure 3 The structure of the electronic device 3 shown does not constitute a limitation of the embodiment of the present invention. The electronic device 3 may also include more or less other hardware or software than shown in the figure, or a different component arrangement.

[0138] In some embodiments, the electronic device 3 is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application-specific integrated circuits, programmable gate arrays, digital processors and embedded devices.

[0139] It should be noted that the electronic device 3 is only an example. Other existing or future electronic products that are suitable for this application should also be included in the scope of protection of this application and included here by reference.

[0140] In some embodiments, the memory 31 stores a computer program that, when executed by the at least one processor 32, implements all or part of the steps in the heterogeneous network intelligent slicing resource scheduling method. The memory 31 includes a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM), or other optical disk storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data. Furthermore, the computer-readable storage medium may primarily include a program storage area and a data storage area. The program storage area may store an operating system, at least one application required for a function, and the like.

[0141] In some embodiments, the at least one processor 32 is the control core (Control Unit) of the electronic device 3. It connects the various components of the electronic device 3 using various interfaces and circuits. It executes programs or modules stored in the memory 31 and accesses data stored in the memory 31 to perform various functions and process data in the electronic device 3. For example, when executing the computer program stored in the memory 31, the at least one processor 32 implements all or part of the steps of the method for intelligent slicing resource scheduling in a heterogeneous network described in the embodiments of this application, or implements all or part of the functions of the apparatus for intelligent slicing resource scheduling in a heterogeneous network. The at least one processor 32 can be composed of an integrated circuit, such as a single packaged integrated circuit or multiple packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and a combination of various control chips.

[0142] In some embodiments, the at least one communication bus 33 is configured to enable communication between the memory 31 and the at least one processor 32. Although not shown, the electronic device 3 may also include a power supply (e.g., a battery) to power various components. Preferably, the power supply can be logically connected to the at least one processor 32 via a power management device, thereby enabling the power management device to manage charging, discharging, and power consumption. The power supply may also include one or more DC or AC power supplies, a recharging device, a power failure detection circuit, a power converter or inverter, a power status indicator, and other components. The electronic device 3 may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be detailed here.

[0143] The above-mentioned integrated unit implemented in the form of a software functional module can be stored in a computer-readable storage medium. The above-mentioned software functional module stored in a storage medium includes a number of instructions for causing an electronic device (which can be a personal computer, electronic device, or network device, etc.) or a processor to execute portions of the methods described in various embodiments of the present application.

[0144] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the module division is only a logical function division, and other division methods may be used in actual implementation.

[0145] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units, and may be located in one place or distributed across multiple network elements. Some or all of the modules may be selected to achieve the purpose of this embodiment based on actual needs.

[0146] The above are all preferred embodiments of the present application, and are not intended to limit the scope of protection of the present application. Therefore, any equivalent changes made based on the structure, shape, and principle of the present application should be included in the scope of protection of the present application.

Claims

1. A method for intelligent slicing resource scheduling in a heterogeneous network, characterized in that: The method comprises: Perform three-dimensional joint modeling of the preset heterogeneous network physical resources in the time domain, frequency domain, and spatial domain to obtain a dynamic resource state tensor; Performing dynamic resource allocation decision processing on the dynamic resource state tensor and the received business demand matrix through a preset deep reinforcement learning model to obtain an initial slice resource allocation matrix; Performing mixed integer linear programming on the initial slice resource allocation matrix and the received URLLC service request queue through a preset preemption decision model to obtain a resource allocation sequence; performing occupancy update processing on the dynamic resource state tensor according to the resource allocation sequence to obtain ground network resource data, and performing resource fusion processing on the ground network resource data and the received satellite beam capacity data to obtain a ground-ground integrated resource pool tensor; Normalizing the preset historical traffic data to obtain a historical traffic input sequence; Performing flow prediction processing on the historical flow input sequence and the space-ground integrated resource pool tensor through a preset long short-term memory network to obtain flow prediction data; Performing slice bandwidth dynamic adjustment processing on the traffic prediction data according to the average traffic in the historical traffic data to obtain a bandwidth pre-adjustment strategy; Performing local resource gap calculation processing according to the bandwidth pre-adjustment strategy and the space-ground integrated resource pool tensor to obtain a resource gap amount; When the resource gap is greater than zero, constructing transaction request data for the resource gap according to a preset transaction priority coefficient to obtain a cross-operator resource transaction request; Performing operator quotation matching processing on the cross-operator resource transaction request according to a preset blockchain smart contract to obtain a resource allocation encrypted certificate; The space-ground integrated resource pool tensor is updated according to the resource allocation encryption certificate and the resource gap amount.

2. The method for scheduling heterogeneous network intelligent slice resources according to claim 1, characterized in that: The performing of three-dimensional joint modeling processing of the preset heterogeneous network physical resources in the time domain, frequency domain, and space domain to obtain a dynamic resource state tensor includes: Performing time slot division processing on the heterogeneous network physical resources according to a preset time window to obtain a time slot set; Performing subcarrier division and priority weighting processing on the heterogeneous network physical resources to obtain a weighted frequency band set; Performing gridding processing on the coverage area of the heterogeneous network physical resources according to preset base station and satellite coverage areas to obtain a set of spatial units; A three-dimensional joint construction process is performed on the time slot set, the weighted frequency band set, and the spatial unit set to obtain the dynamic resource state tensor.

3. The method for scheduling heterogeneous network intelligent slice resources according to claim 1, characterized in that: The performing dynamic resource allocation decision processing on the dynamic resource state tensor and the received business demand matrix by using a preset deep reinforcement learning model to obtain an initial slice resource allocation matrix includes: Performing joint coding processing on the service demand matrix to obtain a state vector; The state vector and the dynamic resource state tensor are subjected to time-domain-frequency-space three-dimensional resource dynamic allocation processing through the deep reinforcement learning model to obtain the initial slice resource allocation matrix.

4. The method for scheduling heterogeneous network intelligent slice resources according to claim 1, characterized in that: The performing mixed integer linear programming processing on the initial slice resource allocation matrix and the received URLLC service request queue by using a preset preemption decision model to obtain a resource allocation sequence includes: Performing urgency sorting on the URLLC service request queue to obtain a priority sequence; Performing mixed integer linear function solving processing on the priority sequence through the preemption decision model to obtain a resource preemption flag vector and a path reconstruction instruction set; The initial slice resource allocation matrix is updated according to the resource preemption flag vector and the path reconstruction instruction set to obtain the resource allocation sequence.

5. The method for scheduling heterogeneous network intelligent slice resources according to claim 1, characterized in that: The step of performing occupancy updating processing on the dynamic resource state tensor according to the resource allocation sequence to obtain ground network resource data, and performing resource fusion processing on the ground network resource data and the received satellite beam capacity data to obtain a ground-ground integrated resource pool tensor includes: Performing resource occupancy extraction processing on the resource allocation sequence to obtain a resource occupancy parameter; updating the dynamic resource state tensor according to the resource occupancy rate parameter to obtain the ground network resource data; Performing virtualization mapping processing on the satellite beam capacity data to obtain a virtual resource block set; The ground network resource data is combined with the virtual resource block set to obtain the ground-ground integrated resource pool tensor.

6. A heterogeneous network intelligent slice resource scheduling device, applied to the heterogeneous network intelligent slice resource scheduling method according to claim 1, characterized in that: The device comprises: A three-dimensional joint module is used to perform three-dimensional joint modeling of the preset heterogeneous network physical resources in the time domain, frequency domain, and spatial domain to obtain a dynamic resource state tensor; A resource allocation module is configured to perform dynamic resource allocation decision processing on the dynamic resource state tensor and the received business demand matrix through a preset deep reinforcement learning model to obtain an initial slice resource allocation matrix; A linear programming module is configured to perform mixed integer linear programming on the initial slice resource allocation matrix and the received URLLC service request queue using a preset preemption decision model to obtain a resource allocation sequence; a resource fusion module, configured to perform occupancy update processing on the dynamic resource state tensor according to the resource allocation sequence to obtain ground network resource data, and perform resource fusion processing on the ground network resource data and the received satellite beam capacity data to obtain a ground-ground integrated resource pool tensor; The traffic prediction module is used to perform traffic prediction processing based on preset historical traffic data and the space-ground integrated resource pool tensor through a preset long-short term memory network to obtain a bandwidth pre-adjustment strategy.

7. An electronic device, characterized in that: The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the heterogeneous network intelligent slicing resource scheduling method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the heterogeneous network intelligent slice resource scheduling method according to any one of claims 1 to 5 are implemented.

Images