FTTR-B network slicing method and device, medium and equipment

Abstract

The invention belongs to the technical field of communication networks, and particularly discloses an FTTR-B network slicing method and device, a medium and equipment, and the method comprises the steps: obtaining the network data of an FTTR-B network; the obtained network data of the FTTR-B network is preprocessed; and inputting the preprocessed network data of the FTTR-B network into a pre-trained FTTR-B network slice model so as to realize dynamic slice management and optimization of the FTTR-B network. According to the invention, high-precision, low-delay and flexible slice management can be carried out on the FTTR-B network.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of communication networks, and particularly relates to an FTTR-B network slicing method, device, medium and equipment. BACKGROUND

[0002] Traditional network slicing technology relies on static configuration or fixed resource allocation, which is difficult to adapt to dynamically changing business demands (such as VR, IoT, security monitoring, etc.) in the FTTR-B scenario. Existing solutions have deficiencies in resource utilization, latency control and business QoS guarantee, especially lacking fine-grained slicing management capabilities in complex home / enterprise environments. Specifically, existing technologies often cannot achieve flexible allocation and dynamic adjustment of resources when dealing with diversified business demands such as high bandwidth, low latency and large-scale connections, resulting in low network resource utilization and difficulty in guaranteeing business service quality. In addition, existing solutions have many deficiencies in slicing priority division, physical layer resource allocation, data link layer slicing management, network layer slicing orchestration, security and reliability guarantee, and slicing life cycle management, which cannot meet the efficient, flexible and reliable slicing management needs of FTTR-B networks. SUMMARY

[0003] In view of the deficiencies in the prior art, the purpose of the present application is to provide an FTTR-B network slicing method, device, medium and equipment, which aims to achieve high-precision, low-latency and flexible slicing management for FTTR-B networks.

[0004] To achieve the above-mentioned purpose, the present application provides the following technical solutions: An FTTR-B network slicing method, the method comprising: acquiring network data of an FTTR-B network; preprocessing the acquired network data of the FTTR-B network; inputting the preprocessed network data of the FTTR-B network into a pre-trained FTTR-B network slicing model to achieve dynamic slicing management and optimization of the FTTR-B network.

[0005] Optionally, the preprocessing of the acquired network data of the FTTR-B network comprises: data cleaning of the network data; data conversion of the cleaned network data; anomaly detection of the data converted network data; integration and synchronization of the anomaly detected network data.

[0006] Optionally, the FTTR-B network slicing model comprises a data input layer, a classification module, a resource allocation and optimization module, an orchestration and edge computing module, a slice prediction module, and a slice monitoring and self-healing module, wherein the data input layer is configured to input preprocessed network data of the FTTR-B network and output a unified standardized tensor; the classification module is configured to classify the preprocessed network data of the FTTR-B network by type, service level, and priority based on the unified standardized tensor; the resource allocation and optimization module is configured to determine an optimal bandwidth resource allocation scheme based on the classification results output by the classification module and the current network state; the orchestration and edge computing module is configured to map the bandwidth resource allocation scheme to an edge node; the slice prediction module is configured to predict the change trend of network resources and key performance indicators of the FTTR-B network based on the bandwidth resource allocation scheme mapped to the edge node; and the slice monitoring and self-healing module is configured to monitor the slice running state, detect abnormalities, and execute a self-healing strategy.

[0007] Optionally, the data input layer adopts a multi-modal feature fusion structure.

[0008] Optionally, the classification module comprises an input layer, an encoding layer, an adaptive attention layer, a shared feature layer, and a multi-head decoder, wherein the input layer is configured to input the unified standardized tensor output by the data input layer; the encoding layer is configured to perform nonlinear mapping and dimension compression on the unified standardized tensor to extract features; the adaptive attention layer is configured to mine the correlations between the features extracted by the encoding layer through a multi-head attention mechanism; the shared feature layer is configured to integrate the outputs of the multi-head attention to form abstract feature representations suitable for multiple classification tasks; and the multi-head decoder is configured to output corresponding prediction results for the abstract feature representations of different classification tasks.

[0009] Optionally, the resource allocation and optimization module comprises a convolutional encoder, an LSTM time series modeling layer, and an Actor-Critic policy output layer, wherein the convolutional encoder is configured to extract spatial correlation features in the prediction results output by the classification module; the LSTM time series modeling layer is configured to process sequence information in the spatial correlation features output by the convolutional encoder; and the Actor-Critic policy output layer is configured to generate resource allocation actions based on the output of the LSTM time series modeling layer and evaluate their long-term returns.

[0010] The application also provides an FTTR-B network slicing device, which comprises an acquisition module configured to acquire network data of an FTTR-B network; a preprocessing module configured to preprocess the acquired network data of the FTTR-B network; and a slice management module configured to input the preprocessed network data of the FTTR-B network into a pre-trained FTTR-B network slicing model to realize dynamic slicing management and optimization of the FTTR-B network.

[0011] Optionally, the preprocessing module comprises: a cleaning submodule for cleaning the network data; a conversion submodule for converting the cleaned network data; an anomaly detection submodule for detecting anomalies in the converted network data; and an integration and synchronization submodule for integrating and synchronizing the network data after anomaly detection.

[0012] The application further provides a storage medium comprising instructions which, when executed on a computer, cause the computer to perform the method of any one of the preceding embodiments.

[0013] The application further provides an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of any one of the preceding embodiments when executing the program.

[0014] Compared with the prior art, the application has the following beneficial effects: The application can realize unified preprocessing, efficient classification, intelligent resource allocation, edge task arrangement, performance trend prediction, and real-time self-healing control of the closed-loop management mechanism for multi-source heterogeneous network data by constructing an intelligent network slicing system based on FTTR-B (fiber to the room enhanced networking). Compared with the traditional static slicing method, the application has high precision, low delay, and adaptive optimization capability, which can effectively improve the network resource utilization, service quality assurance capability, and business scheduling flexibility, and is particularly suitable for meeting the diversified, high reliability, and dynamic communication demands in an enterprise environment. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 is a flowchart of an FTTR-B network slicing method according to an embodiment of the application; Figure 2 is a structural diagram of an FTTR-B network slicing model according to another embodiment of the application; Figure 3 is a structural diagram of an FTTR-B network slicing device according to another embodiment of the application; Figure 4 is a structural diagram of a storage medium according to an embodiment of the application; Figure 5 is a structural diagram of an electronic device according to an embodiment of the application. DETAILED DESCRIPTION

[0016] The specific embodiments of the present application will be described below in detail with reference to the accompanying drawings. Although specific embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0017] It should be noted that certain terms are used in the specification and claims to refer to certain components. Those skilled in the art will understand that the same component can be referred to by different names. The specification and claims of the present application do not distinguish components based on differences in names, but rather on differences in functions. As used throughout the specification and claims, "comprising" or "including" is an open term, which should be interpreted to mean "including, but not limited to." The subsequent description describes preferred embodiments of the present application for the purpose of illustrating the general principles of the present application, and is not intended to limit the scope of the present application. The scope of the present application is defined by the appended claims.

[0018] In order to facilitate the understanding of the embodiments of the present application, the following will be further explained and described with specific embodiments as examples in conjunction with the accompanying drawings, and each drawing does not constitute a limitation on the embodiments of the present application.

[0019] Figure 1 is a flowchart of a network slicing method based on FTTR-B provided by an exemplary embodiment of the present application, as shown in Figure 1 The method comprises the following steps: S100: obtaining network data of the FTTR-B network, wherein the network data comprises, for example, service requirement data, network resource data, dynamic environment data, policy and security data, and cross-domain coordination data, wherein the service requirement data comprises service type (eMBB / URLLC / mMTC), SLA index (bandwidth, latency, jitter, reliability requirement), priority label (such as 5QI level, billing priority); the network resource data comprises physical layer data (fiber link loss, available wavelength, optical power margin), wireless data (base station load rate, spectrum resource utilization rate, MIMO configuration), core network data (NFV resource pool state (vCPU / memory / storage), SDN flow table capacity), edge side data (MEC computing resource, cache state, transmission delay); the dynamic environment data comprises channel quality, terminal mobility and traffic characteristics; the policy and security data comprises slice template, QoS mapping table and security policy; the cross and coordination data comprises transmission network state, cloud resource scheduling policy and billing rules.

[0020] S200: preprocessing the obtained network data of the FTTR-B network; S300: input the pre-processed network data of the FTTR-B network into the pre-trained FTTR-B network slicing model to realize dynamic slicing management and optimization of the FTTR-B network.

[0021] In another example embodiment, in step S200, the pre-processing of the obtained network data of the FTTR-B network includes the following steps: S201: data cleaning of the network data; In this step, first, it is necessary to check whether there are duplicate records in each network data, and ensure the uniqueness of each data through the de-duplication operation. Secondly, the missing part in the data needs to be processed, which can usually be processed in two ways: one is to fill in the missing values by interpolation method, and the other is to delete the records containing a large number of missing values. Then, the abnormal values and error data in the data need to be checked, which can be identified and corrected through statistical methods (such as box plot analysis). Finally, in order to facilitate further analysis, the data needs to be in a unified format, for example, the date format is unified, or the numerical value unit is unified to a standard unit.

[0022] Through the above processing, clean and high-quality data can be provided for subsequent model analysis, avoiding analysis errors or deviations caused by dirty data.

[0023] S202: data conversion of the cleaned network data; In this step, the purpose of data conversion is to convert the cleaned data into a format suitable for analysis and modeling. First, feature extraction is performed on the cleaned data, for example, extracting year, month, day or hour from the timestamp, or extracting proportion or difference from the numerical data. Then, for categorical data (such as gender, region, etc.), it needs to be converted into numerical form through encoding, for example, using one-hot encoding to convert categorical variables into binary vectors, or using label encoding to assign an integer value to each category. Through data conversion, the data can be presented in a more suitable form for analysis, thereby improving the efficiency and accuracy of subsequent analysis.

[0024] S203: anomaly detection of the data-converted network data; In this step, the purpose of anomaly detection is to identify error data, fraudulent behavior or other abnormal behavior. Statistical methods (such as standard deviation) can be used to identify abnormal data points far from the mean, or box plots and other visualization tools can be used to view the distribution of data and mark possible outliers. Through anomaly detection, it helps to improve the robustness of the model and ensure that the model will not be disturbed by abnormal data.

[0025] S204: Integrate and synchronize the network data after anomaly detection.

[0026] In this step, data integration and synchronization are used to integrate and ensure consistency across different data types. During the data integration phase, the first step is to ensure alignment of time, format, and semantics across all data types. For example, when processing data from multiple time zones, the time format needs to be unified; numerical data in different units requires standardization. Secondly, to resolve conflicts between network data, if different types of data have inconsistencies on the same variable (such as different measurement standards or different encoding methods), the appropriate data needs to be selected through priorities or rules. Data synchronization ensures that all data types remain consistent and updated in a timely manner.

[0027] Through data integration and synchronization, different types of data can be integrated into a unified data set to ensure data consistency, integrity and timeliness, thereby providing strong data support for subsequent model analysis.

[0028] In another exemplary embodiment, in step S200, Figure 2 As shown, the FTTR-B network slicing model includes: a data input layer, a classification module, a resource allocation and optimization module, an orchestration and edge computing module, a slice prediction module, and a slice monitoring and self-healing module. The data input layer is used to input pre-processed FTTR-B network data and output a unified standardized tensor; the classification module is used to classify the pre-processed FTTR-B network data by type, service level, and priority based on the unified standardized tensor; the resource allocation and optimization module is used to determine and generate the optimal bandwidth resource allocation plan based on the classification results output by the classification module and the current network status; the orchestration and edge computing module is used to map the bandwidth resource allocation plan to the edge nodes; the slice prediction module is used to predict the changing trends of the network resources and key performance indicators of the FTTR-B network based on the bandwidth resource allocation plan mapped to the edge nodes; and the slice monitoring and self-healing module is used to monitor the operation status of the slice, detect anomalies, and execute self-healing strategies.

[0029] Hereinafter, this application describes in detail the structure of each module mentioned above and the data processing process.

[0030] The data input layer is designed as a multi-modal feature fusion structure. For the network data of the pre-processed FTTR-B network, a corresponding feature embedding layer is constructed, then a fully connected layer is set after each feature embedding layer to perform non-linear mapping and dimension alignment on the output of each feature embedding layer to realize feature dimension unification. Then, a Dropout regularization mechanism is set after each fully connected layer to prevent overfitting, and a ReLU activation function is used to enhance the expression ability of the model. Finally, the features output by each channel are spliced into a unified standardized tensor by a splicing layer to serve as the input of the subsequent module.

[0031] The classification module includes an input layer, an encoding layer, an adaptive attention layer, a shared feature layer, and a multi-head decoder. The input layer is used to input the unified standardized tensor output from the data input layer. The encoding layer includes two fully connected layers. The first fully connected layer is used to map the standardized tensor to a 128-dimensional feature dimension and use a ReLU activation function to improve the non-linear expression ability, and then a Dropout mechanism is used to prevent overfitting. The second fully connected layer further compresses the standardized tensor with a dimension of 128 to 64 to form a stable intermediate expression. The adaptive attention layer uses a multi-head attention mechanism to model the autocorrelation of the feature dimension encoded by the encoding layer, calculates the importance weight of different dimensional features, adaptively strengthens the feature information most relevant to the current classification task, and further mines the association between features to obtain feature representations with context awareness. Next, the enhanced features output by the adaptive attention layer are uniformly mapped to the shared feature layer to capture general abstract feature representations suitable for multiple classification tasks. Finally, the multi-head decoder deploys multiple parallel fully connected classification sub-networks, each of which outputs a prediction result for a specific classification task: including a business type decoding head (using Softmax for mutually exclusive category classification), a service level decoding head (predicting different SLA levels), and a priority label decoding head (supporting multi-label output, which can use a Sigmoid activation).

[0032] Specifically, the business type decoding head is represented as:

[0033] wherein, represents the second layer weight matrix of the business type decoding head; represents the first layer weight matrix of the business type decoding head; represents the first layer bias vector of the business type decoding head; represents the first layer bias vector of the business type decoding head; represents the output vector of the shared feature layer.

[0034] The service level decoding head is denoted as:

[0035] wherein, denotes the output result of the service level decoding head; denotes the second layer weight matrix of the service level decoding head; denotes the first layer weight matrix of the service level decoding head; denotes the first layer bias vector of the service level decoding head; denotes the second layer bias vector of the service level decoding head; denotes the output vector of the shared feature layer.

[0036] The priority tag decoding head is denoted as: wherein, denotes the output result of the priority tag decoding head; denotes the second layer weight matrix of the priority tag decoding head; denotes the first layer weight matrix of the priority tag decoding head; denotes the first layer bias vector of the priority tag decoding head; denotes the second layer bias vector of the priority tag decoding head; denotes the output vector of the shared feature layer.

[0037] The resource allocation and optimization module comprises a convolutional encoder, an LSTM time series modeling layer, and an Actor-Critic policy output layer. The convolutional encoder is used to extract spatial correlation features in the prediction results output by the classification module to enhance the local perception ability of the feature representation. The convolutional encoder comprises two two-dimensional convolutional layers with ReLU activation functions, which are used to extract spatial correlation features from the prediction results output by the classification module. The LSTM time series modeling layer is used to process sequence information in the spatial correlation features output by the convolutional encoder, to mine the dynamic dependence between resource changes and slice requests, and to enhance the memory and trend perception ability of the model for historical decisions. The Actor-Critic policy output layer is used to generate resource allocation actions and evaluate their long-term benefits based on the output of the LSTM time series modeling layer. The Actor-Critic policy output layer comprises an Actor network and a Critic network. The Actor network is used to generate resource allocation actions (such as bandwidth allocation ratio, computing resource mapping, etc.) at the current time based on the output of the LSTM time series modeling layer, and the Critic network is used to predict the long-term return value of the current state. Specifically, the Actor-Critic policy output layer takes the output of the LSTM time series modeling layer as input and further performs intelligent decision optimization. Specifically, the spatial correlation features extracted by the convolutional encoder are input into the LSTM time series modeling layer to capture the dynamic evolution law in the time dimension. The hidden state vector output by the LSTM time series modeling layer is input into the Actor-Critic policy output layer as a high-dimensional representation of the current system state. The Actor network generates specific resource scheduling actions based on the state, such as dynamically allocating a certain type of URLLC business to a MEC node with low load. At the same time, the Critic network estimates the long-term value of the current state to guide the optimization and update of the policy. In this way, the model can adaptively adjust the resource allocation strategy in a dynamic network environment, thereby effectively improving the overall SLA satisfaction rate, resource utilization efficiency, and scheduling response ability.

[0038] The orchestration and edge computing module is configured to transform the scheduling decision output by the resource allocation and optimization module into an executable specific deployment scheme, and complete efficient orchestration and mapping of tasks on edge nodes. The module first receives a resource scheduling action vector output from the Actor network, which contains allocation suggestions for multiple dimensions such as computing resources, bandwidth resources, task types, and priorities. Then, the received action vector is input into a graph neural network (GNN) together with environmental context information such as the current network topology state, edge node resource load, and link delay, to construct a task-node matching association graph. On this basis, the module determines which edge node each task should be scheduled to, what resources it should occupy, and the corresponding start time and life cycle management parameters through task priority sorting, resource adaptability scoring, and multi-objective optimization algorithms (such as heuristic search or reinforcement learning-based strategies). Finally, the module outputs a task scheduling table, an edge node resource allocation list, and corresponding service chain configuration instructions, and feeds the results back to the edge computing platform controller to complete task deployment and running state monitoring.

[0039] In summary, by introducing a flexible graph modeling and adaptive strategy selection mechanism, the module can achieve dynamic adaptation of resources and tasks, ensuring optimal orchestration of tasks in a multi-node, multi-constraint environment, while also considering quality of service (QoS), computing efficiency, and system stability.

[0040] The slice prediction module is used for predicting the resource requirement, life cycle state and key performance indicator (KPI) change trend of future network slices. The slice prediction module includes an input representation layer, a task context encoding layer, a time series modeling layer, a context fusion layer and a multi-head prediction layer. First, the input representation layer receives multi-source data output from the orchestration and edge computing module, for example, including the deployment state of tasks on each edge node, historical resource usage records (such as CPU, memory, bandwidth utilization), service chain topology, task scheduling timestamp and QoS indicators (such as delay, packet loss rate, throughput) during running. This layer embeds and encodes the category information (such as task type, service level), normalizes the numerical information, and constructs a unified time series feature tensor, providing standardized input for subsequent modeling. Next, the task context encoding layer further vectorizes the static attribute information of the task (such as priority, SLA requirement, business type to which it belongs) to construct a task semantic context vector, so as to capture the differences in resource sensitivity and service quality of different tasks and provide prior knowledge for predicting tasks. Subsequently, the time series modeling layer processes the aforementioned time series features through LSTM or Transformer network, learns the dynamic dependence of resource state evolution over time, captures potential periodic fluctuations, burst behaviors and their influence on subsequent system state. Next, the context fusion layer fuses the task semantic vector with the time series representation at each time step, uses splicing or attention mechanism to enhance the prediction model's ability to perceive task characteristics, so that the resource change law under different task types can be modeled. Finally, the multi-head prediction layer performs parallel output prediction on the fused representation, each output head corresponds to a target variable, such as CPU utilization, memory occupation, delay trend, SLA violation risk or slice overload probability at future time steps, and outputs quantifiable prediction results through regression or classification, providing decision basis for subsequent resource reservation, slice rearrangement or elastic expansion and contraction strategy.

[0041] The slice monitoring and self-healing module is used to realize continuous perception, anomaly detection and fault recovery of the network slice running state. The module specifically includes a state aggregation layer, a prediction consistency analysis layer, an anomaly detection layer, a self-healing strategy decision layer and an execution and feedback layer. Firstly, the state aggregation layer receives system running indicators (such as resource utilization of edge nodes, link load, service chain topology, etc.) and future resource and KPI trends output by the slice prediction module in real time, and aligns, fuses and constructs a unified time series state representation of the two types of information as the input of the subsequent layer. Next, the prediction consistency analysis layer compares the prediction value and the actual observation value in the sliding time window, calculates the multi-dimensional deviation index (such as MAE or RMSE), judges whether the current behavior of the system deviates from the predicted trend, and identifies model failure or potential running drift. Subsequently, the anomaly detection layer further analyzes the state sequence on this basis, uses statistical rules or deep anomaly detection methods (such as LSTM autoencoder, Isolation Forest, etc.) to identify possible resource bottlenecks, QoS degradation or SLA violations, and labels each slice with an abnormal label or risk score. For the detected anomalies, the self-healing strategy decision layer generates the most appropriate self-healing actions according to the system state and anomaly type, such as edge node migration, service instance restart, elastic scaling or scheduling reconfiguration, etc. Finally, the execution and feedback layer is responsible for issuing the selected self-healing actions to the edge controller or resource orchestrator, recording the policy execution results and service recovery effects, and feeding back to the policy model for continuous optimization, so that the system has the self-healing ability of fast response, intelligent adaptation and learning evolution.

[0042] In the following, the application will be described by specific data. Table 1 is the network data of the FTTR-B network obtained by the application: Table 1

[0043] The data shown in Table 1 is input into the FTTR-B slice model, and the final output result is shown in Table 2: Table 2

[0044] Based on the slicing results shown in Table 2, the application can successfully identify high-priority URLLC-type service requirements, assign them high service levels and priority labels through the classification module, and then dynamically match reasonable computing and bandwidth resources in the resource allocation module based on the reinforcement learning strategy, ensuring the achievement of low latency and high reliability targets. The orchestration module further deploys tasks to the optimal MEC node through the graph neural network, enabling efficient mapping and life cycle control of service chains; the slicing prediction module estimates the KPI trend in advance to provide decision support for resource reservation and elastic expansion; and the monitoring and self-healing module continuously ensures the stability and robustness of system operation. Overall, the application fully embodies the "perception-decision-execution-feedback" closed-loop control capability, significantly improving the intelligence, flexibility, and service quality assurance capability of FTTR-B network slicing.

[0045] In another example embodiment, as shown in Figure 3 The application also provides an FTTR-B network slicing device, which includes: an acquisition module 100 for acquiring network data of an FTTR-B network; a preprocessing module 200 for preprocessing the acquired network data of the FTTR-B network; and a slicing management module 300 for inputting the preprocessed network data of the FTTR-B network into a pre-trained FTTR-B network slicing model to realize dynamic slicing management and optimization of the FTTR-B network.

[0046] On the basis of the above-mentioned embodiments, with reference to Figure 4 The computer-readable storage medium of the example embodiment of the application is described with reference to Figure 4 The computer-readable storage medium shown is an optical disc 40, which stores a computer program (i.e., a program product) thereon, and the computer program, when run by a processor, will implement the steps described in the above method embodiments, such as acquiring network data of an FTTR-B network; preprocessing the acquired network data of the FTTR-B network; and inputting the preprocessed network data of the FTTR-B network into a pre-trained FTTR-B network slicing model to realize dynamic slicing management and optimization of the FTTR-B network. The specific implementation methods of the steps are not repeated here.

[0047] It should be noted that the computer-readable storage medium includes, but is not limited to, a phase change memory (PRAM), a static random access memory (SRAM), a dynamic random access memory (DRAM), other types of random access memory (RAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a flash memory, or other optical or magnetic storage media, which are not repeated here.

[0048] On the basis of the above-mentioned embodiments, the present application further provides an electronic device, which is described below with reference to Figure 5 An electronic device for file downloading according to an exemplary embodiment of the present application is described.

[0049] Figure 5 A block diagram of an exemplary electronic device 50 suitable for implementing an embodiment of the present application is shown, which can be a computer system or a cloud server. Figure 5 The electronic device 50 shown is merely an example and should not bring any limitation to the functions and usage scope of the embodiments of the present application.

[0050] As shown in Figure 5 , the electronic device 50 includes, but is not limited to, one or more processors or processing units 501, a system memory 502, and a bus 503 that couples various system components including the system memory 502 and the processing unit 501.

[0051] The electronic device 50 typically includes a variety of computer system readable media. Such media can be any available media that is accessible by the electronic device 50 and includes both volatile and non-volatile media, removable and non-removable media.

[0052] The system memory 502 can include computer system readable media in the form of volatile memory, such as random access memory (RAM) 5021 and / or cache memory 5022. The electronic device 50 can further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, ROM 5023 can be used to read-only memory, non-removable, non-volatile magnetic media (e.g., a "hard drive") Figure 5 (not shown in the Figure 5 ). Although not specifically shown, a disk drive can be provided for reading and writing to a removable, non-volatile magnetic disk (e.g., a "floppy disk"), and an optical disk drive can be provided for reading and writing to a removable, non-volatile optical disk (e.g., a CD-ROM, DVD-ROM or other optical media). In these instances, each drive can be connected to the bus 503 by one or more data media interfaces. It is to be appreciated that the system memory 502 can include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of embodiments of the present application.

[0053] Program / utility 5025 having a set of program modules 5024 can be stored in system memory 502 and include operating system software, one or more application programs, other program modules, and program data, each of which implements at least a portion of a network environment as described herein. Program modules 5024 also generally carry out the functions of embodiments of the application as described herein.

[0054] Electronic device 50 can also communicate with one or more external devices 504 such as a keyboard or pointing device, a display, etc. through I / O interface 505. Furthermore, electronic device 50 can communicate with one or more networks such as a local area network (LAN), a wide area network (WAN), and / or a public network such as the Internet, through network adapter 506. As Figure 5 illustrated, network adapter 506 communicates with the other modules of electronic device 50 (e.g., processing unit 501, etc.) through bus 503. It should be appreciated that although not shown, other hardware and / or software modules could be used in conjunction with electronic device 50. Examples include stand-alone applications, other operating systems, component services, etc. ​

[0055] Processing unit 501 performs various function applications and data processing by running programs stored in system memory 502, such as obtaining network data of the FTTR-B network, pre-processing the obtained network data of the FTTR-B network, inputting the pre-processed network data of the FTTR-B network into the pre-trained FTTR-B network slice model to realize dynamic slice management and optimization of the FTTR-B network. The specific implementation of each step is not repeated here. It should be noted that although several units / modules or sub-units / sub-modules of the file concurrent downloading device are mentioned in the above detailed description, such division is only exemplary and not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided into a plurality of units / modules.

[0056] In the description of the present application, it should be noted that the terms "first", "second", "third" are only for the purpose of description and cannot be understood as indicating or implying relative importance.

[0057] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0058] ​In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented by other manners. The above-described device embodiments are only illustrative, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, and for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some communication interfaces, devices or units, which can be electrical, mechanical or other forms.

[0059] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, i.e., they can be located in one place or distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0060] In addition, the functional units in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit.

[0061] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a non-volatile computer readable storage medium executable by a processor. Based on this understanding, the technical solutions of the present application essentially or the parts that contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes a plurality of instructions for executing all or part of the steps of the methods described in the embodiments of the present application by a computer device (which can be a personal computer, a cloud server, or a network device, etc.). The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various program code storage media.

[0062] The above embodiments are only for illustrating the technical concepts and characteristics of the present application, and the purpose is to enable those skilled in the art to understand the content of the present application and implement it, and cannot limit the protection scope of the present application. Any equivalent changes or modifications made according to the spirit and essence of the present application should be covered within the protection scope of the present application.

Claims

1. A FTTR-B network slicing method, characterized in that: The method comprises: Obtain network data for FTTR-B networks; Preprocessing the acquired network data of the FTTR-B network; The preprocessed network data of the FTTR-B network is input into the pre-trained FTTR-B network slicing model to achieve dynamic slicing management and optimization of the FTTR-B network.

2. The method according to claim 1, characterized in that The preprocessing of the acquired FTTR-B network data includes: Perform data cleaning on network data; Perform data conversion on the cleaned network data; Perform anomaly detection on network data after data conversion; Integrate and synchronize network data after anomaly detection.

3. The method according to claim 1, characterized in that The FTTR-B network slicing model includes: Data input layer, classification module, resource allocation and optimization module, orchestration and edge computing module, slice prediction module and slice monitoring and self-healing module, among which, The data input layer is used to input the preprocessed network data of the FTTR-B network and output a unified standardized tensor; The classification module is used to classify the type, service level and priority of the pre-processed FTTR-B network data based on a unified normalized tensor; The resource allocation and optimization module is used to make decisions and generate the optimal bandwidth resource allocation plan based on the classification results output by the classification module and the current network status; The orchestration and edge computing module is used to map the bandwidth resource allocation plan to the edge nodes; The slice prediction module is used to predict the changing trends of network resources and key performance indicators of the FTTR-B network based on the bandwidth resource allocation scheme mapped to the edge nodes; The slice monitoring and self-healing module is used to monitor the slice operation status, detect anomalies and implement self-healing strategies.

4. The method according to claim 3, characterized in that The data input layer adopts a multimodal feature fusion structure.

5. The method according to claim 3, characterized in that The classification module includes: Input layer, encoding layer, adaptive attention layer, shared feature layer and multi-head decoder, where The input layer is used to input the unified normalized tensor output by the data input layer; The encoding layer is used to perform nonlinear mapping and dimension compression on the unified normalized tensor for feature extraction; The adaptive attention layer is used to mine the correlation between features extracted by the encoding layer through a multi-head attention mechanism; The shared feature layer is used to integrate the output of multi-head attention to form an abstract feature representation suitable for multiple classification tasks; The multi-head decoder is used to output the corresponding prediction results for the abstract feature representation of different classification tasks.

6. The method according to claim 3, characterized in that The resource allocation and optimization module includes: Convolutional encoder, LSTM time series modeling layer and Actor-Critic strategy output layer, where The convolutional encoder is used to extract spatial correlation features in the prediction results output by the classification module; The LSTM temporal modeling layer is used to process the sequence information in the spatial correlation features output by the convolutional encoder; The Actor-Critic strategy output layer is used to generate resource allocation actions based on the output of the LSTM timing modeling layer and evaluate their long-term benefits.

7. A FTTR-B network slicing device, characterized in that: The device comprises: An acquisition module is used to obtain network data of the FTTR-B network; A preprocessing module, used for preprocessing the acquired network data of the FTTR-B network; The slice management module is used to input the pre-processed network data of the FTTR-B network into the pre-trained FTTR-B network slicing model to realize dynamic slicing management and optimization of the FTTR-B network.

8. The device according to claim 7, characterized in that The pre-processing module comprises: Cleaning submodule, used to clean network data; The conversion submodule is used to perform data conversion on the cleaned network data; The anomaly detection submodule is used to perform anomaly detection on the network data after data conversion; The integration and synchronization submodule is used to integrate and synchronize network data after anomaly detection.

9. A storage medium, characterized in that: The method comprises instructions, which, when executed on a computer, enable the computer to execute the method according to any one of claims 1 to 6.

10. An electronic device, characterized in that: The electronic device comprises: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to any one of claims 1 to 6 is implemented.

Images