Dynamic allocation method and system for communication network slice resources
By constructing a mapping database between quantum network states and resource templates, the problem of resource allocation latency in existing technologies is solved, achieving sub-millisecond resource scheduling response and improving the network's adaptability to sudden traffic and service quality.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-04
- Publication Date
- 2026-03-10
AI Technical Summary
Existing communication network slicing resource allocation methods rely on predictive models based on historical data, which cannot respond in real time to sudden surges in traffic at the millisecond level, resulting in delays in resource allocation decisions and affecting service quality.
A mapping database between quantum network states and resource templates is constructed. By analyzing historical data offline, the optimal resource configuration template is generated, and high-speed matching is performed during the runtime phase to directly call the pre-set scheme for resource scheduling.
It achieves sub-millisecond resource scheduling response, improves the network's adaptability to sudden traffic surges, ensures the quality of service for latency-sensitive services, and avoids service interruptions and performance degradation.
Smart Images
Figure CN121644366A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of communication technology, specifically relating to a method and system for dynamic allocation of communication network slice resources. Background Technology
[0002] With the evolution of fifth-generation mobile communication (5G) and future network architecture, network slicing technology, as a core enabling means to support diverse vertical industry applications, has shown great potential in fields such as industrial internet, intelligent transportation, telemedicine, and large-scale Internet of Things.
[0003] Network slicing provides customized quality of service (QoS) guarantees for different services by building multiple logically isolated virtual networks on a unified physical infrastructure. Its core lies in dynamically allocating computing, storage, and bandwidth resources according to business needs to meet differentiated performance metrics such as low latency, high reliability, or massive connectivity.
[0004] Dynamic allocation of communication network slice resources is a crucial step in ensuring efficient service. This process requires real-time awareness of service traffic characteristics and adjustment of resource quotas within slices accordingly to ensure that critical applications can maintain stable performance under sudden loads. Ideally, the resource allocation mechanism should have millisecond or even sub-millisecond response capabilities to cope with typical scenarios such as sudden spikes in bitrate for video streaming or concurrent access by massive numbers of IoT devices.
[0005] Existing technologies typically rely on offline predictive models based on historical traffic data for resource pre-allocation, but their decision-making logic lags significantly behind actual traffic changes. Such methods struggle to capture millisecond-level instantaneous traffic fluctuations, leading to resource scheduling delays and consequently causing issues such as video stuttering, lost control commands, or IoT device connection interruptions.
[0006] Meanwhile, centralized prediction architectures, due to the need to aggregate global data and perform centralized computation, further exacerbate decision-making latency, failing to meet the edge's demand for extremely low latency. Therefore, there is an urgent need for a dynamic resource allocation mechanism that can deeply integrate real-time perception and lightweight online learning capabilities to overcome the bottlenecks of existing technologies in terms of response speed and adaptability. Summary of the Invention
[0007] The technical problem to be solved by this invention is that existing communication network slicing resource allocation methods rely on prediction models based on historical data. Such models cannot respond to millisecond-level burst traffic in real time and accurately, resulting in inherent delays in resource allocation decisions, which in turn leads to a decline in service quality.
[0008] To address the aforementioned technical problems, this invention provides a method and system for the dynamic allocation of communication network slice resources. The core idea of this method is to abandon the traditional allocation paradigm of online prediction and real-time calculation, and instead construct a deterministic mapping knowledge base between network state and resource strategy. Specifically, through offline deep analysis of massive historical network state data, a set of finite, representative quantized network states is extracted and solidified using a high-dimensional data clustering algorithm. For each quantized network state, a multi-objective optimization algorithm is used to offline calculate and generate a uniquely corresponding optimal quantized resource template.
[0009] During actual operation, the system collects real-time network status data at an extremely high frequency. Through an efficient approximate nearest neighbor search algorithm, it instantly matches the real-time status to the closest quantized network status in the knowledge base and directly calls its pre-calculated quantized resource template. The system then issues the pre-calculated command to the network infrastructure for execution through a standardized command interface.
[0010] This mechanism transforms complex and time-consuming online optimization problems into simple table lookup and matching operations that can be completed within a defined timeframe, thereby completely eliminating decision-making delays caused by real-time computation and achieving sub-millisecond resource scheduling responses to sudden surges in network traffic.
[0011] This invention provides a method for dynamically allocating communication network slice resources, which includes the following steps: A mapping database between quantum network states and resource templates is constructed, and the construction process is completed offline before the network runs; Real-time acquisition of full-dimensional state data of the network, and synthesis of the full-dimensional state data into a real-time network state vector; In the mapping database, a high-speed matching retrieval is performed based on the real-time network state vector to determine the quantized network state that is closest to the real-time network state vector; Based on the determined quantized network state, extract its unique corresponding quantized resource template from the mapping database; Based on the resource configuration parameters defined in the quantized resource template, resource scheduling instructions are generated and sent to the network infrastructure to complete the dynamic allocation of network slice resources.
[0012] As one embodiment of the present invention, the construction of the mapping database between quantum network states and resource templates specifically includes: Collect historical network operation data, which covers multi-dimensional performance indicators of multiple network slices in different time periods. The multi-dimensional performance indicators include uplink and downlink bandwidth, packet loss rate, transmission latency, latency jitter, number of connected devices, and computing and storage resource utilization rate of the network function virtualization for each slice. The historical operation data is normalized and aligned with timestamps to form a standardized multidimensional historical state dataset; An unsupervised learning clustering algorithm based on noise application is used to perform clustering on the standardized multidimensional historical state dataset. The output of the algorithm is a set of cluster centers, and each cluster center vector is defined as a quantized network state. For each defined quantized network state, it is used as input condition, and a multi-objective particle swarm optimization algorithm is applied to solve the resource allocation optimization problem. The objective function of the optimization problem aims to maximize the service level agreement satisfaction while minimizing resource consumption. The output of the algorithm is the optimal resource configuration vector, which is the quantized resource template corresponding to the quantized network state. All quantized network states and their corresponding quantized resource templates are stored in key-value pairs to form the mapping database.
[0013] As one embodiment of the present invention, the step of acquiring full-dimensional state data of the network in real time and synthesizing the full-dimensional state data into a real-time network state vector specifically includes: By deploying monitoring agents in network forwarding nodes and virtualization infrastructure, real-time performance index data of each network slice are collected synchronously at a sampling frequency of no less than 1000 Hz. The collected multi-dimensional and heterogeneous real-time performance index data are processed and spliced according to the same dimensional order and normalization method as when constructing the quantized network state, thereby generating a real-time network state vector with the same dimensions and scale as the quantized network state vector.
[0014] As one embodiment of the present invention, the step of performing high-speed matching retrieval based on the real-time network state vector in the mapping database to determine the quantized network state closest to the real-time network state vector specifically includes: After the mapping database is constructed, a hierarchical navigable small-world graph index is pre-constructed using all the quantized network state vectors in the database as nodes; Using the real-time network state vector as the query input, a search is performed in the hierarchical navigable small world graph index. By combining greedy traversal with multi-level jumps, the quantized network state vector with the smallest Euclidean distance in the feature space relative to the real-time network state vector is quickly located.
[0015] As one embodiment of the present invention, the step of generating resource scheduling instructions and issuing them to the network infrastructure based on the resource configuration parameters defined in the quantized resource template specifically includes: The quantized resource template is parsed to extract the specific resource quotas defined for each network slice. The resource quotas include the number of CPU cores, memory allocation capacity, physical bandwidth guarantee value, and packet forwarding priority queue weight. The resource quota is encapsulated into a structured data object that conforms to the network infrastructure application programming interface specification, and the structured data object adopts the JavaScript object representation format; The structured data object is sent as an instruction to the software-defined network controller or network function virtualization orchestrator via an application programming interface based on a representational state transition architecture. The software-defined network controller is responsible for parsing the instruction and converting it into specific configuration operations for the underlying physical and virtual resources.
[0016] This invention provides a dynamic allocation system for communication network slice resources, comprising: An offline build module is used to build a mapping database between quantum network states and resource templates; The state awareness module is used to acquire the full-dimensional state data of the network in real time and synthesize the full-dimensional state data into a real-time network state vector. The state matching module is used to perform high-speed matching retrieval based on the real-time network state vector in the mapping database to determine the quantized network state that is closest to the real-time network state vector. The template extraction module is used to extract the unique corresponding quantized resource template from the mapping database based on the determined quantized network state. The instruction execution module is used to generate resource scheduling instructions and send them to the network infrastructure based on the resource configuration parameters defined in the quantized resource template, so as to complete the dynamic allocation of network slice resources.
[0017] In one embodiment of the present invention, the offline construction module includes a data acquisition unit, a state quantization unit, and a template optimization unit. The data acquisition unit collects and preprocesses historical network operation data. The state quantization unit integrates a noise-based density space clustering algorithm to extract quantized network states from historical data. The template optimization unit integrates a multi-objective particle swarm optimization algorithm to calculate and generate the corresponding optimal quantized resource template for each quantized network state.
[0018] In one embodiment of the present invention, the core of the state matching module is a search engine based on a hierarchical navigable small-world graph index. During system initialization, the search engine constructs an index using all quantized network state vectors in the mapping database. At runtime, it receives real-time network state vectors generated by the state-aware module, performs a highly efficient approximate nearest neighbor query operation, and outputs the quantized network state identifier with the highest matching degree.
[0019] Compared with the prior art, the advantages and positive effects of the present invention are as follows: 1. This invention fundamentally solves the response delay problem caused by prediction models and real-time optimization algorithms by transforming the resource allocation decision-making process from complex online real-time calculation to a deterministic offline calculation and online matching process.
[0020] 2. By leveraging a pre-built mapping database of quantum network states and resource templates, the system can obtain the optimal resource allocation scheme by performing only one efficient vector matching operation when facing drastic fluctuations in real-time network traffic, without any time-consuming prediction or solution. This reduces the overall resource scheduling response time from seconds or hundreds of milliseconds to sub-milliseconds, significantly improving the network's adaptability to sudden traffic surges and ensuring the service quality of latency-sensitive services such as high-definition video and industrial IoT, avoiding service interruptions or performance degradation caused by untimely resource allocation. Furthermore, the offline optimization method allows for the use of more computationally complex and powerful global optimization algorithms, ensuring the quality and efficiency of each pre-built resource template. Compared to online heuristic or greedy algorithms, it can find resource allocation schemes closer to Pareto optimality in a wider solution space, thereby achieving higher network resource utilization while maintaining service quality. Attached Figure Description
[0021] Figure 1 This is a schematic diagram of the overall technical architecture of a dynamic allocation method and system for communication network slice resources proposed in this invention. Figure 2 This is a schematic diagram of the core principle framework of the quantum network state and resource template mapping knowledge base in this invention; Figure 3 This is a logical flowchart of the offline construction of the quantum network state and resource template mapping database in this invention; Figure 4 This is a logical flowchart of the real-time network state perception and vector synthesis in this invention. Figure 5 This is a schematic diagram illustrating the principle of the high-speed state matching mechanism based on hierarchical navigable small world graph index in this invention; Figure 6This is a schematic diagram of the multi-level interaction relationship and data flow between the generation and distribution of terminal scheduling instructions to the network infrastructure in this invention. Detailed Implementation
[0022] Please refer to the attached document. Figures 1 to 6 This invention provides a method and system for dynamic allocation of communication network slice resources, aiming to solve the resource allocation response delay problem caused by relying on historical data prediction models in existing technologies. Traditional solutions cannot adjust resource quotas in time when faced with millisecond-level burst traffic, resulting in service quality degradation such as video service interruptions and industrial IoT device disconnections. This embodiment constructs an offline-generated quantum network state and resource template mapping database, and performs high-speed matching between the real-time network state vector and this database during the runtime phase, directly calling the preset optimal resource configuration scheme, thereby achieving sub-millisecond resource scheduling response.
[0023] The dynamic allocation method for communication network slice resources includes the following steps: constructing a mapping database of quantized network states and resource templates; acquiring full-dimensional state data of the network in real time and synthesizing it into a real-time network state vector; performing high-speed matching and retrieval based on the real-time network state vector in the mapping database to determine the nearest quantized network state; extracting the unique corresponding quantized resource template based on the determined quantized network state; generating resource scheduling instructions based on the quantized resource template and issuing them to the network infrastructure.
[0024] First, a mapping database between quantum network states and resource templates is constructed. This construction process is completed before the network officially goes live and is an offline operation. Specifically, historical network operation data is collected, covering multi-dimensional performance indicators of multiple network slices at different time periods. These performance indicators include uplink bandwidth, downlink bandwidth, packet loss rate, transmission latency, latency jitter, number of connected devices, and the utilization rates of computing and storage resources supporting network function virtualization for each slice. All raw data comes from the log interfaces of the network management system, virtualization infrastructure monitoring agent, and software-defined network controller.
[0025] The collected historical operational data is normalized and timestamped. Normalization employs a min-max scaling method, mapping each dimension of the index to a closed interval between 0 and 1 to eliminate numerical biases caused by different physical units. Timestamp alignment is achieved through an interpolation algorithm, ensuring that all indices have corresponding values at the same sampling time, ultimately forming a standardized multidimensional historical state dataset. Each record in this dataset is a high-dimensional vector, with its dimension equal to the total number of the aforementioned performance indices.
[0026] Subsequently, an unsupervised learning clustering algorithm based on noise application was employed to cluster the standardized multidimensional historical state dataset. This algorithm automatically identifies high-density regions in the dataset and removes low-density regions as noise, thereby extracting a set of representative cluster centers. Each cluster center vector is defined as a quantized network state. During the clustering process, the algorithm parameters, including the neighborhood radius and the minimum number of neighborhood points, are determined through grid search based on the dataset size and distribution characteristics to ensure that the clustering results cover the main operating scenarios while avoiding over-subdivision leading to state redundancy.
[0027] For each defined quantized network state, using it as input, a multi-objective particle swarm optimization algorithm is applied to solve the resource allocation optimization problem. The objective function of this optimization problem contains two conflicting sub-objectives: The goal is to maximize Service Level Agreement (SLA) satisfaction while minimizing resource consumption. SLA satisfaction is calculated using a weighted average to determine the compliance rate of each slice across key metrics such as bandwidth, latency, and packet loss rate. Resource consumption is determined by comprehensively considering the total number of CPU cores used, total allocated memory capacity, and total reserved physical link bandwidth. The Particle Swarm Optimization (PSO) algorithm maintains a swarm of particles during iteration, where each particle represents a feasible resource allocation scheme, and its position vector is the resource allocation vector. The algorithm guides particle updates through individual and global optima, ultimately converging to a non-dominated solution on the Pareto front. This solution is the quantized resource template corresponding to the current quantized network state.
[0028] All quantized network states and their corresponding quantized resource templates are stored as key-value pairs in a high-performance key-value database, forming the mapping database. The key is a hash identifier of the quantized network state vector, and the value is the corresponding quantized resource template. This database supports millisecond-level read and write operations and has persistence and snapshot recovery capabilities, ensuring that the integrity of the knowledge base is not affected after a system restart.
[0029] Secondly, real-time, comprehensive network state data is acquired and synthesized into a real-time network state vector. This process is executed by monitoring agents deployed in network forwarding nodes and virtualization infrastructure. The monitoring agents synchronously collect real-time performance metrics data for each network slice at a sampling frequency of no less than 1000 Hz. The collected content is completely consistent with the historical data construction phase, including uplink bandwidth, downlink bandwidth, packet loss rate, transmission latency, latency jitter, number of connected devices, computing resource utilization, and storage resource utilization.
[0030] The collected multi-dimensional, heterogeneous real-time performance metrics data are arranged in the same dimensional order as when constructing the quantized network state and normalized using the same minimum-maximum scaling parameters. These normalization parameters are fixed offline and loaded into the runtime environment along with the mapping database. The processed data is then concatenated into a real-time network state vector with dimensions completely identical to the quantized network state vector. The generation latency of this vector is strictly controlled within 100 microseconds to ensure the timeliness of subsequent matching operations.
[0031] Third, in the mapping database, a high-speed matching retrieval is performed based on the real-time network state vector to determine the quantized network state closest to the real-time network state vector. To achieve efficient retrieval, after the mapping database is constructed, the system uses all the quantized network state vectors in the database as nodes to pre-build a hierarchical navigable small-world graph index. This index structure consists of multiple layers of graphs, with the bottom layer containing all nodes, and the upper layers becoming sparser layer by layer through random sampling, forming a pyramid-shaped topology. In each layer of the graph, nodes are connected to their nearest neighbors through edges, and the weight of the edges is determined by Euclidean distance.
[0032] The real-time network state vector is used as the query input, and a search is performed in the hierarchical navigable small-world graph index. The search process starts from the top layer, using a greedy strategy to select the nearest neighbor node to the query vector and jump to it until no closer node can be found; then it moves to the next layer, repeating the above process in the local neighborhood of the current node, layer by layer down to the bottom layer. Finally, the quantized network state vector with the smallest distance is returned at the bottom layer. The time complexity of the entire search process is logarithmic, with a typical query time of less than 200 microseconds, meeting the sub-millisecond response requirement.
[0033] Fourth, based on the determined quantized network state, extract its unique corresponding quantized resource template from the mapping database. The extraction operation uses the hash identifier of the quantized network state as the key to perform an exact lookup in the key-value database. The database returns a complete quantized resource template, the content of which is a structured vector containing specific resource quotas defined for each network slice. These quotas include the number of CPU cores, memory allocation capacity, physical bandwidth guarantee value, and packet forwarding priority queue weight. All quota values are absolute values, not relative proportions, to ensure that instructions can be executed directly.
[0034] Fifth, based on the resource configuration parameters defined in the quantized resource template, resource scheduling instructions are generated and issued to the network infrastructure. Specifically, the quantized resource template is parsed, the resource quotas for each slice are extracted, and they are encapsulated into a structured data object conforming to the network infrastructure application programming interface specification. This object adopts the JavaScript object notation format, and the field naming follows the Open Network Automation Platform standard, including slice identifier, resource type, allocation value, and effective timestamp.
[0035] The structured data object is sent as an instruction to the software-defined network controller (SDN) or network function virtualization orchestrator via an application programming interface (API) based on a representational state transition architecture. The API uses Hypertext Transfer Protocol Security (HTTP) for encrypted transmission, and the request method is the HTTP placement method. Upon receiving the instruction, the SDN parses the bandwidth and queue weight parameters, generates flow table entries, and distributes them to the programmable switch. The network function virtualization orchestrator parses the CPU and memory parameters and calls the hypervisor interface to adjust the resource limits of virtual machines or containers. The entire distribution process is completed within 500 microseconds, ensuring an end-to-end scheduling latency of less than 1 millisecond.
[0036] The core algorithm formulas involved in the above method steps are as follows: During the state quantization phase, the noise-based density space clustering algorithm determines the core points using the following criteria: for any vector If its The number of points in the neighborhood is no less than ,but The core point. Among them... The neighborhood radius, The minimum number of neighborhood points.
[0037] During the template optimization phase, the position update formula for the multi-objective particle swarm optimization algorithm is: ; ; in and The first The particle in the first Position and velocity in the next iteration This is the best historical position for this particle. As a reference point in the globally non-dominated solution set, Inertial weight and For acceleration coefficient, and A random number between 0 and 1.
[0038] During the state matching phase, the real-time network state vector With the quantum network state vector The formula for calculating the Euclidean distance between them is: ; in For vector dimensions, and The two vectors are respectively in the second place. The components of a dimension.
[0039] In the service level agreement (SLA) satisfaction calculation, the compliance index for the $m$-th slice is defined as: ; in To allocate bandwidth in practice For bandwidth requirements, For actual delay, For the maximum allowable delay, For packet loss rate, , , The weighting coefficients are and satisfy the following conditions: The overall service level agreement satisfaction rate is the weighted average of the compliance indices of all slices.
[0040] Corresponding to the method, the present invention also provides a dynamic allocation system for communication network slice resources. This system includes an offline construction module, a state awareness module, a state matching module, a template extraction module, and an instruction execution module.
[0041] The offline construction module is used to build a mapping database between quantized network states and resource templates. This module includes a data acquisition unit, a state quantization unit, and a template optimization unit. The data acquisition unit interfaces with the network management system and virtualization monitoring interface, and is responsible for pulling, cleaning, and storing historical data. The state quantization unit integrates a density space clustering algorithm engine based on noise applications, receives standardized historical datasets, and outputs a set of quantized network states. The template optimization unit integrates a multi-objective particle swarm optimization algorithm solver to generate a corresponding quantized resource template for each quantized network state. The three units work together to complete the offline construction and persistence of the knowledge base.
[0042] The state-aware module acquires real-time, multi-dimensional network state data and synthesizes it into a real-time network state vector. This module consists of a distributed monitoring agent cluster, deployed at base stations, core network gateways, virtualized hosts, and software-defined network switches. The agents collect performance metrics through polling or event-driven methods and aggregate the data to a central synthesizer via a lightweight message queue. The central synthesizer performs dimension alignment, normalization, and vector concatenation, outputting a standardized real-time network state vector.
[0043] The state matching module is used for high-speed matching and retrieval in the mapping database based on real-time network state vectors. The core of this module is a search engine based on a hierarchical navigable small-world graph index. The index builder loads all quantized network state vectors during system initialization, constructs a multi-layer graph structure, and caches it in memory. The query processor receives the real-time network state vectors, performs greedy traversal and inter-layer jumps, and returns the identifier of the best-matching quantized network state.
[0044] The template extraction module is used to extract the unique corresponding quantized resource template based on the determined quantized network state. This module is implemented as a high-performance key-value query interface, which receives the state identifier as input, reads the corresponding resource template from the mapping database, and outputs it to the instruction execution module.
[0045] The instruction execution module generates resource scheduling instructions based on the quantized resource template and issues them to the network infrastructure. This module includes an instruction wrapper and an interface adapter. The instruction wrapper converts resource quotas into standard JavaScript object representation objects; the interface adapter selects the appropriate representational state transition application interface based on the target infrastructure type (software-defined network controller or network function virtualization orchestrator), executes the instruction issuance, and listens for execution confirmation acknowledgments.
[0046] The modules of the above system are interconnected through an internal message bus and exchange data using a publish-subscribe model. All modules are deployed on a highly available server cluster, supporting hot standby switching and load balancing to ensure the stability and reliability of the system under high-concurrency scenarios.
[0047] This embodiment transforms the complex online optimization problem into a combination of offline computation and online table lookup, completely avoiding the uncertainty and delay caused by real-time prediction and solution. Experiments show that when facing sudden traffic surges, the resource scheduling response time of this solution remains stable within 800 microseconds, which is two orders of magnitude faster than traditional online reinforcement learning-based methods. Simultaneously, the service level agreement violation rate is reduced by 63%, and resource utilization is improved by 21%. This technology is particularly suitable for latency-sensitive applications in 5G and future 6G mobile communication networks, such as augmented reality, remote surgery, and industrial automation.
[0048] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A method for dynamic allocation of communication network slice resources, characterized in that, The application comprises: constructing a mapping database of quantumized network states and resource templates, the construction process being completed offline before network operation; real-time acquisition of full-dimension state data of the network, and integration of the full-dimension state data into a real-time network state vector; in the mapping database, high-speed matching retrieval is performed based on the real-time network state vector to determine a quantumized network state closest to the real-time network state vector; according to the determined quantumized network state, a uniquely corresponding quantumized resource template is extracted from the mapping database; according to resource configuration parameters defined in the quantumized resource template, resource scheduling instructions are generated and issued to network infrastructure to complete dynamic allocation of network slice resources. 2.The method of claim 1, wherein, The construction of the mapping database of quantumized network states and resource templates comprises: collecting network historical operation data, the historical operation data covering multiple network slices in different time periods and multiple-dimension performance indicators, the multiple-dimension performance indicators including uplink and downlink bandwidth, data packet loss rate, transmission delay, delay jitter, number of connected devices, and occupancy rate of computing and storage resources supporting network function virtualization; normalizing and timestamp aligning the historical operation data to form a standardized multi-dimensional historical state data set; performing unsupervised learning clustering on the standardized multi-dimensional historical state data set by using a density space clustering algorithm based on noise application, the output of the algorithm being a set of clustering centers, each clustering center vector being defined as a quantumized network state; for each defined quantumized network state, a multi-objective particle swarm optimization algorithm is applied to solve a resource allocation optimization problem, the objective function of the optimization problem aiming to maximize service level agreement satisfaction degree and minimize resource consumption, the output of the algorithm being an optimal resource configuration vector, which is a quantumized resource template corresponding to the quantumized network state; all quantumized network states and their corresponding quantumized resource templates are stored in the form of key-value pairs to constitute the mapping database. 3.The method of claim 2, wherein, The real-time acquisition of full-dimension state data of the network and the integration of the full-dimension state data into a real-time network state vector comprise: by deploying monitoring agents in network forwarding nodes and virtualized infrastructure, real-time performance indicator data of each network slice is synchronously collected at a sampling frequency of no less than 1000 Hz; the collected multi-dimension, heterogeneous real-time performance indicator data is processed and spliced according to the same dimension order and normalization method as when the quantumized network states are constructed, so as to generate a real-time network state vector consistent in dimension and uniform in scale with the quantumized network state vector.
4. The method of claim 3, wherein, In the mapping database, high-speed matching retrieval is performed based on the real-time network state vector to determine a quantumized network state closest to the real-time network state vector, which comprises: after the mapping database is constructed, all quantumized network state vectors in the database are used as nodes to pre-construct a hierarchical navigable small-world graph index. The real-time network state vector is input as a query to search in the hierarchical navigable small-world graph index, and a quantum network state vector with the minimum Euclidean distance in a feature space from the real-time network state vector is quickly located by combining greedy traversal and multi-layer jumping. 5.The method of claim 4, wherein, According to the resource configuration parameters defined in the quantized resource template, a resource scheduling instruction is generated and sent to the network infrastructure, including: The quantized resource template is parsed to extract the specific resource quota defined for each network slice, including the number of central processing unit cores, memory allocation capacity, physical bandwidth guarantee value, and data packet forwarding priority queue weight; The resource quota is packaged into a structured data object conforming to the network infrastructure application programming interface specification, which adopts the JavaScript Object Notation format; The structured data object is sent as an instruction to a software-defined network controller or network function virtualization orchestrator through a representational state transfer architecture-based application programming interface, which is responsible for parsing and converting the instruction into specific configuration operations on the underlying physical and virtual resources. 6.The method of claim 2, wherein, The historical operation data is normalized and timestamped, including: Each dimension of performance indicators is mapped to the closed interval of 0 to 1 using the min-max scaling method; The timestamp of performance indicator data from different sources is aligned by an interpolation algorithm to ensure that all indicators have corresponding values at the same sampling time, forming the standardized multi-dimensional historical state data set. 7.The method of claim 2, wherein, An unsupervised learning clustering algorithm based on a noise application density space clustering algorithm is used to cluster the standardized multi-dimensional historical state data set, including: The neighborhood radius and minimum neighborhood point number are set as clustering parameters; Core points in high-density areas of the data set are identified, and low-density areas are considered as noise and removed; A set of cluster center vectors is output as the quantized network state. 8.The method of claim 2, wherein, A multi-objective particle swarm optimization algorithm is used to solve the resource allocation optimization problem, including: The service level agreement satisfaction and resource consumption are defined as two conflicting optimization sub-targets; The particle swarm is initialized, and the position vector of each particle represents a feasible resource configuration scheme; The particle is iteratively updated by the individual optimal position and global non-dominated solution set, and finally converges to the non-dominated solution on the Pareto front as the quantized resource template. 9.A system for dynamic allocation of communication network slice resources, characterized in that, including: An offline construction module for constructing a mapping database of quantized network states and resource templates; A state-aware module for real-time acquisition of full-dimensional state data of the network, and integration of the full-dimensional state data into a real-time network state vector; A state matching module for high-speed matching retrieval in the mapping database based on the real-time network state vector to determine the quantum network state closest to the real-time network state vector; A template extraction module for extracting the unique corresponding quantum resource template from the mapping database according to the determined quantum network state; The instruction execution module is configured to generate resource scheduling instructions according to the resource configuration parameters defined in the quantized resource template, and to send the resource scheduling instructions to a network infrastructure to complete dynamic allocation of network slice resources. 10.The system for dynamic allocation of communication network slice resources according to claim 9, characterized in that, The offline construction module comprises: A data acquisition unit is configured to acquire and pre-process network historical operation data. A state quantization unit is configured to integrate a density space clustering algorithm based on noise application, and to extract quantized network states from the historical data. A template optimization unit is configured to integrate a multi-objective particle swarm optimization algorithm, and to calculate an optimal quantized resource template corresponding to each quantized network state.