Network service dynamic optimization method, system, electronic device and medium

Abstract

The present disclosure provides a network service dynamic optimization method, system, electronic device and storage medium to solve the problem that the network service cannot schedule resources and adjust performance indicators. The method comprises: standardizing the vehicle-road cooperation service network indicator to form a performance indicator standard library; corresponding to the performance indicator standard library, constructing a network performance indicator test library containing various service types; using the constructed network performance indicator test library, training the network performance evaluation and optimization model according to the event flow and different service classification; collecting data by the UPF network element according to the performance indicator standard library, and inputting the processed data into the network performance evaluation and optimization model to output the network service optimization strategy; and scheduling resources based on the network service optimization strategy. The present disclosure can realize dynamic self-adaptive adjustment of performance indicator parameters and service resources, and effectively guarantee the network performance of vehicle-road cooperation services.

Description

Technical Field

[0001] This disclosure relates to the field of communication technology, specifically to a method for dynamic optimization of network services, a system for dynamic optimization of network services, an electronic device, and a computer-readable storage medium. Background Technology

[0002] In recent years, V2X (vehicle-to-X) has become a hot topic of innovation, attracting numerous domestic and international giants and startups. The true realization of autonomous driving relies heavily on the crucial support of vehicle-to-infrastructure (V2X). However, due to safety and reliability concerns, related technologies and service capabilities still face many challenges, preventing large-scale commercialization. Network communication is a key technology for V2X. Without low-latency, highly reliable network support, V2X is impossible. The development of V2X communication technology has progressed from DSRC (Dedicated Short-Range Communication) and LTE-V (long-term evolution-vehicle) to 5G. Based on the robust 3GPP (3rd Generation Partnership Project) ecosystem and continuously improving cellular network coverage, the deployment cost of V2X has been significantly reduced. Currently, C-V2X (Cellular-V2X), an evolution of cellular communication technology, has become a de facto communication standard.

[0003] In 5G-V2X technology, leveraging the low latency and high bandwidth characteristics of 5G networks, 5G can be chosen as the channel for vehicles to exchange information with the outside world. Combined with MEC (Mobile Edge Computing) technology, a complete technical solution can be formed. However, because slicing technology has not been widely applied in actual networks, traditional applications and vehicle-to-infrastructure (V2I) applications face resource competition. Network services cannot specifically allocate resources and adjust performance indicators, thus failing to reliably meet the technical requirements of V2I services. Summary of the Invention

[0004] To address the technical problem in existing technologies where network services cannot specifically schedule resources and adjust performance indicators, thus failing to meet the needs of vehicle-road cooperative services, this disclosure provides a network service dynamic optimization method, a network service dynamic optimization system, an electronic device, and a computer-readable storage medium. These methods enable dynamic adaptive adjustment of performance indicator parameters and service resources, effectively ensuring the network performance of vehicle-road cooperative services.

[0005] Firstly, this disclosure provides a method for dynamic optimization of network services, wherein the method...

[0006] The law includes:

[0007] Standardize the network metrics for vehicle-road cooperative services to form a performance metric standard library;

[0008] Corresponding to the performance indicator standard library, a network performance indicator test library containing various service types will be constructed.

[0009] Using the established network performance indicator testing library, network performance evaluation and optimization models were trained according to event flow and different service classifications;

[0010] The UPF (User Plane Function) network element collects data according to the performance indicator standard library, processes the collected data and inputs it into the network performance evaluation and optimization model to output network service optimization strategies;

[0011] Resource scheduling is based on network service optimization strategies.

[0012] Furthermore, the resource scheduling based on network service optimization strategies includes:

[0013] Receive network service optimization strategies reported by network performance evaluation and optimization models;

[0014] Record the corresponding node optimization time point T1 and start the optimization timer;

[0015] Within the time period T1+t, network service optimization strategies for the same node are merged, where t is the preset resource scheduling and optimization period;

[0016] Obtain the resource usage of the corresponding associated nodes and compare it with the network service optimization strategy requirements;

[0017] Based on the comparison results, the UPF resources are optimized by calling the Cloud Resource Dynamic Adjustment Service API (Application Programming Interface).

[0018] Furthermore, the performance metric standard library includes:

[0019] Indicator code, indicator name, indicator definition, data source, calculation method, related indicators, and influence weight;

[0020] The influence weights are key reward factors in the network performance evaluation and optimization model, which are determined through analysis of business requirements.

[0021] Furthermore, the construction of the network performance indicator test library, which includes various service types, includes:

[0022] Collect actual network indicator data for various services in vehicle-road cooperation, form event streams, and identify them according to different service types;

[0023] The data after being labeled is standardized according to the performance indicator standard library to form a network performance indicator test library.

[0024] Furthermore, the network performance evaluation and optimization model is a DRL (Deep Reinforcement Learning) model. By designing the model's reward factor and related functions, and setting the relevant parameters in the model accordingly, the network links are segmented, corresponding dedicated links and forwarding mechanisms are established for different types of service data flows, and network resources are adaptively allocated to dedicated links.

[0025] Furthermore, the data collection according to the performance indicator standard library includes:

[0026] Add corresponding data acquisition probes to the UPF network element, and collect various data at a preset acquisition frequency according to the data source and calculation method of each indicator in the performance indicator standard library, and then label them.

[0027] Furthermore, the process of processing the collected data and inputting it into the network performance evaluation and optimization model includes:

[0028] The data is standardized according to the calculation methods in the performance index standard library, and the data of each node in the region is merged to form a global data system.

[0029] The integrated global data system is used as input to the network performance evaluation and optimization model to output network service optimization strategies in real time.

[0030] Secondly, this disclosure provides a network service dynamic optimization system, the system comprising:

[0031] The indicator standardization module is designed to standardize the indicators of the vehicle-road cooperative business network to form a performance indicator standard library.

[0032] The module is configured to correspond to the performance indicator standard library, and to build a network performance indicator test library containing various service types.

[0033] The model training module is designed to train network performance evaluation and optimization models according to event flow and different service classifications using a built network performance indicator test library.

[0034] The data acquisition module is configured to have UPF network elements collect data according to the performance index standard library, process the collected data and input it into the network performance evaluation and optimization model to output network service optimization strategies.

[0035] The resource scheduling module is configured to perform resource scheduling based on network service optimization strategies.

[0036] Thirdly, this disclosure provides an electronic device including a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the network service dynamic optimization method as described in any of the first aspects.

[0037] Fourthly, this disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the network service dynamic optimization method described in any of the first aspects above.

[0038] Beneficial effects:

[0039] The network service dynamic optimization method, network service dynamic optimization system, electronic equipment and storage medium disclosed herein, in response to the service requirements of vehicle-road cooperative services, collect and analyze indicators such as data service and session connection in UPF in real time, accurately evaluate network service quality, and output optimization schemes by combining network performance evaluation and optimization models. Through the core network signaling protocol, the system achieves dynamic adaptive adjustment of performance indicator parameters and service resources, effectively ensuring the network performance of vehicle-road cooperative services. Attached Figure Description

[0040] Figure 1 This is a flowchart illustrating a method for dynamically optimizing network services according to Embodiment 1 of this disclosure.

[0041] Figure 2 This is a diagram of a network service dynamic adaptive optimization architecture provided in Embodiment 2 of this disclosure;

[0042] Figure 3 This is a schematic diagram of cloud-based resource scheduling and management provided in Embodiment 2 of this disclosure;

[0043] Figure 4 This is a flowchart illustrating a method for dynamically optimizing network services according to Embodiment 2 of this disclosure.

[0044] Figure 5 This is a schematic diagram of a resource scheduling process based on an optimization strategy, provided in Embodiment 2 of this disclosure;

[0045] Figure 6 This is an architecture diagram of a network service dynamic optimization system provided in Embodiment 3 of this disclosure;

[0046] Figure 7 This is an architectural diagram of an electronic device provided in Embodiment 4 of this disclosure. Detailed Implementation

[0047] To enable those skilled in the art to better understand the technical solutions of this disclosure, the disclosure will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments and drawings described herein are merely for explaining the invention and are not intended to limit the invention.

[0048] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence; furthermore, in the absence of conflict, the embodiments and features in the embodiments of this disclosure can be arbitrarily combined with each other.

[0049] The terminology used in the embodiments of this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of this disclosure. The singular forms “a,” “the,” and “the” as used in the embodiments of this disclosure and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0050] In the following description, the use of suffixes such as “module,” “part,” or “unit” to denote elements is solely for the purpose of illustrative purposes and has no specific meaning in itself. Therefore, “module,” “part,” or “unit” may be used interchangeably.

[0051] 5G-V2X typically employs a solution that integrates UPF (User-Defined Flowmeter) with MEC (Multi-access Edge Computing). In actual operation, the UPF processes traffic for vehicle-road cooperative services according to pre-set rules and allocated resources. However, it cannot dynamically and adaptively adjust based on changes in service quality and demand during the service process. When various business processes are large, this will affect the service quality of vehicle-road cooperative services, and in some cases, it will lead to problems such as high latency and traffic congestion, causing serious security risks.

[0052] The following detailed embodiments illustrate the technical solutions of this disclosure and how they solve the aforementioned technical problems in the prior art. It is understood that in the embodiments of this application, the executing entity may perform some or all of the steps in the embodiments of this application. These steps or operations are merely examples, and the embodiments of this application may also perform other operations or variations thereof. Furthermore, the steps may be executed in different orders as presented in the embodiments of this application, and it is not necessary to execute all the operations in the embodiments of this application. Moreover, the following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0053] Figure 1 This is a flowchart illustrating a method for dynamically optimizing network services according to Embodiment 1 of this disclosure, as shown below. Figure 1 As shown, the method includes:

[0054] Step S101: Standardize the network indicators of vehicle-road cooperative services to form a performance indicator standard library;

[0055] By sorting out the key network performance indicators in vehicle-road cooperative services, defining the data sources, calculation methods, and influence weights of each indicator, a performance indicator standard library is formed.

[0056] The performance metrics include network latency, packet loss rate, uplink speed, downlink speed, connection density, and throughput. A unified standard is set for each metric to facilitate the subsequent training of reinforcement learning models (i.e., network performance evaluation and optimization models) and the acquisition of optimal optimization strategies through the models.

[0057] Furthermore, the performance metric standard library includes:

[0058] Indicator code, indicator name, indicator definition, data source, calculation method, related indicators, and influence weight;

[0059] The influence weights are key reward factors in the network performance evaluation and optimization model, which are determined through analysis of business requirements.

[0060] Influence weight is a key reward factor in network performance evaluation and optimization models, playing an important role in model establishment and subsequent self-learning. This parameter is determined by analyzing business requirements, and there are differences for different data requirements in vehicle-road cooperative services.

[0061] The performance indicators are coded and their sources and calculation methods are determined to form a performance indicator standard library, which serves as a unified standard for subsequent performance indicator collection and processing. The definitions of each performance indicator are shown in Table 1 below.

[0062] Table 1: Definitions of Performance Indicators in the Performance Indicator Standard Library

[0063]

[0064] Step S102: Construct a network performance indicator test library that includes various service types, corresponding to the performance indicator standard library;

[0065] The various business types include roadside perception data broadcasting, vehicle perception data reporting, fusion data release, traffic control information release, and other audio-visual entertainment services in vehicle-road cooperative systems. By collecting and labeling data for each business, a training dataset for the model is formed to train the model.

[0066] Furthermore, the construction of the network performance indicator test library, which includes various service types, includes:

[0067] Collect actual network indicator data for various services in vehicle-road cooperation, form event streams, and identify them according to different service types;

[0068] The data after being labeled is standardized according to the performance indicator standard library to form a network performance indicator test library.

[0069] Based on standardized output metrics, actual network metric data is collected for various services to form event streams. These streams are then labeled according to the different service types, facilitating the simulation of network service processes based on the importance of different service types and supporting the training of reinforcement learning models.

[0070] Step S103: Using the established network performance indicator test library, train the network performance evaluation and optimization model according to event flow and different service classifications;

[0071] After the model is designed, the network performance index testing library is used to train, test and validate the model according to event flow and different service categories. During the training process, the function parameters are optimized to improve the model's evaluation capability.

[0072] Furthermore, the network performance evaluation and optimization model is a deep reinforcement learning (DRL) model. By designing the model's reward factor and related functions, and setting the relevant parameters in the model accordingly, the network links are segmented, corresponding dedicated links and forwarding mechanisms are established for different types of service data flows, and network resources are adaptively allocated to dedicated links.

[0073] Based on the definitions and influence weights of various indicators in the standardization process, a reward factor and related functions for the network performance reinforcement learning model DRL-VELink are designed. Based on these, relevant parameters in the model are set to segment links in the network, establishing dedicated links and forwarding mechanisms for different types of service data flows, and adaptively allocating network resources for these dedicated links. This achieves intelligent scheduling of bandwidth, computing, and other resources, reducing competition between different types of flows. DRL-VELink introduces pruning double-Q learning on top of the DDPG (deep deterministic policy gradient) algorithm to address the problem of overestimation of value in the algorithm, optimizing the scheduling policy obtained during training. It also introduces a priority experience replay buffer mechanism and a Noisy Net mechanism to increase the algorithm's convergence speed and overcome the problem of insufficient exploration ability during training.

[0074] Step S104: The UPF network element collects data according to the performance index standard library, processes the collected data and inputs it into the network performance evaluation and optimization model to output network service optimization strategies;

[0075] The model's testing and evaluation provide decision data for the UPF dynamic optimization strategy. Ultimately, the model will output specific optimization solutions that meet the requirements of subsequent dynamic resource scheduling and support integration with cloud-based resource scheduling and management platforms.

[0076] Based on the business process relationships, UPF is a key node for data service forwarding in vehicle-road cooperation, and has a significant impact on reliability, latency, and other aspects. Therefore, dynamic configuration management of UPF resources can effectively utilize resources to serve important services in vehicle-road cooperation.

[0077] Furthermore, the data collection according to the performance indicator standard library includes:

[0078] Add corresponding data acquisition probes to the UPF network element, and collect various data at a preset acquisition frequency according to the data source and calculation method of each indicator in the performance indicator standard library, and then label them.

[0079] Data is collected according to the data sources and calculation methods of each indicator in the performance indicator standard library. Corresponding data collection probes are added to the UPF network element to collect and label data throughput, QoS (Quality of Service), etc. of different services. Data is collected at a frequency of 10Hz and reported to the performance evaluation model service so that the model can evaluate and optimize strategies based on real-time network performance indicator data.

[0080] Furthermore, the process of processing the collected data and inputting it into the network performance evaluation and optimization model includes:

[0081] The data is standardized according to the calculation methods in the performance index standard library, and the data of each node in the region is merged to form a global data system.

[0082] The integrated global data system is used as input to the network performance evaluation and optimization model to output network service optimization strategies in real time.

[0083] Based on real-time data reported by network performance indicator collection probes, the data is standardized according to the calculation methods in the indicator standards. Then, the data from each node in the region is fused to form a global data system. The fused data is provided as input to the network performance evaluation and optimization model service, which outputs optimization strategies in real time and inputs the relevant strategies into the cloud resource scheduling and management platform via API. In the model service, the initial model is the model trained in step S103. During actual online operation, the model parameters are continuously optimized through self-learning. During the self-learning process, the model continuously evaluates and sets reward values ​​based on the actual service quality and the collected network performance indicator values. This feedback information is input into the evaluation scheme of the DRL-VELink model to achieve dynamic feedback on the effect of the current optimization strategy. If the network performance is better than before optimization, a reward is given and the original strategy is enhanced. If the performance deteriorates, a penalty is imposed and the optimization strategy is readjusted. In the dynamic process, the accuracy and reliability of the optimization strategy are continuously improved, and relevant experience is stored in the cache in a timely manner to improve the convergence speed.

[0084] Step S105: Perform resource scheduling based on network service optimization strategies.

[0085] Based on the optimization strategy, specific resource scheduling is carried out to provide targeted data services for vehicle-road cooperative services. The output cycle of the optimization strategy in the model service is set, which is the cycle of resource scheduling and optimization, to avoid continuous fluctuations in service stability caused by excessively rapid resource start-up and shutdown and data migration.

[0086] Furthermore, the resource scheduling based on network service optimization strategies includes:

[0087] Receive network service optimization strategies reported by network performance evaluation and optimization models;

[0088] Record the corresponding node optimization time point T1 and start the optimization timer;

[0089] Within the time period T1+t, network service optimization strategies for the same node are merged, where t is the preset resource scheduling and optimization period;

[0090] Obtain the resource usage of the corresponding associated nodes and compare it with the network service optimization strategy requirements;

[0091] Based on the comparison results, the cloud resource dynamic adjustment service API is invoked to optimize the UPF resources.

[0092] The output period t can be set to 1 minute, 2 minutes, etc., depending on the actual situation. The network performance evaluation and optimization model inputs the relevant strategies into the cloud resource scheduling and management platform via API. Within one period, the network service optimization strategies are merged to optimize UPF resources and ensure the latency and quality of service requirements of vehicle-road cooperative services under the 5G-V2X communication standard.

[0093] This disclosure addresses the service requirements of vehicle-road cooperative services by real-time collection and analysis of metrics such as data service and session connection in the UPF (User-Defined Network Function), accurately assessing network service quality. Combining reinforcement learning algorithms, it outputs optimization schemes and, through the core network signaling protocol, achieves dynamic adaptive adjustment of performance parameters and service resources, effectively ensuring the network performance of vehicle-road cooperative services.

[0094] Example 2:

[0095] like Figures 2 to 4 As shown in Embodiment 2 of this disclosure, a method for dynamic optimization of network services is proposed for vehicle-road cooperative services. Figure 2 The network service dynamically adaptive optimization architecture diagram shows that the UPF connects to connected vehicles via RRU (Remote Radio Unit) / BBU (Building Baseband Unit). Data acquisition probes within the UPF collect and tag data on different service throughput, QoS, etc., at a frequency of 10Hz and report the data to the performance evaluation model service. The performance evaluation model service outputs optimization strategies in real time and inputs these strategies into the cloud resource scheduling and management platform via API. Based on the optimization strategies, the platform calls the cloud resource dynamic adjustment service API to optimize UPF resources, such as... Figure 3 As shown, the UPF calls hardware resources through VIM (Virtualized Infrastructure Manager) and PIM (Physical Infrastructure Manager), enabling the UPF to dynamically and adaptively adjust the allocated resources according to changes in service quality and demand during the business process, and to perform traffic diversion processing for vehicle-road cooperative services.

[0096] The steps of the dynamic optimization method for network services are as follows: Figure 4 As shown, it includes: S10-S50.

[0097] Step S10: Standardization of Vehicle-Road Cooperative Service Network Indicators

[0098] Key network performance indicators in vehicle-road cooperative services were identified, and the data sources, calculation methods, and influence weights of each indicator were defined to form a performance indicator standard library, as shown in Table 2 below:

[0099]

[0100] Among them, the key reward factor in the weighted reinforcement learning model plays an important role in the establishment of the model and subsequent self-learning. This parameter is determined by analyzing business needs, and there are also differences for different data needs in vehicle-road cooperative business.

[0101] Step S20: Network performance evaluation and model training optimization

[0102] 1) Construction of a network performance indicator test library

[0103] Based on standardized output indicators, actual network indicator data are collected for services such as roadside perception data broadcasting, vehicle perception data reporting, fusion data release, traffic control information release, and other audio-visual entertainment in vehicle-road cooperation. This forms an event flow, which is then labeled according to the different service types. This facilitates the simulation of network service business processes based on the importance of different service types, and supports the training of reinforcement learning models.

[0104] 2) Training the network performance evaluation model

[0105] Based on the definitions and influence weights of various indicators in the standardization process, this paper designs the reward factor and related functions for the network performance reinforcement learning model DRL-VELink. Based on this, relevant parameters are set in the model to segment links in the network, establish dedicated links and forwarding mechanisms for different types of service data flows, and adaptively allocate network resources for dedicated links. This achieves intelligent scheduling of bandwidth, computing, and other resources, reducing competition between different types of flows. DRL-VELink introduces pruning double-Q learning on top of the DDPG algorithm to address the problem of overestimation of value in the algorithm, optimizes the scheduling strategy obtained during training, and introduces a priority experience replay buffer mechanism and a Noisy Net mechanism to increase the convergence speed of the algorithm and overcome the problem of insufficient exploration ability during training.

[0106] After the model is designed, the network performance index test library built in the previous step is used to train, test and validate the model according to the event flow and different service categories. During the training process, the function parameters are optimized to improve the model's evaluation capability.

[0107] 3) Output of UPF dynamic optimization strategy

[0108] The purpose of the evaluation is to provide decision data for the UPF dynamic optimization strategy. The final model will output a specific optimization scheme that meets the requirements of subsequent dynamic resource scheduling and supports the connection with the cloud resource scheduling and management platform.

[0109] Step S30: UPF Network Element Service Indicator Collection

[0110] Based on the business process relationships, the UPF (User-Defined Flow) is a key node for data service forwarding in vehicle-road cooperation, significantly impacting reliability and latency. Therefore, dynamic configuration management of UPF resources can effectively utilize resources to serve important services in vehicle-road cooperation. In this step, data is collected according to the data sources and calculation methods of each indicator in the network performance index standardization. Corresponding data collection probes are added to the UPF network element to collect and label data throughput, QoS, and other metrics for different services. Data is collected at a frequency of 10Hz and reported to the performance evaluation model service so that the model can evaluate and optimize strategies based on real-time network performance index data.

[0111] Step S40: Real-time evaluation and reporting of UPF performance

[0112] Based on real-time data reported by network performance indicator collection probes, the data is standardized according to the calculation methods in the indicator standards. Then, the data from each node in the region is fused to form a global data system. The fused data is provided as input to the network performance evaluation and optimization model service, which outputs optimization strategies in real time and inputs the relevant strategies into the cloud resource scheduling and management platform via API. In the model service, the initial model is the model trained in step S20. During actual online operation, the model parameters are continuously optimized through self-learning. During the self-learning process, the model continuously evaluates and sets reward values ​​based on the actual service quality and the collected network performance indicator values. This feedback information is input into the evaluation scheme of the DRL-VELink model to achieve dynamic feedback on the effect of the current optimization strategy. If the network performance is better than before optimization, a reward is given and the original strategy is enhanced. If the performance deteriorates, a penalty is imposed and the optimization strategy is readjusted. In the dynamic process, the accuracy and reliability of the optimization strategy are continuously improved, and relevant experience is stored in the cache in a timely manner to improve the convergence speed.

[0113] The output cycle of the optimization strategy in the model service is 1 minute, that is, the cycle of resource scheduling and optimization is 1 minute, so as to avoid the continuous fluctuations in service stability caused by resource start-up and shutdown and data migration due to excessive speed.

[0114] Step S50: Dynamic Adaptive Scheduling

[0115] Based on optimization strategies, specific resource scheduling is performed to provide targeted data services for vehicle-road cooperative operations. The specific process is as follows: Figure 5As shown, it includes:

[0116] S1: Received optimization strategy A reported by the model service;

[0117] S2: Record the corresponding node optimization time point T1 and start the optimization timer;

[0118] S3: Merge optimization strategies for the same node within the T1+1min time period, repeat this step until the 1min period ends;

[0119] S4: Obtain the resource usage of associated nodes and compare it with the policy requirements;

[0120] S5: Optimize UPF resources by calling the cloud resource dynamic adjustment service API.

[0121] This disclosure proposes a dynamic optimization scheme for network services in a vehicle-road cooperative scenario. It designs a complete adaptive network service performance optimization process and combines real-time monitoring and evaluation of UPF network service quality indicators to effectively ensure the latency and service quality requirements of vehicle-road cooperative services under the 5G-V2X communication standard.

[0122] Example 3

[0123] Figure 6 This is an architecture diagram of a network service dynamic optimization system provided in Embodiment 3 of this disclosure, as follows: Figure 6 As shown, the system includes:

[0124] The indicator standardization module 11 is configured to standardize the indicators of the vehicle-road cooperative business network to form a performance indicator standard library.

[0125] Module 12 is configured to correspond to the performance indicator standard library, and to build a network performance indicator test library containing various service types.

[0126] Model training module 13 is configured to train the network performance evaluation and optimization model according to the event flow and different service classifications using the built network performance index test library.

[0127] The data acquisition module 14 is configured to have the UPF network element collect data according to the performance index standard library, process the collected data and input it into the network performance evaluation and optimization model to output network service optimization strategies.

[0128] Resource scheduling module 15 is configured to perform resource scheduling based on network service optimization strategies.

[0129] Furthermore, the resource scheduling module 15 is specifically configured as follows:

[0130] Receive network service optimization strategies reported by network performance evaluation and optimization models;

[0131] Record the corresponding node optimization time point T1 and start the optimization timer;

[0132] Within the time period T1+t, network service optimization strategies for the same node are merged, where t is the preset resource scheduling and optimization period;

[0133] Obtain the resource usage of the corresponding associated nodes and compare it with the network service optimization strategy requirements;

[0134] Based on the comparison results, the cloud resource dynamic adjustment service API is invoked to optimize the UPF resources.

[0135] Furthermore, the performance metric standard library includes:

[0136] Indicator code, indicator name, indicator definition, data source, calculation method, related indicators, and influence weight;

[0137] The influence weights are key reward factors in the network performance evaluation and optimization model, which are determined through analysis of business requirements.

[0138] Furthermore, the construction module 12 is specifically configured as follows:

[0139] Collect actual network indicator data for various services in vehicle-road cooperation, form event streams, and identify them according to different service types;

[0140] The data after being labeled is standardized according to the performance indicator standard library to form a network performance indicator test library.

[0141] Furthermore,

[0142] The network performance evaluation and optimization model is a deep reinforcement learning (DRL) model. By designing the model's reward factor and related functions, and setting the relevant parameters in the model accordingly, the model can segment links in the network, establish corresponding dedicated links and forwarding mechanisms for different types of service data flows, and adaptively allocate network resources for dedicated links.

[0143] Furthermore, the data acquisition module 14 is specifically configured as follows:

[0144] Add corresponding data acquisition probes to the UPF network element, and collect various data at a preset acquisition frequency according to the data source and calculation method of each indicator in the performance indicator standard library, and then label them.

[0145] Furthermore, the data acquisition module 14 is specifically configured as follows:

[0146] The data is standardized according to the calculation methods in the performance index standard library, and the data of each node in the region is merged to form a global data system.

[0147] The integrated global data system is used as input to the network performance evaluation and optimization model to output network service optimization strategies in real time.

[0148] The network service dynamic optimization system of this disclosure is used to implement the network service dynamic optimization method in method embodiment one and method embodiment two, so the description is relatively simple. For details, please refer to the relevant descriptions in the previous method embodiment one and method embodiment two, which will not be repeated here.

[0149] In addition, such as Figure 7 As shown, Embodiment 4 of this disclosure also provides an electronic device, including a memory 100 and a processor 200. The memory 100 stores a computer program. When the processor 200 runs the computer program stored in the memory 100, the processor 200 executes the various possible methods described above.

[0150] The memory 100 is connected to the processor 200. The memory 100 can be a flash memory, a read-only memory, or another type of memory. The processor 200 can be a central processing unit or a microcontroller.

[0151] Furthermore, embodiments of this disclosure also provide a computer-readable storage medium storing a computer program, which is executed by a processor using the various possible methods described above.

[0152] The computer-readable storage medium includes volatile or non-volatile, removable or non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, computer program modules or other data). Computer-readable storage media includes, but is not limited to, RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory or other memory technologies, CD-ROM (Compact Disc Read-Only Memory), Digital Video Disc (DVD) or other optical disc storage, cartridges, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer.

[0153] It is understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of this disclosure, and this disclosure is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and substance of this disclosure, and these modifications and improvements are also considered to be within the scope of protection of this disclosure.

Claims

1. A method for dynamic optimization of network services, characterized in that, The method includes: Standardize the network metrics for vehicle-road cooperative services to form a performance metric standard library; Corresponding to the performance indicator standard library, a network performance indicator test library containing various service types will be constructed. Using the established network performance indicator testing library, network performance evaluation and optimization models were trained according to event flow and different service classifications; User plane function UPF network elements collect data according to the performance indicator standard library, process the collected data and input it into the network performance evaluation and optimization model to output network service optimization strategies; Resource scheduling is based on network service optimization strategies. The construction of the network performance indicator test library, which includes various service types, includes: Collect actual network indicator data for various services in vehicle-road cooperation, form event streams, and identify them according to different service types; The data after being labeled is standardized according to the performance indicator standard library to form a network performance indicator test library.

2. The method according to claim 1, characterized in that, The resource scheduling based on network service optimization strategies includes: Receive network service optimization strategies reported by network performance evaluation and optimization models; Record the corresponding node optimization time point T1 and start the optimization timer; Within the time period T1+t, network service optimization strategies for the same node are merged, where t is the preset resource scheduling and optimization period; Obtain the resource usage of the corresponding associated nodes and compare it with the network service optimization strategy requirements; Based on the comparison results, the cloud resource dynamic adjustment service application programming interface API is invoked to optimize the UPF resources.

3. The method according to claim 1, characterized in that, The performance metric standard library includes: Indicator code, indicator name, indicator definition, data source, calculation method, related indicators, and influence weight; The influence weights are key reward factors in the network performance evaluation and optimization model, which are determined through analysis of business requirements.

4. The method according to claim 1, characterized in that, The network performance evaluation and optimization model is a deep reinforcement learning (DRL) model. By designing the model's reward factor and related functions, and setting the relevant parameters in the model accordingly, the model can segment links in the network, establish corresponding dedicated links and forwarding mechanisms for different types of service data flows, and adaptively allocate network resources for dedicated links.

5. The method according to claim 3, characterized in that, The data collection according to the performance index standard library includes: Add corresponding data acquisition probes to the UPF network element, and collect various data at a preset acquisition frequency according to the data source and calculation method of each indicator in the performance indicator standard library, and then label them.

6. The method according to claim 3, characterized in that, The process of inputting the collected data into the network performance evaluation and optimization model includes: The data is standardized according to the calculation methods in the performance index standard library, and the data of each node in the region is merged to form a global data system. The integrated global data system is used as input to the network performance evaluation and optimization model to output network service optimization strategies in real time.

7. A network service dynamic optimization system, characterized in that, The system includes: The indicator standardization module is designed to standardize the indicators of the vehicle-road cooperative business network to form a performance indicator standard library. The module is configured to correspond to the performance indicator standard library, and to build a network performance indicator test library containing various service types. The model training module is designed to train network performance evaluation and optimization models according to event flow and different service classifications using a built network performance indicator test library. The data acquisition module is configured to have UPF network elements collect data according to the performance index standard library, process the collected data and input it into the network performance evaluation and optimization model to output network service optimization strategies. The resource scheduling module is configured to perform resource scheduling based on network service optimization strategies. The construction module is specifically configured to collect actual network indicator data for various services in vehicle-road cooperation, form event streams, and identify them according to different service types; and to standardize the identified data according to the performance indicator standard library to form a network performance indicator test library.

8. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the network service dynamic optimization method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the network service dynamic optimization method as described in any one of claims 1-6.

Images