Network intention processing method, computer equipment and readable medium

Abstract

The invention provides a network intention processing method. The method comprises the following steps: determining a weight vector of each influence factor of a to-be-opened service according to network topology information, influence factors, attribute information of the to-be-opened service and an AIGC model; according to the network topology information, the attribute information of the to-be-opened service, the weight vector of each influence factor of the to-be-opened service and the model, determining a whole network service opening configuration total score and path configuration information of the to-be-opened service; the whole network service opening configuration total score is a return of a first track generated after a first action used for intention arrangement is executed, and is used for performing first fine adjustment processing on the model; the problems that the network service opening time is long, the opening operation technology complexity is high, the intention and expectation of opening and using the service by the user cannot be accurately understood, and the optimal scheme of concurrent multi-service resource arrangement and configuration cannot be obtained based on the opening intention of the user can be solved. The invention further provides computer equipment and a readable medium.

Description

Technical Field

[0001] The present disclosure relates to the technical field of self-intelligent optical networks, and particularly relates to a network intention processing method, a computer device, and a computer-readable medium. Background Art

[0002] The global communication industry is moving from the interconnected era and the cloud era towards the intelligent era. Facing challenges such as market competition, digital transformation across industries, and increased network complexity in network planning, optimization, operation and maintenance, reliability, agility, and reduction of OPEX (Operating Expense), the development of optical networks towards intelligence has become inevitable. The rise and development of AI (Artificial Intelligence) software and hardware technologies provide a solid foundation and main means for the realization of the intelligent goal of optical networks.

[0003] Especially since the release of ChatGPT (Chat Generative Pre-trained Transformer), a large model of Artificial Intelligence Generated Content (AIGC), in November 2022, the field of artificial intelligence applications has witnessed a major technological revolution. As an AIGC model, ChatGPT can continuously improve its context semantic understanding and interaction capabilities through continuous training of massive amounts of data, showing infinite potential in numerous application scenarios. The resulting research and application boom has also promoted the development of the entire artificial intelligence industry.

[0004] How to leverage the technical advantages of AIGC generative large model technology and use AIGC large model technology to enhance the core intelligent capabilities of AN OTN (Autonomous Optical Transport Network), and achieve the intelligent goals of intelligent management, ultimate experience, and flexible openness of optical networks, so as to meet the customer needs of improving user experience and reducing operation costs, has become the focus issue of self-intelligence that the industry currently concerns and needs to solve. Currently, the key areas of the industry's focus on the application of AIGC large model in the field of network communication mainly include: intelligent fault diagnosis, network planning & commissioning & optimization solution design and implementation, etc. In particular, how to apply AIGC large model technology to enhance the intelligent capabilities of the intention-based commissioning of OTN network services, the main application scenario of OTN self-intelligent optical networks, has become a hot topic of emerging technologies.

[0005] With the development of the digital economy, the demand of government and enterprise customers for opening government and enterprise services through the OTN network is increasing day by day. Although the OTN network fully has the ability to carry all government and enterprise services, there are the following bottlenecks to be improved in service opening and intelligent operation: there are many links in the government and enterprise service opening process and a large amount of input data, and professional personnel familiar with OTN technology are required to complete the service opening, so government and enterprise customers cannot open services according to their own needs; the service opening time is long (in days), which is different from the rapid opening expected by government and enterprise customers; the network intelligence level is not high, and the time for analyzing and evaluating network resources is long, which affects the service opening time.

[0006] To solve the above bottlenecks, the OTN network control system needs to gradually introduce intelligent functions and transform into an intent-driven intelligent and simple optical network architecture - IBON (Intent-Based Optical Network) that relies on network digital twin simulation and analysis technology and integrates SDN (Software Defined Network) and AI technology applications. On this basis, promote the OTN network service to move towards the construction of a self-intelligent ecological system across the complete life cycle. IBON realizes the intent-based parsing of the expected and target of user service operation through the perception of the OTN network service, ensures that the user intent input is intelligent and accurately understood, and completes the orchestration, optimization, simulation verification, and activation and distribution of IBON services through intelligent technology, realizing the accurate matching of network resource allocation with service operation intent and application scenarios. Summary of the Invention

[0007] The present disclosure aims at the above deficiencies in the prior art and provides a network intent processing method, a computer device, and a computer-readable medium.

[0008] In a first aspect, an embodiment of the present disclosure provides a network intent processing method, including:

[0009] Determine the weight vector of each of the influencing factors of the service to be opened according to the network topology information, the influencing factors, the attribute information of the service to be opened, and a pre-trained model; wherein, the model is an artificial intelligence generated content AIGC model;

[0010] Determine the overall score of the whole-network service activation configuration and the path configuration information of the service to be activated according to the network topology information, the attribute information of the service to be activated, the weight vector of each influencing factor of the service to be activated, and the model; wherein, the overall score of the whole-network service activation configuration is the return of the first trajectory, which is calculated according to a pre-generated prompt template and is used for the first fine-tuning process of the model to obtain the path configuration information of the service to be activated with the highest overall score of the whole-network service activation configuration; the first trajectory is generated after performing the first action for intent orchestration and is used to implement the intent orchestration process of service activation.

[0011] In another aspect, an embodiment of the present disclosure also provides a computer device, including:

[0012] One or more processors;

[0013] A storage device having one or more programs stored thereon;

[0014] When the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the intent network service activation method as described above.

[0015] In another aspect, an embodiment of the present disclosure also provides a computer-readable medium having a computer program stored thereon, wherein the program, when executed, implements the intent network service activation method as described above.

[0016] The network intent processing method provided by the embodiments of the present disclosure determines the weight vector of each influencing factor of the service to be activated according to the network topology information, the influencing factors, the attribute information of the service to be activated, and the pre-trained AIGC model; and determines the overall score of the whole-network service activation configuration and the path configuration information of the service to be activated according to the network topology information, the attribute information of the service to be activated, the weight vector of each influencing factor of the service to be activated, and the model; the overall score of the whole-network service activation configuration is the return of the first trajectory generated after performing the first action for intent orchestration, which is calculated according to a pre-generated prompt template and is used for the first fine-tuning process of the model to obtain the path configuration information of the service to be activated with the highest overall score of the whole-network service activation configuration; the embodiments of the present disclosure can solve problems such as long network service activation time, high technical complexity of activation operations, inability to accurately understand the intent and expectations of users to activate and use services, and inability to obtain the optimal solution for concurrent multi-service resource orchestration and configuration based on the user's activation intent by leveraging the technical advantages of the AIGC large model. At the same time, with the large-scale data analysis ability possessed by the large model, the optimal solution for batch multi-service intent activation resource orchestration and configuration can be obtained efficiently and accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1Schematic diagram of the intent network service activation process provided by an embodiment of the present disclosure;

[0018] Figure 2 Schematic diagram of the principle of intent network service activation provided by an embodiment of the present disclosure;

[0019] Figure 3 Schematic diagram of the design principle of the Prompt template provided by an embodiment of the present disclosure;

[0020] Figure 4 Schematic diagram of the process for calculating the total score of the entire network service activation configuration in the case of activating a single service scenario provided by an embodiment of the present disclosure;

[0021] Figure 5 Schematic diagram of the network topology in the case of activating a single service scenario and multiple service concurrent scenarios provided by a specific example of the present disclosure;

[0022] Figure 6 Schematic diagram of the process for calculating the total score of the entire network service activation configuration in the case of activating multiple concurrent service scenarios provided by an embodiment of the present disclosure;

[0023] Figure 7 Schematic diagram of the principle of performing the first fine-tuning process on the model provided by an embodiment of the present disclosure;

[0024] Figure 8 Schematic diagram of the first fine-tuning process of the AIGC model by the Policy-Gradient algorithm provided by an embodiment of the present disclosure;

[0025] Figure 9 Schematic diagram of the intent network service activation process provided by a specific example of the present disclosure;

[0026] Figure 10 Schematic diagram of the structure of the computer device provided by an embodiment of the present disclosure. Detailed implementation manners

[0027] In the following, example embodiments will be described more fully with reference to the accompanying drawings. However, the example embodiments can be embodied in different forms and should not be construed as limited to the embodiments set forth herein. On the contrary, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0028] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0029] The terms used herein are for describing specific embodiments only and are not intended to limit the present disclosure. As used herein, the singular forms "a" and "the" are also intended to include the plural forms unless the context clearly indicates otherwise. It will also be understood that when the terms "comprises" and / or "consists of" are used in this specification, it specifies the presence of the stated features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0030] The embodiments described herein may be described with reference to plan views and / or cross-sectional views by means of ideal schematic diagrams of the present disclosure. Therefore, the example illustrations may be modified according to manufacturing techniques and / or tolerances. Accordingly, the embodiments are not limited to the embodiments shown in the drawings, but include modifications of configurations formed based on manufacturing processes. Therefore, the regions illustrated in the drawings have schematic properties, and the shapes of the regions shown in the drawings illustrate the specific shapes of the regions of the elements, but are not intended to be restrictive.

[0031] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art. It will also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and the present disclosure, and will not be interpreted as having an idealized or overly formal meaning unless expressly so defined herein.

[0032] The basic process that IBON intends to activate includes four steps:

[0033] (1) Intent input, that is, the description of the service activation intent: the parameter input is intelligent and simple, reducing the technical complexity, and replacing it with the description language expected by government and enterprise customers; establish intent complement mechanisms such as data mining, collection, and KG (Knowledge Graph) reasoning.

[0034] (2) Intent parsing: According to the intent input, parse the SLA (Service Level Agreement) KPI (Key Performance Indicator) optimization strategy for satisfying the activated service, including goals, priorities, and weights.

[0035] (3) Intent orchestration: According to the SLA KPI optimization strategy, adopt multi-factor routing technology to orchestrate and simulate optical and electrical cross-layer resources and performance indicator configurations to meet the intent requirements of user service usage.

[0036] (4) Issuance and verification: Implement the automated activation issuance and verification of intent-based service configurations.

[0037] Intent parsing and intent orchestration are the two most important key steps in the basic process of IBON intent activation. The embodiments of the present disclosure propose a service activation solution implemented using the AIGC large model technology in the self-intelligent OTN (AN OTN) architecture system around these two steps. It should be noted that the embodiments of the present disclosure are described by taking the OTN network as an example, but the solution can be applied to fields such as PTN (Packet Transport Network), POTN (Packet Optical Transport Network), SPN (Secret Private Network), and IP network.

[0038] Figure 1 It is a schematic diagram of the intent network service activation process provided by the embodiments of the present disclosure. Figure 2 It is a schematic diagram of the principle of intent network service activation provided by the embodiments of the present disclosure. As Figure 2 shown, the embodiments of the present disclosure focus on two key steps in the process of intent network service activation in the order of execution, namely intent parsing and intent orchestration, and use the same AIGC model instance - Transformer A to complete the function implementation of these two steps respectively, that is, perform intent parsing and intent orchestration function reasoning, so as to obtain the final OTN network service activation resource orchestration plan (i.e., the routing plan).

[0039] Before implementing these two steps, it is necessary to perform Prompt prompt guidance and Fine-tune fine-tuning processing on Transformer A. As Figure 2 shown, the two steps of implementing intent parsing and intent orchestration functions by Transformer A are respectively designed into two independent DRL (Deep Reinforcement Learning) trajectory (Trajectory) instances, namely DRL Trajectory1 for implementing intent parsing function and DRL Trajectory2 for implementing intent orchestration function. And both DRL Trajectory instances are composed of two states and a single-step action. In DRL Trajectory1, Transformer A implements the reasoning process of intent parsing, that is, executes action a0 from state S0 to state S1, and at this time, the model parameters of Transformer A are the action policy model parameters of this executed action a0. In DRL Trajectory2, Transformer A implements the reasoning process of intent orchestration, that is, executes action a1 from state S1 to state S2, and at this time, the model parameters of Transformer A are the action policy model parameters of this executed action a1.

[0040] Combined withFigure 1 and Figure 2 As shown, the method for opening an intent network service includes the following steps:

[0041] Step S11: Determine the weight vector of each influencing factor of the service to be opened according to the network topology information, influencing factors, attribute information of the service to be opened, and a pre-trained model; where the model is an AIGC model.

[0042] In DRL Trajectory1, the input parameters of the Transformer A model for executing action a0 include: network topology information t0, multiple influencing factors t1 related to the service, and attribute information x of one or more services to be opened i . The Transformer A model is pre-trained, and its output parameters include: the weight vector y of each influencing factor of the service to be opened i , i ∈ [1, svr num , where svr num is the number of services to be opened. That is to say, each service to be opened corresponds to a weight vector y i , and the weight vector y of one service to be opened i includes the weights of each influencing factor of this service.

[0043] Action a0 is the second action for intent parsing. After executing action a0, a second trajectory is generated, and the second trajectory can realize the intent parsing process of service opening.

[0044] In some embodiments, in DRL Trajectory1, the output parameters of the Transformer A model may further include the overall customer satisfaction score y0 of the service to be opened. It should be noted that the overall customer satisfaction score y0 of the service to be opened is the customer satisfaction of all services to be opened, which is the reward of the second trajectory generated by executing action a0 and is used for the second fine-tuning process of the Transformer A model to obtain the weight vector y of each influencing factor of the service to be opened i consistent with the overall customer satisfaction score y0 of the service to be opened.

[0045] By continuously improving the overall customer satisfaction score y0 of the service to be opened and using the backpropagation algorithm, the training and tuning of the action a0 policy model in DRL Trajectory1 can be completed, thereby realizing the Fine-tune fine-tuning process of the Transformer A model in the intent parsing process.

[0046] Step S12: Determine the overall score of the whole-network service activation configuration and the path configuration information of the service to be activated according to the network topology information, the attribute information of the service to be activated, the weight vector of each influencing factor of the service to be activated, and the model. The overall score of the whole-network service activation configuration is the return of the first trajectory, which is calculated according to the pre-generated prompt template and is used for the first fine-tuning process of the model to obtain the path configuration information of the service to be activated with the highest overall score of the whole-network service activation configuration. The first trajectory is generated after executing the first action for intent orchestration and is used to implement the intent orchestration process for service activation.

[0047] In DRL Trajectory2, the input parameters of Transformer A model for executing action a1 include: network topology information t0, the attribute information x of one or more services to be activated i and the weight vector y of each influencing factor of the service to be activated i . The output parameters obtained through the inference of Transformer A model at least include: the overall score x′0 of the whole-network service activation configuration and the path configuration information x′ of the service to be activated i . It should be noted that in DRL Trajectory2, Transformer A model can also output the network topology update information t’0.

[0048] Action a1 is the first action for intent orchestration. After executing action a0, the first trajectory is generated, and the first trajectory can implement the intent orchestration process for service activation. The overall score x′0 of the whole-network service activation configuration is the return of the first trajectory generated by executing action a1. By continuously improving the overall score x′0 of the whole-network service activation configuration and using the reverse gradient propagation algorithm, the training and tuning of the policy model of action a1 in DRL Trajectory2 can be completed, so as to realize the Fine-tune fine-tuning process of Transformer A model in the intent orchestration process.

[0049] The network intent processing method provided by the embodiments of the present disclosure determines the weight vectors of the influencing factors of the service to be activated according to the network topology information, the influencing factors, the attribute information of the service to be activated, and the pre-trained AIGC model; and determines the total score of the whole-network service activation configuration and the path configuration information of the service to be activated according to the network topology information, the attribute information of the service to be activated, the weight vectors of the influencing factors of the service to be activated, and the model; the total score of the whole-network service activation configuration is the return of the first trajectory generated after performing the first action for intent orchestration, calculated according to the pre-generated prompt template, and is used for the first fine-tuning process of the model to obtain the path configuration information of the service to be activated with the highest total score of the whole-network service activation configuration; the embodiments of the present disclosure can solve problems such as long network service activation time, high technical complexity of activation operations, inability to accurately understand the intent and expectations of users to activate and use services, and inability to obtain the optimal solution for concurrent multi-service resource orchestration and configuration based on the user's activation intent by leveraging the technical advantages of the AIGC large model. At the same time, with the large-scale data analysis ability of the large model, the optimal solution for batch multi-service intent activation resource orchestration and configuration can be obtained efficiently and accurately.

[0050] The embodiments of the present disclosure propose an OTN network service activation solution implemented using AIGC generative technology, adopting a combination of natural language Chinese characters and mathematical symbols to construct and form an idea of a self-contained "self-intelligent language" belonging to the field of OTN intelligent activation and operation and maintenance. Through pre-training and Prompt template prompting, by describing the model input and output in "self-intelligent language", the semantic understanding of the model input by AIGC is realized, and the expected model output result is inferred.

[0051] The embodiments of the present disclosure give the specific definitions of the input and output parameters of the AIGC model instance - Transformer A model for realizing intent network activation in the form of "self-intelligent language".

[0052] 1. Definitions of the input and output parameters of the Transformer A model for realizing the intent parsing function.

[0053] In some embodiments, the network topology information t0 may include node information t 0_NodeLst and link information t 0_LnkLst , the node information t 0_NodeLst includes at least one of the following: node type, cross-connect capacity, occupied cross-connect capacity, number of transmission directions, and number of wavelengths in each transmission direction. The link information t 0_LnkLstInclude at least one of the following: type of link, identifier, optical power, optical attenuation, optical signal-to-noise ratio, number of wavelengths of the link; identifier, transmission rate, channel resource occupancy information, power, bit error rate, optical signal-to-noise ratio, optical channel data unit multiplexing relationship, modulation mode, baud rate, spectral efficiency, center frequency, spectral width of each wavelength.

[0054]

[0055]

[0056]

[0057] Therefore, the network topology information t0 may include, but is not limited to: the number and distribution of OTN nodes, the number and distribution of OTN links, the resource scheduling structure characteristics of each node, the resource occupancy information of each link, etc. The resource scheduling structure characteristics of the node may include, but are not limited to: whether it is optical and electrical hybrid scheduling; the cross-connection capacity and bandwidth resource occupancy information of the relevant OXC (Optical Cross-Connect) and DXC (Digital Cross Connect); the number of transmission directions, the number of wavelengths in each direction, the single-wave transmission rate, etc. The resource occupancy information of the link may include, but is not limited to: the distribution and number of OCH (optical channel layer) channels, the distribution and number of ODU (Optical Channel Data Unit) time slot channels created on each OCH channel, the OTN service distribution, number, and occupied bandwidth on each ODU time slot.

[0058] The information description of the current OTN network topology node i is defined as follows:

[0059]

[0060] t 0i_NodeType : The type of node i, the data type is enumeration type, and the current values are OXC, DXC, and OXC&DXC hybrid scheduling, a total of three types.

[0061] t 0i_OptCrssCpcty : The optical cross-connection capacity of node i, the data type is integer, the unit is Tbit / s, and for DXC type nodes, the value of this element is 0.

[0062] t 0i_OptCrssCpctyOccped : The occupied optical cross-connection capacity of node i, the data type is integer, the unit is Tbit / s, and for DXC type nodes, the value of this element is 0.

[0063] t 0i_ElctrcCrssCpcty: The electrical cross-connect capacity of node i, with a real data type and a unit of Tbit / s. For OXC type nodes, the value of this element is 0.

[0064] t 0i_ElctrcCrssCpctyOccped : The occupied electrical cross-connect capacity of node i, with a real data type and a unit of Tbit / s.

[0065] t 0i_TrnsDrctNum : The number of transmission directions of node i, with an integer data type.

[0066] t 0i_TrnsDrct1_WaveNum : The number of wavelengths on transmission direction 1 of node i, with an integer data type.

[0067] The information description of the current OTN network topology link j is defined as follows:

[0068]

[0069] Among them, wavelength 1 refers to a wavelength of link j.

[0070] In some embodiments, the influencing factor t1 includes at least one of the following: delay factor, bandwidth utilization factor, energy consumption factor, cost factor, security factor, computing power factor, hop count factor.

[0071] t1 = [t 11 , t 12 , t 13 , t 14 , t 15 , t 16 , t 17 ,..., t 1k ,..., t 1fn ;

[0072] t 11 is the delay factor, t 12 is the bandwidth utilization factor, t 13 is the energy consumption factor, t 14 is the cost factor, t 15 is the security factor, t 16 is the computing power factor, t 17 is the hop count factor.

[0073]

[0074] Indicates the number of influencing factors considered in the OTN service provisioning and optimized routing calculation in the current OTN network topology.

[0075] In some embodiments, the attribute information x of the service to be provisioned iInclude at least one of the following: source-destination node information, service level agreement level information, application scenario type information, occupied bandwidth information, routing constraint information.

[0076] Attribute information x of the service to be activated i , i ∈ [1, svr num , svr num is the number of services to be activated, and the attribute information x of the service to be activated i is defined as follows:

[0077]

[0078] Weight vector y of each influencing factor of the service to be activated i , i ∈ [1, svr num , is the weight value of each influencing factor of the i-th service to be activated, and the value is a decimal between 0 and 1.

[0079] 2. Definitions of the input and output parameters of Transformer A model for implementing the intent orchestration function.

[0080] The input parameters of Transformer A model for implementing intent orchestration (i.e., action a1) inference using a large model include the network topology information t0 and the attribute information x of the service to be activated in the input parameters of Transformer A model for implementing the intent parsing step function i , and the weight vector y of each influencing factor of the service to be activated in the output parameters of Transformer A model for implementing the intent parsing step function i . The output parameters of Transformer A model for implementing intent orchestration (i.e., action a1) inference using a large model can include network topology update information t'0, overall network service activation configuration total score x'0, and path configuration information x' of the service to be activated i .

[0081] The network topology update information t'0 is the refreshed OTN network topology information after using a large model to orchestrate and configure network resources for OTN services according to the user's intent (i.e., performing action a1), and the data structure of this parameter is the same as that of the network topology information t0.

[0082] The overall network service activation configuration total score x'0 is the evaluation score of the plan for using a large model to orchestrate and configure network resources for the entire OTN network according to the user's intent (i.e., performing action a1).

[0083] Path configuration information x' of the service to be activated i , i ∈ [1, svr num, which is the attribute information of the i-th OTN service after calculating and configuring the impact factors for it, including the configured values of each impact factor on the service path.

[0084] In some embodiments, the path configuration information x′ i includes at least one of the following: source and destination node information, service identification information, service level agreement level information, total score of activation configuration, attribute information corresponding to impact factors, and attribute information corresponding to topology information.

[0085]

[0086] In the embodiments of the present disclosure, the prompt template can wake up and guide the model. Figure 3 This is a schematic diagram of the design principle of the Prompt template provided by the embodiments of the present disclosure. As Figure 3 shown, the prompt template includes a single-service alternative path screening prompt template, a single-service optimized route selection prompt template, a multi-service concurrent alternative path screening prompt template, and a multi-service concurrent optimized route selection prompt template. The guiding ability of the single-service optimized route selection prompt template is generated based on the guiding ability of the single-service alternative path screening prompt template. The guiding ability of the multi-service concurrent alternative path screening prompt template is generated based on the guiding ability of the single-service optimized route selection prompt template. The guiding ability of the multi-service concurrent optimized route selection prompt template is generated based on the guiding ability of the multi-service concurrent alternative path screening prompt template.

[0087] The embodiments of the present disclosure propose a hierarchical Prompt prompt scheme design in the form of "self-intelligent language" to guide the Transformer A model instance to have the ability to reason and implement the OTN network intention orchestration. Considering factors such as the development of the AIGC model's capabilities in the current and future time ranges, the mathematical derivations and function definitions involved in the embodiments of the present disclosure are limited to addition and multiplication operations and do not involve other complex mathematical derivations and calculations. The prompt template has the ability to guide the Transformer A model instance to discover several alternative paths that meet the constraint conditions and the ability to select the optimal path among the alternative paths, rather than prompting and guiding the Transformer A model instance to have the path optimization algorithm ability.

[0088] As Figure 3 shown, the characteristics of the hierarchical Prompt prompt scheme design that enables the AIGC model (here the AIGC model instance is Transformer A) in the embodiments of the present disclosure to have the ability to implement the OTN network intention orchestration are described as follows:

[0089] (1) From top to bottom, design a multi-level inverted pyramid-style Prompt prompt architecture mechanism that goes from shallow to deep, from easy to difficult, and from simple to complex.

[0090] (2) The Prompt template of the upper layer can wake up the large model to learn the relatively easy knowledge and reasoning ability related to a specific scenario field.

[0091] (3) The Prompt template of the lower layer defaults and inherits the knowledge and reasoning ability obtained by waking up the large model by the Prompt template of the upper layer, and serves as the necessary knowledge, providing the premise and foundation for prompting and waking up the large model to learn the knowledge of this layer.

[0092] (4) Progressively layer by layer, from shallow to deep. When the wake-up process of all levels of Prompt templates is completed, the large model will possess complex knowledge and reasoning ability related to a specific scenario field.

[0093] Figure 4 The schematic diagram of the process for calculating the total score of the whole network service activation configuration in the case of activating a single service scenario provided by the embodiments of the present disclosure is as Figure 4 shown. The steps for calculating the total score of the whole network service activation configuration according to the prompt template include:

[0094] Step S21, in the case that the service to be activated is a single service, use the single-service optimized routing prompt template to wake up and guide the model, so as to determine each first alternative path that meets the preset first constraint condition according to the network topology information and the source and destination nodes of the service to be activated.

[0095] The first constraint condition includes at least one of the following: nodes that must be avoided (i.e., nodes to be avoided), bandwidth, delay, security index, energy consumption.

[0096] Step S22, use the single-service optimized routing prompt template to wake up and guide the model, so as to calculate the activation configuration score of each first alternative path according to the evaluation function and weight corresponding to each influencing factor.

[0097] Each influencing factor corresponds to an evaluation function and a weight. For a first alternative path, the activation configuration score of the first alternative path is calculated by using the weighted average method according to the evaluation function and weight corresponding to each influencing factor.

[0098] In the case that the service to be activated is a single service, the path configuration information x i of the service to be activated is the configuration information of the first alternative path corresponding to the highest score among the activation configuration scores of each first alternative path.

[0099] The following combines Figure 5 , and takes a single service to be activated with source and destination nodes A and B as an example for illustration.

[0100] Step A, the Prompt template for screening the single-service activation and optimization alternative paths that meet the first constraint condition wakes up and guides the Transformer A model.

[0101] For Figure 5 the OTN network shown, the OMC (Operation and Maintenance Center) network management issues a single service with source and destination AB. The constraint satisfaction of each first alternative path of this service is shown in Table 1:

[0102] Table 1

[0103]

[0104]

[0105] According to the first constraint condition for screening, the first alternative paths A - D - B, A - B, and A - E - B meet the routing constraint requirements for a single OTN service with source and destination AB and can be used as alternative paths for resource arrangement of this OTN service. The first alternative path A - C - E - B does not meet the routing constraint requirements for this service and therefore cannot be used as an alternative path for resource arrangement of this service. Therefore, according to Figure 5 the current network topology shown, for the opening requirement of a single OTN service with source and destination AB, search and traverse the alternative paths of this service, and screen out the first alternative paths A - D - B, A - B, and A - E - B that meet the first constraint condition.

[0106] Step B, after Transformer A is awakened and guided through the above Step A and has the ability to "screen alternative paths for opening and optimizing a single service that meets the constraint conditions", the Prompt template for optimizing the routing of a single service awakens and guides the Transformer A model, and determines the path configuration information x of the service to be opened i .

[0107] The comprehensive evaluation objective function for the routing of the k-th service is defined as follows:

[0108] obj feval (R k , k ∈ [1, svr num )

[0109] = [ω L · f eval-L (R k ) + ω E · f eval (R k ) + ω S · f eval-S (R K ) + ω C · f eval- (R k ) + ω H f eval-H (Rk )]

[0110] ω L +ω E +ω S +ω C +ω H =1

[0111] obj maxfeval ←max{[ω L f eval- (R k )+ω E ·f eval-E (R k )+ω S ·f eval-S (R k )+ω C f eval-C (R k )+ω H ·f eval-H (R k )],k∈[1,svr num}

[0112] Among them, ω L is the path delay weight, and can take a value of 0.45; ω E is the path energy consumption weight, and can take a value of 0.2; ω S is the path security weight, and can take a value of 0.15; ω C is the path cost weight, and can take a value of 0.15; ω L is the path hop count weight, and can take a value of 0.05. f eval (R k ) is the delay evaluation function of the k-th service alternative path, f eval-E (R k ) is the energy consumption evaluation function of the k-th service alternative path, f eval- (R k ) is the security evaluation function of the k-th service alternative path, f eval-C (R k ) is the cost evaluation function of the k-th service alternative path, f eva (R k ) is the hop count evaluation function of the k-th service alternative path, f eval-L (R k )、f eval-E (R k )、f eval- (R k )、f eva (F k )、f eval-H (R k) takes a positive value within the range of [0, 1].

[0113] According to the above objective optimization function, the three first alternative paths mentioned above are evaluated respectively, and the evaluation results and decision conclusions are shown in Table 2:

[0114] Table 2

[0115]

[0116] Figure 6 It is a schematic diagram of the process for calculating the total score of the whole-network service activation configuration in the scenario of activating multiple concurrent services provided by the embodiments of the present disclosure. As Figure 6 shown, the steps for calculating the total score of the whole-network service activation configuration according to the prompt template include:

[0117] Step S31, in the case where the services to be activated are multiple concurrent services, use the multi-service concurrent alternative path screening prompt template to wake up and guide the model, so as to determine a multi-service concurrent routing scheme in which each service to be activated meets the preset second constraint conditions.

[0118] Each service to be activated has one or more second alternative paths. By arranging and combining the second alternative paths of each service to be activated, multiple multi-service concurrent routing schemes are obtained. Each multi-service concurrent routing scheme includes the second alternative paths of all services to be activated. According to the preset second constraint conditions, the second alternative paths of each service to be activated are screened to obtain one or more multi-service concurrent routing schemes in which all second alternative paths meet the second constraint conditions.

[0119] The second constraint conditions include but are not limited to at least one of the following: nodes that must be avoided (i.e., mandatory avoidance nodes), bandwidth, delay, security metrics, energy consumption, links that must be passed through (i.e., mandatory passing links).

[0120] Step S32, use the multi-service concurrent optimized routing prompt template to wake up and guide the model, so as to calculate the total activation configuration score of each multi-service concurrent routing scheme according to the evaluation function and weight corresponding to each influencing factor.

[0121] In the case where there are multiple multi-service concurrent routing schemes that all meet the second constraint condition, use the multi-service concurrent optimization routing prompt template to wake up and guide the model, and calculate the total opening configuration scores of each multi-service concurrent routing scheme respectively. The total opening configuration score of each multi-service concurrent routing scheme is the sum of the opening configuration scores of the second alternative paths of each service to be opened in the multi-service concurrent routing scheme. Therefore, for a multi-service concurrent routing scheme, calculate the opening configuration scores of the second alternative paths of each service to be opened in this multi-service concurrent routing scheme respectively, and sum up the opening configuration scores of the second alternative paths of each service to be opened in this multi-service concurrent routing scheme to obtain the total opening configuration score of this multi-service concurrent routing scheme.

[0122] In the case where the services to be opened are multiple concurrent services, the path configuration information x of the services to be opened i is the configuration information of the second alternative paths of each service to be opened in the multi-service concurrent routing scheme corresponding to the highest score among the total opening configuration scores of each multi-service concurrent routing scheme.

[0123] The following combines Figure 5 , and takes three concurrent services to be opened with source and destination nodes being AB, BC, and DC as an example for illustration.

[0124] Step C, after Transformer A is woken up and guided through the above Step B and has the ability of "single OTN service optimized routing", the Prompt template for screening multi-service concurrent opening and optimized alternative paths that meet the second constraint condition wakes up and guides the Transformer A model to determine a multi-service concurrent routing scheme in which each service to be opened meets the preset second constraint condition.

[0125] The OMC network management system issues three concurrent services with sources and destinations being AB, BC, and DE respectively. The routing schemes of the three concurrent services, their respective second alternative paths, and the satisfaction of the second constraint condition are as follows. Among them, the multi-service concurrent routing scheme 1 and its second alternative path and the satisfaction of the second constraint condition are shown in Table 3:

[0126] Table 3

[0127]

[0128]

[0129] The multi-service concurrent routing scheme 2 and its second alternative path and the satisfaction of the second constraint condition are shown in Table 4:

[0130] Table 4

[0131]

[0132] The multi-service concurrent routing scheme 3 and its second alternative path and the satisfaction of the second constraint conditions are shown in Table 5 as follows:

[0133] Table 5

[0134]

[0135] The multi-service concurrent routing schemes 1 and 3 meet the requirements of the second constraint conditions for each service's routing and can be used as alternative schemes for the resource orchestration of these multiple concurrent services. The multi-service concurrent routing scheme 2 fails to meet the requirements of the second constraint conditions for each service's routing and thus cannot be used as an alternative scheme for the resource orchestration scenario of these multiple concurrent services. Therefore, according to Figure 5 the current network topology shown, for the three concurrent service opening requirements with source-destination nodes AB, BC, and DC, the multi-service concurrent routing schemes 1 and 3 that meet the preset second constraint conditions are screened out.

[0136] Step D: After Transformer A is awakened and guided through the above Step C, on the basis of having the ability to "screen alternative paths for multi-service concurrent opening and optimization that meet the constraint conditions", the Prompt template for multi-service opening and optimization routing decision wakes up and guides the Transformer A model to calculate the total opening configuration score of each multi-service concurrent routing scheme and determine the configuration information x of the second alternative path of each of the to-be-opened services in the multi-service concurrent routing scheme with the highest score i .

[0137] The comprehensive evaluation objective function for the routing decision of k concurrent services is defined as follows:

[0138]

[0139] where ω L is the path delay weight and can take a value of 0.45; ω E is the path energy consumption weight and can take a value of 0.2; ω S is the path security weight and can take a value of 0.15; ω C is the path cost weight and can take a value of 0.15; ω L is the path hop count weight and can take a value of 0.05. f eval-L (R k ) is the delay evaluation function of the k-th service alternative path, f eval- (R k ) is the energy consumption evaluation function of the k-th service alternative path, f eval- (R k ) is the security evaluation function of the k-th service alternative path, f eval- (R k) is the cost evaluation function for the k-th service alternative path, f eval-H (R k ) is the hop count evaluation function for the k-th service alternative path, f eval- (R k ), f eval-E (R k ), f eva (R k ), f eval-C (F k ), f eval- (R k ) take values as positive numbers between [0, 1].

[0140] According to the above objective optimization function, the above two multi-service concurrent routing schemes are evaluated respectively, and the evaluation results and decision conclusions are shown in Table 6:

[0141] Table 6

[0142]

[0143]

[0144] Figure 7 is the schematic diagram of the principle for the first fine-tuning process of the model provided by this embodiment of the present disclosure, Figure 8 is the schematic diagram of the first fine-tuning process of the AIGC model by the Policy-Gradient algorithm provided by this embodiment of the present disclosure. The following combines Figure 7 and Figure 8 to elaborate in detail on the process of the first fine-tuning process of the model.

[0145] Regarding the entire process of obtaining the network path orchestration scheme as a single-step DRL algorithm inference process, the AIGC large model (including the Adaptor) is the action policy provider and action executor of this single-step DRL algorithm from state S1 to state S2. The θ parameter in the Adaptor, that is, the action policy π θ (a1, s1), can be iteratively optimized through the Policy-Gradient algorithm, so as to obtain the network path orchestration scheme with the highest total score of the opening configuration.

[0146] Combined with Figure 7 and Figure 8 shown, the steps for the first fine-tuning process of the model include:

[0147] Using the DRL action optimization algorithm, the model is used to perform business activation intention orchestration for a preset number of times on multiple concurrent business activation requests in the same group, and preset number of intention orchestration results are obtained. Among them, the steps of performing business activation intention orchestration each time include: inputting the network topology information t0, the attribute information x of the service to be activated i , and the weight vector y of each of the influencing factors of the service to be activated i into the model to obtain the overall network service activation configuration total score x′0 and the path configuration information x′ of the service to be activated i . That is to say, the preset number is τ N (τ N = N), using the Transformer A model to perform τ N times of business activation intention orchestration to obtain τ N intention orchestration results. Calculate the sum average value of the τ N intention orchestration results, and adjust the parameters θ of the adapter (Adaptor) in the model according to the sum average value to obtain the action strategy (π θ (a1, s1)) of the first action a1. The action strategy of the first action is used to determine the path configuration information x′ of the service to be activated with the highest overall network service activation configuration total score i .

[0148] According to x′0 = obj feval (R1,...R k ) = R, it can be concluded that:

[0149] x′0(τ n ) = [obj feval (R1,...R k )](τ n ) = R(τ n ), R(τ n ), n ∈ [1, N], there is τ n = n;

[0150] Among them, R(τ n ) is the reward obtained by the multi-service concurrent routing scheme generated by the τ n -th round of trajectory reasoning, and its value is equal to the activation configuration total score of the multi-service concurrent routing scheme

[0151] Taking the current OTN network topology, the head and tail nodes and constraint conditions of each service to be activated and optimized, the weights of each resource and performance index, etc. as the input of the AIGC model, the model executes the first action a1 for intention orchestration in the form of an action strategy, and obtains the corresponding routing scheme and the overall network service activation configuration total score x′0, thus completing one round of network path orchestration scheme reasoning trajectory

[0152] To clearly illustrate the solutions of the embodiments of the present disclosure, the following is combined with Figure 9 , and a specific example is used to elaborate in detail on the process of intent network service activation. As Figure 9 shown, the process includes the following steps:

[0153] Step 1: Design a Prompt template based on the current network topology information t0 to guide the AIGC model instance Transformer A to perform intent parsing and reasoning on the batch input service activation. The reasoning of Transformer A is the action a0 of the DRL intent parsing trajectory.

[0154] Step 2: Based on the overall customer satisfaction score y0 output by Transformer A and the weight vectors of the influencing factors of each service , use the Policy Gradient algorithm to optimize the action policy of the action a0, that is, fine-tune the intent parsing and reasoning process of Transformer A.

[0155] Step 3: Design a hierarchical Prompt template based on the current network topology information t0 to guide the AIGC model instance Transformer A to perform intent orchestration and reasoning on the batch input services x1, x2... x svrnum and the weight vectors y1, y2... y of the influencing factors of each service svnum . The reasoning of Transformer A is the action a1 of the DRL intent orchestration trajectory.

[0156] Step 4: Based on the overall score x′0 of the whole-network service activation configuration output by Transformer A and the path configuration information x′ of the service to be activated i , use the Policy Gradient algorithm to optimize the action policy of the action a1, that is, fine-tune the intent orchestration and reasoning process of Transformer A.

[0157] Among them, Steps 1 and 2 are used to guide and fine-tune the Transformer A model to enable it to have intent parsing and reasoning capabilities; Steps 3 and 4 are used to guide and fine-tune the Transformer A model to enable it to have intent orchestration and reasoning capabilities.

[0158] The embodiments of the present disclosure also provide a computer device, as Figure 10 shown, including:

[0159] At least one processor 1001;

[0160] A memory 1002 stores at least one program, which, when executed by the at least one processor, enables the at least one processor to implement the intended network service activation method provided in the foregoing embodiments.

[0161] At least one I / O interface 1003 is connected between the processor and the memory and is configured to enable information interaction between the processor and the memory.

[0162] Among them, the processor 1001 is a device with data processing capabilities, including but not limited to a central processing unit (CPU), etc.; the memory 1002 is a device with data storage capabilities, including but not limited to a random access memory (RAM, more specifically such as SDRAM, DDR, etc.), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a flash memory (FLASH); the I / O interface (read / write interface) 1003 is connected between the processor 1001 and the memory 1002 and can enable information interaction between the processor 1001 and the memory 602, including but not limited to a data bus (Bus), etc.

[0163] In some embodiments, the processor 1001, the memory 1002, and the I / O interface 1003 are interconnected through a bus and further connected to other components of the computing device.

[0164] The embodiments of the present disclosure further provide a computer-readable medium storing a computer program, wherein when the program is executed, the intended network service activation method as described above is implemented.

[0165] Those of ordinary skill in the art will understand that all or some of the steps in the methods disclosed above, and the functional modules / units in the devices, can be implemented as software, firmware, hardware, and their appropriate combinations. In the hardware implementation, the division between the functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, one physical component can have multiple functions, or one function or step can be executed by several physical components in cooperation. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include a computer storage medium (or non-transitory medium) and a communication medium (or transitory medium). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory, or other memory technologies, CD-ROM, digital versatile disk (DVD), or other optical disk storage, magnetic cassettes, tapes, magnetic disk storage, or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, as is well known to those of ordinary skill in the art, communication media typically contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and can include any information delivery medium.

[0166] Example embodiments have been disclosed herein, and although specific terms have been used, they are used for and should be construed only as general illustrative meanings and not for purposes of limitation. In some instances, it will be apparent to those skilled in the art that, unless otherwise expressly stated, features, characteristics, and / or elements described in connection with a particular embodiment may be used alone or in combination with features, characteristics, and / or elements described in connection with other embodiments. Accordingly, those skilled in the art will understand that various forms and details may be changed without departing from the scope of the invention as set forth in the appended claims.

Claims

1. A method for intent network service activation, characterized in that Including: Determine the weight vector of each of the influencing factors of the service to be launched according to the network topology information, the influencing factors, the attribute information of the service to be launched, and a pre-trained model; wherein, the model is an artificial intelligence-generated content AIGC model; Determine the overall service activation configuration total score and the path configuration information of the service to be launched according to the network topology information, the attribute information of the service to be launched, the weight vector of each of the influencing factors of the service to be launched, and the model; wherein, the overall service activation configuration total score is the return of the first trajectory, calculated according to a pre-generated prompt template, and is used for the first fine-tuning process of the model to obtain the path configuration information of the service to be launched with the highest overall service activation configuration total score; the first trajectory is generated after executing the first action for intent orchestration and is used to implement the intent orchestration process for service activation.

2. The method according to claim 1, wherein The prompt template can wake up and guide the model, including a single-service alternative path screening prompt template, a single-service optimized route selection prompt template, a multi-service concurrent alternative path screening prompt template, and a multi-service concurrent optimized route selection prompt template; The guiding ability of the single-service optimized route selection prompt template is generated based on the guiding ability of the single-service alternative path screening prompt template, the guiding ability of the multi-service concurrent alternative path screening prompt template is generated based on the guiding ability of the single-service optimized route selection prompt template, and the guiding ability of the multi-service concurrent optimized route selection prompt template is generated based on the guiding ability of the multi-service concurrent alternative path screening prompt template.

3. The method according to claim 2, wherein The steps for calculating the overall service activation configuration total score according to the prompt template include: In the case where the service to be launched is a single service, use the single-service optimized route selection prompt template to wake up and guide the model to determine each first alternative path that meets the preset first constraint condition according to the network topology information and the source and destination nodes of the service to be launched; Use the single-service optimized route selection prompt template to wake up and guide the model to calculate the activation configuration scores of each of the first alternative paths according to the evaluation functions and weights corresponding to the influencing factors.

4. The method according to claim 3, wherein In the case where the service to be launched is a single service, the path configuration information of the service to be launched is the configuration information of the first alternative path corresponding to the highest score among the activation configuration scores of each of the first alternative paths.

5. The method according to claim 2, wherein The steps for calculating the overall service activation configuration total score according to the prompt template include: In the case where the service to be launched is multiple concurrent services, use the multi-service concurrent alternative path screening prompt template to wake up and guide the model to determine a multi-service concurrent routing scheme in which each of the services to be launched meets the preset second constraint condition; wherein, each multi-service concurrent routing scheme includes the second alternative paths of all the services to be launched; Wake up and guide the model by using the multi-service concurrent optimization routing hint template, so as to calculate the total opening configuration score of each multi-service concurrent routing scheme according to the evaluation functions and weights corresponding to the respective influencing factors; wherein, the total opening configuration score of each multi-service concurrent routing scheme is the sum of the opening configuration scores of the second alternative paths of the services to be opened in each multi-service concurrent routing scheme.

6. The method according to claim 5, wherein In the case where the services to be opened are multiple concurrent services, the path configuration information of the services to be opened is the configuration information of the second alternative paths of the services to be opened in the multi-service concurrent routing scheme corresponding to the highest score among the total opening configuration scores of each multi-service concurrent routing scheme.

7. The method according to claim 1, characterized in that The step of performing the first fine-tuning process on the model includes: Adopt the deep reinforcement learning DRL action tuning algorithm, and use the model to perform business opening intention orchestration on the same set of multiple concurrent service opening requests for a preset number of times to obtain a preset number of intention orchestration results; Calculate the sum average value of the preset number of intention orchestration results, and adjust the parameters of the adapter in the model according to the sum average value to obtain the action strategy of the first action, and the action strategy of the first action is used to determine the path configuration information of the service to be opened with the highest total opening configuration score of the entire network service. Wherein, the step of performing business opening intention orchestration each time includes: inputting the network topology information, the attribute information of the service to be opened, and the weight vectors of the respective influencing factors of the service to be opened into the model to obtain the total opening configuration score of the entire network service and the path configuration information of the service to be opened.

8. The method according to claim 1, wherein Before determining the total opening configuration score of the entire network service and the path configuration information of the service to be opened according to the network topology information, the attribute information of the service to be opened, the weight vectors of the respective influencing factors of the service to be opened, and the model, it further includes: Determine the overall customer satisfaction score of the service to be opened according to the network topology information, the influencing factors, the attribute information of the service to be opened, and the pre-trained model; the overall customer satisfaction score of the service to be opened is the return of the second trajectory, which is used to perform the second fine-tuning process on the model so that the weight vectors of the respective influencing factors of the service to be opened obtained are consistent with the overall customer satisfaction score of the service to be opened; the second trajectory is generated after executing the second action for intention parsing, and is used to implement the intention parsing process of service opening.

9. The method according to any one of claims 1-8, characterized in that, The topology information includes node information and link information, and the node information includes at least one of the following: node type, cross-connect capacity, occupied cross-connect capacity, number of transmission directions, number of wavelengths on each transmission direction; The link information includes at least one of the following: type, identifier, optical power, optical attenuation, optical signal-to-noise ratio, number of wavelengths of the link; identifier, transmission rate, channel resource occupancy information, power, bit error rate, optical signal-to-noise ratio, optical channel data unit multiplexing relationship, modulation mode, baud rate, spectral efficiency, center frequency, spectral width of each wavelength.

10. The method according to any one of claims 1-8, characterized in that, The influence factors include at least one of the following: delay factor, bandwidth utilization factor, energy consumption factor, cost factor, security factor, computing power factor, and hop count factor.

11. The method according to any one of claims 1-8, characterized in that, The attribute information includes at least one of the following: source-destination node information, service level agreement level information, application scenario type information, occupied bandwidth information, and routing constraint information.

12. The method according to any one of claims 1-8, characterized in that, The path configuration information includes at least one of the following: source-destination node information, service identifier information, service level agreement level information, activation configuration score, attribute information corresponding to the influence factors, and attribute information corresponding to the topology information.

13. A computer device, comprising: One or more processors; A storage device having stored thereon one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the intended network service activation method according to any one of claims 1-12.

14. A computer-readable medium having a computer program stored thereon, wherein, When the program is executed, it implements the intended network service activation method according to any one of claims 1-12.

Images