Task execution method and apparatus, device, storage medium, and program product

Abstract

The present disclosure provides a task execution method and apparatus, a device, a storage medium, and a program product. A first entity used for decision analysis and a plurality of second entities used for executing tasks are constructed and deployed in a network; and when receiving a task request sent by a demander, the first entity generates a corresponding task execution strategy on the basis of the task request, and sends a task instruction to the corresponding second entity on the basis of the task execution strategy, so that the second entity executes the corresponding task on the basis of the task instruction.

Description

Task execution method, apparatus, device, storage medium and program product

[0001] Cross-reference to Related Applications

[0002] The present disclosure claims priority from Chinese Patent Application No. 202510069701.9 filed on January 16, 2025 in China, the content of which is incorporated herein by reference in its entirety. TECHNICAL FIELD

[0003] The present disclosure relates to the field of communication, and particularly relates to a task execution method, apparatus, device, storage medium and program product. BACKGROUND

[0004] In the related art, a network intelligent function, such as a network data analysis function (NWDAF), has been introduced into a core network, to realize analysis, prediction, optimization and other functions of the network by introducing intelligent means such as data statistics and machine learning (ML) / artificial intelligence (AI) algorithms. In the related art, the network intelligent function is usually directed to a specific task with clear requirements, and a solution with a determined procedure and flow is generated by using technical means such as mathematical formulas and AI models. In this process, the solution is usually realized by an entity carried in one network function (NF) or by entities carried in several independent NFs. However, since the network architecture capable of realizing the network intelligent function, such as the entity carried in one NF or the entities carried in several independent NFs, is relatively simple, and the correlation between the entities is not strong, only simple tasks with clear requirements can be processed. For tasks with uncertain requirements and uncertain solutions, more entities need to be coordinated to analyze, but the network architecture in the related art cannot meet the requirement, resulting in that the intelligent degree of the network function is not high when facing complex tasks, and a suitable solution cannot be provided. SUMMARY

[0005] The purpose of the embodiments of the present disclosure is to provide a task execution method, apparatus, device, storage medium and program product, which can realize autonomous perception, analysis, decision and execution of the network, provide a suitable solution for complex tasks, and ensure efficient use of network resources while meeting network requirements.

[0006] To achieve the above-mentioned purpose, a first task execution method is provided in the embodiments of the present disclosure, which is applied to a first entity, the first entity is a strong intelligent agent, and the method comprises:

[0007] Receiving a task request sent by a demand side;

[0008] Generating a corresponding task execution strategy according to the task request; wherein the task execution strategy includes at least one task instruction corresponding to a second entity, and the second entity is at least one of a strong agent and a weak agent;

[0009] Sending the task instruction to the corresponding second entity according to the task execution strategy, the task instruction being used to instruct the second entity to execute a task corresponding to the task instruction.

[0010] As an improvement of the above scheme, the first entity includes a large model, and the generating of the corresponding task execution strategy according to the task request includes:

[0011] Obtaining corresponding task-related requirements according to the task request;

[0012] Querying network running information related to the task request and obtaining locally stored capability information;

[0013] Generating a task execution strategy corresponding to the task request by using the large model according to the task-related requirements, the network running information and the capability information.

[0014] As an improvement of the above scheme, the task instruction includes at least one of:

[0015] Model name or algorithm name;

[0016] Task execution parameter;

[0017] Task requirement, including transmission trigger condition and transmission path of task execution result;

[0018] Name of upstream entity and / or downstream entity;

[0019] Data acquisition requirement.

[0020] As an improvement of the above scheme, the transmission path includes at least one of:

[0021] Sent by any second entity to the first entity;

[0022] Sent by any second entity to a second entity specified by the first entity.

[0023] As an improvement of the above scheme, the transmission trigger condition includes at least one of:

[0024] Target parameter reaches event trigger threshold value;

[0025] Time reaches reporting time.

[0026] As an improvement of the above scheme, after the task instruction is sent to the corresponding second entity according to the task execution strategy, the method further comprises:

[0027] receiving a task execution result sent by at least one second entity;

[0028] performing an execution effect evaluation on the task execution strategy according to the task execution result;

[0029] when the execution effect of the task execution strategy does not reach an expected target, performing an update operation on the task execution strategy; wherein the update operation comprises at least one of adding a second entity, deleting a second entity, and updating a task instruction of a second entity;

[0030] after the update operation is performed on the task execution strategy, sending a task instruction in the updated task execution strategy to a second entity.

[0031] To achieve the above object, the second task execution method is provided by the embodiment of the present disclosure, which is applied to a second entity, the second entity is at least one of a strong agent and a weak agent, and the method comprises:

[0032] receiving a task instruction sent by a first entity according to a task execution strategy; wherein the first entity is a strong agent, the task execution strategy comprises a task instruction corresponding to at least one second entity, and the task execution strategy is generated by the first entity according to a task request;

[0033] executing a corresponding task according to the task instruction.

[0034] As an improvement of the above scheme, before the task instruction sent by the first entity according to the task execution strategy is received, the method further comprises:

[0035] receiving task configuration information sent by a demand side; wherein the task configuration information comprises capability information for executing a corresponding task, and the capability information corresponds to at least one model or algorithm;

[0036] then, the executing a corresponding task according to the task instruction comprises:

[0037] executing a task corresponding to the task instruction based on the capability information.

[0038] As an improvement of the above scheme, after the task instruction is executed according to the task instruction, the method further comprises:

[0039] send the task execution result to the first entity; wherein the task execution result is used to indicate the first entity to perform an execution effect evaluation on the task execution strategy, and the execution effect evaluation result is used to determine whether to perform an update operation on the task execution strategy; wherein the update operation includes at least one of adding a second entity, deleting a second entity, and updating a task instruction of a second entity.

[0040] As an improvement of the above scheme, the task instruction includes at least one of:

[0041] a task identifier;

[0042] a model identifier or an algorithm identifier;

[0043] a task execution parameter;

[0044] a task requirement, including a transmission trigger condition and a transmission path of the task execution result;

[0045] an identifier of an upstream entity and / or a downstream entity.

[0046] As an improvement of the above scheme, the execution of the corresponding task according to the task instruction includes at least one of the following steps:

[0047] when the task instruction includes a task identifier, indexing to a corresponding model or algorithm according to the task identifier to execute the task by using the model or algorithm;

[0048] when the task instruction includes a model identifier or an algorithm identifier, calling a corresponding model or algorithm according to the model identifier or the algorithm identifier to execute the task by using the model or algorithm;

[0049] when the task instruction includes a task execution parameter, using the task execution parameter as an input of task execution;

[0050] when the task instruction includes a task requirement, controlling the model or algorithm to execute the task according to the task requirement;

[0051] when the task instruction includes an identifier of an upstream entity and / or a downstream entity, receiving data transmitted by a second entity located upstream during the execution of the task, and / or transmitting own data to a second entity located downstream.

[0052] To achieve the above object, the embodiment of the present disclosure further provides a first task execution device applied to a first entity, wherein the first entity is a strong intelligent agent, and the device comprises:

[0053] a task request receiving module, configured to receive a task request sent by a demand party;

[0054] The task execution strategy generation module is configured to generate a corresponding task execution strategy according to a task request, wherein the task execution strategy comprises at least one task instruction corresponding to a second entity, and the second entity is at least one of a strong agent and a weak agent.

[0055] The task instruction sending module is configured to send the task instruction to the corresponding second entity in the task execution strategy, and the task instruction is used to instruct the second entity to perform a task corresponding to the task instruction.

[0056] To achieve the above object, the second task execution device is provided, which is applied to a second entity, and the second entity is at least one of a strong agent and a weak agent. The device comprises:

[0057] The task instruction receiving module is configured to receive a task instruction sent by a first entity according to a task execution strategy, wherein the first entity is a strong agent, the task execution strategy comprises at least one task instruction corresponding to a second entity, and the task execution strategy is generated by the first entity according to a task request.

[0058] The task execution module is configured to perform a corresponding task according to the task instruction.

[0059] To achieve the above object, the task execution device is provided, which comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the computer program is executed by the processor, the task execution method of any one of the above embodiments is implemented.

[0060] To achieve the above object, the computer readable storage medium is provided, which stores a computer program. When the computer program is executed, the device where the computer readable storage medium is located performs the task execution method of any one of the above embodiments.

[0061] To achieve the above object, the computer program product is provided, which comprises a computer program / instruction. When the computer program / instruction is executed by the processor, the task execution method of any one of the above embodiments is implemented.

[0062] The task execution method, device, equipment, storage medium and program product disclosed by the present disclosure construct and deploy a first entity for decision analysis and a plurality of second entities for task execution in a network. When receiving a task request, the first entity analyzes the task request to determine which second entities execute the task. In this process, the first entity splits the task and obtains at least one task instruction, and then sends the task instruction to the second entity capable of executing the task, so that the second entity executes the corresponding task according to the task instruction. Because the first entity can perform decision analysis on the task, the second entity can receive the task instruction sent by the first entity to execute the task, and based on the division of labor and cooperation of the first entity and the second entity, the first entity can analyze the complex task when facing the complex task. In addition, because the second entity is independently deployed and the number is unlimited, each second entity can independently complete its own task, and the first entity can schedule the second entity by using the task instruction, so that even a complex task can be executed by selecting a suitable second entity. Therefore, through the division of labor and cooperation of the first entity and the second entity in the embodiment of the present disclosure, autonomous perception, analysis, decision and execution of the network can be realized, and a suitable solution for a complex task is provided. In addition, because the first entity can meet the autonomous analysis and decision ability of the complex network problem, while bringing large resource consumption, the second entity can meet the rapid execution ability of the task, and needs less resource consumption. Through the combination and interaction mechanism of the two, the network demand can be met while ensuring efficient use of resources. BRIEF DESCRIPTION OF DRAWINGS

[0063] FIG. 1 is a schematic diagram of the overall architecture of a network entity according to an embodiment of the present disclosure;

[0064] FIG. 2 is a structural block diagram of a first entity according to an embodiment of the present disclosure;

[0065] FIG. 3 is a structural block diagram of a second entity according to an embodiment of the present disclosure;

[0066] FIG. 4 is a flowchart of a first task execution method according to an embodiment of the present disclosure;

[0067] FIG. 5 is an example schematic diagram of a task execution method according to an embodiment of the present disclosure;

[0068] FIG. 6 is a flowchart of a second task execution method according to an embodiment of the present disclosure;

[0069] FIG. 7 is a structural block diagram of a first task execution device according to an embodiment of the present disclosure;

[0070] FIG. 8 is a structural block diagram of a second task execution device according to an embodiment of the present disclosure;

[0071] FIG. 9 is a structural block diagram of a task execution device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0072] The technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the scope of protection of the present disclosure.

[0073] The intelligent application scenarios in the related art are relatively simple, and generally are tasks with relatively determined processes. The application cases adopt the way of solving specific problems with specific analysis, and the function extensibility is not strong. However, with the gradual development of the network, realizing an intelligent autonomous network is one of the development goals, that is, realizing a network that can autonomously perceive, autonomously analyze and decide, and autonomously optimize without human intervention. Since the environment, demand, and state of the network have uncertainty, self-management and optimization under complex conditions have high complexity, and therefore the network needs to have higher intelligence, including analysis of user intention, perception of environmental state, thinking of complex problems, generation of optimal strategies, and optimization of its own behavior through interaction and feedback with the environment, and so on. The behavior of this series of thinking chain is similar to the thinking process of human beings. Therefore, how to improve the intelligent degree of the network, and make the network become an intelligent entity that can autonomously perceive, analyze, decide, and execute, has deep thinking ability and rapid decision execution ability, and can autonomously solve problems with non-deterministic answers, has become a problem to be solved urgently. In order to improve the intelligent degree of the network function, the present disclosure provides a task execution method, which can realize autonomous perception, analysis, decision, and execution of the network when facing complex tasks, and provide a suitable solution, which can meet the network demand while ensuring efficient use of network resources.

[0074] Referring to FIG. 1, FIG. 1 is a general architecture schematic diagram of a network entity provided by an embodiment of the present disclosure. A plurality of entities are constructed and deployed in the network, which can be AI agents. The AI agent can be an independent network function, or can be embedded in an existing NF in the form of a component to enhance the intelligent ability of the NF and reduce the processing delay of the task. Through the division and cooperation of multiple AI agents, autonomous perception, analysis, decision, and execution of the network are realized. The overall architecture is divided into a first entity and a second entity, and the first entity and the second entity satisfy a one-to-many relationship. For example, the first entity is a sophisticated AI agent, and the second entity is a task AI agent. In a certain area, one first entity and a plurality of second entities cooperatively process tasks.

[0075] Referring to FIG. 2, FIG. 2 is a structural block diagram of a first entity provided by an embodiment of the present disclosure, the first entity being an analysis agent (Analysis Agent), and the first entity in the embodiment of the present disclosure takes a large model as an intelligent base. The large model, also referred to as an AI large model, refers to a "large parameter" model trained using large-scale data and powerful computing power. For example, the large model can be a large language model (LLM) or a network large model. In FIG. 2, the model defined by the intelligent base is an LLM, which is only an example. In other embodiments, other models such as a network large model can be used. The first entity has the functions of a typical AI agent, including planning, tools, and memory, for realizing complex and comprehensive intent targets that are difficult to complete using tasks with predetermined steps. For example, the intent target is to maximize the service level agreement (SLA) requirements of all users while maximizing network resource utilization, or the intent target is to automatically detect and recover when a network fault occurs to reduce business interruption time. These tasks are difficult to implement using fixed processes and steps and require strong intelligent thinking capabilities. The tools can include querying external information, application programming interfaces (APIs), external resources, knowledge graphs generated based on historical operation and maintenance experience, network configuration information (including the capabilities and deployment information of the task AI agent), and the like to assist in task planning. The memory includes short-term memory and long-term memory, so that the first entity can call past experience and knowledge when facing new or complex situations. The planning is completed based on the data provided by the tools and the memory modules to generate a task execution strategy. The LLM can update and synchronize the model through a large model library.

[0076] Referring to FIG. 3, FIG. 3 is a structural block diagram of a second entity provided by an embodiment of the present disclosure. There can be multiple second entities, which are strong or weak intelligent agents (Execution Agent). Generally, the second entity takes an AI small model or a mathematical logic analysis method as an intelligent base and is used to execute deterministic tasks with clear steps. For example, the second entity can be used for environmental or business perception, performance index analysis, abnormal detection and identification, scheduling optimization strategy, and the like.

[0077] It's important to note that because the first entity deploys a large LLM model, and the problem is complex requiring searching through short and long-term memories, using tools, and planning, it incurs high resource consumption and relatively long latency. Therefore, it's used for handling complex and comprehensive business tasks. Typically, one first entity is used per region for global thinking and planning, generating specific tasks to be executed, similar to the human brain. The second entity, based on a small AI model and mathematical logic code, performs intelligent analysis and reasoning tasks, similar to neurons distributed throughout the body. Because it's based on a small AI model and mathematical logic code, it handles less complex problems, resulting in lower resource consumption and lower inference latency than the first entity. It's suitable for tasks with high latency requirements. Typically, one region has multiple second entities, distributed and deployed closest to the needs or data, or they can be built into a Functional Level (NF) to serve a specific NF. After generating specific tasks, the first entity selects and distributes them to appropriate second entities. When multiple second entities collaborate on a large task, the first entity needs to configure the interaction and connection relationships between tasks when distributing them.

[0078] Referring to Figure 4, which is a flowchart of a first task execution method provided in this embodiment of the present disclosure, the first task execution method is applied to a first entity and includes:

[0079] S11. Receive the task request sent by the requester;

[0080] S12. Generate a corresponding task execution strategy according to the task request; wherein, the task execution strategy includes a task instruction corresponding to at least one second entity, and the second entity is at least one of a strong agent and a weak agent;

[0081] S13. Send the task instruction to the second entity corresponding to the task execution strategy. The task instruction is used to instruct the second entity to execute the task corresponding to the task instruction.

[0082] For example, the task request is a requirement given by the requesting party (such as a network management system) that the network should meet. It typically carries task-related requirements, which are then analyzed by the first entity. During the network configuration phase, the network management system issues configurations to both the first and second entities, configuring comprehensive task requirements for the first entity. For example, a task request might be "maximizing the SLA for all users while ensuring maximum network resource utilization," where the task-related requirements include "maximizing the SLA for all users" and "maximizing network resource utilization"; or, "maximizing the SLA for all users while ensuring optimal network communication quality," where the task-related requirements include "maximizing the SLA for all users" and "optimal network communication quality." In addition to statically receiving task requirements through configuration, the first entity can also receive task requirements through dynamic message requests. The network management system sends task configuration information to the distributed second entity. This task configuration information includes capability information for executing the corresponding task. This capability information corresponds to at least one model or algorithm to configure specific entity functions for the second entity, such as configuring a second entity with anomaly detection capabilities or a second entity with traffic identification capabilities. When a first entity receives a task request, it analyzes the task request to determine which second entities should execute the task. In this process, the first entity analyzes the task and obtains at least one task instruction. Then, it sends the task instruction to the second entity that can execute the task. The second entity executes the task corresponding to the task instruction based on its own capability information.

[0083] In this embodiment of the disclosure, by constructing and deploying a first entity for decision analysis and several second entities for task execution in the network, since the first entity can analyze complex tasks and select suitable second entities, and since the second entities receiving task instructions have corresponding functions, each second entity can autonomously complete tasks suitable for itself. Based on the division of labor and cooperation between the first entity and the second entity, the network can also achieve autonomous perception, analysis, decision-making and execution when facing complex tasks, providing suitable solutions.

[0084] In some embodiments, the task instruction includes at least one of the following 1) to 5):

[0085] 1) Task identifier; including the name and / or identifier (Identity, ID) of the task. Based on this task identifier, the second entity can know which specific tasks need to be performed, and then index the corresponding model or algorithm according to the task identifier to perform the task using the model or algorithm.

[0086] In this embodiment of the disclosure, since the task instruction sent by the first entity to the second entity includes a task identifier, the second entity can clearly identify the relevant information of the model or algorithm to be used based on this task identifier. In this way, the second entity can accurately call the corresponding model or algorithm to execute the task, thereby improving the efficiency and accuracy of task execution.

[0087] 2) Model identifier or algorithm identifier; the model identifier includes the model name and / or ID, and the algorithm identifier includes the algorithm name and / or ID. The second entity deploys the name and / or ID of the model or algorithm, such as an AI model or algorithm for certain specific tasks. The second entity can index the corresponding model based on the model identifier and then call this model to execute the task, and the second entity can index the corresponding algorithm based on the algorithm identifier and then call this algorithm to execute the task.

[0088] In this embodiment of the disclosure, since the task instruction sent by the first entity to the second entity includes a model identifier or an algorithm identifier, the second entity can accurately call the corresponding model or algorithm to execute the task, thereby improving the efficiency and accuracy of task execution.

[0089] 3) Task execution parameters; Task execution parameters include data collection requirements, execution time, and execution object. The second entity collects the data required for the model or algorithm to execute the task within a specified time period according to the task execution parameters, using the task execution parameters as the input for task execution. For example, if the task of the second entity is to monitor business traffic, then the data collection requirement of the second entity is to collect business traffic, the execution time indicates that the current entity needs to execute the data collection task within a certain time period, and the execution object is the traffic of the corresponding business.

[0090] In this embodiment of the disclosure, since the task instruction sent by the first entity to the second entity includes task execution parameters, indicating the data that the second entity needs to collect and when to collect it, the efficiency and accuracy of the second entity's data collection can be improved, thereby improving the efficiency and accuracy of task execution.

[0091] 4) Task Requirements; The task requirements include performance requirements, accuracy requirements, triggering conditions for transmitting task execution results, and transmission paths, etc. The second entity controls the model or algorithm to execute the task according to the task requirements. It should be noted that the task execution result can be an intermediate result generated by the second entity during the task execution process, or a final result generated after the task is completed.

[0092] In this embodiment of the disclosure, since the task instructions sent by the first entity to the second entity include task requirements, the second entity can perform the task in the best possible state, thereby improving the reliability of task execution and thus improving the accuracy of task execution.

[0093] The transmission path includes at least one of the following:

[0094] 4.11) Sent from any second entity to the first entity.

[0095] In this embodiment of the disclosure, the second entity can send the task execution result to the first entity, so that the first entity can collect all the data processing results of the second entity during or after the task execution in a timely manner, and then analyze the specific application process of the second entity, discover potential problems and anomalies of the second entity in a timely manner, and make quick decisions and adjustments.

[0096] 4.12) Sent by any second entity to the second entity specified by the first entity; such as if the first entity instructs the current second entity to send the task execution result to the second entity located downstream of the current second entity.

[0097] In this embodiment of the disclosure, the second entity can send the task execution result to the second entity designated by the first entity, ensuring that the data between the second entities that need to work together can be transmitted accurately, avoiding the situation where data transmission errors cause the task to fail to execute or to be executed incorrectly, thereby improving the task execution process and enhancing the reliability of task execution.

[0098] The transmission triggering condition includes at least one of the following:

[0099] 4.21) The target parameter reaches the event trigger threshold; for example, if the task of the second entity is to monitor service traffic, the target parameter is the service traffic detection value, and the event trigger threshold is the traffic threshold value. When the service traffic detection value reaches the traffic threshold value, the transmission of the task execution result is triggered. At this time, the data carried by the task execution result is "The service traffic detection value has reached the traffic threshold value"; for example, if the task of the second entity is to obtain network bandwidth, the target parameter is the bandwidth detection value, and the event trigger threshold is the bandwidth threshold value. When the bandwidth detection value reaches the bandwidth threshold value, the transmission of the task execution result is triggered. At this time, the data carried by the task execution result is "The bandwidth detection value has reached the bandwidth threshold value";

[0100] In this embodiment, the second entity can report the task execution result after meeting the time trigger threshold. By predefining the trigger threshold, data that is conducive to subsequent analysis can be filtered out. This allows for the collection of effective data from the second entity without the data volume being too large, reducing the load on the first entity. At the same time, it enables the first entity to collect the data processing results of all second entities during or after the task execution in a timely manner. This allows for subsequent analysis of the specific application process of the second entity, timely discovery of potential problems and anomalies, and rapid decision-making and adjustments.

[0101] 4.22) When the reporting time is reached; if the first entity sets the second entity to report the task execution results at regular intervals, the second entity will transmit the task execution results according to the reporting time set in this regular reporting mechanism.

[0102] In this embodiment, the second entity can use a timed reporting mechanism to report the task execution results. The timed reporting method enables all second entities to upload data to the first entity, avoiding data omissions. This allows the first entity to collect the data processing results of all second entities during or after the task execution, and then analyze the specific application process of the second entity, promptly identify potential problems and anomalies, and make quick decisions and adjustments.

[0103] 5) Identification of upstream and / or downstream entities; the identification includes the entity’s name and / or ID, which indicates the information transmission path of the second entity, that the current second entity can receive data transmitted by the upstream second entity, and that the current second entity can transmit data to the downstream second entity.

[0104] In this embodiment of the disclosure, the identifiers of the upstream and / or downstream entities indicate the topological relationship of the second entities, thereby indicating the data flow transmission direction between multiple second entities, enabling the second entities to quickly transmit data to the next node, thereby improving the data processing efficiency during task execution.

[0105] In some embodiments, the first entity carries a large model, which may be a large language model or a large network model, or other models with analytical capabilities. Step S12 specifically includes: obtaining task-related requirements based on the task request; querying network operation information related to the task request and obtaining locally stored capability information; and generating a task execution strategy corresponding to the task request using the large model based on the task-related requirements, network operation information, and capability information.

[0106] In this embodiment of the disclosure, since the large model has powerful analytical capabilities, this disclosure generates task execution strategies by using the large model based on parameters such as task-related requirements, network operation information, and capability information. When faced with complex tasks, it can discover more potential solutions based on these parameters, making the task execution strategy more flexible and adaptable.

[0107] For example, the first entity generates task-related requirements based on the task request, then uses external tools to query network operation information related to the task request, and retrieves locally stored capability information from the memory module, thus determining which capabilities the second entity can provide. The first entity then uses the task-related requirements, network operation information, and capability information to perform task planning based on a large model, generating a task execution strategy.

[0108] For example, taking the task request of "maximizing the SLA for all users while maximizing network resource utilization" as an example, the task request is parsed, and the task-related requirements include "maximizing the SLA for all users" and "maximizing network resource utilization". The system uses tools to query empirical data such as network configuration, online users, network resources, system SLA requirements, historical business patterns and traffic information from the same period, using this data as network operation information, and also retrieves stored capability information, such as the optimization effects of historical optimization strategies, existing AI model names, and algorithm names stored in the memory module. The planning module, based on the large model, task-related requirements, network operation information, and capability information, performs requirement analysis and task decomposition to generate a task execution strategy. Similarly, taking the task request of "maximizing the SLA for all users while ensuring the best network communication quality" as an example, the task request is parsed, and the relevant requirements include "maximizing the SLA for all users" and "the best network communication quality". The network configuration, online users, network bandwidth, system SLA requirements, historical service patterns and traffic information, and other empirical data are queried through tools. These data are used as network operation information, as well as the stored capability information.

[0109] Referring to Figure 5, which is a schematic diagram of an example of a task execution method provided in this disclosure, the example illustrates a task request to "maximize the SLA for all users while ensuring maximum network resource utilization." In this case, the second entity includes a traffic identification entity, a service (QoS) awareness entity, an anomaly detection entity, a problem diagnosis entity, a decision optimization entity, a service status monitoring entity, and a network resource analysis entity. The first entity pre-issues corresponding task instructions to each second entity. After receiving the task instructions, each second entity loads a model / algorithm from the model / algorithm library as needed, performs data collection, sets up upstream and downstream second entities, and executes the task. The collaborative workflow of the first and second entities is as follows:

[0110] 1. Deploy traffic identification entities in user coverage areas with concentrated traffic and requiring key protection. The traffic identification entities call traffic identification models or algorithms to identify traffic in user coverage areas, and then input the traffic identification results into the business awareness entity.

[0111] 2. The business awareness entity calls the business awareness model or algorithm to monitor business traffic and business experience evaluation, uses business traffic and business experience evaluation as the awareness result, and then reports the awareness result to the anomaly detection entity;

[0112] 3. The anomaly detection entity calls the anomaly detection model or algorithm and uses this perception result to assess whether an anomaly has occurred. When the anomaly detection entity detects or predicts an anomaly, it reports the anomaly detection result to the problem diagnosis entity.

[0113] 4. The problem diagnosis entity invokes the problem diagnosis model or algorithm to comprehensively analyze the root cause of the problem based on the reported anomaly detection results. Based on the root cause analysis results, it determines which region or type of decision optimization entity to send the root cause analysis to. If the original region is busy and has limited analysis capabilities, the root cause analysis results can be sent to an idle decision optimization entity deployed in another region for analysis. Alternatively, if multiple decision optimization entities are divided according to processing capacity or processing type, the capability requirements of the entity needing to perform decision optimization are determined based on the root cause analysis results, or the decision optimization entity capable of handling this type of event is identified, and then a suitable decision optimization entity is selected for analysis.

[0114] 5. After receiving the root cause analysis results, the decision optimization entity requests service information and resource information from the service status monitoring entity and the network resource analysis entity, and calls the decision optimization model or algorithm to determine the Quality of Service (QoS) optimization strategy based on the service information, resource information and root cause analysis results. At the same time, depending on the type of QoS optimization strategy, it determines whether the QoS optimization strategy needs to be verified and optimized by the network twin entity (which is also a type of second entity).

[0115] 6. The first entity can collect the task execution results of each second entity and decide whether to adjust the tasks, models, etc. of the second entities.

[0116] During task execution, the second entity, according to the task instructions, reports the task execution results to the first entity when the event trigger threshold is reached or the reporting time set in the timed reporting mechanism is reached, or interacts with the second entity designated by the first entity. For example, business awareness entities, anomaly detection entities, problem diagnosis entities, decision optimization entities, and network twin entities of specific areas or users report relevant information (task execution results) to the first entity according to the first entity's requirements through event triggering or timed reporting mechanisms. Reporting parameters may include reporting time, reporting event type, event ID, event description, etc.

[0117] In this embodiment, by conducting in-depth analysis of task requests to obtain task-related requirements, the large model can more accurately grasp the core objectives and specific requirements of the task, avoiding execution errors or inefficiencies caused by misunderstandings of the task. Combining network operation information and locally stored capability information, the large model can comprehensively understand the currently available second entity, thereby formulating a more practical task execution strategy, optimizing the allocation of second entity resources, and improving resource utilization efficiency. Furthermore, leveraging the large model's powerful learning and reasoning capabilities, combined with locally stored capability information, more potential solutions can be discovered based on existing resources and capabilities, making the task execution strategy more flexible and adaptable, enabling the first entity to handle various complex tasks.

[0118] Furthermore, after performing step S13, the method further includes: receiving a task execution result sent by at least one second entity; evaluating the execution effect of the task execution strategy based on the task execution result; when the execution effect of the task execution strategy does not meet the expected target, performing an update operation on the task execution strategy; wherein the update operation includes at least one of adding a second entity, deleting a second entity, and updating the task instructions of the second entity; after performing the update operation on the task execution strategy, sending the task instructions in the updated task execution strategy to the second entity.

[0119] For example, after receiving task execution results from at least one second entity, the first entity initiates further task analysis as needed. Based on the task execution results, it evaluates the effectiveness of the task execution strategy. The large model deployed in the first entity can perform data comparison (comparing with historical data), analysis, and diagnosis based on the collected task execution results to determine the execution effect of each second entity. Then, based on the execution effect, it decides whether to adjust the task execution strategy. When the execution effect of the task execution strategy reaches the expected goal, the first entity does not need to perform any processing on the task execution strategy and simply waits for all second entities to complete their respective tasks. When the execution effect of the task execution strategy does not reach the expected goal, an update operation is performed on the task execution strategy, such as adding new second entities, deleting redundant second entities, or updating the task instructions of second entities. After performing the update operation, the task instructions in the updated task execution strategy need to be resent to the second entities. Furthermore, the first entity can use this task execution result as input to the large model, allowing the large model to understand the task execution capability of this second entity, thereby continuously learning and optimizing the task execution strategy generated by the large model.

[0120] In this embodiment, the first entity can promptly analyze the rationality of the task execution strategy based on the task execution results of the second entity and quickly adjust and update it. This avoids prolonged pauses and repeated trial and error caused by incorrect deployment of the second entity during task execution, enabling the task to proceed more efficiently. Furthermore, by analyzing the task execution results, the first entity can understand the work efficiency of each second entity, thereby allocating tasks more accurately and further improving overall task execution efficiency. Simultaneously, the first entity can promptly identify problems during task execution, such as data acquisition errors or data processing lag, and adjust the task execution strategy in a timely manner to prevent further escalation of the problems, thus ensuring accurate task execution.

[0121] The task execution method disclosed in this disclosure constructs and deploys a first entity for decision analysis and several second entities for task execution within a network. Since the first entity can perform task decision analysis, and the second entities can receive task instructions from the first entity to execute the task, based on the division of labor and cooperation between the first and second entities, even with complex tasks, the first entity can still analyze them. Furthermore, because the second entities are independently deployed and their number is unlimited, each second entity can autonomously complete its own task. The first entity uses task instructions to uniformly schedule the second entities, ensuring that even complex tasks can be executed by selecting appropriate second entities. Therefore, through the division of labor and cooperation between the first and second entities in this disclosure, autonomous perception, analysis, decision-making, and execution of the network can be achieved, providing suitable solutions for complex tasks. Additionally, while the first entity can satisfy the autonomous analysis and decision-making capabilities for complex network problems, it also incurs significant resource consumption, whereas the second entities can satisfy the rapid execution capabilities of tasks with less resource consumption. The combination and interaction mechanism of the two entities can ensure efficient resource utilization while meeting network requirements.

[0122] Referring to Figure 6, which is a flowchart of a second task execution method provided in this embodiment, the second task execution method is applied to a second entity, which is at least one of a strong agent and a weak agent. The second task execution method includes:

[0123] S21. Receive a task instruction sent by a first entity according to a task execution strategy; wherein, the first entity is a strong agent, the task execution strategy includes a task instruction corresponding to at least one second entity, and the task execution strategy is generated by the first entity according to a task request;

[0124] S22. Execute the corresponding task according to the task instruction.

[0125] In some embodiments, before receiving the task instruction sent by the first entity according to the task execution strategy, the method further includes: receiving task configuration information sent by the requester; wherein the task configuration information includes capability information for executing the corresponding task, and the capability information corresponds to at least one model or algorithm;

[0126] Then, executing the corresponding task according to the task instruction includes:

[0127] Execute the task corresponding to the task instruction based on the capability information.

[0128] In some embodiments, after executing the corresponding task according to the task instruction, the method further includes: sending the task execution result to a first entity; wherein the task execution result is used to instruct the first entity to evaluate the execution effect of the task execution strategy, and the execution effect evaluation result is used to determine whether to perform an update operation on the task execution strategy; wherein the update operation includes at least one of the task instructions for adding a second entity, deleting a second entity, and updating a second entity.

[0129] In some embodiments, the task instructions include at least one of the following:

[0130] Task identifier;

[0131] Model identifier or algorithm identifier;

[0132] Task execution parameters;

[0133] Task requirements, including the triggering conditions and transmission path for transmitting task execution results;

[0134] Identification of upstream and / or downstream entities.

[0135] In some embodiments, step S22, executing the corresponding task according to the task instruction, includes at least one of the following steps:

[0136] When the task instruction includes a task identifier, the corresponding model or algorithm is indexed according to the task identifier so as to execute the task using the model or algorithm;

[0137] When the task instruction includes a model identifier or an algorithm identifier, the corresponding model or algorithm is invoked according to the model identifier or algorithm identifier to perform the task using the model or algorithm;

[0138] When the task instruction includes task execution parameters, the task execution parameters are used as the input for task execution;

[0139] When the task instruction includes task requirements, the control model or algorithm executes the task according to the task requirements;

[0140] When the task instruction includes the identifiers of an upstream entity and / or a downstream entity, the entity receives data transmitted by a second entity located upstream during the execution of the task, and / or transmits its own data to a second entity located downstream.

[0141] It is worth noting that the working process of the second task execution method described in this embodiment can refer to the working process of the second entity in the first task execution method described in the above embodiment, and will not be repeated here.

[0142] Referring to Figure 7, which is a structural block diagram of a first task execution device 100 provided in this embodiment of the present disclosure, the task execution device 100 is applied to a first entity, the first entity being a strong intelligent agent, and the task execution device 100 includes:

[0143] Task request receiving module 11 is used to receive task requests sent by the requester.

[0144] The task execution strategy generation module 12 is used to generate a corresponding task execution strategy according to the task request; wherein, the task execution strategy includes a task instruction corresponding to at least one second entity, and the second entity is at least one of a strong agent and a weak agent;

[0145] The task instruction sending module 13 is used to send the task instruction to the second entity corresponding to the task execution strategy, wherein the task instruction is used to instruct the second entity to execute the task corresponding to the task instruction.

[0146] In some embodiments, the task execution device 100 further includes:

[0147] The task execution result receiving module is used to receive the task execution result sent by at least one second entity.

[0148] The execution effect evaluation module is used to evaluate the execution effect of the task execution strategy based on the task execution results;

[0149] The task execution strategy adjustment module is used to perform an update operation on the task execution strategy when the execution effect of the task execution strategy does not achieve the expected goal; wherein, the update operation includes at least one of the task instructions of adding a second entity, deleting a second entity, and updating a second entity;

[0150] Then, the task instruction sending module 13 is further configured to: after performing the update operation on the task execution strategy, send the task instructions in the updated task execution strategy to the second entity.

[0151] It is worth noting that the working process of each module in the task execution device 100 described in this embodiment can be referred to the working process of the first entity in the first task execution method described in the above embodiment, and will not be repeated here.

[0152] Referring to Figure 8, which is a structural block diagram of a second task execution device 200 provided in an embodiment of this disclosure, the task execution device 200 is applied to a second entity, the second entity being at least one of a strong agent and a weak agent, and the task execution device 200 includes:

[0153] The task instruction receiving module 21 is used to receive task instructions sent by a first entity according to a task execution strategy; wherein, the first entity is a strong agent, the task execution strategy includes task instructions corresponding to at least one second entity, and the task execution strategy is generated by the first entity according to a task request;

[0154] The task execution module 22 is used to execute the corresponding task according to the task instruction.

[0155] In some embodiments, the task execution device 200 further includes:

[0156] The task configuration information receiving module is used to receive task configuration information sent by the requester; wherein, the task configuration information includes capability information for executing the corresponding task, and the capability information corresponds to at least one model or algorithm;

[0157] Then, the task execution module 22 is used to execute the task corresponding to the task instruction based on the capability information.

[0158] In some embodiments, the task execution device 200 further includes:

[0159] A task execution result sending module is used to send the task execution result to a first entity; wherein the task execution result is used to instruct the first entity to evaluate the execution effect of the task execution strategy, and the execution effect evaluation result is used to determine whether to perform an update operation on the task execution strategy; wherein the update operation includes at least one of the task instructions for adding a second entity, deleting a second entity, and updating a second entity.

[0160] It is worth noting that the working process of each module in the task execution device 200 described in this embodiment can be referred to the working process of the second entity in the first task execution method described in the above embodiment, and will not be repeated here.

[0161] Referring to Figure 9, which is a structural block diagram of a task execution device 300 provided in an embodiment of this disclosure, the task execution device 300 includes a processor 31, a memory 32, and a computer program stored in the memory 32 and executable on the processor 31. When the processor 31 executes the computer program, it implements the steps in the above-described task execution method embodiments, such as steps S11-S13, S21-S22, etc.

[0162] For example, the computer program may be divided into one or more modules / units, which are stored in the memory 32 and executed by the processor 31 to complete the present disclosure. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the task execution device 300.

[0163] The task execution device 300 may include, but is not limited to, a processor 31 and a memory 32. Those skilled in the art will understand that the schematic diagram is merely an example of the task execution device 300 and does not constitute a limitation on the task execution device 300. It may include more or fewer components than illustrated, or combine certain components, or use different components. For example, the task execution device 300 may also include input / output devices, network access devices, buses, etc.

[0164] The processor 31 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor 31 is the control center of the task execution device 300, connecting all parts of the task execution device 300 via various interfaces and lines.

[0165] The memory 32 can be used to store the computer programs and / or modules. The processor 31 implements various functions of the task execution device 300 by running or executing the computer programs and / or modules stored in the memory 32 and calling the data stored in the memory 32. The memory 32 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 32 may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0166] Wherein, if the modules / units integrated in the task execution device 300 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by the processor 31, it can implement the steps of the various method embodiments described above. Wherein, the computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0167] This disclosure also provides a computer program product, including computer instructions, which, when executed by a processor, implement the task execution method as described in the above embodiments.

[0168] The above description represents the preferred embodiments of this disclosure. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this disclosure, and these improvements and modifications are also considered to be within the scope of protection of this disclosure.

Claims

1. A task execution method applied to a first entity, wherein the first entity is a strong agent, the method comprising: Receive task requests sent by the requester; A task execution strategy is generated based on the task request; wherein the task execution strategy includes a task instruction corresponding to at least one second entity, and the second entity is at least one of a strong agent and a weak agent; The task instruction is sent to the second entity corresponding to the task execution strategy, and the task instruction is used to instruct the second entity to execute the task corresponding to the task instruction.

2. The task execution method as described in claim 1, wherein, The first entity includes a large model, and the step of generating a corresponding task execution strategy based on the task request includes: Obtain task-related requirements based on the task request; Query network operation information related to the task request and obtain locally stored capability information; Based on the task-related requirements, network operation information, and capability information, the large model is used to generate a task execution strategy corresponding to the task request.

3. The task execution method as described in claim 1, wherein, The task instructions include at least one of the following: Task identifier; Model identifier or algorithm identifier; Task execution parameters; Task requirements, including the triggering conditions and transmission path for transmitting task execution results; Identification of upstream and / or downstream entities.

4. The task execution method as described in claim 3, wherein, The transmission path includes at least one of the following: Send from any second entity to the first entity; Sends from any second entity to the second entity specified by the first entity.

5. The task execution method as described in claim 3, wherein, The transmission triggering condition includes at least one of the following: The target parameter has reached the event trigger threshold; The reporting time has been reached.

6. The task execution method as described in claim 1, wherein, After sending the task instruction to the corresponding second entity according to the task execution strategy, the method further includes: Receive the task execution result sent by at least one second entity; The execution effect of the task execution strategy is evaluated based on the task execution results. When the execution effect of the task execution strategy fails to achieve the expected goal, an update operation is performed on the task execution strategy; wherein, the update operation includes at least one of the task instructions for adding a second entity, deleting a second entity, and updating a second entity; After performing the update operation on the task execution strategy, the task instructions in the updated task execution strategy are sent to the second entity.

7. A task execution method applied to a second entity, wherein the second entity is at least one of a strong agent and a weak agent, the method comprising: Receive a task instruction sent by a first entity according to a task execution strategy; wherein the first entity is a strong agent, the task execution strategy includes task instructions corresponding to at least one second entity, and the task execution strategy is generated by the first entity according to a task request; Execute the corresponding task according to the task instructions.

8. The task execution method as described in claim 7, wherein, Before receiving the task instruction sent by the first entity according to the task execution strategy, the method further includes: Receive task configuration information sent by the requester; wherein, the task configuration information includes capability information for executing the corresponding task, and the capability information corresponds to at least one model or algorithm; Then, executing the corresponding task according to the task instruction includes: Execute the task corresponding to the task instruction based on the capability information.

9. The task execution method as described in claim 7, wherein, After executing the corresponding task according to the task instruction, the method further includes: The task execution result is sent to the first entity; wherein the task execution result is used to instruct the first entity to evaluate the execution effect of the task execution strategy, and the execution effect evaluation result is used to determine whether to perform an update operation on the task execution strategy; wherein the update operation includes at least one of the task instructions for adding a second entity, deleting a second entity, and updating a second entity.

10. The task execution method as described in claim 7, wherein, The task instructions include at least one of the following: Task identifier; Model identifier or algorithm identifier; Task execution parameters; Task requirements, including the triggering conditions and transmission path for transmitting task execution results; Identification of upstream and / or downstream entities.

11. The task execution method as described in claim 10, wherein, The step of executing the corresponding task according to the task instruction includes at least one of the following steps: When the task instruction includes a task identifier, the corresponding model or algorithm is indexed according to the task identifier so as to execute the task using the model or algorithm; When the task instruction includes a model identifier or an algorithm identifier, the corresponding model or algorithm is invoked according to the model identifier or algorithm identifier to perform the task using the model or algorithm; When the task instruction includes task execution parameters, the task execution parameters are used as the input for task execution; When the task instruction includes task requirements, the control model or algorithm executes the task according to the task requirements; When the task instruction includes the identifiers of an upstream entity and / or a downstream entity, the entity receives data transmitted by a second entity located upstream during the execution of the task, and / or transmits its own data to a second entity located downstream.

12. A task execution device applied to a first entity, the first entity being a strong intelligent agent, the device comprising: The task request receiving module is used to receive task requests sent by the requester. A task execution strategy generation module is used to generate a task execution strategy based on a task request; wherein, the task execution strategy includes a task instruction corresponding to at least one second entity, and the second entity is at least one of a strong agent and a weak agent; The task instruction sending module is used to send the task instruction to the second entity corresponding to the task execution strategy, wherein the task instruction is used to instruct the second entity to execute the task corresponding to the task instruction.

13. A task execution apparatus applied to a second entity, the second entity being at least one of a strong agent and a weak agent, the apparatus comprising: The task instruction receiving module is used to receive task instructions sent by a first entity according to a task execution strategy; wherein, the first entity is a strong agent, the task execution strategy includes task instructions corresponding to at least one second entity, and the task execution strategy is generated by the first entity according to a task request; The task execution module is used to execute the corresponding task according to the task instruction.

14. A task execution device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the task execution method as claimed in any one of claims 1 to 11.

15. A computer-readable storage medium, wherein, The computer-readable storage medium stores a computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the task execution method as described in any one of claims 1 to 11.

16. A computer program product comprising a computer program / instructions that, when executed by a processor, implement the task execution method as described in any one of claims 1 to 11.

Images