Intelligent agent scheduling methods, devices, equipment, media and program products

By monitoring the operational metrics of intelligent agents and generating temporary scheduling factors, abnormal intelligent agents can be identified and handled, thus solving the problem of inflexible scheduling methods in existing technologies and improving the scheduling flexibility and stability of the system.

CN123114644APending Publication Date: 2026-11-17CHINA UNITED NETWORK COMM GRP CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611097187.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-22
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing technologies, the scheduling methods for intelligent agents are not flexible enough and cannot be adjusted flexibly during task execution, resulting in system fragility and high maintenance costs.

Method used

By monitoring the operational metrics of intelligent agents, temporary scheduling factors are determined, abnormal intelligent agents are identified, and anomaly handling strategies are generated, including rescheduling, reducing computing resources, or verifying task execution results.

Benefits of technology

It improves the flexibility of agent scheduling, avoids the problem of inflexible adjustment during task execution, enhances system stability, and reduces operation and maintenance costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123114644A_ABST
    Figure CN123114644A_ABST
Patent Text Reader

Abstract

This application provides a method, apparatus, device, medium, and program product for intelligent agent scheduling, relating to the field of artificial intelligence technology, and is used to improve the flexibility of intelligent agent scheduling. The specific technical solution is as follows: At least one intelligent agent executes sub-tasks within a target task, with each intelligent agent corresponding to at least one sub-task; during the execution of the target task, based on the operational metrics of each intelligent agent, a temporary scheduling factor is determined for each intelligent agent, representing the degree of abnormality in the intelligent agent's operational metrics; based on the temporary scheduling factor of each intelligent agent, an abnormal intelligent agent is identified from the at least one intelligent agent, and an exception handling strategy is generated for the abnormal intelligent agent. The exception handling strategy includes at least one of the following: rescheduling the intelligent agent for the sub-task corresponding to the abnormal intelligent agent, reducing the computing resources of the abnormal intelligent agent, and verifying the task execution result of the abnormal intelligent agent. This application is applied to scenarios requiring intelligent agent scheduling.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to an intelligent agent scheduling method, apparatus, device, medium, and program product. Background Technology

[0002] With the rapid development of artificial intelligence technology, it has become increasingly important to invoke at least one intelligent agent to perform target tasks through electronic devices.

[0003] In related technologies, electronic devices break down a target task into at least one subtask. Then, based on the task type of the subtask and the trust score of each agent, an agent is determined to execute each subtask. After the agent completes the assigned subtask, the trust score of the agent is updated based on the task execution result.

[0004] However, in the above scheme, because the scheduling of agents relies on a trust score, and the update of the trust score depends on the task execution result of the agent, but the trust score of the agent remains unchanged throughout the task execution process, the electronic device cannot flexibly adjust the scheduling strategy for the agent during task execution. Thus, the scheduling method for agents in the related technologies is not flexible enough. Summary of the Invention

[0005] This application provides an agent scheduling method, apparatus, device, medium, and program product to improve the flexibility of agent scheduling.

[0006] In a first aspect, embodiments of this application provide an agent scheduling method, the method comprising: executing sub-tasks in a target task through at least one agent, each agent corresponding to at least one sub-task; determining a temporary scheduling factor for each agent based on the operational metrics of each agent, the temporary scheduling factor representing the degree of abnormality of the operational metrics of the agent; determining an abnormal agent from the at least one agent based on the temporary scheduling factor of each agent, and generating an exception handling strategy for the abnormal agent, the exception handling strategy comprising at least one of the following: rescheduling the agent for the sub-task corresponding to the abnormal agent, reducing the computing resources of the abnormal agent, and verifying the task execution result of the abnormal agent.

[0007] The technical solution provided in this application brings at least the following beneficial effects: In this solution, a temporary scheduling factor is determined based on the operating indicators of the intelligent agent to characterize the degree of abnormality of the operating indicators of each intelligent agent. Then, based on the temporary scheduling factor and the abnormal intelligent agent, a processing strategy for the abnormal intelligent agent is determined, which avoids the problem that electronic devices cannot flexibly adjust the scheduling strategy for intelligent agents during task execution and improves the flexibility of the scheduling method for intelligent agents.

[0008] One possible implementation is that the above-mentioned abnormal agent generates an exception handling strategy, including: determining a second score corresponding to the abnormal agent based on the above-mentioned temporary scheduling factor and the first score corresponding to the abnormal agent, wherein the second score is the difference between the above-mentioned first score and the above-mentioned temporary scheduling factor; and determining an exception handling strategy for the abnormal agent based on the second score corresponding to the abnormal agent.

[0009] Another possible implementation involves the following operational metrics: performance metrics, behavioral metrics, and safety metrics. The performance metrics characterize the efficiency and computational resource utilization of the agent in processing subtasks; the behavioral metrics characterize the quality of the subtask processing results output by the agent; and the safety metrics characterize the number of times the agent involves the first data during subtask processing. The determination of a temporary scheduling factor for each agent based on its operational metrics includes: determining the performance metric volatility of each agent based on its performance metrics; determining the effect score of each agent in executing subtasks based on its behavioral metrics; determining the first number of times each agent involves the first data during subtask processing based on its safety metrics; and determining the temporary scheduling factor for each agent based on its performance metric volatility, effect score, and first number of times.

[0010] Another possible implementation, the above method further includes: when the first score update condition is met, updating the first score corresponding to the abnormal intelligent agent based on the above temporary scheduling factor and the above abnormal intelligent agent's operating indicators, the above first score update condition includes at least one of the following: the above target task is completed, the scheduling strategy of the above abnormal intelligent agent is adjusted, and the number of abnormal intelligent agents exceeds the second threshold.

[0011] Another possible implementation is that, based on the aforementioned temporary scheduling factor and the operational metrics of the aforementioned abnormal agent, updating the first score corresponding to the aforementioned abnormal agent includes: determining a third score based on the operational metrics of the aforementioned abnormal agent, wherein the third score is the effect score of the aforementioned abnormal agent in performing the sub-task corresponding to the aforementioned target task; and updating the first score corresponding to the aforementioned abnormal agent based on the aforementioned third score and the aforementioned temporary scheduling factor.

[0012] Secondly, embodiments of this application provide an agent scheduling device, comprising: an execution module and a determination module; the execution module is configured to execute sub-tasks in a target task through at least one agent, each agent corresponding to at least one sub-task; the determination module is configured to, during the execution of the target task by the execution module, determine a temporary scheduling factor for each agent based on the operating indicators of each agent, the temporary scheduling factor representing the degree of abnormality of the operating indicators of the agent; and, based on the temporary scheduling factor of each agent, determine an abnormal agent from the at least one agent; the generation module is configured to generate an exception handling strategy for the abnormal agent, the exception handling strategy including at least one of the following: rescheduling the agent for the sub-task corresponding to the abnormal agent, reducing the computing resources of the abnormal agent, and verifying the task execution result of the abnormal agent.

[0013] One possible implementation is that the above-mentioned generation module is specifically used to: determine a second score corresponding to the above-mentioned abnormal intelligent agent based on the above-mentioned temporary scheduling factor and the first score corresponding to the above-mentioned abnormal intelligent agent, wherein the second score is the difference between the above-mentioned first score and the above-mentioned temporary scheduling factor; and determine an abnormal handling strategy for the above-mentioned abnormal intelligent agent based on the second score corresponding to the above-mentioned abnormal intelligent agent.

[0014] Another possible implementation involves the following operational metrics: performance metrics, behavioral metrics, and safety metrics. The performance metrics characterize the efficiency and computational resource utilization of the agent in processing subtasks; the behavioral metrics characterize the quality of the subtask processing results output by the agent; and the safety metrics characterize the number of times the agent involves the first data during subtask processing. Specifically, the determining module is used to: determine the performance metric volatility of each agent based on its performance metrics; determine the effect score of each agent in executing subtasks based on its behavioral metrics; determine the first number of times each agent involves the first data during subtask processing based on its safety metrics; and determine the temporary scheduling factor for each agent based on its performance metric volatility, effect score, and first number of times.

[0015] In another possible implementation, the above device further includes: an update module; the update module is used to update the first score corresponding to the abnormal intelligent agent based on the temporary scheduling factor and the operation index of the abnormal intelligent agent when the first score update condition is met, the first score update condition includes at least one of the following: the target task is completed, the scheduling strategy of the abnormal intelligent agent is adjusted, and the number of abnormal intelligent agents exceeds the second threshold.

[0016] Another possible implementation is that the above-mentioned update module is specifically used to: determine a third score based on the above-mentioned abnormal agent's operating indicators, the third score being the effect score of the above-mentioned abnormal agent in executing the sub-task corresponding to the above-mentioned target task; and update the first score corresponding to the above-mentioned abnormal agent based on the above-mentioned third score and the above-mentioned temporary scheduling factor.

[0017] Thirdly, this application provides an electronic device comprising: a processor and a memory; the memory stores a program or instructions executable on the processor, wherein the program or instructions, when executed by the processor, implement the method of the first aspect described above.

[0018] Fourthly, this application provides a readable storage medium on which a program or instructions are stored, which, when executed by a computer, implement the method of the first aspect described above.

[0019] Fifthly, this application provides a computer program product stored in a storage medium, which, when executed by a computer, implements the method described in the first aspect.

[0020] In a sixth aspect, embodiments of this application provide a chip including a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the method described in the first aspect.

[0021] The beneficial effects of the second to sixth aspects mentioned above are described in the corresponding description of the first aspect and will not be repeated here. Attached Figure Description

[0022] Figure 1 A schematic diagram of the network architecture for an application of an agent scheduling method provided in this application embodiment;

[0023] Figure 2 A flowchart illustrating an agent scheduling method provided in an embodiment of this application;

[0024] Figure 3 A flowchart illustrating another agent scheduling method provided in an embodiment of this application;

[0025] Figure 4 A flowchart illustrating another agent scheduling method provided in this application embodiment;

[0026] Figure 5 A flowchart illustrating an agent scheduling method provided in an embodiment of this application;

[0027] Figure 6 A flowchart illustrating another agent scheduling method provided in an embodiment of this application;

[0028] Figure 7 This is a schematic diagram of the structure of an intelligent agent scheduling device provided in an embodiment of this application;

[0029] Figure 8 A schematic diagram of another intelligent agent scheduling device provided in an embodiment of this application;

[0030] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0031] The intelligent agent scheduling method, apparatus, device, medium, and program products provided in this application will now be described in detail with reference to the accompanying drawings.

[0032] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0033] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0034] The terms "at least one," "at least one of," etc., used in the specification and claims of this application refer to any one, any two, or a combination of two or more of the included items. For example, at least one of a, b, and c can mean: "a," "b," "c," "a and b," "a and c," "b and c," and "a, b, and c," where a, b, and c can be single or multiple. Similarly, "at least two" refers to two or more items, and its meaning is similar to that of "at least one."

[0035] In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0036] The embodiments of this application provide an intelligent agent scheduling method, apparatus, device, medium, and program product that can be applied to demand scheduling intelligent agent scenarios.

[0037] In related technologies, multi-agent systems mainly focus on how to collaboratively complete complex tasks, such as information extraction and specialized reasoning, but generally lack endogenous governance mechanisms for the long-term stability and security of the system itself. Existing systems suffer from problems such as static scheduling, inability to perceive agent performance degradation, and lack of dynamic optimization, resulting in system fragility and high operation and maintenance costs.

[0038] While existing multi-agent collaborative systems can allocate tasks through preset rules, workflow diagrams, or historical scores, these solutions are prone to frequent fluctuations in credibility scores if each fluctuation directly impacts the credibility score. Furthermore, evaluations are only conducted after the task is completed, making it impossible to trigger timely scheduling adjustments and anomaly handling during task execution. Thus, the scheduling methods for agents in these technologies are not flexible enough.

[0039] To address the aforementioned technical problems, this application provides an agent scheduling method, apparatus, device, medium, and program product. The method involves at least one agent executing sub-tasks within a target task, with each agent corresponding to at least one sub-task. During the execution of the target task, a temporary scheduling factor is determined for each agent based on its operational metrics. This temporary scheduling factor characterizes the degree of abnormality in the agent's operational metrics. Based on the temporary scheduling factor, an abnormal agent is identified from the at least one agent, and an exception handling strategy is generated for this abnormal agent. This exception handling strategy includes at least one of the following: rescheduling the agent for the sub-task corresponding to the abnormal agent, reducing the computational resources of the abnormal agent, and verifying the task execution result of the abnormal agent. In this solution, a temporary scheduling factor characterizing the degree of abnormality in the operational metrics of each agent is determined. Then, based on the temporary scheduling factor and the abnormal agent, a handling strategy for the abnormal agent is determined. This avoids the problem that electronic devices cannot flexibly adjust the scheduling strategy for agents during task execution, thus improving the flexibility of agent scheduling.

[0040] The following description, in conjunction with the accompanying drawings, details the intelligent agent scheduling method, apparatus, device, medium, and program products provided in the embodiments of this application.

[0041] Figure 1 This paper illustrates a network architecture for an agent scheduling method application provided in an embodiment of this application. For example... Figure 1 As shown, the network architecture includes an intelligent agent scheduling device 101 and a terminal device 102. The intelligent agent scheduling device 101 and the terminal device 102 are interconnected.

[0042] In some embodiments, the intelligent agent scheduling device 101 may be a server, a computer, or a processor or processing unit within a server or computer. The server may be a single server or a server cluster consisting of multiple servers. It should be noted that the specific device form of the intelligent agent scheduling device 101 is not limited in the embodiments of this application. Figure 1 The example shown is a single server, namely the intelligent agent scheduling device 101.

[0043] In some embodiments, the terminal device may be a mobile phone, tablet computer, laptop computer, handheld computer, in-vehicle electronic device, mobile internet device (MID), augmented reality (AR) / virtual reality (VR) device, robot, wearable device, personal computer (PC), ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc., and the embodiments of this application do not specifically limit it. Figure 1 The example shown is a mobile phone, with terminal device 102 as an example.

[0044] It should be noted that the network architecture described in the embodiments of this application is for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and does not constitute a limitation on the technical solutions provided in the embodiments of this application. As network architectures evolve, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0045] See Figure 2 This is a flowchart illustrating an agent scheduling method provided in an embodiment of this application. Figure 2 As shown, the agent scheduling method provided in this application embodiment can be implemented by the above-mentioned agent scheduling device, specifically including the following steps 201 to 203.

[0046] Step 201: The agent scheduling device executes sub-tasks in the target task through at least one agent.

[0047] In some embodiments, each agent corresponds to at least one subtask.

[0048] In some embodiments, the intelligent agent scheduling device can determine the aforementioned target task based on the task requirements input by the user.

[0049] In some embodiments, the agent scheduling device can divide the target task into multiple subtasks, and then assign a corresponding agent to each subtask to execute the subtask.

[0050] Step 202: During the execution of the target task, the agent scheduling device determines the temporary scheduling factor for each agent based on the operating indicators of each agent.

[0051] In some embodiments, the above-mentioned operating metrics include at least one of the following: performance metrics, behavioral metrics, and security metrics.

[0052] In some embodiments, the aforementioned performance indicators characterize the efficiency and computational resource utilization of the agent in processing sub-tasks, the aforementioned behavioral indicators characterize the quality of the sub-task processing results output by the agent, and the aforementioned security indicators characterize the number of times the agent involves the first data during the processing of sub-tasks.

[0053] In some embodiments, the intelligent agent scheduling device can write the collected performance indicators, behavioral indicators, and safety indicators into an operational indicator status table, and simultaneously record indicator values, update timestamps, change trends, anomaly tags, and pending settlement identifiers. For continuous indicators such as response time and resource utilization, updates are performed according to the sampling period.

[0054] In some embodiments, the aforementioned operational metrics include performance metrics, behavioral metrics, and security metrics. The performance metrics characterize the efficiency and computational resource utilization of the agent in processing subtasks; the behavioral metrics characterize the quality of the subtask processing results output by the agent; and the security metrics characterize the number of times the agent accesses the first data during the processing of subtasks. For example, in combination with... Figure 2 ,like Figure 3 As shown, step 202 above can be implemented through steps 202a and 202d.

[0055] Step 202a: The agent scheduling device determines the performance index volatility of each agent based on the performance index of each agent.

[0056] In some embodiments, the aforementioned performance metric volatility characterizes the degree of fluctuation in the performance of the agent.

[0057] Step 202b: The agent scheduling device determines the performance score of each agent in performing sub-tasks based on the behavioral indicators of each agent.

[0058] In some embodiments, the above-mentioned behavioral indicators are used to reflect the stability of the agent's output content and execution behavior.

[0059] Step 202c: The agent scheduling device determines the first number of times each agent involves the first data during the processing of subtasks based on the safety indicators of each agent.

[0060] In some embodiments, the first data mentioned above refers to sensitive data.

[0061] Step 202d: The agent scheduling device determines the temporary scheduling factor for each agent based on the performance index volatility, effect score, and first number of each agent.

[0062] In some embodiments, the intelligent agent scheduling device reads the running indicator status table during task execution and generates a temporary scheduling factor based on short-term fluctuations in the indicators. The temporary scheduling factor is used for current task migration, parallel review, rate limiting and degradation, or adjustment of subsequent task allocation ratios, but is not directly written into the long-term credibility score. The long-term credibility score is still uniformly updated by the credibility assessment entity after the settlement triggering conditions are met.

[0063] The temporary scheduling factor can be expressed as follows:

[0064] (Formula 1)

[0065] in , , These represent the standardized processing of performance indicators, behavioral indicators, and safety indicators, respectively. , , As weight, Let be the normalization function. The scheduling objective function can be further expressed as Equation 2 below:

[0066] (Formula 2)

[0067] in For long-term credibility, For resource consumption, Under load conditions As a risk penalty item, This is a temporary scheduling factor during execution.

[0068] In this way, the agent scheduling device can determine the temporary scheduling factor for each agent based on the performance index volatility, effect score, and first count of each agent. Then, based on the temporary scheduling factor and the first score corresponding to the abnormal agent, a second score representing the real-time reliability of the abnormal agent is determined. Finally, the processing strategy for the abnormal agent is determined based on the second score. This avoids the problem that electronic devices cannot flexibly adjust the scheduling strategy for agents during task execution, and improves the flexibility of the agent scheduling method.

[0069] In some embodiments, the aforementioned temporary scheduling factor characterizes the degree of abnormality in the agent's operational metrics.

[0070] Step 203: The agent scheduling device determines the abnormal agent from the above-mentioned at least one agent based on the temporary scheduling factor of each agent, and generates an abnormal handling strategy for the abnormal agent.

[0071] In some embodiments, the intelligent agent scheduling device continuously updates the above three types of indicators through a security monitoring module. For continuous indicators such as response time and resource utilization, the security monitoring module updates according to a preset sampling period; for event-type indicators such as sensitive content hit rate, the security monitoring module updates according to the occurrence of events; for behavioral indicators such as semantic offset and output length deviation, the security monitoring module updates according to task stage or output segment. Based on semantic offset, output length standardization deviation, and sensitive content hit rate, the security monitoring module constructs a comprehensive anomaly scoring model, which can be expressed as the following formula 3:

[0072] (Formula 3)

[0073] when If the threshold is exceeded, it is considered an abnormal event.

[0074] In some embodiments, combined with Figure 2 ,like Figure 4 As shown, step 203 above can be implemented through steps 203a and 203b.

[0075] Step 203a: The agent scheduling device determines the second score corresponding to the abnormal agent based on the temporary scheduling factor and the first score corresponding to the abnormal agent.

[0076] In some embodiments, the second score is the difference between the first score and the temporary scheduling factor.

[0077] In some embodiments, the second score represents the real-time reliability of the anomalous agent.

[0078] Step 203b: The agent scheduling device determines the abnormal handling strategy for the abnormal agent based on the second score corresponding to the abnormal agent.

[0079] In some embodiments, agent scheduling may determine the abnormal agent handling strategy as reducing the computational resources of the abnormal agent if the second score is greater than the third threshold and less than the fourth threshold.

[0080] In some embodiments, if the agent scheduling is greater than the second threshold and less than the fifth threshold, the abnormal handling strategy for the abnormal agent can be determined as verifying the task execution result of the abnormal agent.

[0081] In some embodiments, agent scheduling may determine the abnormal handling strategy for abnormal agents as rescheduling agents for the subtasks corresponding to the abnormal agents if the second score is greater than the fifth threshold.

[0082] In this way, the agent scheduling device can determine a temporary scheduling factor that represents the degree of abnormality of the operating indicators of each agent based on the agent's operating indicators. Then, based on the temporary scheduling factor and the first score corresponding to the abnormal agent, a second score representing the real-time reliability of the abnormal agent is determined. Finally, the processing strategy for the abnormal agent is determined based on the second score. This avoids the problem that electronic devices cannot flexibly adjust the scheduling strategy for agents during task execution, and improves the flexibility of the agent scheduling method.

[0083] In some embodiments, the above-mentioned exception handling strategy includes at least one of the following: rescheduling the agent for the subtask corresponding to the abnormal agent, reducing the computing resources of the abnormal agent, and verifying the task execution result of the abnormal agent.

[0084] This application provides an agent scheduling method. Based on the agent's operating indicators, a temporary scheduling factor is determined to characterize the degree of abnormality of each agent's operating indicators. Then, based on the temporary scheduling factor and the abnormal agent, a handling strategy for the abnormal agent is determined. This avoids the problem that electronic devices cannot flexibly adjust the scheduling strategy for agents during task execution, and improves the flexibility of agent scheduling.

[0085] In some embodiments, the agent scheduling method provided in this application may further include the following step 204.

[0086] Step 204: When the first score update condition is met, the agent scheduling device updates the first score corresponding to the abnormal agent based on the temporary scheduling factor and the operating indicators of the abnormal agent.

[0087] In some embodiments, the first score update condition includes at least one of the following: the target task is completed, the scheduling strategy of the abnormal agent is adjusted, or the number of abnormal agents exceeds the second threshold.

[0088] In some embodiments, the first score mentioned above may also be referred to as the long-term credibility score.

[0089] In some embodiments, the agent scheduling device integrates the execution results of one or more agents to generate a final response and present it to the user. The system simultaneously collects explicit and implicit feedback. Explicit feedback includes direct user evaluations, while implicit feedback includes performance indicator change records, behavioral indicator change records, security indicator change records, and anomaly analysis results generated by the security monitoring module. The system reports the task execution results, user feedback, indicator change records, and anomaly handling results to the trust assessment entity. The trust assessment entity first determines whether the trustworthiness settlement trigger conditions are met. Trigger conditions include task phase completion, task execution end, anomaly event handling completion, consecutive anomaly count exceeding a threshold, or user / human feedback being returned.

[0090] If the credibility settlement trigger condition is not met, the credibility assessment entity will only save the indicator change record and the pending settlement identifier, and will not update the credibility score; if the credibility settlement trigger condition is met, the credibility assessment entity will calculate and update the agent's long-term credibility score based on the task execution results, result quality, user feedback, indicator change records and anomaly handling results within the settlement period.

[0091] In this way, the agent scheduling device can update the first score corresponding to the abnormal agent only when the first score update condition is met, based on the temporary scheduling factor and the operating index of the abnormal agent, thus avoiding the waste of computing resources and system instability caused by frequent updates of the first score corresponding to the agent.

[0092] In some embodiments, step 204 can be implemented through steps 204a and 204b.

[0093] Step 204a: Under the condition that the first score update condition is met, the intelligent agent scheduling device determines the third score based on the above-mentioned abnormal intelligent agent's operating indicators.

[0094] In some embodiments, the third score is the effect score of the abnormal agent performing the sub-task corresponding to the target task.

[0095] In some embodiments, the third score is used to quantify the effectiveness of the abnormal agent in performing the sub-tasks corresponding to the target task.

[0096] Step 204a: The agent scheduling device updates the first score corresponding to the abnormal agent based on the third score and the temporary scheduling factor.

[0097] In some embodiments, the agent scheduling device may subtract the temporary scheduling factor from the first score and then add the third score to the resulting value to update the first score corresponding to the abnormal agent.

[0098] In this way, the agent scheduling device can update the first score of abnormal agents based on the temporary scheduling factor and the first score, thereby reducing the probability of subsequent abnormal agents being scheduled and improving the efficiency of scheduling agents to perform tasks.

[0099] The methods of the embodiments of this application will be described below through specific examples.

[0100] like Figure 5 As shown, the implementation process of the method provided in this application embodiment includes the following S1 to S6:

[0101] S1. Task Request and Intelligent Scheduling Decision. The master agent receives the request, and the dynamic scheduling strategy engine queries the long-term credibility score to determine the target agent.

[0102] In some embodiments, a user submits a business request to the master agent. Upon receiving the request, the master agent does not directly route it to a specific functional agent, but instead initiates a scheduling consultation with the dynamic scheduling strategy engine. The dynamic scheduling strategy engine then queries the trust assessment entity for a list of dynamic trust scores for all currently available functional agents. The trust assessment entity returns this list (e.g., knowledge-based question-answering agent: 95 points, process-based case agent: 85 points). The dynamic scheduling strategy engine determines the optimal target functional agent based on a preset strategy (e.g., selecting the highest score) and returns its ID to the master agent.

[0103] S2. Task Execution and Parallel Safety Monitoring. The target agent executes tasks, and the safety monitoring module continuously collects performance, behavioral, and safety metrics.

[0104] In some embodiments, the master control agent assigns specific subtasks to the target functional agent selected in step 1. The functional agent then begins executing the task. Simultaneously, a security monitoring module is asynchronously activated, continuously monitoring the operational behavior of the functional agent and collecting performance metrics, behavioral metrics, and security metrics throughout the task execution process. Performance metrics reflect the agent's response efficiency and resource consumption; behavioral metrics reflect the stability of the agent's output and execution behavior; and security metrics reflect security risks during the agent's execution.

[0105] S3. Write to the operation indicator status table. Update indicator values, timestamps, trends, anomaly tags, and pending settlement identifiers according to sampling period, task stage, output segment, or abnormal event.

[0106] In some embodiments, during task execution, the security monitoring module writes the collected performance metrics, behavioral metrics, and security metrics into the running metric status table, and synchronously records the metric values, update timestamps, change trends, anomaly tags, and pending settlement identifiers. For continuous metrics such as response time and resource utilization, updates are made according to the sampling period; for event-based metrics such as sensitive content hits, updates are made according to the occurrence of the event; and for behavioral metrics such as semantic offset and output length deviation, updates are made according to the task stage or output segment. This stage forms an execution process metric status stream, mainly used to record and reflect the current running state of the agent, and does not directly trigger long-term credibility score updates.

[0107] S4. In-process scheduling and exception handling. The dynamic scheduling strategy engine reads short-term indicator status to generate temporary scheduling factors; the handling and coordination module performs rate limiting, degradation, review, isolation, or migration.

[0108] In some embodiments, the dynamic scheduling strategy engine reads the runtime indicator status table, generates a temporary scheduling factor based on the short-term indicator status, and adjusts the allocation ratio of the current or subsequent tasks, triggers other agents for review, switches the execution object, or performs task migration accordingly. The handling and coordination module performs handling actions such as rate limiting, degradation, review, isolation, or stopping the allocation of new tasks based on the anomaly label and anomaly type, and writes the handling results to the anomaly handling record. This stage uses indicator status to serve in-process scheduling and anomaly handling, but the temporary scheduling factor is not directly written to the long-term credibility score, thereby avoiding frequent changes in credibility due to single indicator fluctuations.

[0109] S5. Determine the credibility settlement trigger conditions. Trigger conditions include stage completion, task end, abnormal loop completion, continuous abnormal exceeding the threshold, or user or manual feedback.

[0110] In some embodiments, if the settlement conditions are not met, only the indicator change record and the pending settlement identifier are saved, and the credibility score is not updated. If the settlement conditions are met, the indicator change record, task results, user feedback, and handling results within the settlement period are read, and the long-term credibility is updated.

[0111] In some embodiments, the target functional agent returns the subtask execution results to the master agent. The master agent integrates the execution results of one or more agents, generates a final response, and presents it to the user. The system collects both explicit and implicit feedback. Explicit feedback includes direct user evaluations, while implicit feedback includes performance indicator change records, behavioral indicator change records, security indicator change records, and anomaly analysis results generated by the security monitoring module.

[0112] In some embodiments, the master control agent, the security monitoring module, and the handling and coordination module report the task execution results, user feedback, indicator change records, and anomaly handling results to the trust assessment entity, respectively. The trust assessment entity first determines whether the trust settlement trigger conditions are met. The trigger conditions include task phase completion, task execution end, anomaly event handling completion, consecutive anomaly count exceeding a threshold, or user / human feedback return.

[0113] In some embodiments, if the credibility settlement trigger condition is not met, the credibility assessment entity only saves the indicator change record and the pending settlement identifier, and does not update the credibility score; if the credibility settlement trigger condition is met, the credibility assessment entity calculates and updates the long-term credibility score of the target functional agent based on the task execution results, result quality, user feedback, indicator change records and anomaly handling results within the settlement period.

[0114] It is understandable that this method allows the system to distinguish between indicator status updates and credibility score updates during the execution process, making credibility updates more stable while retaining the real-time scheduling capability during task execution.

[0115] S6. Closed-loop self-optimization. The credibility update event is fed back to the dynamic scheduling strategy engine to optimize subsequent task allocation strategies.

[0116] In some embodiments, after completing a long-term trustworthiness update, the trust assessment entity will update the event notification to the dynamic scheduling strategy engine. The dynamic scheduling strategy engine adjusts its internal scheduling strategy based on the latest global trustworthiness status and the indicator status records generated during execution, forming a closed loop of "monitoring—indicator status update—in-process scheduling and handling—trustworthiness settlement—strategy optimization." When the next user request arrives, the system allocates tasks based on the updated trustworthiness information and the optimized scheduling strategy, thereby achieving continuous self-optimization.

[0117] Thus, the agent scheduling method proposed in this application monitors agent behavior in real time, rapidly isolates and coordinates anomalies, prevents the spread of faults in individual agents, and ensures the continuity of the overall system service. Scheduling is based on both long-term trust scores and short-term performance indicators, prioritizing high-value tasks for stable and trustworthy agents. When performance indicators become abnormal, timely review, migration, or degradation is triggered, thereby improving overall system performance and user experience. A quantitative trust evaluation system is established for agents, and key outputs are stored on the blockchain, upgrading the agent system from a tool to a trusted business partner.

[0118] It should be noted that the descriptions of each step S1 to S6 in this embodiment can be found in the descriptions in the above embodiments, and will not be repeated here.

[0119] The agent scheduling method proposed in this application will be described in detail below.

[0120] The purpose of this application is to provide an intelligent agent collaborative management method and system that separates and coordinates the updating of execution process indicators, the periodic settlement update of credibility, the dynamic scheduling adjustment and the handling of anomalies, so that the system can respond to changes in performance, behavior and security indicators during task execution, and avoid frequent fluctuations in credibility score due to single indicator fluctuations.

[0121] For example, such as Figure 6 As shown, this application proposes an intelligent agent collaborative management method and system based on dynamic trust assessment. By constructing an independent intelligent agent governance layer outside of the functional intelligent agents, it realizes real-time monitoring, quantitative assessment, dynamic scheduling, and anomaly handling of the multi-agent system operation process, thereby forming a closed-loop management mechanism. The system as a whole adopts a layered architecture design, consisting of a master intelligent agent, a functional intelligent agent layer, and an intelligent agent governance layer. The modules interact with each other through service interfaces and message queues to ensure the scalability and low coupling of the system.

[0122] During task execution, the safety monitoring module continuously collects and updates the agent's performance, behavioral, and safety metrics. These metrics do not immediately trigger a trust score update with every change; instead, they are first used as state data during operation to support the dynamic scheduling engine in adjusting scheduling and to support the handling and coordination modules in handling anomalies. The trust assessment entity updates the agent's trustworthiness based on changes in the aforementioned metrics, task execution results, and feedback information after the task phase is completed, task execution ends, or anomaly handling is finished.

[0123] In this way, the present invention distinguishes between indicator updates and credibility updates during the execution process, enabling the system to adjust the scheduling or handling strategy in a timely manner according to indicator changes during task execution, while avoiding frequent changes in credibility score due to single indicator fluctuations.

[0124] Furthermore, the security monitoring module maintains an operational indicator status table for each functional agent. This table includes at least performance indicator fields, behavioral indicator fields, security indicator fields, indicator update timestamps, indicator change trends, anomaly tags, and pending settlement identifiers. The operational indicator status table is used to generate an indicator status stream during execution. This stream can be read in real-time by the dynamic scheduling strategy engine and the handling and coordination module; however, changes in this stream do not necessarily lead to an immediate update of the credibility score.

[0125] The operational indicator state vector can be represented by the following formula 4:

[0126] (Formula 4)

[0127] in Let i represent the set of performance metrics for the i-th agent at time t. Represents a set of behavioral indicators. Represents a set of safety indicators. Indicates the most recent update timestamp. Indicates the trend of indicator changes. Indicates an exception label. This indicates a pending settlement flag. This state vector is continuously updated during task execution, but it is not equivalent to an update of the credibility score.

[0128] Furthermore, the trust assessment entity maintains a trust settlement flow independent of the indicator status flow. The trust settlement flow is initiated only when preset settlement trigger conditions are met. These trigger conditions include task phase completion, task execution completion, abnormal event handling completion, consecutive abnormal occurrences exceeding a threshold, or feedback from humans / users. This forms a dual-flow processing mechanism of "in-process indicator status update" and "phased trust settlement update."

[0129] During system operation, the master control agent acts as a unified entry point to receive user requests and sends scheduling decision requests to the dynamic scheduling strategy engine. The scheduling strategy engine invokes the trust assessment entity to obtain the trust scores of each functional agent and selects the optimal execution object based on a preset optimization objective function. The objective function is defined as shown in Formula 5 below:

[0130] (Formula 5)

[0131] in This represents the credibility of the i-th agent. Indicates resource consumption. Indicates the current load status. These are weight parameters. Scheduling strategies include not only selection mechanisms based on maximum reliability, but also load balancing strategies based on probability allocation and those based on... Exploring and utilizing strategies for greedy methods, when the random value is less than If a random selection is performed, the agent with the largest objective function is selected; otherwise, the agent with the largest objective function is selected.

[0132] The dynamic scheduling strategy engine reads the running indicator status table during task execution and generates temporary scheduling factors based on short-term fluctuations in the indicators. These temporary scheduling factors are used for current task migration, parallel review, rate limiting and degradation, or adjustment of subsequent task allocation ratios, but are not directly written into the long-term credibility score. The long-term credibility score is still uniformly updated by the credibility assessment entity after the settlement triggering conditions are met.

[0133] The temporary scheduling factor can be expressed as the following formula 6:

[0134] (Formula 6)

[0135] in , , These represent the standardized processing of performance indicators, behavioral indicators, and safety indicators, respectively. , , As weight, This is the normalization function. The scheduling objective function can be further expressed as shown in Formula 7 below:

[0136] (Formula 7)

[0137] in For long-term credibility, For resource consumption, Under load conditions As a risk penalty item, This is a temporary scheduling factor during execution.

[0138] During task execution, the safety monitoring module performs parallel monitoring of the target agent. Monitoring includes performance metrics, behavioral metrics, and safety metrics. Performance metrics include response time, timeout rate, resource utilization, and task throughput, reflecting the agent's processing efficiency and resource consumption during execution. Behavioral metrics calculate stability using semantic vectors, with the semantic offset defined by the following formula:

[0139] (Formula 8)

[0140] Simultaneously, statistical modeling is performed on the output length, and the standardized deviation is calculated. The security metric is obtained by detecting sensitive content to determine the sensitivity hit rate. The security monitoring module continuously updates the above three types of indicators. For continuous indicators such as response time and resource utilization, the security monitoring module updates according to a preset sampling period; for event-based indicators such as sensitive content hit rate, the security monitoring module updates according to the occurrence of the event; for behavioral indicators such as semantic offset and output length deviation, the security monitoring module updates according to the task stage or output segment. Based on semantic offset, output length standardization deviation, and sensitive content hit rate, the security monitoring module constructs a comprehensive anomaly scoring model, which can be expressed as the following formula 9:

[0141] (Formula 9)

[0142] when If the threshold is exceeded, it is considered an abnormal event.

[0143] The security monitoring module generates indicator change records simultaneously when updating the above indicators. For continuous indicators, it records the sampled value, moving average, variance, and trend direction; for event-based indicators, it records the event type, occurrence time, number of hits, and risk level; for behavioral indicators, it records the task stage, output segment, semantic offset, and length deviation. These records are written to the running indicator status table and archived to the credibility settlement stream according to the pending settlement identifier.

[0144] The credibility assessment entity quantifies the agent based on multi-dimensional indicators, and the credibility is defined by the following formula 10:

[0145] (Formula 10)

[0146] in Let k be the evaluation index function. These correspond to the weights. Evaluation metrics include task success rate. Results quality score and abnormality penalty items The overall credibility is calculated as follows:

[0147] During trustworthiness updates, the trust assessment entity does not simply read a single indicator value, but rather reads the indicator change records, task results, user feedback, and handling results within the settlement period to form the settlement period evaluation quantity. For indicator changes that are short-term fluctuations but do not form a sustained anomaly, they only affect the temporary scheduling factor; only for sustained anomalies, major security incidents, or handling failures, they are included in the trustworthiness update as anomaly penalty items.

[0148] The phased credibility settlement update can be expressed as the following formula 11:

[0149] (Formula 11)

[0150] in This represents the long-term reliability after the end of the nth settlement period. This represents the comprehensive evaluation metric formed by task success rate, result quality, user feedback, and metric change records within the settlement period, where H_i(n) represents the anomaly penalty. To preserve the historical credibility coefficient, This represents the exception penalty coefficient. This formula is only executed when the settlement trigger condition is met; during task execution, the indicator state vector... It can be updated multiple times. It remains unchanged.

[0151] (Formula 12)

[0152] To reflect the dynamic changes in the system, a weight update mechanism and a time decay mechanism are introduced. The weight update employs an exponential smoothing method. ,in For smoothing coefficients, This represents the contribution of the current cycle indicator. Historical scores are updated using a time decay function, expressed as Formula 13 below:

[0153] (Formula 13)

[0154] When the monitoring module detects an abnormal event, the handling and coordination module triggers a response mechanism. The system handles the abnormal agent according to the type and severity of the abnormality; when the abnormality reaches the isolation condition, its state is marked as unschedulable, and its credibility is attenuated using the following formula 14:

[0155] (Formula 14)

[0156] The scheduling policy engine then re-executes the scheduling decision, assigning the task to other available agents for continued execution. Simultaneously, the system records abnormal events, generating immutable audit logs for subsequent tracing and analysis.

[0157] The entire system achieves continuous optimization through a cyclical mechanism of "monitoring-evaluation-scheduling-handling". When a new task arrives, the system makes scheduling decisions based on the latest updated credibility information, thereby achieving dynamic optimization of the performance and security of the multi-agent system.

[0158] It should be noted that the various method embodiments described above, or the various possible implementations of the various method embodiments, can be executed individually, or, provided there is no conflict, they can be combined with each other. The specific implementation can be determined according to actual usage requirements, and this application embodiment does not impose any limitations on this. It can be seen that the above mainly introduces the solutions provided by the embodiments of this application from a methodological perspective. To achieve the above functions, this application embodiment provides corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the modules and algorithm steps of the various examples described in the embodiments disclosed herein, this application embodiment can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed by hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0159] This application embodiment can divide the intelligent agent scheduling device into functional modules according to the above method example. For example, each function can be divided into its own functional module, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module. Optionally, the module division in this application embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0160] In some embodiments, this application also provides an agent scheduling device. The agent scheduling device may include one or more functional modules for implementing the agent scheduling method of the above method embodiments.

[0161] For example, Figure 7 This is a schematic diagram of the structure of an intelligent agent scheduling device provided in an embodiment of this application. Figure 7 As shown, the intelligent agent scheduling device 900 includes an execution module 901 and a determination module 902.

[0162] The execution module 901 is configured to execute sub-tasks in the target task through at least one agent, with each agent corresponding to at least one sub-task. The determination module 902 is configured to, during the execution of the target task by the execution module 901, determine a temporary scheduling factor for each agent based on the operating indicators of each agent, wherein the temporary scheduling factor characterizes the degree of abnormality of the operating indicators of the agent; and, based on the temporary scheduling factor of each agent, determine an abnormal agent from the at least one agent. The generation module is configured to generate an exception handling strategy for the abnormal agent, wherein the exception handling strategy includes at least one of the following: rescheduling the agent for the sub-task corresponding to the abnormal agent, reducing the computing resources of the abnormal agent, and verifying the task execution result of the abnormal agent.

[0163] The intelligent agent scheduling device provided in this application determines a temporary scheduling factor that characterizes the degree of abnormality of the operating indicators of each intelligent agent based on the operating indicators of the intelligent agent. Then, based on the temporary scheduling factor and the abnormal intelligent agent, a processing strategy for the abnormal intelligent agent is determined. This avoids the problem that electronic devices cannot flexibly adjust the scheduling strategy for intelligent agents during task execution and improves the flexibility of the scheduling method for intelligent agents.

[0164] One possible implementation is that the above-mentioned generation module is specifically used to: determine a second score corresponding to the above-mentioned abnormal intelligent agent based on the above-mentioned temporary scheduling factor and the first score corresponding to the above-mentioned abnormal intelligent agent, wherein the second score is the difference between the above-mentioned first score and the above-mentioned temporary scheduling factor; and determine an abnormal handling strategy for the above-mentioned abnormal intelligent agent based on the second score corresponding to the above-mentioned abnormal intelligent agent.

[0165] Another possible implementation is that the aforementioned operational metrics include performance metrics, behavioral metrics, and safety metrics. The performance metrics characterize the efficiency and computational resource utilization of the agent in processing subtasks, the behavioral metrics characterize the quality of the subtask processing results output by the agent, and the safety metrics characterize the number of times the agent involves the first data during the processing of subtasks. The determining module 902 is specifically used for: determining the performance metric volatility of each agent based on its performance metrics; determining the effect score of each agent in executing subtasks based on its behavioral metrics; determining the first number of times each agent involves the first data during the processing of subtasks based on its safety metrics; and determining the temporary scheduling factor of each agent based on its performance metric volatility, effect score, and first number of times.

[0166] Another possible implementation method, combined with Figure 7 ,like Figure 8 As shown, the above-mentioned intelligent agent scheduling device 900 further includes: an update module 903; the update module 903 is used to update the first score corresponding to the abnormal intelligent agent based on the above-mentioned temporary scheduling factor and the operation index of the abnormal intelligent agent when the first score update condition is met. The first score update condition includes at least one of the following: the target task is completed, the scheduling strategy of the abnormal intelligent agent is adjusted, and the number of abnormal intelligent agents exceeds the second threshold.

[0167] Another possible implementation is that the update module 903 is specifically used to: determine a third score based on the operation indicators of the abnormal agent, the third score being the effect score of the abnormal agent in performing the sub-task corresponding to the target task; and update the first score corresponding to the abnormal agent based on the third score and the temporary scheduling factor.

[0168] It should be noted that the intelligent agent scheduling device can implement all the processes implemented in the above method embodiments and achieve the same beneficial effects. To avoid repetition, it will not be described again here.

[0169] In the case where the functions of the integrated modules described above are implemented in hardware, this application provides a possible structural schematic diagram of the electronic device involved in the above embodiments. For example... Figure 9 As shown, the electronic device 90 includes: a processor 92, a communication interface 93, and a bus 94. Optionally, the electronic device 90 may also include a memory 91.

[0170] Processor 92 may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 92 may be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 92 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0171] Communication interface 93 is used to connect with other devices via a communication network. This communication network can be Ethernet, wireless access network, wireless local area network (WLAN), etc.

[0172] The memory 91 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.

[0173] As one possible implementation, the memory 91 can exist independently of the processor 92. The memory 91 can be connected to the processor 92 via a bus 94 and is used to store instructions or program code. When the processor 92 calls and executes the instructions or program code stored in the memory 91, it can implement the intelligent agent scheduling method provided in the embodiments of this application.

[0174] In another possible implementation, memory 91 can also be integrated with processor 92.

[0175] Bus 94 can be an Extended Industry Standard Architecture (EISA) bus, etc. Bus 94 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0176] Through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the service calling device can be divided into different functional modules to complete all or part of the functions described above.

[0177] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described intelligent agent scheduling method embodiment and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0178] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0179] This application also provides a readable storage medium storing a program or instructions that, when executed by a computer, implement the intelligent agent scheduling method provided in the above embodiments. It is understood that all or part of the processes in the above method embodiments can be executed by computer instructions instructing related hardware; the readable storage medium can be any of the foregoing embodiments or memory; the readable storage medium can also be an external storage device of the service invocation device, such as a pluggable hard drive, Smart MediaCard (SMC), Secure Digital (SD) card, flash card, etc., equipped on the service invocation device. Further, the readable storage medium can include both internal storage units of the service invocation device and external storage devices. The readable storage medium is used to store the computer program and other programs and data required by the service invocation device. The readable storage medium can also be used to temporarily store data that has been output or will be output.

[0180] This application also provides a computer program product, which is stored in a storage medium and implements the agent scheduling method provided in the above embodiments when the computer program product is executed by a computer.

[0181] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0182] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0183] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A method for scheduling intelligent agents, characterized in that, include: Each agent executes a subtask in the target task, with each agent corresponding to at least one subtask. During the execution of the target task, a temporary scheduling factor is determined for each agent based on the operational metrics of each agent. The temporary scheduling factor characterizes the degree of abnormality in the operational metrics of the agent. Based on the temporary scheduling factor of each agent, an abnormal agent is determined from the at least one agent, and an exception handling strategy is generated for the abnormal agent. The exception handling strategy includes at least one of the following: rescheduling the agent for the subtask corresponding to the abnormal agent, reducing the computing resources of the abnormal agent, and verifying the task execution result of the abnormal agent.

2. The agent scheduling method according to claim 1, characterized in that, The generation of anomaly handling strategies for the anomalous agent includes: Based on the temporary scheduling factor and the first score corresponding to the abnormal agent, a second score corresponding to the abnormal agent is determined, and the second score is the difference between the first score and the temporary scheduling factor. Based on the second score corresponding to the abnormal agent, an anomaly handling strategy for the abnormal agent is determined.

3. The agent scheduling method according to claim 1, characterized in that, The operational metrics include performance metrics, behavioral metrics, and security metrics. The performance metrics characterize the efficiency of the agent in processing sub-tasks and the utilization rate of computing resources. The behavioral metrics characterize the quality of the sub-task processing results output by the agent. The security metrics characterize the number of times the agent involves the first data during the processing of sub-tasks. The determination of the temporary scheduling factor for each agent based on its operational metrics includes: Based on the performance metrics of each agent, determine the volatility of the performance metrics of each agent. Based on the behavioral metrics of each agent, determine the performance score of each agent in performing sub-tasks; Based on the security metrics of each agent, determine the first number of times each agent involves the first data during the processing of subtasks; Based on the performance metrics volatility, effect score, and first-time count of each agent, a temporary scheduling factor is determined for each agent.

4. The agent scheduling method according to claim 1, characterized in that, The method further includes: If the first score update condition is met, the first score corresponding to the abnormal agent is updated based on the temporary scheduling factor and the operation index of the abnormal agent. The first score update condition includes at least one of the following: the target task is completed, the scheduling strategy of the abnormal agent is adjusted, and the number of abnormal agents exceeds the second threshold.

5. The agent scheduling method according to claim 1, characterized in that, The step of updating the first score corresponding to the abnormal agent based on the temporary scheduling factor and the operating indicators of the abnormal agent includes: Based on the operational metrics of the abnormal agent, a third score is determined, which is the effect score of the abnormal agent in executing the sub-task corresponding to the target task. Based on the third score and the temporary scheduling factor, update the first score corresponding to the abnormal agent.

6. A smart agent scheduling device, characterized in that, include: Execution module and determination module; The execution module is used to execute sub-tasks in the target task through at least one intelligent agent, with each intelligent agent corresponding to at least one sub-task. The determining module is used to determine a temporary scheduling factor for each intelligent agent based on the operating indicators of each intelligent agent during the execution of the target task by the execution module. The temporary scheduling factor characterizes the degree of abnormality of the operating indicators of the intelligent agent. Furthermore, based on the temporary scheduling factor for each agent, abnormal agents are identified from the at least one agent; The generation module is used to generate an exception handling strategy for the abnormal agent determined by the determining module. The exception handling strategy includes at least one of the following: rescheduling the agent for the subtask corresponding to the abnormal agent, reducing the computing resources of the abnormal agent, and verifying the task execution result of the abnormal agent.

7. The intelligent agent scheduling device according to claim 6, characterized in that, The generation module is specifically used for: Based on the temporary scheduling factor and the first score corresponding to the abnormal agent, a second score corresponding to the abnormal agent is determined, and the second score is the difference between the first score and the temporary scheduling factor. Based on the second score corresponding to the abnormal agent, an anomaly handling strategy for the abnormal agent is determined.

8. The intelligent agent scheduling device according to claim 6, characterized in that, The operational metrics include performance metrics, behavioral metrics, and security metrics. The performance metrics characterize the efficiency of the agent in processing sub-tasks and the utilization rate of computing resources. The behavioral metrics characterize the quality of the sub-task processing results output by the agent. The security metrics characterize the number of times the agent involves the first data during the processing of sub-tasks. The determining module is specifically used for: Based on the performance metrics of each agent, determine the volatility of the performance metrics of each agent. Based on the behavioral metrics of each agent, determine the performance score of each agent in performing sub-tasks; Based on the security metrics of each agent, determine the first number of times each agent involves the first data during the processing of subtasks; Based on the performance metrics volatility, effect score, and first-time count of each agent, a temporary scheduling factor is determined for each agent.

9. The intelligent agent scheduling device according to claim 6, characterized in that, The device further includes: an update module; The update module is used to update the first score corresponding to the abnormal agent based on the temporary scheduling factor and the operation index of the abnormal agent when the first score update condition is met. The first score update condition includes at least one of the following: the target task is completed, the scheduling strategy of the abnormal agent is adjusted, and the number of abnormal agents exceeds the second threshold.

10. The intelligent agent scheduling device according to claim 6, characterized in that, The update module is specifically used for: Based on the operational metrics of the abnormal agent, a third score is determined, which is the effect score of the abnormal agent in executing the sub-task corresponding to the target task. Based on the third score and the temporary scheduling factor, update the first score corresponding to the abnormal agent.

11. An electronic device, characterized in that, It includes a processor and a memory, the memory storing programs or instructions that can run on the processor, the programs or instructions being executed by the processor to implement the agent scheduling method as described in any one of claims 1 to 5.

12. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a computer, implement the agent scheduling method as described in any one of claims 1 to 5.

13. A computer program product, characterized in that, The computer program product is stored in a storage medium, and when executed by a computer, the computer program product implements the agent scheduling method as described in any one of claims 1 to 5.