Multi-agent cooperative control method based on large model instruction scheduling

Abstract

The invention relates to the technical field of multi-agent cooperative control, in particular to a multi-agent cooperative control method based on large-model instruction scheduling, which comprises the following steps of: analyzing task load fluctuation in time sequence data based on real-time data acquisition and task state recording of multiple agents; and performing comparison processing on the trend change rate and the task chain state characteristics to generate a task execution load trend fluctuation value. In the invention, the resource allocation imbalance between the tasks is solved through the real-time weight calculation of the resource allocation rule, the resource scheduling balance is improved, the adaptability of task execution is improved through the dynamic allocation of the task chain and the optimized distribution of the nodes, the execution conflict is reduced, and the execution efficiency is improved. Real-time adjustment and optimization of time delay and resource competition enhance the stability of the cooperative relationship and the task consistency, a multi-round resource income evaluation and optimization mechanism shortens the task period, and the global cooperative consistency and the task execution efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of multi-agent cooperative control technology, and in particular to a multi-agent cooperative control method based on large model instruction scheduling. Background Technology

[0002] The field of multi-agent cooperative control technology encompasses methods and mechanisms for distributed control of complex system tasks through interaction and collaboration among multiple agents. The core components of this technology include task decomposition of multi-agent systems, distributed decision-making mechanisms, communication and coordination methods between agents, and global system optimization strategies. The overall technology involves the development of algorithm- and model-based multi-agent cooperative control frameworks, covering techniques such as task scheduling, path planning, resource allocation, and real-time dynamic adjustment to address the needs of complex scenarios with multiple objectives, constraints, and environments.

[0003] Among them, the multi-agent cooperative control method based on large-scale model instruction scheduling refers to designing control mechanisms and methods suitable for multi-agent systems by utilizing the instruction generation capabilities of large-scale pre-trained models. This patent covers task allocation methods based on agent behavior analysis, dynamic task scheduling strategies combining large-scale model inference results, and methods for generating multi-agent communication protocols through specific model rules. Specifically, it focuses on establishing an efficient multi-agent task cooperative working mode by combining the high-dimensional data features output by large models and employing distributed task partitioning and resource scheduling schemes, thereby avoiding internal system conflicts or resource competition.

[0004] Existing technologies fall short in dynamic monitoring and response to task chains, relying heavily on fixed task chain structures and failing to adapt quickly to dynamic changes, easily leading to delayed or interrupted task responses. Resource allocation mechanisms lack real-time weight analysis capabilities, failing to dynamically adjust resource distribution according to different task requirements, easily resulting in resource overload or idleness, and reduced resource utilization. The establishment of agent collaboration relationships lacks adequate handling of time delays and resource contention issues, easily causing task execution conflicts and decreased collaboration efficiency. The allocation process of resource benefits fails to incorporate dynamic quantitative evaluation and real-time adjustment, unable to adapt to multi-objective requirements in complex task scenarios, leading to extended task completion times and insufficient system collaboration consistency. These problems restrict the global optimization capabilities and task completion efficiency of multi-agent collaborative systems in complex environments. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a multi-agent cooperative control method based on large model instruction scheduling.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a multi-agent cooperative control method based on large model instruction scheduling, comprising the following steps: S1: Based on real-time data acquisition and task status recording of multiple agents, analyze the task load fluctuation in time series data, and generate task execution load trend fluctuation value by comparing the trend change rate with the task chain state characteristics. S2: Based on the task execution load trend fluctuation value, analyze the agent's resource usage, evaluate it through resource differences and task processing capacity fluctuations, formulate resource allocation rules for load distribution, and generate resource allocation priority weight values; S3: Based on the resource allocation priority weight value, decompose the sub-task nodes in the complex task chain, perform dynamic allocation on the sub-tasks, adjust the distribution of task nodes, and generate a sub-task distribution balance mapping matrix. S4: Based on the subtask distribution equilibrium mapping matrix, analyze the agent cooperation relationship, adjust the time delay and resource contention contradiction in the task chain, optimize the cooperation network relationship, and generate a cooperation node relationship weight matrix. S5: Based on the collaborative node relationship weight matrix, perform multiple rounds of resource revenue quantitative evaluation, analyze the fluctuation of node resource revenue distribution, adjust the resource allocation scheme, and generate a resource allocation revenue optimization index matrix; S6: Based on the resource allocation benefit optimization index matrix, update the collaborative node path execution scheme, analyze the task completion time reduction rate and consistency status, establish collaborative consistency results, and generate global collaborative consistency status index values.

[0007] As a further aspect of the present invention, the task execution load trend fluctuation value includes the task load change amplitude, the task chain state change trend, and the time series fluctuation coefficient; the resource allocation priority weight value includes resource usage weight, task processing capacity weight, and resource distribution balance weight; the sub-task distribution balance mapping matrix includes sub-task node priority distribution, agent load balance state, and dynamic task node distribution adjustment; the collaborative node relationship weight matrix includes time delay relationship weight, resource conflict coordination weight, and node collaboration optimization weight; the resource allocation benefit optimization index matrix includes resource benefit deviation value, resource benefit fluctuation coefficient, and resource allocation optimization weight; and the global collaboration consistency state index value includes task completion time reduction index, agent collaboration consistency index, and global collaboration state evaluation value.

[0008] As a further aspect of the present invention, based on real-time data acquisition and task status recording by multiple agents, the specific steps for analyzing task load fluctuations in time-series data and generating task execution load trend fluctuation values ​​by comparing the trend change rate with task chain state characteristics are as follows: S101: Based on real-time data acquisition and task status recording of multiple agents, the task status data is segmented according to the time series, each segment of task load data is extracted, the start and end positions of the data are marked according to the time interval, the change amplitude of each segment of load data is calculated, the calculation results are stored as a data sequence, and task load time segment data is generated. S102: Based on the task load time segmentation data, perform correlation analysis on the load change amplitude according to time period, calculate the time difference value of the change amplitude, compare the change amplitude with the difference trend of adjacent time periods, filter the interval data where the load fluctuation value is greater than the change threshold, and extract relevant features in combination with task chain status data to generate task chain status fluctuation feature data. S103: Based on the task chain state fluctuation characteristic data, the fluctuation characteristic values ​​are sorted by time series, and the corresponding load fluctuation intervals are matched according to the task chain state. The time overlap between the task chain state characteristics and the load fluctuation is analyzed. The fluctuation trend parameter is calculated by the trend change difference value to obtain the task execution load trend fluctuation value.

[0009] As a further aspect of the present invention, based on the task execution load trend fluctuation value, the resource usage of the intelligent agent is analyzed, and an assessment is conducted through resource differences and task processing capacity fluctuations. The specific steps for formulating resource allocation rules for load distribution and generating resource allocation priority weight values ​​are as follows: S201: Based on the task execution load trend fluctuation value, statistically analyze the resource usage data of the agent during task execution, decompose the resource usage type and quantity corresponding to the task load, classify and organize them, calculate the corresponding ratio value of task load and resource usage, analyze the impact of task load changes on resource usage, and generate resource usage and task load correlation data. S202: Based on the resource usage and task load correlation data, the resource usage differences in the task load are statistically segmented, and the resource differences are analyzed in time segments in combination with the task processing capacity fluctuation data. The correlation between the resource difference and the task processing capacity fluctuation is calculated, the relevant resource types and corresponding tasks are screened, and resource allocation rule data is generated. S203: Based on the resource allocation rule data, sort the resource types according to task type and resource usage efficiency, calculate the correspondence between resource allocation priority and task type, compare the resource type sorting with the task allocation requirements, allocate the resource weight required for the task and mark the priority, and obtain the resource allocation priority weight value.

[0010] As a further aspect of the present invention, based on the resource allocation priority weight value, the specific steps for decomposing sub-task nodes in a complex task chain, dynamically allocating sub-tasks, adjusting the distribution of task nodes, and generating a sub-task distribution equilibrium mapping matrix are as follows: S301: Based on the resource allocation priority weight value, decompose the sub-task nodes in the complex task chain, extract the resource requirements and execution dependency data of the sub-tasks, analyze the resource competition between task nodes, calculate the node task dependency strength and resource occupation distribution value, group and sort the sub-tasks according to priority, and generate sub-task node priority distribution data. S302: Based on the subtask node priority distribution data, dynamically adjust the execution order of subtasks, extract the execution time and available resource range of each task node, calculate the resource time balance between task nodes, optimize resource scheduling to balance node time distribution, and generate dynamic allocation results for subtask execution. S303: Based on the dynamic allocation results of the subtask execution, the time interval and resource distribution of the task nodes are statistically analyzed, the node distribution balance and resource load matching are analyzed, the resource allocation scheme is adjusted, and a distribution mapping matrix between task nodes is established to generate a subtask distribution balance mapping matrix.

[0011] As a further aspect of the present invention, the resource time balance calculation formula is specifically as follows: ; in This represents the resource-time balance of subtask node p. Let be the time consumption of subtask node p on the q-th type of resource. Let be the consumption of subtask node p on the q-th type of resource. This is the time distribution adjustment factor. For resource consumption adjustment coefficient, Indicates the type of resource.

[0012] As a further aspect of the present invention, the specific steps for analyzing agent collaboration relationships, adjusting time delays and resource contention conflicts in the task chain, optimizing collaboration network relationships, and generating a collaboration node relationship weight matrix based on the subtask distribution equilibrium mapping matrix are as follows: S401: Based on the subtask distribution equilibrium mapping matrix, extract task node collaboration data and time interval distribution, analyze the dependency chain strength and resource conflict data between nodes, calculate the impact value of resource conflict on collaboration efficiency, determine the degree of dependence of task delay on subsequent nodes, and statistically analyze the fluctuation range of resource occupation caused by collaboration conflict, and generate agent collaboration dependency data. S402: Based on the agent collaboration dependency data, adjust the contradiction between resource contention and time distribution in the task chain, analyze the overlap range of resource usage time of conflicting nodes, adjust resource allocation and task time node distribution, balance resource contention and task dependency time distribution, and reorder the resource allocation time of conflicting nodes to generate collaborative relationship optimization data. S403: Based on the optimized data of the cooperation relationship, the resource sharing status and cooperation frequency of the task nodes are statistically analyzed, the relationship between resource utilization and cooperation intensity between nodes is analyzed, the resource usage weight value between task nodes is calculated, and the cooperation frequency is analyzed. The resource weight matrix of the nodes in the intelligent agent cooperation network is constructed, and the cooperation node relationship weight matrix is ​​generated.

[0013] As a further aspect of the present invention, the fluctuation range of resource usage refers to the range of numerical changes in the amount of the same resource used within the task execution cycle, which measures the stability of resource usage.

[0014] As a further aspect of the present invention, the specific steps for performing multiple rounds of quantitative evaluation of resource returns based on the collaborative node relationship weight matrix, analyzing the fluctuations in node resource return distribution, adjusting the resource allocation scheme, and generating a resource allocation return optimization index matrix are as follows: S501: Based on the collaborative node relationship weight matrix, statistically analyze the resource allocation data and revenue changes of the nodes, extract the change values ​​of node resource usage and revenue, analyze the time series characteristics of node revenue distribution, calculate the revenue change range and distribution fluctuation amplitude, determine the consistency of the time distribution of node resource revenue, and generate quantitative data of resource revenue. S502: Based on the resource revenue quantification data, calculate the matching degree value between the fluctuation range of resource revenue between nodes and the resource allocation ratio, statistically analyze the time coverage of node resource allocation, analyze the imbalance between revenue and resource allocation ratio, adjust the resource allocation weight to optimize the node revenue distribution, re-label the allocation weight data, and generate resource allocation adjustment scheme data. S503: Based on the resource allocation adjustment scheme data, analyze the impact of the adjusted resource allocation scheme on revenue fluctuations, statistically analyze the changes in node revenue and resource allocation weights after adjustment, calculate the time balance value of revenue distribution, construct the correspondence matrix between node resource revenue and allocation ratio, and generate a resource allocation revenue optimization index matrix.

[0015] As a further aspect of the present invention, the specific steps for updating the collaborative node path execution scheme based on the resource allocation benefit optimization index matrix, analyzing the task completion time reduction rate and consistency status, establishing collaborative consistency results, and generating global collaborative consistency status index values ​​are as follows: S601: Based on the resource allocation benefit optimization index matrix, analyze the correspondence between resource allocation and execution order in the collaborative node path, count the resource usage and task completion time of the path nodes, extract the impact data of resource allocation on task completion time, adjust the resource allocation and execution order of the path nodes, and generate collaborative node path execution scheme data. S602: Based on the collaborative node path execution scheme data, calculate the reduction ratio of completion time in the task path, statistically analyze the consistency parameters of time distribution between nodes, analyze the matching degree of resource allocation and time consistency between nodes, adjust the coordination of time distribution and resource allocation of the task path, re-optimize the time distribution relationship between path nodes, and generate collaborative consistency results. S603: Based on the aforementioned collaboration consistency results, analyze the adjusted node resource allocation and global time distribution data, calculate the balance of global resource allocation on node task completion time, statistically analyze the impact of resource utilization on collaboration status, construct a global collaboration model of resource and time allocation, and generate global collaboration consistency status index values.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, the imbalance in resource allocation among tasks is solved by real-time weight calculation of resource allocation rules, which improves the balance of resource scheduling. The dynamic allocation of task chains and the optimized distribution of nodes improve the adaptability of task execution and reduce execution conflicts. The real-time adjustment and optimization of time delay and resource contention enhance the stability of collaborative relationships and task consistency. The multi-round resource benefit evaluation and optimization mechanism shortens the task cycle and improves the global collaboration consistency and task execution efficiency. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the steps of the present invention; Figure 2 This is a flowchart of steps S1 of the present invention; Figure 3 This is a flowchart of steps S2 of the present invention; Figure 4 This is a flowchart of steps S3 of the present invention; Figure 5 This is a flowchart of step S4 of the present invention; Figure 6 This is a flowchart of steps S5 of the present invention; Figure 7 This is a flowchart of step S6 of the present invention. Detailed Implementation

[0019] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0020] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0021] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0022] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0023] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0024] Please see Figure 1 A multi-agent cooperative control method based on large model instruction scheduling includes the following steps: S1: Based on real-time data acquisition and task status recording of multiple agents, analyze the task load fluctuation in time series data, and generate task execution load trend fluctuation value by comparing the trend change rate with the task chain state characteristics. S2: Based on the task execution load trend fluctuation value, analyze the agent's resource usage, evaluate through resource differences and task processing capacity fluctuations, formulate resource allocation rules for load distribution, and generate resource allocation priority weight values; S3: Based on the resource allocation priority weight value, decompose the sub-task nodes in the complex task chain, perform dynamic allocation on the sub-tasks, adjust the distribution of task nodes, and generate a sub-task distribution balance mapping matrix. S4: Based on the subtask distribution equilibrium mapping matrix, analyze the agent cooperation relationship, adjust the time delay and resource contention contradiction in the task chain, optimize the cooperation network relationship, and generate a cooperation node relationship weight matrix. S5: Based on the weight matrix of collaborative node relationships, perform multiple rounds of quantitative evaluation of resource revenue, analyze the fluctuation of node resource revenue distribution, adjust the resource allocation scheme, and generate a resource allocation revenue optimization index matrix; S6: Based on the resource allocation benefit optimization index matrix, update the collaborative node path execution plan, analyze the task completion time reduction rate and consistency status, establish collaborative consistency results, and generate global collaborative consistency status index values.

[0025] The task execution load trend fluctuation value includes the task load change amplitude, task chain state change trend, and time series fluctuation coefficient. The resource allocation priority weight value includes resource usage weight, task processing capacity weight, and resource distribution balance weight. The subtask distribution balance mapping matrix includes subtask node priority distribution, agent load balance status, and dynamic task node distribution adjustment. The collaborative node relationship weight matrix includes time delay relationship weight, resource conflict coordination weight, and node collaboration optimization weight. The resource allocation benefit optimization index matrix includes resource benefit deviation value, resource benefit fluctuation coefficient, and resource allocation optimization weight. The global collaboration consistency status index value includes task completion time reduction index, agent collaboration consistency index, and global collaboration status evaluation value.

[0026] Please see Figure 2 The specific steps of S1 are as follows: S101: Based on real-time data acquisition and task status recording of multiple agents, the task status data is segmented according to the time series, each segment of task load data is extracted, the start and end positions of the data are marked according to the time interval, the change amplitude of each segment of load data is calculated, the calculation results are stored as a data sequence, and task load time segment data is generated. The task status data is segmented according to the time series, and the task load data for each segment is extracted according to the formula: ; Calculate the magnitude of change for each load data segment, where, Let i be the load value of the i-th segment. Let i be the timestamp of the i-th segment. This represents the range of load variation in that segment. and These are the timestamps for the (i+1)th and ith segments, respectively. This formula is used to calculate the magnitude of change in the load data for each segment.

[0027] In this formula, Let represent the load value of the i-th segment. Calculating its variation requires considering its differences within the time interval. The load value is measured in real time by sensors or an intelligent agent, and the timestamp is automatically generated by the system. By calculating the ratio of the load value change to the time interval, the load variation amplitude of each segment is obtained. Assume the load value of the first segment is . The load value of the second segment is The timestamps are respectively and Then we can substitute it into the formula:

[0028] ; This formula can be used to obtain the variation range of the load for each segment.

[0029] S102: Based on the task load time segmentation data, perform correlation analysis on the load change amplitude according to time period, calculate the time difference value of the change amplitude, compare the change amplitude with the difference trend of adjacent time periods, filter the interval data where the load fluctuation value is greater than the change threshold, and extract relevant features by combining the task chain status data to generate task chain status fluctuation feature data. By calculating the time difference of load variation, the variation range of each load data segment is first calculated. Then, by comparing the trend of load variation in adjacent time periods, intervals with variation ranges exceeding a set threshold are filtered out. This process involves two main operations: first, calculating the variation range of load data in different time periods and obtaining the difference value by comparing the changes in adjacent segments; second, filtering out intervals with large load fluctuations based on the set threshold. These data reflect abnormal fluctuations or critical periods of change in the task. The processed data is then used for subsequent analysis of task chain status and fluctuation characteristics extraction.

[0030] S103: Based on the task chain state fluctuation characteristic data, the fluctuation characteristic values ​​are sorted by time series, and the corresponding load fluctuation intervals are matched according to the task chain state. The time overlap between the task chain state characteristics and the load fluctuation is analyzed. The fluctuation trend parameter is calculated through the trend change difference value to obtain the task execution load trend fluctuation value. First, the fluctuation characteristic values ​​are sorted by time series. By calculating the change amplitude of each fluctuation characteristic, and then matching the corresponding load fluctuation range with the task chain state, the correlation between the fluctuation characteristic values ​​and the task state is analyzed. By comparing the overlap between the fluctuation values ​​and states in different time periods, the trend fluctuation value of the load is obtained, further reflecting the load trend fluctuation during task execution. This process ensures that changes in the task chain state can be effectively compared with load fluctuations and provides data support for subsequent task state optimization.

[0031] Please see Figure 3 The specific steps of S2 are as follows: S201: Based on the trend fluctuation value of task execution load, statistically analyze the resource usage data of the agent during task execution, decompose the resource usage type and quantity corresponding to the task load, classify and organize them, calculate the corresponding ratio value of task load and resource usage, analyze the impact of task load changes on resource usage, and generate data on the correlation between resource usage and task load. By statistically analyzing the resource usage data of intelligent agents, we first collect data on the types and quantities of various resources involved in task execution. Then, we categorize and organize the data according to the usage of different resources, and further calculate the corresponding ratios between task load and the usage of various resources. Through this calculation, we can analyze the impact of changes in task load on the use of different resources, help identify which resource consumption is closely related to changes in task load, and generate correlation data between task load and resource usage, providing valuable information for subsequent task optimization and resource allocation.

[0032] S202: Based on the correlation data between resource usage and task load, perform segmented statistics on the differences in resource usage in the task load, combine the fluctuation data of task processing capacity to perform time segmented analysis on the resource differences, calculate the correlation between the amount of resource differences and the fluctuation of task processing capacity, filter the relevant resource types and corresponding tasks, and generate resource allocation rule data. First, the resource usage differences in the task load are statistically segmented, and the changes in resource usage over different time periods are analyzed. Combined with task processing capacity fluctuation data, the resource differences are segmented over time. By calculating the correlation between the resource usage difference in each segment and the task processing capacity fluctuation, the relationship between resource consumption and task processing capacity can be assessed. This allows for the selection of resource types with strong correlations, which are then matched with tasks. Finally, resource allocation rules are generated based on this data to ensure reasonable resource scheduling and successful task completion.

[0033] S203: Based on resource allocation rule data, sort resource types according to task type and resource usage efficiency, statistically analyze the correspondence between resource allocation priority and task type, compare the resource type sorting with task allocation requirements, allocate resource weights required for tasks and mark priorities, and obtain resource allocation priority weight values. Resource types are sorted according to task type and resource utilization efficiency. The sorting priority of resource types is calculated, and then, based on task requirements, the resource types are ranked according to the formula: ; Calculate the resource weights required for the task, where, This indicates the priority of the i-th resource. Let be the utilization efficiency of the i-th type of resource. This formula assigns task requirements for the i-th resource type. It also determines the allocation priority for each resource type, providing a basis for subsequent task allocation.

[0034] Detailed explanation of the formula and its calculation derivation: In this formula, The utilization efficiency of resource type i can be obtained through real-time monitoring and historical data. This represents the task's requirement for resource type i, typically determined by the task's complexity and workload. By calculating the ratio of each resource's utilization efficiency to the task's requirement, the resource's priority is determined, thus deciding the order of resource allocation. Assume the utilization efficiency of resource 1 is... The required quantity for Task 1 is Then we can substitute it into the formula:

[0035] ; The calculation results show that resource 1 has a lower priority. In practical applications, these priority values ​​are compared with other resources to determine the most suitable resource allocation order for the current task, ensuring that the task is completed on time and effectively.

[0036] Please see Figure 4 The specific steps of S3 are as follows: S301: Based on the resource allocation priority weight value, decompose the sub-task nodes in the complex task chain, extract the resource requirements and execution dependency data of the sub-tasks, analyze the resource competition between task nodes, calculate the node task dependency strength and resource occupation distribution value, group and sort the sub-tasks according to priority, and generate sub-task node priority distribution data. First, the resource requirements and execution dependencies of each subtask are extracted to analyze resource competition among task nodes. To analyze resource competition, the resource utilization ratio of each task node during resource usage needs to be calculated, thereby assessing the resource contention between tasks. These calculations yield the task dependency strength and resource utilization distribution value for each node, providing data support for prioritizing and grouping task nodes. Finally, subtasks are grouped and sorted according to priority, generating subtask node priority distribution data to help allocate resources more rationally.

[0037] S302: Based on the priority distribution data of subtask nodes, dynamically adjust the execution order of subtasks, extract the execution time and available resource range of each task node, calculate the resource time balance between task nodes, optimize resource scheduling to balance the node time distribution, and generate dynamic allocation results for subtask execution. The specific formula for calculating resource time balance is as follows: ; in This represents the resource-time balance of subtask node p. Let be the time consumption of subtask node p on the q-th type of resource. Let be the consumption of subtask node p on the q-th type of resource. This is the time distribution adjustment factor. For resource consumption adjustment coefficient, Indicates the type of resource.

[0038] In this formula, This represents the resource-time balance of subtask node p. It measures the balance of resource and time distribution by calculating the time consumption and resource usage of node p on different resources.

[0039] This represents the time consumed by subtask node p on the q-th type of resource, in hours or minutes. This parameter is obtained by recording the resource time consumed during the subtask's execution. For example, node p might consume 4 hours on CPU and 3 hours on storage resources, which is typically monitored by a task management system during task execution.

[0040] This represents the resource consumption of subtask node p on resource type q, typically expressed in MB, GB, or CPU cycles. This data is derived by monitoring resource consumption (such as CPU, memory, and storage). For example, node p might use 10GB of storage and 4GB of memory.

[0041] This is a time distribution adjustment factor used to reflect the importance of time consumption in the overall resource scheduling. Its value is determined by the complexity of the task, resource contention, and scheduling optimization objectives. For example, for CPU-intensive tasks, Higher values ​​(e.g., 0.8) may be used for memory-intensive tasks, while lower values ​​(e.g., 0.5) may be used for memory-intensive tasks.

[0042] This is a resource consumption adjustment factor used to balance resource usage and scheduling. The setting of this parameter is based on resource utilization efficiency and the scheduling requirements of different resources. For example, for storage resources, It could take a value of 0.6, but for computing resources, The setting may be 0.4, reflecting the different weights of different resources in task scheduling.

[0043] Numerical example calculation: Assuming that subtask node p has two types of resources (CPU and storage), we need to calculate the resource time balance of node p.

[0044] The given data is as follows: Hours (CPU time consumption) Hours (Storage time consumption) GB (resources consumed by the CPU) GB (Storage resources consumed) (Time distribution adjustment factor) (Resource consumption adjustment coefficient) Substitute into the formula to calculate: ; First, calculate the molecule: ; Molecular calculation results: ; Next, calculate the denominator: ; Calculation result of denominator: ; Final calculation results: ; Parameter explanation: The value of 0.123 represents the resource-time balance of subtask node p, indicating that there is a certain imbalance between the resources consumed and the time consumed by node p, which may require further optimization of resource scheduling and task execution order.

[0045] This indicates the time consumed by the task across various resources (such as CPU and storage time).

[0046] This indicates the amount of resources a task consumes (such as CPU and storage resources).

[0047] and These are adjustment coefficients for time and resource consumption, which are optimized and adjusted based on the nature of the task and the use of resources.

[0048] The results indicate that the resource and time allocation of subtask node p is relatively unbalanced. Optimizing resource allocation and execution order can improve task execution efficiency and reduce resource waste.

[0049] S303: Based on the dynamic allocation results of subtask execution, the time interval and resource distribution of task nodes are statistically analyzed, the node distribution balance and resource load matching are analyzed, the resource allocation scheme is adjusted, and a distribution mapping matrix between task nodes is established to generate a subtask distribution balance mapping matrix. Based on the dynamic allocation results of subtask execution, the time intervals and resource distribution of task nodes are statistically analyzed to determine the balance of node distribution and resource load matching, according to the formula: ; Calculate the distribution balance of task node i, where: Let be the balance degree of node i, representing the balance between node time and resource distribution. Let be the time consumption of node i on the k-th resource. Let i be the weight of node i in the k-th resource (e.g., the priority or efficiency of the resource type). This represents the total resource consumption of node i.

[0050] In this formula, This represents the time consumed by node i on the k-th resource, which can usually be obtained from the task execution log, such as the time node i spends on resources like CPU and memory. This indicates the priority or utilization efficiency of node i on the k-th resource, which may be determined by system configuration, task type, or load conditions. The total resource usage of node i can be calculated by the system based on the resource consumption of node i at different stages.

[0051] By calculating the weighted time of all task nodes across different resource consumption levels, we can obtain the degree of matching between node time and resource consumption. Assume node 1 has three types of resources (e.g., CPU, memory, and storage), and their time consumption is as follows: Hour, Hour, Hour, weights are respectively , , Total resource usage is Then we can substitute it into the formula:

[0052] ; The results indicate that node 1 is relatively balanced in terms of resource usage and time distribution, but there is still room for optimization. This can be achieved by calculating the resource usage and time distribution of all nodes. The value can be used to construct a distribution mapping matrix between task nodes, thereby generating a subtask distribution balanced mapping matrix, providing a scientific basis for resource allocation.

[0053] Please see Figure 5 The specific steps of S4 are as follows: S401: Based on the subtask distribution equilibrium mapping matrix, extract task node collaboration data and time interval distribution, analyze the dependency chain strength and resource conflict data between nodes, calculate the impact of resource conflict on collaboration efficiency, determine the degree of dependence of task delay on subsequent nodes, and statistically analyze the fluctuation of resource occupation caused by collaboration conflict to generate agent collaboration dependency data. First, data extraction is performed on the collaborative relationships between task nodes. Specifically, the frequency of collaboration, number of data interactions, and duration of interactions for each node are obtained from the task scheduling records. The collaborating objects and their execution order are also analyzed to clarify the specific position and role of each node in the collaborative network. When extracting the time interval distribution between task nodes, the task start and completion times recorded in the task logs are combined. The execution time of each task node is compared with the execution times of its upstream and downstream nodes to form a set of time differences. For example, for two consecutive nodes in a task chain, the time difference is calculated by comparing their task completion times. By determining the start and end times, the time interval between them can be obtained. This information reflects the time delay between nodes in the task chain. Further segmentation and statistical analysis of the time interval data is performed to extract task pairs with intervals higher than the average task interval for subsequent analysis. When analyzing the strength of dependency chains between nodes, the completion status of each task node is first identified. This is then cross-analyzed with the amount of resources released by the node and the dependence of its downstream tasks on those resources to quantify the importance of the task for the initiation of subsequent tasks. Nodes with strong dependencies are prioritized as critical nodes. Extraction of resource conflict data requires statistical analysis of each node's performance in... Overlapping usage of the same resource category is identified by examining resource allocation records during task scheduling. Node groups using the same resource type (such as CPU, memory, communication links, etc.) at overlapping times are marked, and the ratio of overlap duration to total usage time is calculated. If multiple tasks contend for a resource within a certain period, leading to decreased execution efficiency, this resource is marked as a conflicting resource and recorded in the conflict dataset. When analyzing the impact of resource conflicts on collaboration efficiency, task dependency and resource conflict intensity must be jointly evaluated. If a node is in a segment with a high frequency of resource conflicts, and its collaboration with other nodes is... If the dependency relationship is strong, it is determined that resource contention has a significant impact on collaboration efficiency. By retrospectively analyzing the collaboration path execution of such nodes, it is determined whether waiting or blocking due to resource contention frequently occurs during task execution. Subsequently, the determination of the degree of dependency of task delay on subsequent nodes is based on task scheduling logs. The completion time delay of the preceding task is compared with the start delay of the subsequent task, and a baseline delay threshold is set. If the delay duration exceeds the threshold and is highly consistent with the delay of the downstream task start time, a delay dependency relationship is determined. The frequency of node combinations that occur in this case is accumulated to form a dependency distribution map.Finally, when statistically analyzing the fluctuation range of resource usage caused by collaborative conflicts, it is necessary to perform a difference analysis on the resource usage data of agents across multiple task time windows. If the usage value of a certain resource type fluctuates significantly over multiple periods, it indicates that the current resources are unevenly distributed among different nodes or that there is high-frequency contention. By extracting the range of resource usage changes for each resource type across different windows, the stability level of resource usage is summarized. This leads to the integration of a resource fluctuation data set reflecting collaborative stability. Combined with all the above analyses, a collaborative dependency data structure that comprehensively characterizes the task dependencies and conflict levels among multiple agents is constructed.

[0054] S402: Based on agent collaboration dependency data, adjust the contradiction between resource contention and time distribution in the task chain, analyze the overlap range of resource usage time of conflicting nodes, adjust resource allocation and task time node distribution, balance resource contention and task dependency time distribution, and reorder the resource allocation time of conflicting nodes to generate collaborative relationship optimization data. First, the conflict between resource contention and time distribution in the task chain is addressed by analyzing the overlap in resource usage time among conflicting nodes, identifying nodes with overlapping resource contention and time constraints. Next, based on this information, resource allocation and task time distribution are adjusted to achieve a balance between resource contention and task dependency time distribution. Finally, the resource allocation times of conflicting nodes are reordered to generate optimized collaboration data. This data helps optimize task execution order, reduce resource conflicts, and improve overall task execution efficiency.

[0055] S403: Based on the data optimization of cooperation relationships, the resource sharing and cooperation frequency of task nodes are statistically analyzed, the relationship between resource utilization and cooperation intensity between nodes is analyzed, the resource usage weight value between task nodes is calculated, and the cooperation frequency is analyzed. The resource weight matrix of nodes in the intelligent agent cooperation network is constructed, and the cooperation node relationship weight matrix is ​​generated. Based on the data optimization of collaborative relationships, the resource sharing and collaboration frequency of task nodes are statistically analyzed, and the relationship between resource utilization and collaboration intensity between nodes is analyzed according to the formula: ; Calculate the resource usage weight values ​​between task nodes, where, Let i be the resource weight value between node i and node j. Let i be the resource utilization rate of node i. Let be the frequency of cooperation of node j. Let represent the cooperation strength between node i and node j. This formula is used to analyze the resource usage weights between nodes and construct the resource weight matrix of nodes in the agent cooperative network.

[0056] In this formula, The resource utilization rate of node i can be obtained by monitoring the ratio of resource consumption to total resources in real time. This indicates the collaboration frequency of node j, that is, the number of times node j collaborates with other nodes, which is usually provided by the collaboration records during task execution. Let be the collaboration strength between node i and node j, representing the strength of the dependency relationship between the two nodes, obtained through task scheduling data between nodes. The resource utilization weight value is calculated using this formula, which helps in analyzing and optimizing the collaborative network structure between nodes.

[0057] Assume that the resource utilization rate of node 1 is The cooperation frequency of node 2 is The cooperation strength between node 1 and node 2 is Then substitute it into the formula: ; The results indicate that the resource utilization weight value between node 1 and node 2 is 1, suggesting a relatively balanced cooperative relationship between them. In practical applications, these weight values ​​can be compared with the weight values ​​of other nodes to optimize resource allocation and cooperation strategies among nodes, thereby improving the overall system's collaborative efficiency.

[0058] Please see Figure 6 The specific steps of S5 are as follows: S501: Based on the collaborative node relationship weight matrix, statistically analyze the resource allocation data and revenue changes of nodes, extract the changes in node resource usage and revenue, analyze the time series characteristics of node revenue distribution, calculate the range of revenue changes and distribution fluctuation amplitude, determine the consistency of the time distribution of node resource revenue, and generate quantitative data of resource revenue. First, the resource usage and corresponding revenue of each node are extracted and analyzed to determine their changes over time. This data allows for the calculation of the time-series characteristics of node revenue distribution and analysis of its fluctuations, thus revealing the range of revenue variation. Next, the consistency of the time distribution of node resource revenue is assessed to reveal the stability of resources and revenue during task execution. Finally, quantitative data on resource revenue is generated, providing a basis for subsequent resource allocation and optimization.

[0059] S502: Based on the quantitative data of resource revenue, calculate the matching degree between the fluctuation range of resource revenue between nodes and the resource allocation ratio, statistically analyze the time coverage of node resource allocation, analyze the imbalance between revenue and resource allocation ratio, adjust the resource allocation weight to optimize the distribution of node revenue, re-label the allocation weight data, and generate resource allocation adjustment scheme data. First, the time coverage of node resource allocation is statistically analyzed. Further analysis of the difference between revenue and resource allocation ratios identifies situations of uneven resource allocation. Based on these analysis results, the weights of resource allocation are adjusted to optimize node revenue distribution, ensuring more rational resource allocation and reducing resource waste. The allocation weight data is relabeled, and resource allocation adjustment scheme data is generated. This data will help improve resource allocation strategies and enhance overall system performance.

[0060] S503: Based on the resource allocation adjustment scheme data, analyze the impact of the adjusted resource allocation scheme on revenue fluctuations, statistically analyze the changes in node revenue and resource allocation weight after adjustment, calculate the time balance value of revenue distribution, construct the correspondence matrix between node resource revenue and allocation ratio, and generate a resource allocation revenue optimization index matrix. Based on the data from the resource allocation adjustment plan, analyze the impact of the adjusted resource allocation plan on revenue fluctuations, calculate the time equilibrium value of revenue distribution, and apply the formula: ; Calculate the correspondence between resource revenue and allocation ratio among task nodes, where, This represents the balance of revenue between node i and node j. Let be the revenue of node i. Assign weights to the resources of node j. Let be the time difference between node i and node j. This formula is used to generate a resource allocation benefit optimization index matrix.

[0061] In this formula, The amount of revenue representing node i is usually calculated through revenue feedback during task execution. The resource weights assigned to node j can be determined based on the resource scheduling strategy and the node's priority settings. The time difference between node i and node j represents the execution timing difference between the two nodes, reflecting their task intervals. This formula provides an indicator of the balance of revenue among nodes, allowing us to evaluate the effectiveness of adjustments to the resource allocation scheme.

[0062] Assume the revenue of node 1 is The resource allocation weight of node 2 is The time difference between node 1 and node 2 is Then substitute it into the formula: ; The results show that the balance score between resource allocation and revenue between node 1 and node 2 is 8. By comparing the balance values ​​between different nodes, the resource allocation scheme can be further optimized to improve the efficiency of resource use and the stability of revenue.

[0063] Please see Figure 7 The specific steps of S6 are as follows: S601: Based on the resource allocation benefit optimization index matrix, analyze the correspondence between resource allocation and execution order in the collaborative node path, count the resource usage and task completion time of the path nodes, extract the impact data of resource allocation on task completion time, adjust the resource allocation and execution order of the path nodes, and generate collaborative node path execution plan data. First, the relationship between resources and time is analyzed by extracting resource usage and task completion time from path nodes. To optimize task paths, the impact of resource allocation on task completion time is calculated, and resource allocation and execution order of path nodes are adjusted accordingly. Through this data and analysis, collaborative node path execution plan data is generated, thereby optimizing task execution efficiency and resource utilization.

[0064] S602: Based on the collaborative node path execution scheme data, calculate the reduction ratio of completion time in the task path, statistically analyze the consistency parameters of time distribution between nodes, analyze the matching degree of resource allocation and time consistency between nodes, adjust the coordination of time distribution and resource allocation of the task path, re-optimize the time distribution relationship between path nodes, and generate collaborative consistency results. First, the consistency parameters of time distribution among nodes are statistically analyzed to determine the degree of matching between resource allocation and time consistency. Based on these analyses, the coordination between the time distribution and resource allocation of the task path is further adjusted to ensure a more balanced resource allocation. Finally, the time distribution relationship between path nodes is re-optimized to generate a collaborative consistency result, thereby improving the execution efficiency and collaborative effect of the entire task path.

[0065] S603: Based on the results of collaboration consistency, analyze the adjusted node resource allocation and global time distribution data, calculate the balance of global resource allocation on node task completion time, statistically analyze the impact of resource utilization on collaboration status, construct a global collaboration model of resource and time allocation, and generate global collaboration consistency status index values. Based on the collaboration consistency results, the adjusted node resource allocation and global time distribution data are analyzed to calculate the balance of global resource allocation on node task completion time, according to the formula: ; Calculate the balance of node resource and time allocation in the global collaboration model, where: This represents the global collaborative balance between node i and node j. This represents the amount of resource consumed by node i in the p-th case. This represents the weight or efficiency of the p-th resource. This represents the time spent by node i on the q-th type of resource.

[0066] This formula is used to calculate the global collaboration consistency index value and optimize the balance of task paths and resource allocation.

[0067] In this formula, This represents the amount of resource consumed by node i on the p-th type, which is usually obtained through a resource monitoring system; It is the weight or efficiency of the p-th type of resource, reflecting the use value or allocation priority of the resource, and is set according to the different nature of the task; It represents the time consumed by node i on the q-th type of resource, obtained through the time record during task execution.

[0068] By using weighted summation, and combining the consumption and weights of different resources, the overall resource and time balance of each node is calculated. Assume that node 1's consumption of the three resources is as follows: , , The corresponding resource weights are respectively , , The time consumption of node 1 is as follows: , , Then substitute it into the formula:

[0069] ; The result shows that the global collaboration balance of node 1 is 0.275. This calculation allows us to evaluate the efficiency of task execution and optimize node resource allocation and time usage, thereby improving the efficiency and consistency of the entire collaborative task.

[0070] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A multi-agent cooperative control method based on large model instruction scheduling, characterized in that, Includes the following steps: S1: Based on real-time data acquisition and task status recording of multiple agents, analyze the task load fluctuation in time series data, and generate task execution load trend fluctuation value by comparing the trend change rate with the task chain state characteristics. S2: Based on the task execution load trend fluctuation value, analyze the agent's resource usage, evaluate it through resource differences and task processing capacity fluctuations, formulate resource allocation rules for load distribution, and generate resource allocation priority weight values; S3: Based on the resource allocation priority weight value, decompose the sub-task nodes in the complex task chain, perform dynamic allocation on the sub-tasks, adjust the distribution of task nodes, and generate a sub-task distribution balance mapping matrix. S4: Based on the subtask distribution equilibrium mapping matrix, analyze the agent cooperation relationship, adjust the time delay and resource contention contradiction in the task chain, optimize the cooperation network relationship, and generate a cooperation node relationship weight matrix. S5: Based on the collaborative node relationship weight matrix, perform multiple rounds of resource revenue quantitative evaluation, analyze the fluctuation of node resource revenue distribution, adjust the resource allocation scheme, and generate a resource allocation revenue optimization index matrix; S6: Based on the resource allocation benefit optimization index matrix, update the collaborative node path execution scheme, analyze the task completion time reduction rate and consistency status, establish collaborative consistency results, and generate global collaborative consistency status index values.

2. The multi-agent cooperative control method based on large model instruction scheduling according to claim 1, characterized in that, The task execution load trend fluctuation value includes the task load change amplitude, task chain state change trend, and time series fluctuation coefficient. The resource allocation priority weight value includes resource usage weight, task processing capacity weight, and resource distribution balance weight. The subtask distribution balance mapping matrix includes subtask node priority distribution, agent load balance status, and dynamic task node distribution adjustment. The collaborative node relationship weight matrix includes time delay relationship weight, resource conflict coordination weight, and node collaboration optimization weight. The resource allocation benefit optimization index matrix includes resource benefit deviation value, resource benefit fluctuation coefficient, and resource allocation optimization weight. The global collaboration consistency status index value includes task completion time reduction index, agent collaboration consistency index, and global collaboration status evaluation value.

3. The multi-agent cooperative control method based on large model instruction scheduling according to claim 1, characterized in that, Based on real-time data acquisition and task status recording by multiple agents, the specific steps for analyzing task load fluctuations in time-series data and generating task execution load trend fluctuation values ​​by comparing the trend change rate with task chain state characteristics are as follows. S101: Based on real-time data acquisition and task status recording of multiple agents, the task status data is segmented according to the time series, each segment of task load data is extracted, the start and end positions of the data are marked according to the time interval, the change amplitude of each segment of load data is calculated, the calculation results are stored as a data sequence, and task load time segment data is generated. S102: Based on the task load time segmentation data, perform correlation analysis on the load change amplitude according to time period, calculate the time difference value of the change amplitude, compare the change amplitude with the difference trend of adjacent time periods, filter the interval data where the load fluctuation value is greater than the change threshold, and extract relevant features in combination with task chain status data to generate task chain status fluctuation feature data. S103: Based on the task chain state fluctuation characteristic data, the fluctuation characteristic values ​​are sorted by time series, and the corresponding load fluctuation intervals are matched according to the task chain state. The time overlap between the task chain state characteristics and the load fluctuation is analyzed. The fluctuation trend parameter is calculated by the trend change difference value to obtain the task execution load trend fluctuation value.

4. The multi-agent cooperative control method based on large model instruction scheduling according to claim 1, characterized in that, Based on the task execution load trend fluctuation value, the resource usage of the intelligent agent is analyzed. An assessment is conducted using resource differences and task processing capacity fluctuations. The specific steps for formulating resource allocation rules for load distribution and generating resource allocation priority weight values ​​are as follows: S201: Based on the task execution load trend fluctuation value, statistically analyze the resource usage data of the agent during task execution, decompose the resource usage type and quantity corresponding to the task load, classify and organize them, calculate the corresponding ratio value of task load and resource usage, analyze the impact of task load changes on resource usage, and generate resource usage and task load correlation data. S202: Based on the resource usage and task load correlation data, the resource usage differences in the task load are statistically segmented, and the resource differences are analyzed in time segments in combination with the task processing capacity fluctuation data. The correlation between the resource difference and the task processing capacity fluctuation is calculated, the relevant resource types and corresponding tasks are screened, and resource allocation rule data is generated. S203: Based on the resource allocation rule data, sort the resource types according to task type and resource usage efficiency, calculate the correspondence between resource allocation priority and task type, compare the resource type sorting with the task allocation requirements, allocate the resource weight required for the task and mark the priority, and obtain the resource allocation priority weight value.

5. The multi-agent cooperative control method based on large model instruction scheduling according to claim 1, characterized in that, Based on the resource allocation priority weight values, the specific steps for breaking down sub-task nodes in a complex task chain, dynamically allocating sub-tasks, adjusting the distribution of task nodes, and generating a sub-task distribution equilibrium mapping matrix are as follows: S301: Based on the resource allocation priority weight value, decompose the sub-task nodes in the complex task chain, extract the resource requirements and execution dependency data of the sub-tasks, analyze the resource competition between task nodes, calculate the node task dependency strength and resource occupation distribution value, group and sort the sub-tasks according to priority, and generate sub-task node priority distribution data. S302: Based on the subtask node priority distribution data, dynamically adjust the execution order of subtasks, extract the execution time and available resource range of each task node, calculate the resource time balance between task nodes, optimize resource scheduling to balance node time distribution, and generate dynamic allocation results for subtask execution. S303: Based on the dynamic allocation results of the subtask execution, the time interval and resource distribution of the task nodes are statistically analyzed, the node distribution balance and resource load matching are analyzed, the resource allocation scheme is adjusted, and a distribution mapping matrix between task nodes is established to generate a subtask distribution balance mapping matrix.

6. The multi-agent cooperative control method based on large model instruction scheduling according to claim 5, characterized in that, The specific formula for calculating the resource time balance is as follows: ； in This represents the resource-time balance of subtask node p. Let be the time consumption of subtask node p on the q-th type of resource. Let be the consumption of subtask node p on the q-th type of resource. This is the time distribution adjustment factor. For resource consumption adjustment coefficient, Indicates the type of resource.

7. The multi-agent cooperative control method based on large model instruction scheduling according to claim 1, characterized in that, Based on the aforementioned subtask distribution equilibrium mapping matrix, the specific steps for analyzing agent collaboration relationships, adjusting time delays and resource contention conflicts in the task chain, optimizing collaboration network relationships, and generating a collaboration node relationship weight matrix are as follows: S401: Based on the subtask distribution equilibrium mapping matrix, extract task node collaboration data and time interval distribution, analyze the dependency chain strength and resource conflict data between nodes, calculate the impact value of resource conflict on collaboration efficiency, determine the degree of dependence of task delay on subsequent nodes, and statistically analyze the fluctuation range of resource occupation caused by collaboration conflict, and generate agent collaboration dependency data. S402: Based on the agent collaboration dependency data, adjust the contradiction between resource contention and time distribution in the task chain, analyze the overlap range of resource usage time of conflicting nodes, adjust resource allocation and task time node distribution, balance resource contention and task dependency time distribution, and reorder the resource allocation time of conflicting nodes to generate collaborative relationship optimization data. S403: Based on the optimized data of the cooperation relationship, the resource sharing status and cooperation frequency of the task nodes are statistically analyzed, the relationship between resource utilization and cooperation intensity between nodes is analyzed, the resource usage weight value between task nodes is calculated, and the cooperation frequency is analyzed. The resource weight matrix of the nodes in the intelligent agent cooperation network is constructed, and the cooperation node relationship weight matrix is ​​generated.

8. The multi-agent cooperative control method based on large model instruction scheduling according to claim 7, characterized in that, The fluctuation range of resource usage refers to the range of numerical changes in the amount of the same resource used within the task execution cycle, which measures the stability of resource usage.

9. The multi-agent cooperative control method based on large model instruction scheduling according to claim 1, characterized in that, Based on the aforementioned collaborative node relationship weight matrix, the specific steps for performing multiple rounds of quantitative evaluation of resource returns, analyzing the fluctuations in node resource return distribution, adjusting the resource allocation scheme, and generating a resource allocation return optimization index matrix are as follows: S501: Based on the collaborative node relationship weight matrix, statistically analyze the resource allocation data and revenue changes of the nodes, extract the change values ​​of node resource usage and revenue, analyze the time series characteristics of node revenue distribution, calculate the revenue change range and distribution fluctuation amplitude, determine the consistency of the time distribution of node resource revenue, and generate quantitative data of resource revenue. S502: Based on the resource revenue quantification data, calculate the matching degree value between the fluctuation range of resource revenue between nodes and the resource allocation ratio, statistically analyze the time coverage of node resource allocation, analyze the imbalance between revenue and resource allocation ratio, adjust the resource allocation weight to optimize the node revenue distribution, re-label the allocation weight data, and generate resource allocation adjustment scheme data. S503: Based on the resource allocation adjustment scheme data, analyze the impact of the adjusted resource allocation scheme on revenue fluctuations, statistically analyze the changes in node revenue and resource allocation weights after adjustment, calculate the time balance value of revenue distribution, construct the correspondence matrix between node resource revenue and allocation ratio, and generate a resource allocation revenue optimization index matrix.

10. The multi-agent cooperative control method based on large model instruction scheduling according to claim 1, characterized in that, Based on the resource allocation benefit optimization index matrix, the specific steps for updating the collaborative node path execution scheme, analyzing the task completion time reduction rate and consistency status, establishing collaborative consistency results, and generating global collaborative consistency status index values ​​are as follows: S601: Based on the resource allocation benefit optimization index matrix, analyze the correspondence between resource allocation and execution order in the collaborative node path, count the resource usage and task completion time of the path nodes, extract the impact data of resource allocation on task completion time, adjust the resource allocation and execution order of the path nodes, and generate collaborative node path execution scheme data. S602: Based on the collaborative node path execution scheme data, calculate the reduction ratio of completion time in the task path, statistically analyze the consistency parameters of time distribution between nodes, analyze the matching degree of resource allocation and time consistency between nodes, adjust the coordination of time distribution and resource allocation of the task path, re-optimize the time distribution relationship between path nodes, and generate collaborative consistency results. S603: Based on the aforementioned collaboration consistency results, analyze the adjusted node resource allocation and global time distribution data, calculate the balance of global resource allocation on node task completion time, statistically analyze the impact of resource utilization on collaboration status, construct a global collaboration model of resource and time allocation, and generate global collaboration consistency status index values.

Images