Collaborative management and execution system based on artificial intelligence workflow

Abstract

This invention relates to the field of workflow automation technology, specifically to a collaborative management and execution system based on artificial intelligence workflows. The system includes a role behavior representation module, a collaboration relationship characterization module, a collaboration evolution identification module, a task matching and adjustment module, and a collaboration deviation correction module. By capturing role behavior characteristics and task execution performance through data, the system can synchronously map changes in behavioral relationships during task flow, continuously mining the intensity and trends of collaboration. Combined with comprehensive analysis of collaboration status and task suitability, it drives intelligent optimization of task and role allocation order, dynamically adjusts task scheduling structure and allocation priorities, and automatically provides feedback correction for scheduling deviations. With data support, the process proactively adapts to changes in the collaboration chain, simultaneously improving collaboration efficiency and task flow response capabilities. The system's scenario adaptability and collaboration accuracy are significantly enhanced in diverse business environments.

Description

Technical Field

[0001] This invention relates to the field of workflow automation technology, and in particular to a collaborative management and execution system based on artificial intelligence workflow. Background Technology

[0002] Workflow automation involves the automatic execution and management of task flows in a digital environment through predefined rules and logic control, and is widely used in various industry scenarios such as enterprise management, government systems, manufacturing scheduling, and information services. Among them, traditional workflow collaborative management and execution systems refer to a type of system that manages and automatically executes the collaborative process of task flow. It is mainly used for scenarios involving task coordination, resource allocation, and status tracking among multiple participants. Typically, task nodes are statically divided through manual configuration or preset rules, and the process direction is guided by fixed logical path judgments.

[0003] In existing technologies, task flow mainly relies on static configuration and predetermined rules. Role behaviors and collaborative relationships are fixed within fixed task node divisions and logical judgments. There is a lack of continuous monitoring of changes in participant behavior and collaborative status during task execution. When facing multi-department collaboration or task priority adjustment needs, the system processing flow has the problem of not being able to reflect the actual collaboration dynamics in a timely manner. As a result, task allocation and node paths cannot be flexibly adjusted according to the actual performance of roles. Response delays and imbalances in the division of labor in collaboration are easily amplified. Task feedback and resource utilization are often limited by static process paths, making it difficult to adapt to changing organizational scenarios and business needs. This leads to a decline in management efficiency and limited space for process optimization. Summary of the Invention

[0004] To address the technical problems existing in the prior art, embodiments of the present invention provide a collaborative management and execution system based on artificial intelligence workflow. The technical solution is as follows: On the one hand, it provides a collaborative management and execution system based on artificial intelligence workflows, including: The role behavior representation module is based on the task response terminal. It analyzes the task interaction performance of each department role at the workflow node, tracks the role response according to the task number, associates the behavior status with the node flow sequence, and judges the status change of adjacent stages to obtain the role behavior feature vector set. The collaboration relationship characterization module compares the response actions of the task intersection role group at the node based on the set of role behavior feature vectors, analyzes the behavior sequence and response consistency, summarizes the changes in interaction frequency, evaluates the degree of collaboration, and obtains the collaboration relationship strength index. Based on the collaborative relationship strength index, the collaborative evolution identification module judges the trend of role relationship changes, analyzes relationship consistency, and associates the role behavior of collaborative trends with task performance to obtain the collaborative change trend factor. Based on the collaboration change trend factor, the task matching and adjustment module filters the suitability status of tasks to be assigned, optimizes the matching of roles and task nodes, adjusts the mapping relationship, and reassigns tasks according to priority order to obtain the task assignment priority ranking. The collaborative deviation correction module sorts the tasks based on the task allocation priority, judges the role task response performance, analyzes feedback data, sorts out behaviors that are inconsistent with the changed state, synchronizes and corrects the scheduling according to the deviation performance, and obtains the role scheduling deviation correction result.

[0005] On the other hand, the set of role behavior feature vectors includes task response time, task execution success rate and task execution delay; the strength index of collaborative relationship includes collaboration frequency, response consistency and interaction strength; the collaborative change trend factor includes relationship stability, trend persistence and collaborative adaptability; the task allocation priority ranking includes role adaptation priority, task urgency and execution capability; and the role scheduling offset correction result includes scheduling deviation amount, execution delay correction and behavior adjustment factor.

[0006] On the other hand, the character behavior representation module includes: The task response tracking submodule analyzes the acquired role response action data based on the task response terminal, determines whether the task number corresponding to the response action is consistent, compares the response action and behavior type of each role one by one, arranges the response actions under the same task number in chronological order, and obtains the role task response feature set. The node order alignment submodule adjusts the arrangement of response actions of each node according to the node flow order based on the character task response feature set, compares the changes in the order of response actions between nodes, analyzes the correspondence of response action sequences between nodes, determines the response order offset of each character between different nodes, and obtains the node response order difference. The behavior state comparison submodule determines the associated role response records based on the difference in the node response order, classifies the behavior state labels of each task node, optimizes the state switching trajectory of each role between adjacent nodes, and obtains a set of role behavior feature vectors by comparing the combination of node behavior states and sorting out the change path.

[0007] On the other hand, the collaborative relationship characterization module includes: The node behavior comparison submodule analyzes the behavior action data of each role on the same task node based on the role behavior feature vector set, compares the different types of actions performed by the role on the task node, determines whether the role behavior order is consistent with the task flow order, rearranges the response action order according to the node flow relationship, compares the differences in behavior types between roles, and obtains the node behavior coordination deviation. The response consistency analysis submodule analyzes the sequential consistency of the role response actions on the behavior time axis based on the node behavior coordination deviation amount and the order of each role's behavior response in the collaborative nodes. It determines whether the response sequences between roles are synchronized, quantitatively evaluates the matching degree of response behaviors, and obtains the multi-role response sequence difference degree. The collaboration frequency summarization submodule determines the distribution of the multi-role response sequence differences in the task flow path, counts the response interaction frequency of each pair of roles at the collaboration node, analyzes the impact of frequency changes on role collaboration behavior, summarizes the interaction intensity of each pair of roles in task execution, and obtains the collaboration relationship strength index.

[0008] On the other hand, the cooperative evolution identification module includes: The relationship change judgment submodule analyzes the collaboration strength of each role in different task nodes based on the collaboration relationship strength index, compares the increase and decrease trends of role collaboration relationships between nodes, analyzes the stability of collaboration relationships, and determines the role relationship change trend to obtain the relationship change trend analysis results. Based on the analysis results of the relationship change trend, the behavioral feature correlation analysis submodule determines the persistence of each role's behavioral features during task execution, evaluates the matching degree between the role's behavioral patterns and task execution performance at each node, and checks the consistency between the role's behavioral changes and collaboration trends to obtain the correlation degree of the role's behavioral features. The collaboration trend factor integration submodule analyzes the adaptability between each role's behavior and task execution based on the correlation degree of the role's behavioral characteristics, evaluates the stability of the collaboration trend at different task stages, judges the changes in collaboration behavior, and obtains the collaboration change trend factor.

[0009] On the other hand, the task matching adjustment module includes: The task adaptation and filtering submodule obtains the adaptation data of the task to be assigned and the role based on the collaboration change trend factor, compares the task requirements and the role capabilities, determines whether the role meets the task requirements, evaluates the adaptation status by combining the role's current capabilities and the task requirements, filters out roles and tasks that meet the task requirements, and obtains a set of role-task adaptation information. The role-task matching optimization submodule analyzes the compatibility of each role and task node based on the role-task adaptation information set, determines the matching priority of roles and tasks, and adjusts the matching of roles and tasks according to the role's execution capabilities and task requirements to obtain the role-task matching optimization result. The task allocation and sorting submodule analyzes the urgency of task nodes and the execution capability of roles based on the role-task matching optimization results, calculates the adaptation priority of roles and tasks, and re-sorts tasks according to task urgency and role execution capability to obtain the task allocation priority ranking.

[0010] On the other hand, the cooperative deviation correction module includes: The role feedback analysis submodule sorts the tasks based on the task allocation priority, obtains the role's feedback data during task execution, analyzes the role's task response performance, determines whether the role has deviated from the expected state during execution, compares the difference between the role's feedback data and the expected state, identifies the inconsistent parts in task execution, and obtains the role's task execution feedback data. The deviation processing submodule, based on the task execution feedback data of the role, processes the deviation performance of the role in task execution, calculates the degree of deviation between the behavioral differences of each role and the predetermined task execution state, determines the frequency and scope of deviation, and classifies the deviation behavior to obtain the role execution deviation performance data. The scheduling correction submodule analyzes the changes in the role scheduling process based on the role execution deviation performance data, adjusts the matching relationship between the role and the task node, corrects the task execution process according to the role deviation data, optimizes the role scheduling order, and obtains the role scheduling offset correction result.

[0011] On the other hand, the roles of each department refer to the roles that participate in different tasks in the workflow, representing personnel from various departments or functions. The task interaction performance refers to the interactive behavior of the roles during the task execution process, including responding, executing tasks, and providing feedback.

[0012] On the other hand, the role response action refers to the action or behavior of a role when accepting, executing or processing a task, and the node flow sequence refers to the flow order of tasks from one node to the next in the workflow.

[0013] On the other hand, the task intersection role group refers to a group of roles that participate together in the same task or workflow. Roles collaborate at different stages of the task. The response consistency refers to the consistency of the behavior and response of multiple roles on the same task node, that is, whether they execute the task synchronously or as expected.

[0014] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: By capturing role behavior characteristics and task execution performance through data, the system can synchronously map changes in behavioral relationships during task flow, continuously explore the intensity and trends of collaboration, and drive intelligent optimization of task and role allocation order by combining comprehensive analysis of collaboration status and task adaptability. It can dynamically adjust task scheduling structure and allocation priority, automatically provide feedback and correction for scheduling deviations, and proactively adapt the process to changes in the collaboration chain with data support. Collaboration efficiency and task flow response capabilities are improved simultaneously, and the system's scenario adaptability and collaboration accuracy in diverse business environments are significantly enhanced. Attached Figure Description

[0015] 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.

[0016] Figure 1 This is a schematic diagram of the system of the present invention; Figure 2 This is a schematic diagram of the system framework of the present invention; Figure 3 This is a flowchart of the character behavior representation module of the present invention; Figure 4 A flowchart of the collaborative relationship characterization module of this invention; Figure 5 This is a flowchart of the collaborative evolution identification module of the present invention; Figure 6 This is a flowchart of the task matching and adjustment module of the present invention; Figure 7 This is a flowchart of the collaborative deviation correction module of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] 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.

[0019] This invention provides a collaborative management and execution system based on artificial intelligence workflow, such as... Figure 1 As shown, the system includes: The role behavior representation module is based on the task response terminal. It analyzes the task interaction performance of each department's role in the workflow process, tracks the role's response action item by item according to the task number, aligns the response behavior with the node flow sequence, continuously compares the role's behavior state at different nodes, judges the state change pattern of the same role in adjacent stages, and obtains the role behavior feature vector set. The collaboration relationship characterization module compares the response actions of task intersection role groups at the same node based on the set of role behavior feature vectors, analyzes the behavior sequence and response consistency of role collaboration nodes, performs inductive processing on the differences in collaboration sequence and the frequency of interaction in task flow, evaluates the change process of the closeness of role collaboration, and obtains the collaboration relationship strength index. The collaboration evolution identification module judges the trend of role relationship changes based on the collaboration relationship strength index, analyzes the consistency of relationship changes on the task timeline, makes correlation judgments on the behavior of roles that maintain the collaboration trend, and integrates the role's behavioral characteristics with task execution performance to obtain the collaboration change trend factor. The task matching and adjustment module uses collaboration change trend factors to filter the compatibility between the status and the tasks to be assigned within the task allocation unit, optimizes the matching process between roles and task nodes, maps roles and task execution numbers that meet the requirements of the adjusted status, and reorganizes the role allocation relationship according to the adaptation priority order to obtain the task allocation priority ranking. The collaborative deviation correction module sorts tasks based on task allocation priority, judges the role's task response performance, analyzes the role's feedback data during the task phase, sorts out the parts of behavior that are inconsistent with the existing change state, synchronizes according to the deviation performance, corrects the role scheduling change process, and obtains the role scheduling deviation correction result.

[0020] The set of role behavior feature vectors includes task response time, task execution success rate, and task execution delay. The indicators of collaborative relationship strength include collaboration frequency, response consistency, and interaction intensity. The factors of collaborative change trend include relationship stability, trend persistence, and collaboration adaptability. The priority ranking of task allocation includes role adaptation priority, task urgency, and execution capability. The results of role scheduling offset correction include scheduling offset amount, execution delay correction, and behavior adjustment factor.

[0021] In the role behavior representation module, departmental roles refer to different roles participating in different tasks in the workflow, usually representing personnel or system nodes from different departments or functions; task interaction performance refers to the interactive behavior of roles during task execution, such as responding, executing tasks, and providing feedback; role response actions refer to the actions or behaviors of roles when accepting, executing, or processing tasks, including starting a task, response time, and completing a task; node flow sequence refers to the flow order of tasks from one node to the next in the workflow, reflecting the task execution path; different nodes refer to the various task execution points or stages in the workflow, where roles may perform different tasks or activities at each node; behavioral state refers to the behavioral pattern or execution state of a role at a certain task node, such as whether the task is completed on time or the execution status of the task; state change pattern refers to the change pattern or rule of the behavioral state of a role between different nodes or over time, reflecting the changing trend of the role in the workflow.

[0022] In the collaboration relationship characterization module, the task intersection role group refers to a group of roles that participate together in the same task or workflow, and the roles collaborate at different stages of the task; the role collaboration node refers to each task node in the workflow that involves role collaboration, at which roles need to cooperate with each other to complete the task; response consistency refers to the consistency of the behavior and response of multiple roles at the same task node, that is, whether they execute the task synchronously or as expected; collaboration order difference refers to whether there are differences in the order in which roles execute tasks in the workflow, reflecting the asynchronous situation of role collaboration; interaction frequency refers to the frequency of interaction between roles in the workflow, which is usually related to the progress of task completion and the dependency relationship between roles; inductive processing refers to the comprehensive induction of the interaction patterns and order between roles to identify the rules of role collaboration; the process of change in the degree of closeness refers to the process of change in the degree of closeness of the collaboration relationship between roles (such as high frequency of interaction, collaboration quality, etc.) during task execution.

[0023] In the collaboration evolution identification module, the role relationship change trend refers to the direction of change in the collaborative relationship between roles, which is a trend of gradual strengthening, weakening, or stabilization of the collaborative relationship; the consistency of relationship change refers to the degree of consistency in the change of role relationships, which is manifested in the consistency of the collaborative relationship change (for example, showing similar collaborative trends in multiple tasks); maintaining the collaborative trend refers to the continuous collaborative trend shown by roles in multiple task cycles, indicating that the collaborative relationship between roles remains consistent or stabilizes over time; role behavior refers to the specific behavioral patterns shown by roles in the workflow, such as task response time, task completion rate, etc.; behavioral characteristics refer to the specific indicators or attributes describing role behavior, such as response time, task execution success rate, interaction frequency, etc.; task execution performance refers to the performance of roles in the process of executing tasks, such as whether the task is completed on time and the quality of the task.

[0024] In the task matching and adjustment module, "state and task allocation unit" refers to the matching relationship between a role and a task allocation unit in a specific state. The task allocation unit usually refers to the task management system or scheduling system. "Task node" refers to a specific execution point or stage of a task in the workflow, where a role needs to complete specific work. "Change state" refers to changes in role behavior or task state, such as changes in role response time or task completion status. "Task execution number" refers to a unique identifier for a task in the workflow, used to track the task execution process. "Adaptation priority" refers to prioritizing roles and tasks based on role behavior evolution factors or task requirements to ensure that the right role completes the right task. "Role allocation relationship" refers to the allocation and matching relationship between roles and tasks during task execution.

[0025] In the Collaborative Deviation Correction module, Role Task Response Performance refers to the role's response performance after accepting a task, such as response time and timeliness of task completion; Role Feedback Data refers to the feedback information provided by the role during task execution, including completion status, execution effect, and delays; Inconsistent Behavior with Existing Changed State refers to the part where the role's actual execution behavior differs from the previously defined expected behavior or state; Deviation Performance refers to the degree of deviation of the role's behavior, such as delays in task completion or low task execution quality, indicating that the role has failed to perform as expected; Role Scheduling Change refers to changes in the role's task scheduling status in the workflow, such as adjusting task priorities or reassigning roles, and correcting scheduling based on actual performance.

[0026] like Figure 2 and Figure 3 As shown, the character behavior representation module includes: The task response tracking submodule analyzes the acquired role response action data based on the task response terminal, determines whether the task number corresponding to the response action is consistent, compares the response action and behavior type of each role one by one, arranges the response actions under the same task number in chronological order, and obtains the role task response feature set. Sort each response action by time field to create a response time sequence. Extract role number and task number, and determine if adjacent records belong to the same role. If so, check if the task number matches. If both match, group them into the same task behavior set; otherwise, create a new set and record again. Extract the action type field and label it as receiving task, starting execution, or submitting feedback. Record the occurrence time of each action and calculate the total duration from receiving feedback to the start of the action. For example, if role X receives task T200 at 10:00, starts execution at 10:05, and submits feedback at 10:20, the total duration is 20 minutes, and the delay time is 5 seconds. The time interval is calculated and incorporated into the task behavior sequence. A sorting sequence is established for all records under the task number, and behavior labels are arranged in chronological order to form a behavior record table. Then, all records are traversed to determine whether the interval between any two behaviors exceeds a set value. If it exceeds 5 minutes, the behavior is considered to be interrupted. The number of interruptions and their duration are recorded. A behavior response frequency table for each role in all tasks is established, recording data such as behavior delay, interval interruption, and feedback duration. These are integrated into a role task response feature set, with a separate set generated for each role, containing attributes such as time parameters, behavior type, and response efficiency, to provide a basis for subsequent process analysis.

[0027] The node order alignment submodule is based on the character task response feature set. It adjusts the arrangement of response actions of each node according to the node flow order, compares the changes in the order of response actions between nodes, analyzes the correspondence of response action sequences between nodes, determines the response order offset of each character between different nodes, and obtains the node response order difference. The task standard node order is set to a fixed sequence, such as nodes 1 to 3 being submission, approval, and archiving respectively. Then, the actual action nodes of the role are sorted by time, and their positions in the standard order are compared. If they are inconsistent, the order offset is recorded. For example, if role A completes node 3 before node 2, the node order offset value is 1. This process continues to iterate through all action records, counting the total number of offset actions. For example, if 10 response actions are recorded and 3 are offset actions, the offset rate is 30%. The offset direction is also recorded to determine whether the action occurred prematurely or delayed. For each response action, its occurrence time is compared with the time of the previous action. If the current action occurs earlier than the previous action, it is recorded as a time sequence conflict. For example, if the time of node 2 is 10:15 and the time of node 1 is 10:20, it is a time reversal behavior. The number of such behaviors in the entire task sequence is counted. Then, a comparison matrix is ​​built for the time sequence and node sequence of each role in each task, marking the number of offsets, offset direction, number of time conflicts, etc. The execution position of the role's behavior in the task flow process is compared with the standard sequence. The difference in node response sequence of each role is obtained by summarizing the differences. This is used to identify whether there is a sequential deviation in the role's execution behavior, which in turn affects the accuracy of cooperation.

[0028] The behavior state comparison submodule determines the associated role response records based on the difference in node response order, classifies the behavior state labels of each task node, optimizes the state switching trajectory of each role between adjacent nodes, and obtains the role behavior feature vector set by comparing the combination of node behavior states and sorting out the change path. Each response record's corresponding action within a node is labeled with a status tag. These tags are categorized into four states: No Response, Responding, Completed, and Abnormal Feedback. No Response means no record; Responding means there was a start but no feedback; Completed means on-time feedback; and Abnormal Feedback means delayed or failed feedback. Abnormal feedback is determined by whether the feedback time exceeds a specified threshold. For example, if the average feedback time for a node task is 15 minutes and the delay tolerance is set to 5 minutes, then exceeding 20 minutes is considered abnormal. After labeling the role status within each task node, the states are arranged sequentially by node to form a role status trajectory sequence. For instance, if role B's status across three nodes is Responding, Completed, and Abnormal Feedback, the path would be recorded as Responding - Completed. - Anomaly feedback is used to establish a state sequence for each role. By comparing different role path combinations, it is analyzed whether the state transitions between adjacent nodes are repeated. If the same state combination appears 7 times in 10 roles, the proportion is 70%, which is judged as the dominant path. The number of times the dominant path is included in a role's path is calculated as the ratio of the total number of state pairs to obtain the state path consistency ratio. If 3 out of 5 state groups in a role's path are consistent with the dominant path, the consistency is 60%. The consistency of state paths for all roles is statistically analyzed. Combined with the type distribution, switching frequency, and dominant path proportion of each state, the role behavior feature vector set is output to provide support for behavior pattern recognition and collaborative optimization.

[0029] like Figure 2 and Figure 4 As shown, the collaborative relationship characterization module includes: The node behavior comparison submodule analyzes the behavior action data of each role on the same task node based on the role behavior feature vector set, compares the different types of actions performed by the role on the task node, determines whether the order of role behavior is consistent with the order of task flow, rearranges the order of response actions according to the node flow relationship, compares the differences in behavior types between roles, and obtains the node behavior coordination deviation. Filter by node number to identify role behavior sequences with overlapping task numbers, then categorize actions by behavior type (e.g., receiving, executing, feedback) and count the frequency of each action at each role node. For multiple roles' behavior records under the same node, sort by action occurrence time and determine if their sequence matches the set flow. For example, if the node flow is node 1 receive, node 2 execute, node 3 feedback, and role B executes first and then provides feedback at node 2, the sequence matches; if B provides feedback first and then executes, the sequence is inconsistent and recorded as reverse behavior. Count the total number of role records with inconsistent action sequences. If the same role has 3 incorrect sequence responses out of 10, the sequence matching rate is 70%. Then, re-sort all response behaviors according to the standard node flow order. The new arrangement is then compared with the distribution of behavior types among the roles. The parts with different behavior types under the same node are extracted. For example, role A performs a feedback operation in node 3, role B only receives the task without feedback, and role C provides delayed feedback. The frequency of missing or abnormal feedback behavior is counted. If a role is missing feedback actions more than 3 times in node 3, it is marked as a significant behavior difference. The threshold for judging behavior type difference is set as the number of missing actions in the standard behavior of a node exceeds one-third of the total number of behavior items. If a node should have three types of actions but a role only has one, its behavior difference is judged as a serious deviation. Combining the order offset and the number of behavior type differences, the degree of behavior coordination of each role on each node is output. The overall number of behavior offsets and the number of difference actions are calculated. Based on this, the dispersion of the behavior consistency of the role on the node is counted and summarized as the node behavior coordination deviation.

[0030] The response consistency analysis submodule is based on the node behavior coordination deviation. According to the order of each role's behavior response in the collaborative nodes, it analyzes the sequential consistency of the role's response actions on the behavior time axis, determines whether the response sequences between roles are synchronized, and quantitatively evaluates the degree of matching of response behaviors to obtain the multi-role response sequence difference. The responses of each role in the collaborative nodes are categorized. First, the times of occurrence are sorted, and the responses are aligned along a timeline. The time interval difference between multi-role response sequences is calculated. If two roles perform the same action at the same node with a time difference within 3 minutes, it is marked as a consistent response. If the time difference exceeds 3 minutes but does not exceed 8 minutes, it is marked as partially consistent. If the time difference exceeds 8 minutes, it is judged as an inconsistent response. For example, if role A submits feedback at node 2 at 10:00, role B at 10:02, and role C at 10:10, then A and B are consistent, and C is inconsistent. After categorizing the time differences for each pair of roles, the number of consistent responses, partially consistent responses, and inconsistent responses are summarized by node, and the consistency ratio is calculated. For example, if three roles have a total of 9 mutual response matches, 6 are consistent, 2 are partially consistent, and 1 is inconsistent. If there is a discrepancy, the consistency score is 6 divided by 9, which equals 0.667. Further, it is used to determine whether the overall response sequence is synchronized. The synchronization judgment benchmark is a consistency score of 0.75 or higher. If the response order between roles shows continuous matching in multiple tasks, it is considered that the response sequence is synchronized. A two-dimensional comparison matrix is ​​constructed for the response action type and occurrence time, respectively. It is checked whether the response sequence position of any two roles remains stable at each node. For example, in the approval task with task number T310, role D always responds before role E. If this feature occurs more than 3 times, role D is judged to take precedence over E in the response sequence. Such response order consistency is accumulated and scored. The matching degree of all roles on all task nodes is normalized and statistically analyzed to form the multi-role response sequence difference, which is used to measure the consistency difference of the behavior time sequence of multiple subjects at collaborative nodes.

[0031] The collaboration frequency summarization submodule determines the distribution of responses in the task flow path based on the multi-role response sequence differences, counts the response interaction frequency of each pair of roles at the collaboration node, analyzes the impact of frequency changes on role collaboration behavior, summarizes the interaction intensity of each pair of roles in task execution, and obtains the collaboration relationship strength index. The differences are mapped based on the distribution of nodes along the task flow path. First, the number of times each pair of roles responds together at each collaborative node is counted to construct a collaboration frequency matrix for the roles. If roles A and B both exhibit feedback behavior at node 3 with a time difference of less than 3 minutes, the interaction frequency of that node is incremented by 1. All task numbers and node numbers are traversed, and the total number of response interactions for each pair of roles is summarized. For example, roles F and G appear at collaborative nodes 12 times in 5 tasks, of which 8 are synchronous responses and 4 are asynchronous behaviors with a time difference of less than 5 minutes, recording a total frequency of 12. Next, the distribution of each frequency value across all role pairs is analyzed, and a threshold for classifying the strength of collaboration frequencies is set. Values ​​are categorized as follows: more than 10 interactions are considered high frequency, 5 to 10 interactions are considered medium frequency, and less than 5 interactions are considered low frequency. For example, if roles H and J only respond together 3 times, it is classified as low frequency interaction. The percentage of role pairs within each frequency range is recorded, and the impact of frequency changes on role collaboration behavior is analyzed. If the number of high-frequency collaborations for a certain role pair decreases over time, it is marked as a weakening of collaboration frequency; conversely, it is marked as an strengthening of collaboration frequency. The relationship between role collaboration frequency and response consistency is compared. If the response difference for high-frequency collaboration is less than 0.3, it is judged as a close collaboration relationship; if it is greater than 0.6, it is judged as a weak collaboration relationship. The collaboration frequency distribution, time period distribution, response consistency, and other parameters of all role pairs are integrated to generate a collaboration relationship strength index.

[0032] like Figure 2 and Figure 5 As shown, the collaborative evolution identification module includes: The relationship change judgment submodule analyzes the collaboration strength of each role in different task nodes based on the collaboration relationship strength index, compares the increase and decrease trends of role collaboration relationships between nodes, analyzes the stability of collaboration relationships, and determines the role relationship change trend to obtain the relationship change trend analysis results. First, using the task number as the primary key, we retrieve the statistical values ​​of interaction frequency, response consistency, and interaction intensity between roles at different nodes. We compare the collaboration intensity values ​​of the same role across multiple nodes and calculate the increase or decrease between nodes. If the collaboration intensity of nodes 1 to 3 is 4, 6, and 3 respectively, then nodes 1 to 2 represent an increase, and nodes 2 to 3 represent a decrease. We assign positive and negative labels to each change interval, compiling a complete collaboration trend sequence. Next, we determine the number of trend changes for each role pair. If two changes in the same direction occur within three consecutive nodes, the trend is marked as stable; otherwise, it is marked as fluctuating. We set a threshold of 2 for the collaboration intensity change amplitude; that is, if the change value is greater than or less than 2, it is considered a drastic change. For example, for a role... The strengths of A and B at nodes 1, 2, and 3 are 3, 6, and 7, respectively. This indicates a continuous increase with a change range of 3 and 1, both exceeding the threshold, and is marked as a stable reinforcing relationship. Conversely, if the strengths of C and D are 5, 3, and 2, this indicates a continuous decrease, and is judged as a weakening relationship. The trend patterns of the same character in different tasks are then summarized to determine the frequency of trend occurrence. If the stable reinforcing pattern occurs in more than 60% of the total number of tasks, the overall trend is marked as reinforcing. Next, the trend change ratio is calculated and the location of nodes with drastic changes is recorded. For example, if characters E and F show 6 stable reinforcing times, 3 fluctuations, and 1 drastic decrease in 10 tasks, the node with the drastic decrease is located at task number T509, forming the relationship change trend analysis results.

[0033] The behavioral feature correlation analysis submodule, based on the relationship change trend analysis results, determines the persistence of each role's behavioral features during task execution, assesses the matching degree between the role's behavioral patterns and task execution performance at each node, and checks the consistency between the role's behavioral changes and collaboration trends to obtain the correlation degree of the role's behavioral features. Grouping by role, we extract behavioral feature vector sets for each role at each task node. We examine the continuous changes in behavioral latency, task completion time, and execution success rate for each role at different nodes. If the latency in three consecutive tasks is 2 minutes, 3 minutes, and 2 minutes respectively, the completion time is 12 minutes, 13 minutes, and 12 minutes respectively, and the success rate is 100%, 100%, and 100% respectively, then the role's behavior is considered stable. If the latency fluctuation exceeds 5 minutes or the success rate differs by more than 30% across tasks, it is marked as behavioral fluctuation. The threshold for behavioral feature fluctuation is set as follows: a latency change exceeding 5 minutes or a success rate change exceeding 30% is considered unsatisfactory. For stability, the behavior patterns of each role at its collaboration nodes are time-line aligned to check whether they are consistent with changes in collaboration trends. For example, when the collaboration intensity increases, does the role's behavior fluctuation decrease? If role H's delay time decreases from 6 minutes to 2 minutes and completion time decreases from 15 minutes to 11 minutes within the intensity increase range, it is judged as behavior improvement. If the opposite is true, it is judged as behavior inconsistency. Then, the trend of role behavior change is compared with the trend of collaboration relationship change, and the ratio of the number of trend consistency to the total number of tasks is calculated. If the consistency ratio is higher than 70%, it is determined that the behavior characteristics and collaboration trends are highly correlated, and the correlation degree of role behavior characteristics is obtained.

[0034] The Collaboration Trend Factor Integration Submodule analyzes the adaptability between each role's behavior and task execution based on the correlation degree of role behavior characteristics, evaluates the stability of collaboration trends at different task stages, judges the changes in collaboration behavior, and obtains collaboration change trend factors. To assess the stability of collaboration trends across different task phases, the following formula is used: ; Calculate the collaboration stability index value, determine the changes in collaboration behavior, and obtain the collaboration change trend factor; in, Representing the Collaboration stability index values ​​for each role at different task stages Representing the The character in the first Task response time for each task phase Representing the The character in the first Task execution delay in each task phase This represents the total number of mission phases that the character participated in. Collaboration stability index value ( The value refers to the average absolute difference between the sum (absolute value) of the differences between the task response time and the task execution delay of the same role across multiple task stages, divided by the number of stages. This indicator measures the average deviation between the response time and the actual task execution delay of a role in different task stages. If the response time and execution delay of a role are basically consistent across all task stages, the value approaches 0, indicating that its collaborative behavior is stable and consistent. If the difference between stages is large, the value is high, reflecting that its behavior fluctuates greatly and is less stable during the collaborative process.

[0035] Extract the first from the system task log. Each character The behavioral data for each task stage is recorded, along with the task response time. and task execution delay The task response time is the time taken for the role to confirm and respond after receiving the task, and the task execution delay is the time elapsed from response confirmation to task completion; the records for each stage are as follows: Phase 1 , ; Phase 2 , ; Phase 3 , ; Phase 4 , ; Phase 5 ,

[0036] Input the raw data into the formula for calculation, first calculating the difference between each item: ; ; ; ; ; Sum the differences: ; Take the absolute value and divide by the number of stages. : ; For each stage , Using minimum-maximum normalization, the following settings are defined: right Minimum value maximum value ; right Minimum value maximum value .

[0037] The normalized response time value is: ; ; ; ; ; The normalized execution latency value is: ; ; ; ; ; Then, the difference between the normalized values ​​is calculated: ; ; ; ; ; Sum the above five items: ; Take the absolute value and substitute it into the formula to calculate the normalized stability index value: ; Set the range of values ​​for the collaboration stability index: High stability: This range indicates that the collaborative behavior of a role changes very little between task phases, and the stability is high. Typically, the difference between the response time and execution delay of a role in a task is small, the consistency of collaborative behavior is strong, and the role's behavior remains highly consistent, which is suitable for stable task allocation.

[0038] Moderate stability: This range indicates that the character's collaborative behavior between task phases fluctuates to some extent, but remains within an acceptable range. The character may occasionally exhibit longer or shorter response times, but overall performance is stable and suitable for most tasks; however, minor fluctuations should be monitored.

[0039] Low stability: This range indicates that the collaborative behavior of roles fluctuates significantly, with noticeable differences in response time and execution latency, which may lead to delays in task execution or imbalances in collaboration. In this case, it is necessary to optimize the matching of roles and tasks, or adjust the task execution of roles to ensure efficient collaboration.

[0040] The calculation result is It belongs to the high stability range ( ),because The result is below 0.15, indicating that Role 1 exhibits strong consistency in collaborative behavior during the multi-task execution phase and demonstrates a highly stable collaborative pattern. Therefore, it does not produce significant negative impacts during task matching and allocation and can be considered a stable collaborative role.

[0041] like Figure 2 and Figure 6 As shown, the task matching adjustment module includes: The task matching and filtering submodule obtains the matching data of the tasks to be assigned and the roles based on the collaboration change trend factor, compares the task requirements and the role capabilities, determines whether the role meets the task requirements, evaluates the matching situation by combining the role's current capabilities and task requirements, filters out the roles and tasks that meet the task requirements, and obtains the role-task matching information set. Extract the task number, required capability tags, expected completion time, and urgency level of the currently pending tasks. Simultaneously, read the latest behavioral characteristic records and task execution performance data of each role. Role capabilities are evaluated based on their average completion time, success rate, and delay frequency over the last 5 tasks. Task requirements are defined by setting capability requirement ranges; for example, task T800 requires a completion time of no more than 15 minutes, a success rate higher than 90%, and no more than one delay. Then, compare the role capability data with these requirement values ​​item by item. If role Q has an average completion time of 13 minutes, a success rate of 92%, and zero delays in the last 5 tasks, it is considered a perfect match. If the completion time is 17 minutes, it does not meet the time requirement and is judged as a mismatch. A weight value is set for each requirement, with completion time weighted at 0.4, success rate at 0.4, and delay frequency at 0.2. If a role's ability deviates slightly but the overall score exceeds 80, it is still considered a match. All roles are sorted by their ability matching values ​​with the current task, and roles with scores exceeding 80 are included in the candidate set. For example, if role R scores 84 and role S scores 77, only role R is included in the candidate set. All tasks to be assigned are combined with roles that meet the conditions, and the task number, role number, matching score, and deviation of each indicator are obtained. The output is a set of role-task matching information.

[0042] The role-task matching optimization submodule analyzes the compatibility of each role and task node based on the role-task adaptation information set, determines the matching priority of roles and tasks, and adjusts the matching of roles and tasks according to the role's execution capabilities and task requirements to obtain the role-task matching optimization results. The adaptation data records between each role and task node were analyzed one by one. First, the task number was associated with the sequence of task nodes to determine whether the role had the ability to execute consecutive nodes. Then, each ability indicator was standardized and scored, with the execution ability score range set from 0 to 100. The task requirement intensity was scored according to completion time limit, failure tolerance, and response timeliness. For example, if the response time limit for task node N5 is set to 10 minutes, and the response time of role U in the last three times was 8 minutes, 9 minutes, and 12 minutes, with an average of 9.6 minutes, it meets the standard and scores 90 points. If it exceeds 1 minute, 10 points are deducted, and if there are consecutive timeouts, 20 points are deducted. Then, the adaptation level of the role in this task node is calculated, and the matching priority threshold is set at 85 points. If the score is higher than the threshold, the role will be considered as having a higher adaptation level. The value is prioritized for matching, and the role is marked as a priority candidate for that node. If two roles have similar scores but one has a positive trend factor and the other has a downward trend, the one with the increasing trend is selected first. The matching values ​​of all roles on all task nodes are compared to find all high-priority roles and their matching task number combinations. For example, if role W matches task T900 nodes 1 and 2 with scores of 92 and 95 respectively, then role W has a priority of 1 in this task. All matching priorities are classified and processed, conflicting nodes are marked, and low-priority roles are adjusted to secondary tasks or reserved for later assignment. The five items of information are output: task number, task node, role number, matching score, trend label, and priority level, forming the role task matching optimization result.

[0043] The task allocation and sorting submodule analyzes the urgency of task nodes and the execution capability of roles based on the role-task matching optimization results, calculates the matching priority between roles and tasks, and re-sorts tasks according to task urgency and role execution capability to obtain the task allocation priority ranking. The urgency level is read according to the task node, and the urgency level range is set from 1 to 5, where level 1 is the most urgent, with a corresponding processing time of less than 15 minutes, and level 5 is a general task, with a corresponding processing time that can be extended to more than 1 hour. The processing time limit field of each task node is read and converted into an urgency level. For example, task T1000 node 2 requires a completion time of 12 minutes, which is determined to be urgency level 1, and node 3 requires a completion time of 35 minutes, which is determined to be urgency level 3. Then, the character execution ability score in the character task matching result is read. If character X scores 95 and the trend factor is positive, then a weight factor of 1.1 is assigned. If character Y scores 90 but the trend is downward, then... With a weight of 0.9, the execution ability score is multiplied by the weight factor to obtain the comprehensive ability score. For example, the comprehensive score of role X is 104.5 and that of role Y is 81. For each task node, the comprehensive ability score is multiplied by the node urgency level to obtain the role-task adaptation priority value. For example, if the node urgency level is 1, the priority value of role X is 104.5 and that of role Y is 81. Then, the adaptation priority values ​​of all roles and task nodes are sorted. The higher the priority value, the more suitable the role is for the task node. Based on this, a task allocation order table is generated, and the task number, node number, candidate role, role comprehensive ability, task urgency level, adaptation priority value and sort number are output to form the task allocation priority ranking.

[0044] like Figure 2 and Figure 7 As shown, the cooperative deviation correction module includes: The role feedback analysis submodule sorts tasks by priority, obtains feedback data of roles during task execution, analyzes role task response performance, determines whether there is any deviation behavior of roles during execution, compares the differences between role feedback data and expected state, identifies inconsistencies in task execution, and obtains role task execution feedback data. The task records are indexed by task number and node number. Fields such as feedback time, completion status, execution time, delay duration, and exception flags are extracted. Then, the feedback data is compared item by item with the expected task execution status. The expected status is based on the node completion time, response time threshold, and behavior status defined in the allocation sorting. For example, task T1200 node 2 requires completion within 15 minutes and a response time of no more than 3 minutes. If role L's feedback record shows a execution time of 18 minutes and a response time of 5 minutes, then it deviates from the expectation in both dimensions and is judged as a behavior deviation record. A difference comparison table is built for the feedback data of each role, and the time difference, status discrepancy label, and behavior interruption label are marked. For each discrepancy item, a delay threshold of 5 minutes is set. If this value is exceeded, it is recorded as an explicit deviation. Records with abnormal feedback, response timeouts, and failure to execute in the order of nodes are summarized one by one. A feedback difference statistics item is constructed for each role, and the number and percentage of deviations under the current task are counted. For example, if role M executes 10 task nodes, and 3 of them exceed the response time threshold and 2 fail to respond in the order, the total number of deviations is 5, accounting for 50%, which is recorded as a medium deviation level. If the number of deviations exceeds 60% of the total number of nodes, it is marked as a severe deviation. Finally, the role number, task number, number of deviations, deviation type, comparison difference, and deviation level are output to form the role task execution feedback data.

[0045] The Deviation Analysis submodule organizes the deviation performance of each role in task execution based on the role's task execution feedback data. It calculates the degree of deviation between each role's behavioral differences and the predetermined task execution state, determines the frequency and scope of deviations, and classifies them according to the deviation behavior to obtain the role's execution deviation performance data. Extract deviation behavior records for each role in each task, item by item, by role number. First, extract fields such as behavior time deviation, matching of feedback status with expected status, consistency of execution order, and whether feedback is missing. Calculate the behavioral difference for each record. For example, if role P is delayed by 9 minutes at node 3 of task T1300, the deviation level is set as severe. If role Q is delayed by only 2 minutes, it is set as mild. The deviation severity grading standard is set as follows: delay exceeding 8 minutes is severe, 4 to 8 minutes is moderate, and less than 4 minutes is mild. After summarizing all deviation records within each role's task sequence, the frequency of each level is calculated. If role R has 10 deviation behaviors in the task sequence, including 3 severe behaviors and 4 moderate behaviors... If there are 3 mild instances, the deviation frequency is judged to be 100%. The deviation level is calculated by combining frequency and severity. The severe frequency ratio is 30% as the judgment criterion. If it exceeds 30%, it is marked as a serious deviation behavior. Then, each deviation behavior is classified according to the deviation type. The classification dimensions are set as four categories: time offset, status mismatch, execution order misalignment, and feedback missing. Each deviation record is classified with a corresponding category label according to the behavioral characteristics. For example, if role S has a response time of 0 and a feedback status of failure in task T1350, it is classified as status mismatch. The deviation behaviors of all roles are statistically analyzed in a structured manner, and the role number, task node, deviation category, deviation duration, frequency distribution, and deviation level are output to form the role execution deviation performance data.

[0046] The scheduling correction submodule analyzes changes in the role scheduling process based on role execution deviation performance data, adjusts the matching relationship between roles and task nodes, corrects the task execution process based on role deviation data, optimizes the role scheduling order, and obtains the role scheduling offset correction result. The matching of roles and task nodes in the scheduling process is re-evaluated. First, records with a deviation level of moderate or higher are extracted to determine if the role's suitability for the current task node is still valid. If role T has two consecutive severe deviations at node 2 in task T1400, its priority allocation qualification at that node is revoked, and it is downgraded to a waitlist status. Then, other roles with similar scores and mild or no deviation levels at the same node are searched as replacements. If role U has a score of 88 points at the same node and only one mild deviation record, role U is assigned to that node as the executor. For each replacement action, the affected node number and the position before and after the adjustment are recorded. The color code and deviation difference are used to reconstruct the role scheduling order table according to the task nodes, adjust the original scheduling order, and move roles with continuous deviations to the back or remove them from the original node matching table. At the same time, the number of replacement nodes is recorded. If a role has more than 3 nodes adjusted in total, it is marked as an unstable role and its allocation priority is reduced in subsequent scheduling. The distribution density of roles on nodes is recalculated for the task execution path after each round of adjustment to avoid a single role being concentrated on multiple key nodes. The adjusted task number, node number, corrected role, replacement basis, deviation level, correction method and scheduling order number are output to form the role scheduling offset correction result.

[0047] 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 protection of the described technical solutions.

Claims

1. A collaborative management and execution system based on artificial intelligence workflow, characterized in that, The system includes: The role behavior representation module is based on the task response terminal. It analyzes the task interaction performance of each department role at the workflow node, tracks the role response according to the task number, associates the behavior status with the node flow sequence, and judges the status change of adjacent stages to obtain the role behavior feature vector set. The collaboration relationship characterization module compares the response actions of the task intersection role group at the node based on the set of role behavior feature vectors, analyzes the behavior sequence and response consistency, summarizes the changes in interaction frequency, evaluates the degree of collaboration, and obtains the collaboration relationship strength index. Based on the collaborative relationship strength index, the collaborative evolution identification module judges the trend of role relationship changes, analyzes relationship consistency, and associates the role behavior of collaborative trends with task performance to obtain the collaborative change trend factor. Based on the collaboration change trend factor, the task matching and adjustment module filters the suitability status of tasks to be assigned, optimizes the matching of roles and task nodes, adjusts the mapping relationship, and reassigns tasks according to priority order to obtain the task assignment priority ranking. The collaborative deviation correction module sorts the tasks based on the task allocation priority, judges the role task response performance, analyzes feedback data, sorts out behaviors that are inconsistent with the changed state, synchronizes and corrects the scheduling according to the deviation performance, and obtains the role scheduling deviation correction result.

2. The collaborative management and execution system based on artificial intelligence workflow according to claim 1, characterized in that, The set of role behavior feature vectors includes task response time, task execution success rate, and task execution delay. The strength index of the collaborative relationship includes collaboration frequency, response consistency, and interaction strength. The collaboration change trend factor includes relationship stability, trend persistence, and collaboration adaptability. The task allocation priority ranking includes role adaptation priority, task urgency, and execution capability. The role scheduling offset correction result includes scheduling deviation, execution delay correction, and behavior adjustment factor.

3. The collaborative management and execution system based on artificial intelligence workflow according to claim 1, characterized in that, The role behavior representation module includes: The task response tracking submodule analyzes the acquired role response action data based on the task response terminal, determines whether the task number corresponding to the response action is consistent, compares the response action and behavior type of each role one by one, arranges the response actions under the same task number in chronological order, and obtains the role task response feature set. The node order alignment submodule adjusts the arrangement of response actions of each node according to the node flow order based on the character task response feature set, compares the changes in the order of response actions between nodes, analyzes the correspondence of response action sequences between nodes, determines the response order offset of each character between different nodes, and obtains the node response order difference. The behavior state comparison submodule determines the associated role response records based on the difference in the node response order, classifies the behavior state labels of each task node, optimizes the state switching trajectory of each role between adjacent nodes, and obtains a set of role behavior feature vectors by comparing the combination of node behavior states and sorting out the change path.

4. The collaborative management and execution system based on artificial intelligence workflow according to claim 1, characterized in that, The collaborative relationship characterization module includes: The node behavior comparison submodule analyzes the behavior action data of each role on the same task node based on the role behavior feature vector set, compares the different types of actions performed by the role on the task node, determines whether the role behavior order is consistent with the task flow order, rearranges the response action order according to the node flow relationship, compares the differences in behavior types between roles, and obtains the node behavior coordination deviation. The response consistency analysis submodule analyzes the sequential consistency of the role response actions on the behavior time axis based on the node behavior coordination deviation amount and the order of each role's behavior response in the collaborative nodes. It determines whether the response sequences between roles are synchronized, quantitatively evaluates the matching degree of response behaviors, and obtains the multi-role response sequence difference degree. The collaboration frequency summarization submodule determines the distribution of the multi-role response sequence differences in the task flow path, counts the response interaction frequency of each pair of roles at the collaboration node, analyzes the impact of frequency changes on role collaboration behavior, summarizes the interaction intensity of each pair of roles in task execution, and obtains the collaboration relationship strength index.

5. The collaborative management and execution system based on artificial intelligence workflow according to claim 1, characterized in that, The cooperative evolution identification module includes: The relationship change judgment submodule analyzes the collaboration strength of each role in different task nodes based on the collaboration relationship strength index, compares the increase and decrease trends of role collaboration relationships between nodes, analyzes the stability of collaboration relationships, and determines the role relationship change trend to obtain the relationship change trend analysis results. Based on the analysis results of the relationship change trend, the behavioral feature correlation analysis submodule determines the persistence of each role's behavioral features during task execution, evaluates the matching degree between the role's behavioral patterns and task execution performance at each node, and checks the consistency between the role's behavioral changes and collaboration trends to obtain the correlation degree of the role's behavioral features. The collaboration trend factor integration submodule analyzes the adaptability between each role's behavior and task execution based on the correlation degree of the role's behavioral characteristics, evaluates the stability of the collaboration trend at different task stages, judges the changes in collaboration behavior, and obtains the collaboration change trend factor.

6. The collaborative management and execution system based on artificial intelligence workflow according to claim 1, characterized in that, The task matching and adjustment module includes: The task adaptation and filtering submodule obtains the adaptation data of the task to be assigned and the role based on the collaboration change trend factor, compares the task requirements and the role capabilities, determines whether the role meets the task requirements, evaluates the adaptation status by combining the role's current capabilities and the task requirements, filters out roles and tasks that meet the task requirements, and obtains a set of role-task adaptation information. The role-task matching optimization submodule analyzes the compatibility of each role and task node based on the role-task adaptation information set, determines the matching priority of roles and tasks, and adjusts the matching of roles and tasks according to the role's execution capabilities and task requirements to obtain the role-task matching optimization result. The task allocation and sorting submodule analyzes the urgency of task nodes and the execution capability of roles based on the role-task matching optimization results, calculates the adaptation priority of roles and tasks, and re-sorts tasks according to task urgency and role execution capability to obtain the task allocation priority ranking.

7. The collaborative management and execution system based on artificial intelligence workflow according to claim 1, characterized in that, The collaborative deviation correction module includes: The role feedback analysis submodule sorts the tasks based on the task allocation priority, obtains the role's feedback data during task execution, analyzes the role's task response performance, determines whether the role has deviated from the expected state during execution, compares the difference between the role's feedback data and the expected state, identifies the inconsistent parts in task execution, and obtains the role's task execution feedback data. The deviation processing submodule, based on the task execution feedback data of the role, processes the deviation performance of the role in task execution, calculates the degree of deviation between the behavioral differences of each role and the predetermined task execution state, determines the frequency and scope of deviation, and classifies the deviation behavior to obtain the role execution deviation performance data. The scheduling correction submodule analyzes the changes in the role scheduling process based on the role execution deviation performance data, adjusts the matching relationship between the role and the task node, corrects the task execution process according to the role deviation data, optimizes the role scheduling order, and obtains the role scheduling offset correction result.

8. The collaborative management and execution system based on artificial intelligence workflow according to claim 1, characterized in that, The roles in each department refer to the roles that participate in different tasks in the workflow, representing personnel from various departments or functions. The task interaction performance refers to the interactive behaviors of the roles during task execution, including responding, executing tasks, and providing feedback.

9. The collaborative management and execution system based on artificial intelligence workflow according to claim 1, characterized in that, The role response action refers to the action or behavior of a role when accepting, executing or processing a task, and the node flow sequence refers to the flow order of tasks from one node to the next in the workflow.

10. The collaborative management and execution system based on artificial intelligence workflow according to claim 1, characterized in that, The task intersection role group refers to a group of roles that participate together in the same task or workflow. Roles collaborate at different stages of the task. The response consistency refers to the consistency of the behavior and response of multiple roles at the same task node, that is, whether they execute the task synchronously or as expected.

Images