Task priority adaptive adjustment system and method for complex projects

Abstract

This application relates to the field of task scheduling technology, and in particular to a system and method for adaptive adjustment of task priorities for complex projects. The proposed scheme involves acquiring multiple task datasets for the target project, extracting multi-dimensional features of time, resources, and dependencies, constructing multiple dependency graphs, and dividing tasks into first-level, second-level, and third-level tasks. The priority of each level of task is calculated sequentially, and a global task execution order is generated by combining the time urgency of the first-priority tasks, the resource requirements of the second-priority tasks, and the dependencies of the third-priority tasks. When adjustment conditions are triggered, priorities are adjusted sequentially according to the priority hierarchy, and the task execution order is dynamically updated. This application can improve the task scheduling efficiency and global adaptability of complex projects, thereby increasing productivity and resource utilization.

Description

Technical Field

[0001] This application relates to the field of task scheduling technology, and in particular to a system and method for adaptive adjustment of task priorities for complex projects. Background Technology

[0002] In the task management and scheduling of complex projects, especially in dynamic environments involving multiple tasks, resources, and constraints, effectively determining the execution order of tasks has long been a challenging problem. Current technologies typically calculate task priorities based on a single-dimensional static model, such as prioritizing tasks based on time urgency or resource requirements. However, this single-dimensional priority calculation method struggles to fully reflect the multidimensional attributes and global interrelationships of tasks in complex project environments, resulting in scheduling schemes that often lack flexibility and global optimization capabilities.

[0003] The problem raised in this background technology is that existing scheduling technologies lack a dynamic adjustment mechanism for priorities, resulting in task execution order being unable to adapt to actual production needs. To solve the above problems, this application designs a task priority adaptive adjustment system and method for complex projects. Summary of the Invention

[0004] The technical problem this application aims to solve is to address the shortcomings of existing technologies by providing a system and method for adaptively adjusting task priorities for complex projects. This system acquires multiple task datasets from the target project, extracts multi-dimensional features of time, resources, and dependencies, constructs multiple dependency graphs, and categorizes tasks into first-level, second-level, and third-level tasks. The priority of each level of task is calculated sequentially, and a global task execution order is generated by combining the time urgency of the first-priority tasks, the resource requirements of the second-priority tasks, and the dependencies of the third-priority tasks. When adjustment conditions are triggered, priorities are adjusted sequentially according to the priority hierarchy, and the task execution order is dynamically updated. This application can improve the task scheduling efficiency and global adaptability of complex projects, thereby enhancing productivity and resource utilization.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] An adaptive task priority adjustment method for complex projects, the method comprising:

[0007] Based on the target project, obtain initial data for multiple tasks and generate multiple task datasets;

[0008] The tasks are stratified according to the task dataset, and multiple priorities are calculated in sequence. Each of the multiple priorities represents the priority of a feature dimension. The task execution order is determined according to the multiple priorities.

[0009] The triggering conditions are adjusted according to the preset priorities, and the multiple priorities are adjusted to obtain the corresponding update priorities;

[0010] Update the task execution order according to the update priority.

[0011] The step of stratifying tasks based on the task dataset includes:

[0012] Extract data features from the task dataset, wherein the data features include a first dependency graph, a second dependency graph, and a third dependency graph, wherein the first dependency graph, the second dependency graph, and the third dependency graph have different dependency dimensions;

[0013] Based on the data characteristics, the tasks are divided into first-layer tasks, second-layer tasks, and third-layer tasks according to the dependency dimensions of the first dependency graph, the second dependency graph, and the third dependency graph.

[0014] The extraction of data features from the task dataset includes:

[0015] Extract the time field from the task dataset, and calculate the first dependency graph based on the time field;

[0016] Task elements are matched from the task dataset using predefined task execution rules, and a second dependency graph is calculated based on the task elements.

[0017] The directed associations between the task datasets are extracted, and a third dependency graph is calculated using a topological sorting algorithm.

[0018] The sequential calculation of multiple priorities includes:

[0019] Calculate the first priority based on the time contribution of the first-level task to the target project;

[0020] The second priority is calculated based on the task requirements of the second-level task in resource allocation, combined with the first priority.

[0021] The third priority is calculated based on the execution order of the third-level tasks in the task dependency chain, combined with the first priority and the second priority.

[0022] The calculation of the first priority includes:

[0023] Define the urgency, importance, and completion indicators for the first-level tasks;

[0024] The urgency indicators, importance indicators, and completion indicators are converted into fuzzy judgment matrices in the form of triangular fuzzy numbers, and the fuzzy judgment matrices are subjected to fuzzy consistency tests.

[0025] After the matrix consistency check is passed, the fuzzy judgment matrix is ​​transformed into a fuzzy weight vector, and the first priority is calculated based on the fuzzy weight vector.

[0026] The calculation of the second priority in conjunction with the first priority includes:

[0027] Initialize a reinforcement learning environment, wherein the reinforcement learning environment includes an agent, actions, a state space, and a reward;

[0028] The agent is iteratively updated based on the action, state space, and reward, wherein the iterative update includes policy update, the update step size is based on the first priority, and the update is performed according to the policy error;

[0029] After the number of iterations reaches a preset number or the strategy error is less than the error threshold, a second priority is calculated based on the strategy error.

[0030] The calculation of the third priority by combining the first priority and the second priority includes:

[0031] A task dependency chain is constructed based on the third dependency graph, wherein each node in the task dependency chain represents a third-level task, and the directed edges between nodes represent pre- or post-dependencies between tasks.

[0032] Calculate the initial priority based on the node's first priority, second priority, and node depth;

[0033] The initial priority of the current node is updated through a propagation mechanism, and the initial priority of each node is set as the third priority.

[0034] The adjustment of the multiple priorities to obtain the corresponding update priority includes:

[0035] Based on the triggered adjustment conditions, if the first priority is adjusted, the first priority is adjusted first, and then the second and third priorities are adjusted in sequence to generate an updated priority.

[0036] If the second priority is adjusted, the third priority is adjusted after the second priority is adjusted, and an updated priority is generated.

[0037] If the third priority is adjusted, the third priority of the remaining third-level tasks upstream is updated based on the position of the adjusted third-level task in the task dependency chain, and an updated priority is generated.

[0038] When multiple adjustment conditions are triggered simultaneously, the priorities are adjusted in the order of first priority, second priority, and third priority to generate an updated priority.

[0039] An adaptive task priority adjustment system for complex projects, comprising a task hierarchy module, a priority calculation module, a priority update module, and a task scheduling module;

[0040] The task layering module is used to obtain initial data for multiple tasks based on the target project, generate multiple task datasets, and layer the tasks according to the task datasets.

[0041] The priority calculation module is used to calculate the priority corresponding to each layer based on the layering results;

[0042] The priority update module is used to adjust the triggering conditions according to the preset priority, adjust the multiple priorities, and obtain the corresponding update priority;

[0043] The task scheduling module is used to generate a global task scheduling scheme based on priority, or to adjust the task execution order based on the updated priority.

[0044] Compared with the prior art, the beneficial effects of the present invention are:

[0045] 1. This invention obtains multi-dimensional data features of tasks, divides tasks into layers, calculates the priority of each layer of tasks in turn, generates a global task execution order, and combines a dynamic adjustment mechanism to realize real-time updates of task priorities. This can effectively solve the problems of single task priority calculation, lack of scheduling flexibility and global optimization capability in the prior art.

[0046] 2. This invention comprehensively considers multiple characteristics such as the time urgency of tasks, resource requirements, and dependencies, making the task scheduling scheme more in line with actual production needs; and when the project changes dynamically, it can adjust priorities and update task order in sequence, improving the adaptability and global optimization capability of the scheduling scheme. Attached Figure Description

[0047] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0048] Figure 1 This is a flowchart illustrating the task priority adaptive adjustment method for complex projects according to Embodiment 1 of the present invention.

[0049] Figure 2 This is a schematic diagram of the priority calculation process of the MES system in Embodiment 1 of the present invention;

[0050] Figure 3 This is a schematic diagram of the MES system priority adjustment process in Embodiment 1 of the present invention;

[0051] Figure 4 This is a schematic diagram of the data feature extraction process in Embodiment 1 of the present invention;

[0052] Figure 5 This is a module diagram of a task priority adaptive adjustment system for complex projects according to Embodiment 2 of the present invention. Detailed Implementation

[0053] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0054] Example 1:

[0055] Please see Figure 1 The present invention provides an embodiment of a method for adaptively adjusting task priorities for complex projects, the specific steps of which are as follows:

[0056] S1: Obtain initial data for multiple tasks based on the target project, and generate multiple task datasets;

[0057] In this embodiment, the target project can be a production order received by the Manufacturing Execution System (MES), or other types of complex task management projects, such as supply chain planning, engineering construction projects, or R&D management projects, without limitation.

[0058] In this embodiment, the initial data can be set according to the specifics of the target project. For example, for an MES system, the initial data includes the task details of production orders, delivery time, resource requirements (equipment, personnel, and materials), and dependencies between tasks; for a supply chain planning project, the initial data may include purchase orders, logistics time, warehouse capacity, and supplier constraints; for an engineering construction project, the initial data may include construction node plans, required machinery and equipment and personnel arrangements, and cross-stage task coordination relationships. Specifically, the initial data for the target project can be summarized into three core dimensions: time constraints, resource requirements, and dependencies.

[0059] In this embodiment, the target project can be divided into tasks, breaking down the overall project into several sub-tasks with independent execution conditions. The sub-tasks after task division also include a task dataset with time nodes, resource allocation requirements, and inter-task dependencies.

[0060] S2: Based on the task dataset, the tasks are stratified, and multiple priorities are calculated sequentially;

[0061] In this embodiment, the purpose of task layering is to refine task management in complex projects by clarifying the role and characteristics of tasks within the overall project, thereby providing a structured foundation for dynamic priority calculation. The core logic of layering lies in decomposing tasks according to three dimensions: time, resources, and dependencies, to ensure that the scheduling scheme not only meets the global goals but also efficiently utilizes local resources.

[0062] It is important to note that the task layering in this embodiment is applied to each subtask rather than the overall target project. The first layer of tasks primarily considers the impact of tasks on the global objective. For example, in an MES system, the first layer of tasks represents the delivery time and key process planning for each production order. The second layer of tasks mainly analyzes resources, calculating the priority of each task in resource allocation. The third layer of tasks calculates priorities based on task dependency chains and execution order, focusing on the logical relationships between tasks and local scheduling order.

[0063] Specifically, layering is a structured modeling approach for complex scheduling scenarios. After layering, tasks are prioritized at three levels, which can meticulously reflect their multidimensional characteristics in the global, resource, and dependency chain contexts. For example, in a MES system, time-sensitive tasks with low resource utilization have high factory-level priority but low production line-level priority, allowing the scheduling system to accurately allocate resources.

[0064] S3: Adjust the triggering conditions according to the preset priorities, adjust the multiple priorities, and obtain the corresponding update priority;

[0065] In this embodiment, priority adjustment is based on the sequential adjustment and dynamic adaptation of multi-level priorities. It is driven by trigger conditions and updated sequentially in combination with the logic of hierarchical priorities. Trigger conditions may include, but are not limited to, changes in global targets, abnormal resource status, and changes in task status.

[0066] In this embodiment, when the triggering condition is activated, the priority of the first-level tasks is adjusted first. For example, in a scenario where the global goal changes, the time urgency weight of the tasks is adjusted, the priority of the tasks in the time dimension is recalculated, and thus the priority of the first-level tasks is updated. The adjustment of the priority of the first-level tasks provides a guiding benchmark for subsequent priority adjustments based on the global goal.

[0067] In this embodiment, the priorities of the second-level tasks are adjusted based on the priorities of the first-level tasks. The second-level task priorities, combined with the adjusted results of the first-level tasks, are then optimized for resource allocation. Adjusting resource priorities may include reassessing the matching relationship between tasks and resources; for example, reducing the resource weight of tasks with urgent resource needs but low time priority, while reallocating resources to tasks with higher time priority. The adjustment results are then used to dynamically evaluate resource allocation efficiency and conflict situations through a resource utilization optimization model, ensuring the global rationality of resource allocation.

[0068] In this embodiment, the priorities of the third-level tasks are adjusted based on the updated priorities of the second-level tasks. The third-level tasks primarily depend on the execution order of tasks and the logical relationships within the dependency chains. The adjustment process is based on the directed graph structure of the task dependency chains, dynamically optimizing the task order within the dependency chains by analyzing the priorities of preceding and succeeding tasks. For example, when the resource priority of a task is significantly increased, the execution order of its subsequent tasks will be automatically adjusted to reduce logical conflicts during scheduling. Through the priority propagation mechanism of the dependency chains, the consistency between the logical order of tasks and the global objective is ensured.

[0069] S4: Update the task execution order according to the update priority.

[0070] Please see Figure 2 and Figure 3These are schematic diagrams illustrating the priority calculation process and priority adjustment process of the MES system according to an embodiment of the present invention. Specifically, this embodiment proposes an adaptive task priority adjustment method for complex projects. Taking an MES system as an example, after receiving a production order through the factory's MES system, the production order is decomposed into multiple production tasks and a task dataset is generated. By extracting multi-dimensional features from the task dataset, a time-based task dependency graph, a resource-based task dependency graph, and a sequence-based task dependency graph are constructed. The production tasks are divided into three levels: factory-level tasks, production line-level tasks, and workstation-level tasks, and the priority of each level of tasks is calculated sequentially. The factory-level priority is calculated by defining time urgency, order importance, and completion indicators, and then converting them into fuzzy weight vectors using fuzzy mathematics. The production line-level priority is based on reinforcement learning, dynamically optimizing the matching of tasks and resources by constructing an agent and resource allocation strategy. The workstation-level priority is calculated by constructing a directed graph through task dependency chains, combining the factory-level priority, production line-level priority, and node depth, and dynamically updating the initial priority through a feature propagation mechanism. After priority calculation, the task execution order is determined based on multiple priorities. When changes in task status or resource conditions are detected, priorities are adjusted sequentially based on triggering adjustment conditions to ensure the dynamic adaptability of the global scheduling target. When the triggering condition involves global target adjustment, factory-level priorities are adjusted first, followed by production line-level priorities and then workstation-level priorities. When the triggering condition involves resource allocation issues or changes in task status, relevant priorities are adjusted in a hierarchical manner and propagated to other tasks. The updated priorities are used to adjust and update the task execution order in real time. Ultimately, this achieves global optimization and dynamic adaptation of task scheduling in complex projects, improving production efficiency and resource utilization.

[0071] For example, an MES system receives a production order containing multiple closely related subtasks. These subtasks involve different time points (such as production deadlines), resource requirements (such as equipment, materials, and personnel), and dependencies between tasks (such as a task requiring the completion of a preceding task). Traditional technologies, when faced with such complex projects, often rely solely on static priorities, neglecting the dynamic changes of multi-dimensional factors. In this embodiment, the subtasks of the production order are processed in layers. The first layer calculates priorities based on the time dimension, ensuring that time-sensitive tasks are completed first. The second layer calculates priorities based on the resource dimension, adjusting resource allocation schemes in conjunction with the first layer's priorities to maximize resource utilization efficiency. The third layer calculates priorities based on dependencies, optimizing the consistency of task execution order.

[0072] For example, suppose a production order contains three sub-tasks: sub-task A needs to be completed first to meet time requirements, sub-task B requires reallocation of equipment due to resource constraints, and sub-task C depends on the completion of sub-task A to start. In this case, firstly, the time-based priority is calculated, and the global priority of sub-task A is increased; then, the priority is adjusted based on the resource-based priority, and resources are reallocated for sub-task B; finally, the start time of sub-task C is replanned according to the dependencies to ensure logical consistency of scheduling and efficient use of resources.

[0073] Furthermore, when changes in task status or resource conditions are detected, this embodiment can dynamically adjust priorities based on triggering conditions. For example, assuming a delay in the completion time of subtask A, a recalculation of the time-dimensional priority will be triggered, thereby adjusting its impact on subsequent subtasks B and C. Next, the resource-dimensional priority will update the resource allocation weights of the tasks based on the time adjustment result, ensuring dynamic adaptability of resource allocation. Finally, the dependency-dimensional priority will adjust the execution order of all related tasks through optimized propagation of the task dependency chain. This sequential adjustment mechanism enables task scheduling to quickly respond to changes in the dynamic environment, improving production efficiency and the reliability of order delivery.

[0074] The specific steps of S2 are as follows:

[0075] S2.1: Extract the data features of the task dataset, where the data features include the first dependency graph, the second dependency graph, and the third dependency graph;

[0076] It is understood that in this embodiment, the first dependency graph can be a time-based task dependency graph, the second dependency graph can be a resource-based task dependency graph, and the third dependency graph can be a sequential task dependency graph.

[0077] Specifically, by constructing a first dependency graph, a second dependency graph, and a third dependency graph, the time, resource, and dependency characteristics of tasks are visualized and structured. The first dependency graph extracts task time nodes and related constraint information based on the time field, such as start time, deadline, and estimated duration. It then uses dynamic programming to calculate the time association weights between tasks, forming a time urgency relationship. The second dependency graph extracts resource requirement information from the dataset using predefined task association rules, analyzes the matching relationships between tasks and equipment, personnel, and materials, and combines IoT technology to obtain real-time resource availability and load rates, generating a dynamic mapping relationship between tasks and resources. The third dependency graph constructs a directed graph of tasks based on pre- or post-dependencies, identifies critical paths through topological sorting, and analyzes the execution logic of tasks within the dependency chain. Through the construction of these three dependency graphs, the time urgency, resource requirements, and execution sequence characteristics of tasks can be accurately described.

[0078] Please see Figure 4 The data feature extraction process of this embodiment of the invention is shown in the following diagram. The specific steps of S2.1 are as follows:

[0079] S2.1.1: Extract the time field from the task dataset, and calculate the first dependency graph based on the time field;

[0080] Specifically, the time fields extracted from the task dataset include constraint information such as the task's start time, end time, time limit, earliest start time, and latest finish time. These time fields can be automatically extracted using predefined parsing rules, such as identifying time keywords from the task description using natural language processing techniques, or directly reading standardized time fields from a database.

[0081] In this embodiment, the first dependency graph is a directed acyclic graph, where nodes represent tasks and edges represent the temporal dependencies between tasks. The temporal dependencies are calculated using a dynamic programming algorithm and quantified based on the time span between tasks (such as the difference between the earliest start time and the latest finish time of a task).

[0082] S2.1.2: Match task elements from the task dataset using predefined task execution rules, and calculate a second dependency graph based on the task elements;

[0083] Specifically, by analyzing the types of resources required for a task and their allocation constraints, a mapping relationship between tasks and resources is established. Task elements include, but are not limited to, information such as equipment requirements, personnel requirements, and material requirements. This information is extracted from the task dataset and correlated through predefined task execution rules. For example, equipment requirements can be automatically extracted from the task description or resource allocation table, personnel requirements can be determined based on job information and process requirements, and material requirements are dynamically obtained from the inventory database using IoT technology.

[0084] In this embodiment, nodes in the second dependency graph represent tasks, while edges represent the mapping relationship between tasks and resources. By analyzing resource availability, resource load rate, and resource sharing, weights are added to the dependencies between tasks and resources. The weight values ​​reflect the urgency or degree of resource consumption of the task. To avoid resource conflicts, an optimization algorithm based on conflict resolution rules is used during the construction process to adjust the allocation order between tasks and resources.

[0085] S2.1.3: Extract the directed associations between the task datasets and calculate the third dependency graph using the topological sorting algorithm;

[0086] Specifically, directed relationships between tasks can be obtained by analyzing task preconditions, postconditions, and execution order. For example, logical information such as "must be started after a certain task is completed" or "triggered after a certain task is completed" can be extracted from the task description, and the relationship information can be automatically identified through natural language processing (NLP) and keyword rule extraction techniques.

[0087] In this embodiment, a directed graph (DAG) of task dependency chains is constructed, where nodes represent tasks and edges represent pre- or post-task logical relationships. A topological sorting algorithm is used to sort the directed graph and generate an execution sequence of tasks.

[0088] S2.2: Divide tasks into first-level tasks, second-level tasks, and third-level tasks based on data characteristics;

[0089] It is understood that in this embodiment, if it is for an MES system, the first layer of tasks can be factory-level tasks, the second layer of tasks can be production line-level tasks, and the third layer of tasks can be workstation-level tasks.

[0090] Furthermore, in supply chain management projects, the first-level tasks can be logistics scheduling tasks, primarily reflecting the time objectives and delivery requirements of the entire supply chain. For example, if a supply chain needs to complete the end-to-end delivery of orders at a specific time, procurement planning and transportation scheduling tasks are prioritized in the first level. The second-level tasks can be regional warehousing tasks, mainly focusing on resource allocation, such as adjusting warehouse capacity and optimizing inbound and outbound efficiency. The third-level tasks can be picking or sorting tasks, reflecting the local execution sequence of tasks in material handling. This hierarchical approach allows supply chain management to be refined to the operations of each warehousing node based on global scheduling, improving efficiency and accuracy.

[0091] Furthermore, in R&D projects, the first layer of tasks can be tasks from major R&D phases, such as the conceptual design phase, prototype development phase, or testing phase. These tasks focus on the overall project timeline and key milestone delivery requirements. The second layer of tasks can be task allocation among various R&D teams, such as hardware design tasks, software development tasks, or testing tasks for the testing team, primarily reflecting resource allocation and team collaboration. The third layer of tasks can be specific R&D activities, such as designing a module, writing a piece of code, or running specific test cases, mainly demonstrating the logical sequence and dependencies between tasks. This layered approach can effectively organize the R&D process, ensuring collaborative efficiency and consistency of task objectives among teams.

[0092] Specifically, task layering aims to decompose the scheduling problem of complex projects into manageable sub-problems, enabling priority calculation to be carried out at different levels. The first layer of tasks is based on a time-dependent task graph, representing the contribution of tasks to the global time objective, with priority centered on the time urgency of the tasks. The second layer of tasks is based on a resource-dependent task graph, reflecting the needs and conflicts of tasks for resource allocation, refining scheduling constraints through the resource dimension. The third layer of tasks is based on a sequential task dependency graph, describing the position and logical order of tasks in the dependency chain, playing a key role in the local consistency of the scheduling scheme and the optimization of execution logic. This decouples multi-dimensional scheduling objectives into independent optimization problems at different levels, reducing the complexity of scheduling calculations through a divide-and-conquer approach.

[0093] S2.3: Calculate the first priority based on the time contribution of the first-level tasks to the target project;

[0094] In this embodiment, the calculation of the first priority is quantitatively analyzed using three core factors: urgency index, importance index of the target project, and task completion index. The urgency index measures the urgency of the task time, reflecting the dynamic ratio of remaining time to the deadline; the importance index of the target project is evaluated based on the task's contribution to the overall project goal, highlighting the task's strategic position in the overall project; the completion index quantifies the ratio of completed to incomplete parts of the task, reflecting the actual impact of the task's current execution status on scheduling.

[0095] Specifically, to address the issues of weight allocation bias caused by subjective judgment and the inability to scientifically handle indicator ambiguity in traditional priority calculations, this embodiment employs fuzzy mathematics to construct a fuzzy judgment matrix, quantifying the relative importance of indicators through triangular fuzzy methods. Fuzzy computation introduces consistency checks and defuzzification techniques to ensure the logical rationality and scientific allocation of indicator weights, while dynamically adjusting weights to adapt to real-time changes in task status.

[0096] The specific steps of S2.3 are as follows:

[0097] S2.3.1: Define the urgency, importance, and completion indicators for the first-level tasks;

[0098] In this embodiment, the urgency index is quantified by calculating the ratio of the remaining time of a task to its deadline. Specifically, the remaining time represents the time difference between the current time and the task's deadline. For example, if the remaining time of a task is close to the deadline, its urgency index value will increase significantly, ensuring that the scheduling system can prioritize time-sensitive tasks.

[0099] In this embodiment, importance metrics are measured from the perspective of the target project, specifically quantified based on the target project's contribution to the overall business objectives. For example, in a manufacturing project, certain tasks may be on the critical path of order delivery, and their importance metrics will be higher because delays in these tasks will directly affect the overall delivery target. For a supply chain management project, importance metrics may be related to the order fulfillment capabilities of core suppliers; tasks related to key suppliers will have higher importance metric values ​​to ensure that the supply of core materials is not affected by delays.

[0100] In this embodiment, the completion rate index is used to measure the gap between the actual progress of a task and the expected progress. For example, when the task completion rate is low, the completion rate index is low, thereby prompting the scheduling system to prioritize such tasks.

[0101] S2.3.2: Convert the urgency indicator, importance indicator, and completion indicator into a fuzzy judgment matrix in the form of triangular fuzzy numbers, perform a fuzzy consistency test on the fuzzy judgment matrix, and calculate the fuzzy consistency index.

[0102] Specifically, the urgency indicator, importance indicator, and completion indicator are defined as triangular fuzzy numbers. ,in, This represents the minimum possible value of the indicator, reflecting its weight in the least important case. This represents the maximum possible value of the indicator, reflecting its weight in the most important scenario. It represents the most likely value of the indicator, reflecting the weight of the indicator under typical conditions;

[0103] Construct language evaluation indicators, categorizing language evaluation into equally important, slightly important, significantly important, very important, and extremely important, with corresponding fuzzy numerical values ​​in descending order. , , , and .

[0104] Furthermore, based on the task's time urgency (urgency indicator), the criticality of the target project (importance indicator), and the task completion progress (completion indicator), a triangular fuzzy judgment matrix is ​​constructed by comparing each other pairwise using expert experience and combining it with language evaluation scales. The specific form is as follows:

[0105] ,

[0106] Where M represents the triangular fuzzy judgment matrix, Each element in the matrix represents the relative importance of index i and index j, and is assigned a value according to the language evaluation scale.

[0107] For example, if time urgency is slightly more important than the criticality of the target project, it is assigned the value "slightly more important," and the corresponding triangular fuzzy number is... The inverse relation is denoted as If time urgency is significantly more important than task completion schedule, it is assigned the value "significantly important," and its corresponding triangular fuzzy number is... The inverse relation is denoted as If the criticality of the target project is slightly more important than the task completion schedule, it is assigned the value "slightly important", and its corresponding triangular fuzzy number is . The inverse relation is denoted as Finally, the triangular fuzzy judgment matrix is ​​denoted as:

[0108] .

[0109] By constructing a triangular fuzzy judgment matrix, the relative importance of time urgency, target project criticality, and task completion progress can be quantified.

[0110] Furthermore, the purpose of the fuzzy consistency test is to ensure that the comparison relationships in the judgment matrix satisfy the consistency principle, that is, if urgency is more important than importance, and importance is more important than completion, then urgency should be more important than completion. First, the fuzzy maximum eigenvalue of the fuzzy judgment matrix is ​​calculated. The fuzzy maximum eigenvalue is compared with the dimension n of the fuzzy judgment matrix to calculate the fuzzy consistency weight of each indicator. The consistency ratio is calculated based on the fuzzy consistency weight and the random consistency index RI. Here, RI is a value retrieved from a predefined table based on the dimensions of the matrix. For example, when n=3, RI=0.58. When CR is less than 0.1, the judgment matrix passes the consistency check; otherwise, the triangular fuzzy judgment matrix needs to be readjusted.

[0111] S2.3.3: After the matrix consistency check passes, the fuzzy judgment matrix is ​​transformed into a fuzzy weight vector, and the first priority is calculated based on the fuzzy weight vector;

[0112] Specifically, the summation is performed on each column of the fuzzy judgment matrix to obtain the fuzzy sum of each column. Each element in the matrix is ​​then divided by the fuzzy sum of its column, thus completing the matrix normalization process.

[0113] Furthermore, in the normalized fuzzy matrix, the mean of each row is calculated as the fuzzy weight vector for that indicator. For example, the fuzzy weight values ​​for time urgency, the criticality of the target project, and the task completion progress are calculated by averaging the minimum, most likely, and maximum possible values ​​of each row, respectively, to form the fuzzy weight vector.

[0114] Furthermore, the fuzzy weight vector is defuzzified using the centroid method, transforming the triangular fuzzy number into definite values ​​for urgency, criticality, and completion. The first priority is then calculated based on these three single definite values, using the following formula:

[0115] ,

[0116] in, Indicates first priority. Indicators of urgency for the first-level tasks. Indicators representing the importance of the first-level tasks. This indicates the completion rate of the first-level task. The urgency level of the urgency indicator is determined by its definite value. This indicates the key, definitive value of the importance indicator. This indicates the defined value of the completion rate of the completion indicator.

[0117] S2.4: Calculate the second priority based on the task requirements of the second-level tasks in resource allocation, combined with the first priority;

[0118] For example, taking a MES system as an example, the calculation of the second-layer task priority needs to solve the following problems: competition for resource allocation, dependencies between multiple tasks, and global coordination of time objectives. The action of the production task agent is defined as selecting suitable resources, while the action of the resource agent is defined as allocating resources to suitable tasks. In the state space of reinforcement learning, the resource requirements of the production task (such as equipment type, material requirements, and personnel configuration) and the real-time load of resources (such as current occupancy rate and available time window) are used as core state variables. The reward function is designed to maximize resource utilization and the timeliness of task completion. For example, when the resource allocation efficiency of the production task is high or the time urgency objective is met, the agent will receive a higher positive reward, while if resource allocation conflicts occur or resources cannot meet the task requirements, the agent will be penalized.

[0119] The specific steps of S2.4 are as follows:

[0120] S2.4.1: Initialize the reinforcement learning environment, wherein the reinforcement learning environment includes an agent, actions, a state space, and a reward;

[0121] In this embodiment, the initialization of the reinforcement learning environment is to simulate the interaction between tasks and resource allocation, and to improve resource utilization efficiency and the accuracy of task priority calculation through dynamic optimization. In the reinforcement learning environment, it is defined as a resource allocation optimization model, where each task and resource is considered an agent. Actions are defined as the task agent selecting the required resources based on the current resource state, and the resource agent responding to task allocation requests and adjusting its load. The state space describes dynamic information such as current resource availability, task resource requirements, and load conditions, while the reward is defined as the degree to which the optimization objective is achieved, such as maximizing resource utilization and improving the efficiency of high-priority task completion. This environment construction transforms the resource allocation problem into a dynamic optimization problem, allowing agents to continuously learn and improve their strategies, thereby achieving improvements in global resource utilization and task completion efficiency.

[0122] S2.4.2: The agent is iteratively updated based on the action, state space and reward, wherein the iterative update includes policy update, the update step size is based on the first priority and is updated according to the policy error;

[0123] In this embodiment, the iterative update of the agent is accomplished through a reinforcement learning algorithm, incorporating the task's first priority weight to optimize its resource selection strategy. In each iteration, the task agent selects the optimal resource allocation action based on the current state, while the resource agent responds to the allocation decision based on the load state and provides real-time feedback rewards. The size of the reward depends on the rationality of the resource allocation, such as the degree of resource load balance and the completion efficiency of high-priority tasks. During iteration, the agent adjusts its decision parameters based on the policy error (i.e., the deviation between the current allocation strategy and the optimal strategy), and the update step size is based on the first priority, with higher-priority tasks having larger step sizes to accelerate the strategy optimization process. For example, in an example scenario, if a task has a high time priority but the current resource selection strategy is inefficient, the system will increase the step size to accelerate the convergence process of the task's strategy, ensuring rapid optimization of high-priority tasks.

[0124] S2.4.3: After the number of iterations reaches a preset number or the strategy error is less than the error threshold, calculate the second priority based on the strategy error.

[0125] In this embodiment, the magnitude of the strategy error reflects the quality of the resource allocation strategy. High-priority tasks, due to the influence of their weights, usually converge to a better strategy, thus increasing their second priority. By combining the strategy error with the first priority to calculate the second priority, dynamic optimization of task priorities can be achieved. For example, in a production line scenario, a task with a high first priority, after multiple rounds of iterative optimization of the allocation strategy, will have significantly improved resource matching efficiency, thus its second priority will be higher than other tasks. This method effectively ensures both time urgency and resource allocation efficiency, while improving the overall scheduling efficiency of the production line.

[0126] The formula for calculating the second priority is:

[0127] ,

[0128] in, Indicates second priority. The weighting coefficients represent the impact of policy error. This represents the weighting coefficient indicating the impact of resource utilization. This represents a reinforcement learning algorithm. Representing the state space, Represents the action space. Represents the reward function, This indicates the resource utilization rate of the second-level task;

[0129] The formula for calculating the update step size is as follows:

[0130] ,

[0131] in, Indicates the update step size. Indicates the current step size. This represents the basic update rate in reinforcement learning algorithms. Represents the value function at the current step size. Next, execute the strategy. The cumulative expected reward that can be obtained This represents the policy gradient at the current step size;

[0132] S2.5: Calculate the third priority based on the execution order of the third-level tasks in the task dependency chain, combined with the first and second priorities;

[0133] The specific steps of S2.5 are as follows:

[0134] S2.5.1: Construct a task dependency chain based on the third dependency graph, where each node in the task dependency chain represents a third-level task, and the directed edges between nodes represent pre- or post-dependencies between tasks.

[0135] In this embodiment, the task dependency chain is a directed graph structure built based on the third dependency graph, used to describe the execution logic and dependency relationships between tasks at the third layer. Each task node represents a specific subtask, and the directed edges between nodes represent the dependency constraints between tasks. For example, if task B can only start after task A is completed, then there is a directed edge from task A to task B. The dependency chain is constructed based on the task logic relationships centrally parsed from the task dataset, combined with a topological sorting algorithm to generate an ordered chain of task execution. Furthermore, to improve the accuracy of the dependency chain, the dependency relationships are dynamically updated according to the actual progress of the tasks, nodes corresponding to completed tasks are deleted, and the pre-dependencies of related tasks are adjusted.

[0136] S2.5.2: Calculate the initial priority based on the node's first priority, second priority, and node depth;

[0137] Specifically, the global importance of each task in the dependency chain is quantified by combining the first priority, second priority, and node depth. The first priority reflects the task's time urgency, the second priority reflects its resource requirements and allocation status, and the node depth represents the task's hierarchical position in the dependency chain. Node depth is calculated based on the dependency chain structure and determined using either a depth-first search (DFS) or breadth-first search (BFS) algorithm. A greater depth indicates that the task is further from the final goal, and the stronger the weakening factor on its initial priority. In priority calculation, these three factors are weighted and combined using weighted coefficients to form the task's initial priority.

[0138] S2.5.3: Traverse the directed graph to extract the longest path, and update the initial priority of each node in the longest path according to the average initial priority of the longest path;

[0139] In this embodiment, the extraction of the longest path aims to identify the critical path in the task dependency chain that has the greatest impact on global scheduling. The longest path is extracted by traversing the directed graph, accumulating the task node depth of each path during traversal, and selecting the chain with the longest path length as the critical path. The initial priority of each node on the critical path is adjusted based on the average initial priority of the path to ensure that tasks on the critical path are completed with priority. For example, if the longest path includes tasks A, B, and C, and their initial priorities are 0.8, 0.6, and 0.7 respectively, then its average initial priority is 0.7, and the priority of each task node on the critical path will be adjusted accordingly. The priority update method can be achieved by setting an enhancement factor, such as weighting the average priority of the critical path to the initial priority of the node, ensuring that critical path tasks have a higher priority during scheduling. Alternatively, a non-linear priority enhancement strategy can be used to strengthen the influence of critical path tasks. For example, on the task nodes of the longest path, an exponential priority growth method is used, dynamically adjusting the priority value based on the node's initial priority and the global importance index of the path.

[0140] Specifically, for each node on the path, its priority depends not only on its initial priority but also on an adaptive adjustment based on the priority distribution of other nodes in the path, ensuring that tasks on the overall critical path are completed with higher priority. This avoids the risk of overall path delays due to the low priority of a single node, and enhances the global impact of critical path tasks through a non-linear boosting strategy, thereby further improving the global goal achievement rate and local optimization effect of the task scheduling scheme.

[0141] S2.5.4: Each node receives feature information from its predecessor node and updates the initial priority according to the feature information, wherein the feature information includes the first priority, second priority and initial priority of the predecessor node;

[0142] In this embodiment, task priority updates are based on a feature propagation mechanism of preceding nodes in the task dependency chain. Each task node receives feature information from its direct preceding node, including the preceding node's first priority, second priority, and initial priority. For example, if task A is the preceding node of task B, the priority update formula for task B can be dynamically adjusted based on the feature information of task A. For instance, the time urgency weight of task A can be passed to task B, and a weighted correction can be made by combining task B's own features. The principle of this feature propagation mechanism is that each task in the task chain is affected by the state of its preceding tasks. Feature propagation can dynamically reflect the transitivity of task states in the dependency chain. For example, if the time priority of preceding task A is increased due to a delay, the priorities of subsequent tasks B and C will also be adjusted accordingly to ensure the consistency of the scheduling logic. Through feature propagation between tasks, the task priorities in the dependency chain can be dynamically optimized, ensuring that the scheduling scheme can adapt to changes in task states in real time.

[0143] S2.5.5: After feature propagation is complete, the initial priority of each node is set as the third priority.

[0144] In this embodiment, after feature propagation is completed for all tasks in the dependency chain, the initial priority of each task node is formally determined as the third priority. At this point, the third priority not only reflects the task's time urgency, resource status, and dependency chain position, but also incorporates the influence of other tasks during feature propagation, forming a dynamically adjusted priority result. For example, when the time and resource status of task C's predecessors A and B change, feature propagation can transmit this change to task C in real time, ensuring that task C's third priority remains consistent with the actual scheduling requirements.

[0145] The specific steps for S3 are as follows:

[0146] S3.1: Based on the triggered adjustment conditions, if the first priority needs to be adjusted, the first priority is adjusted first, and then the second and third priorities are adjusted in sequence to generate an updated priority.

[0147] In this embodiment, if the triggering condition involves a change in the global objective (e.g., the delivery time of a production order is brought forward or delayed), the first priority needs to be adjusted. Specifically, the adjustment of the first priority is centered on the time dimension, dynamically adjusting the priority by analyzing the deadline, remaining time, and task progress status in the task's time field. For example, when the delivery time of a production order is brought forward, the time urgency weight of the related tasks will increase, thus recalculating the first priority. The adjusted first priority is directly used as input for the second priority, re-evaluating the task's priority in resource allocation. Through resource mapping and a dynamic allocation model, the resource load is updated in conjunction with the adjusted first priority, for example, prioritizing the allocation of critical equipment and personnel for urgent tasks. Subsequently, based on the updated second priority, the third priority is adjusted. In the dependency chain, the logical order of tasks is affected by the linkage of the first two levels of priority. For example, when a task needs to be executed earlier due to an increase in time priority, the dependencies of its subsequent tasks need to be dynamically adjusted. This layer-by-layer adjustment mechanism ensures the global consistency of priority adjustment, effectively avoiding task execution conflicts caused by a single level of adjustment, thereby improving the robustness and global optimization level of the scheduling scheme.

[0148] S3.2: If the second priority needs to be adjusted, adjust the second priority first, then adjust the third priority, and generate an updated priority.

[0149] In this embodiment, if the triggering condition involves resource allocation issues (e.g., equipment failure, resource overload, or insufficient personnel), the second priority needs to be adjusted. Specifically, by monitoring resource availability and load rate in real time and combining this with a resource dependency graph, the matching relationship between tasks and resources is reassessed. The adjustment process dynamically reduces the priority of tasks with significant resource conflicts while increasing the priority of tasks with high resource utilization efficiency. The adjusted second priority directly affects the update of the third priority, and through the optimized propagation of the dependency chain, the execution order of tasks is replanned. For example, when a task needs to reallocate resources due to equipment priority adjustments, the execution time of its subsequent dependent tasks also needs to be adjusted to ensure the consistency of the logical order of the dependency chain. By dynamically optimizing resource allocation, not only can resource conflicts be avoided, but also the execution of time-sensitive tasks can be prioritized in situations of resource scarcity, thereby improving resource utilization efficiency and the flexibility of task scheduling.

[0150] S3.3: If the third priority needs to be adjusted, after adjusting the third priority, update the third priority of the remaining third-level tasks upstream according to the position of the adjusted third-level task in the task dependency chain, and generate the updated priority.

[0151] In this embodiment, if the triggering condition involves a change in task status (e.g., task delay, decreased completion rate, or broken dependency chain), the third priority needs to be adjusted. Specifically, through directed graph analysis of the task dependency chain, the predecessor and successor tasks of the affected task are identified, and their priorities are re-evaluated. For example, when a critical task is delayed, the priority of its successor tasks is automatically increased, and the critical path in the dependency chain is recalculated to optimize the task execution order. Simultaneously, the task adjustment propagates upstream, such as adjusting the remaining priority of predecessor tasks, to avoid impacting the global plan on local adjustments. Through this dependency chain optimization propagation mechanism, it can be ensured that the adjusted priority maintains logical consistency globally, especially in scenarios with frequent dynamic changes, enabling rapid adaptation to environmental changes and ensuring the stability and reliability of task scheduling.

[0152] S3.4: When multiple adjustment conditions are triggered simultaneously, the priorities are adjusted in the order of first priority, second priority, and third priority to generate an updated priority.

[0153] In this embodiment, when multiple adjustment conditions are triggered simultaneously, such as earlier order delivery times, resource allocation conflicts, and task delays occurring concurrently, the adjustment priority follows a logic from first priority to third priority. First, the first priority is adjusted with the global goal as the core, for example, by dynamically updating the time urgency weight to ensure the achievement of the global time goal. The adjusted first priority is then passed as input to the second priority to dynamically optimize resource allocation strategies, for example, by prioritizing resource allocation for urgent tasks to resolve resource conflicts. Finally, the third priority is updated by combining the first two priority levels, and the task execution order is re-planned through dependency chain analysis to ensure the consistency of the adjustment results in task logic. For example, when a production line task is delayed due to equipment failure, resources are reallocated by adjusting the time priority, and the logical order of related tasks in the dependency chain is adjusted simultaneously to ensure that the impact of the delay is minimized. This sequential adjustment method ensures global consistency in priority adjustment, avoids priority conflicts that may occur when multiple adjustment conditions are triggered simultaneously, and guarantees the flexibility and robustness of the scheduling scheme.

[0154] In this embodiment, determining the task execution order and updating the task execution order based on multiple priorities specifically includes:

[0155] A weighted model based on three priorities—time, resources, and dependencies—determines the overall priority of each task. The weight of the time priority determines the task's importance in meeting the overall delivery goals, while the weight of the resource priority is dynamically adjusted based on the current resource status (such as equipment occupancy and material inventory). Dependency priorities further calibrate the logical order between tasks, thereby ensuring the rationality and efficiency of the final execution order. For example, in a production order within an MES system, if subtask A has a higher time priority than subtask B, but its resource priority is lower, the overall priority model will combine the priority values ​​of both and dynamically adjust their execution order according to the project's overall optimization goals.

[0156] The specific process of updating the task execution order is based on a multi-level, progressive logic. First, globally urgent tasks are recalculated based on time-based priority adjustments. For example, when an order delivery time is brought forward, the system immediately increases the priority of urgent tasks and propagates this adjustment backward in the dependency chain, reordering the execution order of related tasks. Next, resource-based priority adjustments optimize the allocation of critical resources. For instance, through dynamic allocation algorithms, high-priority tasks are assigned to readily available equipment or materials, thus avoiding scheduling delays caused by resource conflicts. Finally, based on a task dependency adjustment mechanism, the task order is updated in the directed graph of the dependency chain. For example, when a task is delayed, the system readjusts the order of its subsequent tasks through graph topology updates, ensuring the logical consistency of the entire scheduling scheme.

[0157] Example 2:

[0158] Please see Figure 5 This invention provides an embodiment of a task priority adaptive adjustment system for complex projects, the system comprising a task layering module, a priority calculation module, a priority update module, and a task scheduling module;

[0159] The task layering module is used to obtain initial data for multiple tasks based on the target project, generate multiple task datasets, and layer the tasks according to the task datasets.

[0160] The priority calculation module is used to calculate the priority corresponding to each layer based on the layering results;

[0161] The priority update module is used to adjust the triggering conditions according to the preset priority, adjust the multiple priorities, and obtain the corresponding update priority;

[0162] The task scheduling module is used to generate a global task scheduling scheme based on priority, or to adjust the task execution order based on the updated priority.

[0163] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for adaptively adjusting task priorities for complex projects, characterized in that: The method includes: Based on the target project, obtain initial data for multiple tasks and generate multiple task datasets; The tasks are stratified according to the task dataset, and multiple priorities are calculated in sequence. Each of the multiple priorities represents the priority of a feature dimension. The task execution order is determined according to the multiple priorities. The triggering conditions are adjusted according to the preset priorities, and the multiple priorities are adjusted to obtain the corresponding update priorities; Update the task execution order according to the update priority; The step of stratifying tasks based on the task dataset includes: Extract data features from the task dataset, wherein the data features include a first dependency graph, a second dependency graph, and a third dependency graph, wherein the first dependency graph, the second dependency graph, and the third dependency graph have different dependency dimensions; Based on the data characteristics, the tasks are divided into first-layer tasks, second-layer tasks, and third-layer tasks according to the dependency dimensions of the first dependency graph, the second dependency graph, and the third dependency graph. The sequential calculation of multiple priorities includes: Calculate the first priority based on the time contribution of the first-level task to the target project; The second priority is calculated based on the task requirements of the second-level task in resource allocation, combined with the first priority. The third priority is calculated based on the execution order of the third-level tasks in the task dependency chain, combined with the first priority and the second priority. The calculation of the second priority in conjunction with the first priority includes: Initialize a reinforcement learning environment, wherein the reinforcement learning environment includes an agent, actions, a state space, and a reward; The agent is iteratively updated based on the action, state space, and reward, wherein the iterative update includes policy update, the update step size is based on the first priority, and the update is performed according to the policy error; After the number of iterations reaches a preset number or the strategy error is less than the error threshold, a second priority is calculated based on the strategy error. The calculation of the third priority by combining the first priority and the second priority includes: A task dependency chain is constructed based on the third dependency graph, wherein each node in the task dependency chain represents a third-level task, and the directed edges between nodes represent pre- or post-dependencies between tasks. Calculate the initial priority based on the node's first priority, second priority, and node depth; The initial priority of the current node is updated through a propagation mechanism, and the initial priority of each node is set as the third priority.

2. The task priority adaptive adjustment method for complex projects according to claim 1, characterized in that, The extraction of data features from the task dataset includes: Extract the time field from the task dataset, and calculate the first dependency graph based on the time field; Task elements are matched from the task dataset using predefined task execution rules, and a second dependency graph is calculated based on the task elements. The directed associations between the task datasets are extracted, and a third dependency graph is calculated using a topological sorting algorithm.

3. The task priority adaptive adjustment method for complex projects according to claim 1, characterized in that, The calculation of the first priority includes: Define the urgency, importance, and completion indicators for the first-level tasks; The urgency indicators, importance indicators, and completion indicators are converted into fuzzy judgment matrices in the form of triangular fuzzy numbers, and the fuzzy judgment matrices are subjected to fuzzy consistency tests. After the matrix consistency check is passed, the fuzzy judgment matrix is ​​transformed into a fuzzy weight vector, and the first priority is calculated based on the fuzzy weight vector.

4. The task priority adaptive adjustment method for complex projects according to claim 1, characterized in that, The adjustment of the multiple priorities to obtain the corresponding update priority includes: Based on the triggered adjustment conditions, if the first priority is adjusted, the first priority is adjusted first, and then the second and third priorities are adjusted in sequence to generate an updated priority. If the second priority is adjusted, the third priority is adjusted after the second priority is adjusted, and an updated priority is generated. If the third priority is adjusted, the third priority of the remaining third-level tasks upstream is updated based on the position of the adjusted third-level task in the task dependency chain, and an updated priority is generated.

5. The task priority adaptive adjustment method for complex projects according to claim 1, characterized in that, When multiple adjustment conditions are triggered simultaneously, adjusting the multiple priorities to obtain the corresponding update priority further includes: The priorities are adjusted sequentially in the order of first priority, second priority, and third priority to generate an updated priority.

6. A task priority adaptive adjustment system for complex projects, used to implement the task priority adaptive adjustment method for complex projects according to any one of claims 1-5, characterized in that, The system includes a task layering module, a priority calculation module, a priority update module, and a task scheduling module; The task layering module is used to obtain initial data for multiple tasks based on the target project, generate multiple task datasets, and layer the tasks according to the task datasets. The priority calculation module is used to calculate the priority corresponding to each layer based on the layering results; The priority update module is used to adjust the triggering conditions according to the preset priority, adjust the multiple priorities, and obtain the corresponding update priority; The task scheduling module is used to generate a global task scheduling scheme based on priority, or to adjust the task execution order based on the updated priority.

Images