Multi-task real-time scheduling optimization method based on dependent perception information age

Abstract

The invention provides a multi-task real-time scheduling optimization method based on dependency perception information age, and provides a brand-new optimization scheme for overcoming the defects of a traditional information age model in the aspects of processing multi-task dependency and multi-source heterogeneity. By defining a dependency perception information age model, the comprehensive influence of the freshness of task data and dependency task completion delay is comprehensively quantified, and the scheduling performance of the system in a complex industrial scene is remarkably improved. The method combines the dependency relationship of the task chain, dynamically optimizes the scheduling strategy, preferentially processes the critical path task, and adjusts the task priority and the resource allocation strategy in real time. Under a discrete time linear time-invariant system framework, a linear quadratic Gaussian control method is combined to evaluate the overall performance of a scheduling scheme. In addition, a multi-agent deep reinforcement learning framework is utilized, the task sampling frequency and the resource allocation and transmission strategy are optimized, and the robustness and overall performance of the system are further improved.

Description

Technical Field

[0001] The present invention relates to the field of industrial Internet task scheduling and real-time optimization, and more specifically, to a multi-task real-time scheduling optimization method based on dependency-aware age of information. Background Art

[0002] With the rapid development of industrial Internet technology, industrial production is gradually transforming from informatization to intelligence. In this process, the requirements for task scheduling and information processing become more complex, not only need to adapt to diverse task types, but also need to handle the dependency relationships between tasks and the integration requirements of multi-source heterogeneous data. In modern industrial scenarios, the real-time nature of tasks and the collaborative scheduling between multiple tasks are of great significance for improving production efficiency and system stability.

[0003] Currently, many industrial scenarios have problems such as uneven resource allocation, low overall resource utilization, and insufficient real-time performance in task scheduling. For example, the demands for network bandwidth and computing resources vary significantly among different tasks, and there are competitive and sharing relationships in resource occupancy. The impact of multi-task dependency relationships and heterogeneous data sources on the overall system performance has not been fully quantified and optimized. Designing an efficient real-time scheduling optimization method under the conditions of complex multi-task dependencies and resource constraints is an important technical challenge.

[0004] Traditional task scheduling methods usually evaluate the timeliness of single tasks, only focusing on the data update frequency of tasks or a single completion time constraint, and it is difficult to comprehensively reflect the interaction between task dependencies and global performance. The age of information is used to measure the information timeliness of tasks, but in multi-task dependency and multi-source heterogeneous scenarios, it is difficult to comprehensively consider the delay propagation effect in the task chain and the heterogeneous characteristics in multi-source data fusion, and the system performance optimization faces bottlenecks. Summary of the Invention

[0005] Aiming at the performance optimization problems in current industrial Internet task scheduling caused by task dependencies, multi-source heterogeneity, and real-time requirements, the present invention proposes a multi-task real-time scheduling optimization method based on dependency-aware age of information. By combining the update status of task data with the dependency relationships of task chains, a task timeliness evaluation index is constructed to optimize the collaborative scheduling of multi-tasks, providing an efficient solution for complex scenarios in the industrial Internet. By defining a dependency-aware age of information model, combining the dependency relationships between multi-tasks and data timeliness, the task scheduling strategy and resource allocation scheme are optimized to improve the overall performance and real-time nature of the task chain. This method is applicable to complex scenarios such as industrial manufacturing, Internet of Things, and edge computing that require high real-time performance and multi-task collaborative scheduling.

[0006] The technical solution adopted by the present invention to achieve the above object is: a multi-task real-time scheduling optimization method based on dependence-aware information age, comprising the following steps:

[0007] 1) Dependent task scenario modeling: By analyzing the dependence relationship between tasks in the task chain, establish a dependent task scenario and define the attributes of each task;

[0008] 2) Dependence-aware information age calculation: Based on the attributes of the task, calculate the dependence-aware information age of the task to quantify the relationship between the data state of the current task and the completion time of its dependent tasks;

[0009] 3) Scheduling strategy optimization: Based on the dependence-aware information age value of the task, optimize the scheduling strategy of multi-tasks and dynamically adjust the task priorities;

[0010] 4) Control performance evaluation: Using the discrete-time linear time-invariant system framework and combining with the evaluation of the global performance of the scheduling strategy, obtain the control cost equivalent function;

[0011] 5) Multi-agent deep reinforcement learning optimization: According to the dependence-aware information age of the task and the control cost equivalent function, optimize the scheduling strategies of multiple task nodes through the multi-agent deep reinforcement learning framework.

[0012] In step 1), the attributes include the data collection time of the task, the set of dependent tasks and the completion time of the dependent tasks.

[0013] In step 1), the dependent task scenario modeling includes at least one of the following modeling methods:

[0014] 1.1) Serial dependence modeling:

[0015] Based on the task chain, determine the serial dependence relationship of the tasks and generate a serial task dependence graph; where the start time of the task satisfies: t start，i is the start time of task i, t end，j is the end time of the dependent task j, D i represents the set of dependent tasks of task i; the dependent task refers to the previous task that must be completed before a certain task is executed;

[0016] 1.2) Parallel dependence modeling:

[0017] Based on the task chain, determine the parallel dependence relationship of the tasks, generate a parallel task dependence graph, and label the multi-path dependence relationship of the tasks; the completion time of the dependent task is: t dep，i is the dependent completion time of task i;

[0018] 1.3) Hybrid dependence modeling:

[0019] Determine the serial and parallel dependency relationships of tasks based on the task chain, and generate a mixed dependency task chain diagram; the mixed dependency task chain diagram includes a serial task dependency diagram and a parallel task dependency diagram.

[0020] In step 2), the calculation of the dependency-aware information age is as follows:

[0021] Calculate the dependency-aware information age of the task, and its value is the difference between the current time, the latest data collection time of the task, and the latest completion time of its dependent tasks:

[0022]

[0023] Among them, is the dependency-aware information age, k is the current time, and τ i (k) is the latest data collection time of task i, and D i is the set of dependent tasks of task i, and C j (k) is the completion time of the dependent task j ∈ D i .

[0024] In step 3), the optimization of the scheduling strategy includes the following steps:

[0025] Based on the dependency relationships of tasks in the dependent task scenario modeling, use the topological sorting algorithm to dynamically adjust the task priorities to identify the key tasks that affect the global performance in the critical path, and schedule them as the task nodes to be preferentially scheduled;

[0026] Among them, the dynamic adjustment of task priorities is as follows:

[0027]

[0028] Among them, P i is the task priority, R i is the resource requirement of task i, is the delay cost of task i,.f represents the priority sorting function in the topological sorting algorithm, which is used to assign priorities to tasks according to task attributes including dependency-aware information age, resource requirement, and delay cost.

[0029] In step 4), the control performance evaluation includes the following steps:

[0030] (1) For the system composed of the terminal device and the edge server, construct a discrete-time linear time-invariant system model to describe the state evolution of the task; the evolution relationship of the task state is as follows:

[0031] x i，k+1 = A i x i，k + B iu i，k + ω i，k

[0032] where is the system task status vector, used to characterize the status of the terminal device, is the control input vector generated by the edge server, used to characterize the control quantity of the terminal device, and the constant matrix is the system matrix, is the input matrix of the edge server, i d represents the state space dimension of the task, i n represents the control input vector space dimension; is the zero-mean Gaussian random exogenous interference noise;

[0033] (2) Use the linear quadratic Gaussian function to evaluate the control cost equivalent function of each task system, and the calculation is as follows:

[0034]

[0035] where W i，N and W i are the system state weight matrices, U i is the control input weight matrix, is used to measure the control consumption, is used to measure the state error of the actual system estimated based on the available measurement values; is used to express the final error, represents the expected value operation; k represents the k-th time step, and N represents the total number of time steps;

[0036] (3) According to the evolution relationship of the task status and the control cost equivalent function, construct the control cost equivalent function based on the age of dependence-aware information:

[0037]

[0038] where, e i,k is the control system estimation error, is the system task status estimation vector, is the age of dependence-aware information;

[0039]

[0040] where, J C，i is the control cost equivalent function, represents the expected value operation, ∑ i,k is the covariance of e i，k , S i is the solution of the Riccati equation, and Tr(·) represents the trace operation of the matrix.

[0041] In step 5), the multi-agent deep reinforcement learning optimization includes the following steps:

[0042] 5.1) State space definition: Define the state space in multi-agent deep reinforcement learning

[0043]

[0044] where s i (k) represents the state of task i at the k-th time step, Φ i (k) is the data size, C i，m，k (k) is the computing resource requirement, B i,m,k (k) is the bandwidth, h i，m，k (k) is the channel power gain, is the set of dependent tasks of the task, is the set of subsequent tasks of the task, is the age of information for dependency awareness;

[0045] 5.2) Action space design: Construct the action space of multi-agent deep reinforcement learning to present the policies of all tasks:

[0046]

[0047] where a i (k) represents the executed action, is the decision on the matching of computing resource types, is the task division ratio, p i,m,k (k) is the transmission power of the sensor in the terminal device, f i，m，k (k) is the computing resource allocation policy of the edge server, τ i，m，k (k) is the task sampling frequency;

[0048] 5.3) The reward function is:

[0049] where J C，i is the control cost, J N，i is the network energy consumption cost, ρ i is the weight factor of the age of information reward term for dependency awareness, Δ th is the threshold of the age of information for dependency awareness, α ∈ [0, 1] and β ∈ [0, 1] are the corresponding weight factors, and α + β = 1;

[0050] 5.4) Multi-agent collaborative optimization: According to the obtained computing resource allocation policy f i，m，k (k) of the edge server and the task sampling frequency τ i，m，k(k) to optimize the scheduling strategy of multiple task nodes.

[0051] A multi-task real-time scheduling optimization system based on dependency-aware information age, comprising:

[0052] A dependency task scenario construction module for establishing a dependency task scenario and defining the attributes of each task by analyzing the dependency relationships between tasks in a task chain;

[0053] A dependency-aware information age calculation module for calculating the dependency-aware information age of a task based on the attributes of the task to quantify the relationship between the data state of the current task and the completion time of its dependent tasks;

[0054] A scheduling strategy optimization module for optimizing the scheduling strategy of multi-tasks and dynamically adjusting task priorities based on the dependency-aware information age values of tasks;

[0055] A control performance evaluation module for obtaining a control cost equivalent function by using a discrete-time linear time-invariant system framework and combining the evaluation of the global performance of the scheduling strategy;

[0056] A multi-agent deep reinforcement learning optimization module for optimizing the scheduling strategy of multiple task nodes through a multi-agent deep reinforcement learning framework according to the dependency-aware information age of tasks and the control cost equivalent function.

[0057] A computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the multi-task real-time scheduling optimization method based on dependency-aware information age is implemented.

[0058] The present invention has the following beneficial effects and advantages:

[0059] 1. By introducing a dependency-aware information age model, the present invention comprehensively quantifies the data freshness of tasks and the timeliness of dependency relationships, and optimizes task priorities and scheduling strategies. Compared with traditional methods, it can effectively reduce the impact of delay propagation in the task chain on the global performance, and significantly improve the real-time performance of complex task scenarios and the global optimization ability of the system.

[0060] 2. In a multi-source heterogeneous scenario, the present invention reduces the uncertainty of data synchronization by dynamically adjusting the data source sampling frequency, transmission delay, and resource allocation strategy, and improves the efficiency and accuracy of data fusion tasks. At the same time, combined with a multi-agent deep reinforcement learning framework, it optimizes the resource allocation and task scheduling scheme, and improves the resource utilization efficiency.

[0061] 3. The present invention is based on a discrete-time linear time-invariant system model and an error feedback mechanism to ensure the stability and fault tolerance of task scheduling in a dynamically changing environment. It is applicable to complex scenarios such as industrial manufacturing, the Internet of Things, and edge computing, and has good scalability and application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 is a schematic diagram of the method of the present invention;

[0063] Figure 2 is an overall system architecture diagram considered by the present invention. SPECIFIC IMPLEMENTATION METHODS

[0064] The following further elaborates on the present invention in conjunction with the drawings and examples.

[0065] The present invention provides a multi-task real-time scheduling optimization method based on dependency-aware age of information, and provides a new optimization solution for the deficiencies of traditional age of information models in dealing with multi-task dependencies and multi-source heterogeneity. By defining a dependency-aware age of information model, it comprehensively quantifies the combined impact of the freshness of task own data and the completion delay of dependent tasks, significantly improving the scheduling performance of the system in complex industrial scenarios. This method combines the dependency relationships of task chains, dynamically optimizes the scheduling strategy, preferentially processes tasks on the critical path, and simultaneously adjusts the task priority and resource allocation strategy in real time. Under the framework of a discrete-time linear time-invariant system, the linear quadratic Gaussian control method is used to evaluate the global performance of the scheduling scheme. In addition, a multi-agent deep reinforcement learning framework is used to optimize the task sampling frequency, resource allocation, and transmission strategy, further improving the robustness and global performance of the system. The present invention is applicable to complex scenarios such as industrial Internet and edge computing, meeting the requirements of high real-time, multi-source heterogeneous data fusion, and task chain optimization.

[0066] For multi-dependency process tasks executed at the terminal in an industrial control scenario, the present invention performs task modeling, parameter calculation, and numerical analysis in an edge server, and uses a multi-agent deep reinforcement learning framework to optimize the task sampling frequency, resource allocation, and transmission strategy. Figure 1 The following shows a schematic diagram of the method of the present invention, and the specific implementation process is as follows:

[0067] Step 1: Dependency task scenario modeling:

[0068] For an industrial control scenario, based on multi-dependency process tasks executed at the terminal, a dependency relationship modeling in any of the following ways is constructed.

[0069] Step 1.1: Series dependency modeling:

[0070] Model the serial relationship in the task chain and analyze the relationship between the direct predecessor tasks and successor tasks of the tasks. During the modeling process, assign the completion time of the predecessor tasks to each task to ensure that the start time of the successor tasks occurs after the completion of the predecessor tasks. By sorting out the task order in the task chain, generate a clearly structured serial task dependency graph that accurately reflects the sequential dependency relationship in the task chain. The start time of the task satisfies: where t start，i is the start time of task i, t end，j is the end time of the dependent task j, and D i represents the set of dependent tasks of task i.

[0071] Step 1.2: Parallel Dependency Modeling:

[0072] For the parallel relationship in the task chain, analyze the multiple predecessor tasks on which the task depends and their mutual relationship. For each task, calculate the maximum value of the completion times among all the predecessor tasks as the dependent completion time of this task, that is: where t dep，i is the dependent completion time of task i. Relying on the in-depth analysis of the multi-path dependencies between tasks, generate a parallel task dependency graph, mark the parallel relationships and dependency paths between tasks, and ensure the logical consistency of parallel tasks in the dependency chain.

[0073] Step 1.3: Hybrid Dependency Modeling:

[0074] When dealing with a hybrid dependency task chain, consider the structural characteristics of both serial and parallel at the same time to construct a task chain model. By analyzing each task in the hybrid dependency task chain, clarify its dependency relationship with the predecessor tasks and parallel tasks. Combine the dependency relationship with the overall structure of the task chain to generate a complete hybrid dependency task chain model.

[0075] Step 2: Dependence-Aware Information Age Calculation:

[0076] Step 2.1: Definition of Dependence-Aware Information Age:

[0077] Calculate the dependence-aware information age of the task, which is the difference between the current time and the latest data collection time of the task and the latest completion time of its dependent tasks:

[0078]

[0079] where k is the current time, τ i (k) is the latest data collection time of task i, D i is the set of dependent tasks of task i, and C j (k) is the completion time of the dependent task j ∈ D i .

[0080] Step 2.2: Global Delay Propagation Effect Quantification:

[0081] Evaluate the data freshness of the current task and the timeliness of the dependency relationship using the dependency-aware information age value. By comparing the actual completion time and the ideal completion time of the task chain, measure the real-time deviation of the task chain, and dynamically adjust the task priority P and resource allocation strategy in combination with the evaluation results to ensure that the overall real-time requirements of the task chain are met.

[0082] Step 2.3: Multi-source Heterogeneous Data Analysis:

[0083] Based on the calculation results of the dependency-aware information age, comprehensively analyze the delay characteristics and update frequencies of different data sources, optimize the data fusion strategy, and dynamically adjust the sampling frequency and transmission delay of each data source.

[0084] Step 3: Scheduling Strategy Optimization:

[0085] Step 3.1: Critical Path Identification:

[0086] Based on the dependency relationship of the task chain, use the topological sorting algorithm to identify the critical tasks that affect the global performance in the critical path. By analyzing the contribution degree of tasks in the task chain to the global delay, clarify the task nodes that need to be scheduled preferentially in the critical path.

[0087] Step 3.2: Dynamically Adjust Task Priority:

[0088] According to the dependency-aware information age value and the global dependency relationship of the task chain, update the task priority in real time. Give priority to completing the critical path tasks, reduce the propagation effect of the global delay, and optimize the resource allocation strategy. The calculation formula for the task priority P is:

[0089]

[0090] where R i is the resource requirement of task i, is the delay cost of task i. f represents the priority sorting function in the topological sorting algorithm. Compared with the existing priority sorting function, the index adds the dependency-aware information age, which assigns priorities to tasks according to task attributes (dependency-aware information age, resource requirement, delay cost).

[0091] Step 3.3: Multi-task Collaboration Optimization:

[0092] Considering the series and parallel relationships in the task chain comprehensively, balance the resource allocation by dynamically optimizing the scheduling order to improve the overall completion efficiency of the task chain and the system performance.

[0093] Step 4: Control Performance Evaluation:

[0094] Step 4.1: Control System Modeling:

[0095] For the overall architecture of the system as Figure 2 shown, considering that there are multiple sensor networks and multiple edge servers in the overall system. There are multiple sensor networks and multiple edge servers in the physical space. Multiple sensor networks generate tasks with series and parallel dependencies, and these tasks are assigned to different edge servers through a scheduling optimization strategy to ensure the efficient operation of the overall system. Therefore, a discrete-time linear time-invariant system model is constructed to describe the state evolution of the tasks. Combining the dependency relationship and completion time constraint of the tasks, the change of the task state during the scheduling process is calculated. Among them, each sensor network includes multiple terminal devices, and each edge server covers one or more sensor networks. For a system composed of multiple terminal devices (robots) and multiple edge servers, the evolution relationship of the task state is as follows:

[0096] x i，k+1 = A i x i,k + B i u i，k + ω i，k

[0097] Where, is the system task state vector, is the control input vector generated by the edge control decision unit, and the constant matrix is the system matrix, is the input matrix of the control unit. The stochastic process is a zero-mean Gaussian random exogenous interference noise with covariance Ξ. For the actual robot control scenario, x i，k is the position and speed of the robot, etc., and u i，k is the steering angle and acceleration of the robot, etc. For the temperature control scenario, x i，k is the temperature information of the system, and u i，k is the information such as the fan speed of the heat dissipation device, etc.

[0098] Step 4.2: Global Performance Index Calculation:

[0099] Evaluate the scheduling scheme based on the global performance index and calculate the control cost of the system. Use the linear quadratic Gaussian function to evaluate the control cost of each task system, and the calculation is as follows:

[0100]

[0101] Where, W i，N and W i are the system state weight matrices, and U i is the control input weight matrix, is used to measure control consumption, is used to measure the state error of the actual system estimated based on available measurement values. What is used to express is the terminal error. represents the expected value operation, and N represents the total number of time steps.

[0102] Step 4.3: Error feedback and optimal control analysis:

[0103] Perform error feedback and optimal control analysis on the execution effect of the scheduling strategy, and adjust the task priority and resource allocation strategy in real time. Combining error accumulation, propagation analysis, and optimal control theory, optimize the stability and timeliness of global scheduling to ensure the robustness and efficiency of the system in a dynamically changing environment.

[0104] The system state evolution equation of the control system is:

[0105]

[0106] The estimated state equation of the control system is:

[0107]

[0108] Furthermore, the estimated error of the control system is obtained as:

[0109]

[0110] According to the covariance definition: and combined with the optimal control law the control cost equivalent function can be obtained:

[0111] where, ∑ i，k is the covariance of e i，k L i is the gain matrix, S i is the solution of the Riccati equation, Tr(·) represents the trace operation of the matrix, and r represents the r-th time step.

[0112] The specific implementation process of the multi-agent algorithm in the present invention is as follows.

[0113] Step 5: Multi-agent deep reinforcement learning optimization:

[0114] Step 5.1: State space definition:

[0115] Define the state space in multi-agent deep reinforcement learning The state space describes the running state of the task and the edge computing and communication resources. At each decision period k, the state s of the i-agent i(k) is characterized by data size, computing resource requirements, bandwidth, channel power gain, task dependencies (predecessor tasks and successor tasks), and dependency-aware age of information, i.e.:

[0116]

[0117] Among them, Φ i (k) is the data size, C i,m,k (k) is the computing resource requirement, B i,m,k (k) is the bandwidth, h i，m，k (k) is the channel power gain, is the set of predecessor tasks of the task, is the set of successor tasks of the task, is the dependency-aware age of information. The dependent task is for the current task, and the successor task represents the next task or multiple tasks of the current task, which is for the current task.

[0118] Step 5.2: Action space design:

[0119] Design the action space of multi-agent deep reinforcement learning The action space presents the policies of all agents. At each decision epoch k, the i-agent executes the action a i (k) according to the entire state s i (k). The action space a i (k) is given by the actions of computing resource type matching decision, task division ratio, sensor system transmission power, edge server computing resource allocation policy, and task sampling frequency as follows:

[0120]

[0121] Among them, is the computing resource type matching decision, is the task division ratio, p i,m,k (k) is the sensor system transmission power, f i，m，k (k) is the edge server computing resource allocation policy, τ i，m，k (k) is the task sampling frequency.

[0122] Step 5.3: Reward function design:

[0123] The reward function is: Among them, J C，i is the control cost, and J N，i is the network energy consumption cost, ρ i is the weight factor of the dependency-aware age of information reward term, Δ this the threshold depending on the age of the sensed information. Additionally, α ∈ [0, 1] and β ∈ [0, 1] are the corresponding weight factors, and α + β = 1. Optimize the multi-task scheduling strategy based on the reward function, and dynamically adjust the scheduling strategy by evaluating the task completion rate, system energy consumption, and the change in the age value of the sensed information, so as to balance the timeliness of tasks and resource utilization.

[0124] Step 5.4: Multi-agent collaborative optimization:

[0125] Obtain the computing resource allocation strategy f i，m，k (k), the terminal task sampling frequency τ i，m，k (k) and other agent actions, and collaboratively optimize the scheduling strategies of multiple task nodes under the multi-agent deep reinforcement learning framework. Through the information interaction between agents, achieve the global resource allocation optimization of task nodes, and improve the overall completion efficiency of the task chain and system performance.

[0126] Finally, it should be noted that: The above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: They can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope defined by the claims of the present invention.

Claims

1. A multi-task real-time scheduling optimization method based on age-dependent perception information, characterized in that: The following steps are involved: 1) Dependent task scenario modeling: By analyzing the dependencies between tasks in the task chain, dependent task scenarios are established and the attributes of each task are defined; 2) Dependency-aware information age calculation: Based on the attributes of the task, the dependency-aware information age of the task is calculated to quantify the relationship between the data state of the current task and the completion time of its dependent tasks; 3) Scheduling strategy optimization: Based on the age value of task dependency perception information, optimize the scheduling strategy of multiple tasks and dynamically adjust task priorities; 4) Control performance evaluation: Using the discrete-time linear time-invariant system framework and combining it with the global performance evaluation of the scheduling strategy, the control cost equivalent function is obtained; 5) Multi-agent deep reinforcement learning optimization: Based on the task dependency perception information age and control cost equivalent function, the scheduling strategy of multiple task nodes is optimized through the multi-agent deep reinforcement learning framework.

2. According to claim 1, a multi-task real-time scheduling optimization method based on age-dependent perception information is characterized in that: In step 1), the attributes include the data collection time of the task, the dependent task set and the dependent task completion time.

3. According to claim 1, a multi-task real-time scheduling optimization method based on age-dependent perception information is characterized in that: In step 1), the dependent task scenario modeling includes at least one of the following modeling methods: 1.1) Serial Dependency Modeling: Determine the serial dependency of tasks based on the task chain and generate a serial task dependency graph; the start time of the task satisfies: t start，i >t end，j , t start，i is the start time of task i, t end，j is the end time of dependent task j, D i represents the set of dependent tasks of task i; the dependent tasks represent the predecessor tasks that must be completed before a certain task is executed; 1.2) Parallel Dependency Modeling: Determine the parallel dependencies of tasks based on the task chain, generate a parallel task dependency graph, and mark the multi-path dependencies of tasks; The dependent task completion time is: t dep，i =max(t end，j ), t dep，i is the dependent completion time of task i; 1.3) Hybrid Dependency Modeling: The serial dependency and parallel dependency of the tasks are determined based on the task chain, and a mixed dependency task chain graph is generated; the mixed dependency task chain graph includes a serial task dependency graph and a parallel task dependency graph.

4. The multi-task real-time scheduling optimization method based on age-dependent perception information according to claim 1 is characterized in that: In step 2), the age calculation based on the perceived information is as follows: Calculate the age of the task's dependent perception information, which is the difference between the current time and the task's latest data collection time and the latest completion time of its dependent tasks: in, is the age of dependent perception information, k is the current time, τ i (k) is the latest data collection time of task i, D i is the set of dependent tasks of task i, C j (k) is the dependent task j∈D i completion time.

5. The multi-task real-time scheduling optimization method based on age-dependent perception information according to claim 1 is characterized in that: In step 3), the scheduling strategy optimization includes the following steps: Based on the dependency relationship of tasks in the dependent task scenario modeling, the topological sorting algorithm is used to dynamically adjust the task priority to identify the key tasks that affect the global performance in the critical path and schedule them as the priority task nodes; The task priority is dynamically adjusted as follows: Among them, P i is the task priority, R i is the resource requirement of task i, is the delay cost of task i, and f represents the priority sorting function in the topological sorting algorithm, which is used to assign priorities to tasks based on task attributes including the age of dependency-aware information, resource requirements, and delay cost.

6. The multi-task real-time scheduling optimization method based on age-dependent perception information according to claim 1 is characterized in that: In step 4), the control performance evaluation comprises the following steps: (1) For the system composed of terminal devices and edge servers, a discrete-time linear time-invariant system model is constructed to describe the state evolution of tasks. The evolution relationship of task states is as follows: x i，k+1 =A i x i，k +B i u i，k +ω i，k in is the system task state vector, which is used to characterize the state of the terminal device. is the control input vector generated by the edge server, which is used to represent the control amount of the terminal device. The constant matrix is the system matrix, is the input matrix of the edge server, i d Represents the state space dimension of the task, i n Represents the dimension of the control input vector space; is zero-mean Gaussian random external interference noise; (2) A linear quadratic Gaussian function is used to evaluate the control cost equivalent function of each task system, which is calculated as follows: Where W i，N and W i is the system state weight matrix, U i is the control input weight matrix, is used to measure and control consumption, It is used to measure the state error of the actual system based on the available measurement values; It is used to express the final error, represents the expected value operation; k represents the kth time step, and N represents the total time step; (3) According to the evolution relationship of task status and control cost equivalent function, a control cost equivalent function based on the age of dependent perception information is constructed: Among them, e i，k is the control system estimation error, is the system task state estimation vector, To rely on perceived information age; Among them, J C，i is the control cost equivalence function, represents the expected value operation, Σ i，k for e i，k The covariance of S i is the solution of the Riccati equation, and Tr(·) represents the trace operation of the matrix.

7. The multi-task real-time scheduling optimization method based on age-dependent perception information according to claim 1 is characterized in that: In step 5), the multi-agent deep reinforcement learning optimization comprises the following steps: 5.1) State Space Definition: Defining the State Space in Multi-Agent Deep Reinforcement Learning Among them, s i (k) represents the state of task i at the kth time step, φ i (k) is the data size, C i，m，k (k) is the computing resource requirement, B i，m，k (k) is the bandwidth, h i，m，k (k) is the channel power gain, is the dependent task set of the task, is the set of subsequent tasks of the task, To rely on perceived information age; 5.2) Action Space Design: Constructing the Action Space for Multi-agent Deep Reinforcement Learning To render the policies for all tasks: Among them, a i (k) indicates execution of an action, To calculate resource type matching decisions, is the task division ratio, p i，m，k (k) is the transmission power of the sensor in the terminal device, f i，m，k (k) is the edge server computing resource allocation strategy, τ i，m，k (k) is the task sampling frequency; 5.3) The reward function is: Among them, J C,i To control costs, J N，i is the network energy consumption cost, ρ i is the weight factor of the age reward item that depends on the perceived information, Δ th is the threshold value that depends on the age of the perceived information, α∈[0,1] and β∈[0,1] are the corresponding weight factors, and α+β=1; 5.4) Multi-agent collaborative optimization: According to the obtained edge server computing resource allocation strategy f i，m，k (k), task sampling frequency τ i，m，k (k) Optimize the scheduling strategy of multiple task nodes.

8. A multi-task real-time scheduling optimization system based on age-dependent perception information, characterized in that: include: The dependent task scenario building module is used to establish dependent task scenarios and define the attributes of each task by analyzing the dependency relationships between tasks in the task chain; A dependency-aware information age calculation module is used to calculate the dependency-aware information age of a task based on the attributes of the task, so as to quantify the relationship between the data state of the current task and the completion time of its dependent tasks; Scheduling strategy optimization module, which is used to optimize the scheduling strategy of multiple tasks and dynamically adjust the task priority based on the age value of task dependency perception information; The control performance evaluation module is used to obtain the control cost equivalent function by utilizing the discrete-time linear time-invariant system framework and combining the evaluation of the global performance of the scheduling strategy; The multi-agent deep reinforcement learning optimization module is used to perceive the information age and control the cost equivalence function according to the task dependency, and optimize the scheduling strategy of multiple task nodes through the multi-agent deep reinforcement learning framework.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, a multi-task real-time scheduling optimization method based on the age of dependent perceived information as described in any one of claims 1 to 7 is implemented.

Images