A joint optimization scheduling method with edge scheduling freedom degree as core

Abstract

A kind of joint optimization scheduling method with edge scheduling freedom as core, first, task priority scheduling sequencing method is used, and the test task of edge perception node is scheduled according to the privacy degree of test task, the deadline constraint of task, the adaptability of task and processor and other characteristics Task priority scheduling sequencing;Then, for the two optimization objectives of minimum overall power consumption and longest comprehensive endurance, different edge scheduling freedom parameters are calculated to control the degree of edge perception node scheduling task, to ensure the global optimization effect;Finally, to ensure the solution speed of task optimization scheduling, according to task priority scheduling sequencing and edge scheduling freedom, the self-scheduling of test task in edge perception node is carried out, and the task that fails to complete scheduling in edge perception node is uploaded to edge computing node for comprehensive optimization scheduling, to complete the joint optimization scheduling of test task;Therefore, on the basis of ensuring the solution speed of task optimization, the global optimization effect is fully considered.

Description

Technical Field

[0001] This invention belongs to the technical field of task optimization scheduling, and in particular relates to a joint optimization scheduling method with edge scheduling degrees of freedom as the core. Background Technology

[0002] In a heterogeneous multi-core distributed intelligent testing system, edge sensing nodes that can solve some simple test computing tasks and edge computing nodes that can execute some relatively complex intelligent testing tasks complement each other.

[0003] The rapid increase in the types and number of heterogeneous multi-core distributed intelligent testing tasks has brought new challenges to task scheduling strategies. A reasonable scheduling strategy should fully consider the system's resource constraints, energy limitations, and the real-time requirements of intelligent testing tasks, in order to significantly improve the resource utilization of the intelligent testing system, extend its runtime, and reduce the overall latency of intelligent testing tasks. However, current task scheduling strategies cannot simultaneously guarantee the speed of task optimization and scheduling solutions and the global optimization effect of task scheduling. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, the technical problem to be solved by this invention is to provide a joint optimization scheduling method with edge scheduling degrees of freedom as the core, which can fully consider the global optimization effect while ensuring the speed of task optimization solution.

[0005] The technical solution of this invention is: a joint optimization scheduling method with edge scheduling degrees of freedom as its core, which includes the following steps:

[0006] (1) Edge sensing node EDa sorts the intelligent test tasks on its own node using a task priority scheduling and sorting method;

[0007] (2) Based on the two optimization objectives that intelligent testing systems usually focus on, namely minimizing overall power consumption and maximizing overall battery life, the corresponding edge scheduling degree of freedom parameter α was designed;

[0008] (3) Based on the completion of task priority scheduling comprehensive sorting and edge scheduling freedom, the intelligent test tasks on the edge sensing nodes are rapidly self-scheduled.

[0009] (4) After each task is scheduled, the scheduling strategy is written into the decision matrix to reduce the number of unknown decisions to be optimized in the decision matrix;

[0010] (5) After the local task scheduling is completed, the local task decision and the undecided tasks to be scheduled are sent to the edge computing node for comprehensive optimization and scheduling.

[0011] Step (1) includes the following sub-steps:

[0012] (1.1) Edge-aware node EDa acquires local processor resources Pna and locally occurring tasks T. EDa According to the test system information, the edge sensing node cannot obtain the load status of the edge computing node or information about other edge sensing nodes;

[0013] (1.2) Sort the intelligent test tasks that are prioritized for local execution to form a priority local execution queue;

[0014] (1.3) Sort the smart test tasks that are prioritized for uninstallation and execution to form a priority uninstallation and execution queue;

[0015] (1.4) Based on the sorting results of priority local execution and priority unload execution, the priority of intelligent test tasks is comprehensively sorted.

[0016] This invention first employs a task-priority scheduling and sorting method, prioritizing test tasks appearing at edge sensing nodes based on characteristics such as the privacy level of the test tasks, deadline constraints, and compatibility between the tasks and processors. Then, it calculates different edge scheduling degree-of-freedom parameters for two optimization objectives: minimizing overall power consumption and maximizing overall battery life. This controls the degree of task scheduling at edge sensing nodes, ensuring overall optimization effectiveness. Finally, to guarantee the speed of task optimization scheduling, test tasks are self-scheduled at edge sensing nodes based on the task-priority scheduling and edge scheduling degrees of freedom. Tasks that cannot be scheduled at edge sensing nodes are uploaded to edge computing nodes for comprehensive optimization scheduling, completing the joint optimization scheduling of test tasks. Therefore, it can fully consider the overall optimization effect while ensuring the speed of task optimization solution. Attached Figure Description

[0017] Figure 1 The flowchart below illustrates the joint optimization scheduling method based on edge scheduling degrees of freedom according to the present invention.

[0018] Figure 2 This is an architecture diagram of the joint optimization scheduling method based on edge scheduling degrees of freedom according to the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] To make the description of this disclosure more detailed and complete, illustrative descriptions of embodiments and specific examples of the present invention are provided below; however, these are not the only forms of implementing or utilizing the specific examples of the present invention. The embodiments cover features of multiple specific examples and methods and steps for constructing and operating these specific examples, and their order. However, other specific examples may also be used to achieve the same or equivalent functions and order of steps.

[0021] like Figure 1 As shown, this joint optimization scheduling method, which focuses on edge scheduling degrees of freedom, includes the following steps:

[0022] (1) Edge sensing node EDa sorts the intelligent test tasks on its own node using a task priority scheduling and sorting method;

[0023] (2) Based on the two optimization objectives that intelligent testing systems usually focus on, namely minimizing overall power consumption and maximizing overall battery life, the corresponding edge scheduling degree of freedom parameter α was designed;

[0024] (3) Based on the completion of task priority scheduling comprehensive sorting and edge scheduling freedom, the intelligent test tasks on the edge sensing nodes are rapidly self-scheduled.

[0025] (4) After each task is scheduled, the scheduling strategy is written into the decision matrix to reduce the number of unknown decisions to be optimized in the decision matrix;

[0026] (5) After the local task scheduling is completed, the local task decision and the undecided tasks to be scheduled are sent to the edge computing node for comprehensive optimization and scheduling.

[0027] Step (1) includes the following sub-steps:

[0028] (1.1) Edge-aware node EDa acquires local processor resources Pna and locally occurring tasks T. EDa According to the test system information, the edge sensing node cannot obtain the load status of the edge computing node or information about other edge sensing nodes;

[0029] (1.2) Sort the intelligent test tasks that are prioritized for local execution to form a priority local execution queue;

[0030] (1.3) Sort the smart test tasks that are prioritized for uninstallation and execution to form a priority uninstallation and execution queue;

[0031] (1.4) Based on the sorting results of priority local execution and priority unload execution, the priority of intelligent test tasks is comprehensively sorted.

[0032] This invention first employs a task-priority scheduling and sorting method, prioritizing test tasks appearing at edge sensing nodes based on characteristics such as the privacy level of the test tasks, deadline constraints, and compatibility between the tasks and processors. Then, it calculates different edge scheduling degree-of-freedom parameters for two optimization objectives: minimizing overall power consumption and maximizing overall battery life. This controls the degree of task scheduling at edge sensing nodes, ensuring overall optimization effectiveness. Finally, to guarantee the speed of task optimization scheduling, test tasks are self-scheduled at edge sensing nodes based on the task-priority scheduling and edge scheduling degrees of freedom. Tasks that cannot be scheduled at edge sensing nodes are uploaded to edge computing nodes for comprehensive optimization scheduling, completing the joint optimization scheduling of test tasks. Therefore, it can fully consider the overall optimization effect while ensuring the speed of task optimization solution.

[0033] Preferably, the sorting method in step (1.2) is as follows:

[0034] (1.2.1) For intelligent testing tasks with privacy and data security requirements, local execution is given priority, and these tasks are prioritized as the first tier of task ranking, represented as: Tlf1=[Tlf11,Tlf12,...],Tlf1∈T EDa ;

[0035] (1.2.2) For tasks whose execution time uploaded to the edge computing node cannot meet the minimum execution time requirement Tfsmin, they are prioritized for local execution and are placed in the second tier of task ranking. This part of the tasks is represented as: Tlf2=[Tlf21,Tlf22,...],Tlf2∈T EDa ;

[0036] (1.2.3) For intelligent test tasks suitable for the local processor, local execution is given priority, and these tasks are placed in the third tier of task ranking. The task fitness (Tfit) is used to evaluate the suitability of the task for execution on different processors. This parameter is calculated as follows:

[0037] Calculate the total energy consumption Te required to execute each task on different processors. i,j Including task T j In processor P i Energy consumption during execution, and communication consumption when tasks are offloaded to other nodes;

[0038] For Te i,j The values ​​in each column are reordered in descending order, and the sorted index is assigned to the task fitness Tfiti. At this point, the task fitness Tfiti is... i,j Represents task T j The suitability for execution on different processors; the larger the number, the better it is suitable for execution on that processor.

[0039] Different processors accelerate tasks to varying degrees; task adaptability takes into account the differences in acceleration across different processors.

[0040]

[0041] The new task fitness is represented as:

[0042]

[0043] Where Pmaxr represents the maximum possible difference in task performance across different processors. At this point, the task fitness Tfit... i,j It can provide a more refined representation of how well a task is suited for execution on different processors;

[0044] According to Tfit i,j The difference ΔTfit between the maximum fitness of processors on local nodes and edge computing nodes. j Arranged in descending order, the larger the difference, the more suitable it is for priority local scheduling;

[0045] The sorted tasks are represented as Tlf3 = [Tlf31, Tlf32, ...];

[0046] The final priority local execution queue is formed as Tlf = [Tlf1, Tlf2, Tlf3].

[0047] Preferably, the sorting method in step (1.3) is as follows:

[0048] For tasks whose execution time on local nodes cannot meet the minimum execution time requirement Tfsmin, they are preferentially offloaded to edge computing nodes for execution and placed in the first tier of task sorting. These tasks are represented as Tof1 = [Tof11, Tof12, ...];

[0049] For intelligent testing tasks suitable for edge computing nodes, they should be prioritized for offloading and execution, and placed in the second tier of task sorting, according to Tfit. i,j The difference ΔTfit between the maximum fitness of processors on edge computing nodes and local nodes. j Arranged in descending order, the larger the difference, the more suitable it is for priority edge computing node scheduling;

[0050] The tasks are sorted according to their fitness on the edge computing nodes. Tasks with higher fitness are executed on the edge computing nodes first. These tasks are represented as Tof2 = [Tof21, Tof22, ...]; and finally, a priority local offloading queue is formed as Tof = [Tof1, Tof2].

[0051] Preferably, in step (1.4):

[0052] First, sort the tasks that must be executed locally or must be unloaded. Prioritize the first and second tiers of the locally executed sorting queue and prioritize the first tier of the unloaded execution sorting queue. There is no order between locally executed and unloaded tasks. This is represented as: Tsf1 = [Tlf1, Tlf2, Tof1].

[0053] Based on the priority of local execution ΔTfit j Perform a unified sort, ΔTfit j The larger the absolute value, the greater the difference between the local and edge processors, and the better the task scheduling effect; according to |ΔTfit j Sort the remaining tasks in descending order, and the sorting result is Tsf2;

[0054] The final task priority scheduling comprehensive sorting queue is: Tsf = [Tsf1, Tsf2].

[0055] Preferably, step (2) includes:

[0056] (2.1) To achieve the optimization goal of minimizing overall power consumption, those with higher energy efficiency ratios are given more priority for local execution.

[0057] First, the total power consumption Ptp of different tasks executed on different processors is calculated at the edge computing node. i,j ;

[0058] Then calculate a certain processor P. i Energy consumption P required to perform all tasks sum :

[0059]

[0060] Finally, the edge processor priority scheduling factor P pr Represented as:

[0061]

[0062] When P pr When P is ≥0, the larger the value, the greater the energy consumption for executing all tasks, and the more suitable the tasks on it are for priority offloading and scheduling. pr When |P| is ≤0, the smaller the value, the less energy is consumed in executing all tasks, and the tasks on it are more suitable for local execution first. pr The larger the value, the more suitable it is for self-scheduling on edge sensing nodes; select the maximum priority scheduling coefficient of the processor on the edge sensing node. The edge scheduling degree of freedom parameter α of the edge sensing node is expressed as:

[0063]

[0064] Where β is a scheduling coefficient that controls the task scheduling power of the edge computing node to control the edge sensing node. In practical applications, it can be adjusted according to the actual situation. The larger the parameter is, the more tasks the edge sensing node can schedule, and the faster the overall optimization scheduling speed will be, but the final optimization result may be worse.

[0065] (2.2) For scheduling scenarios with optimal overall battery life, the energy consumption of edge sensing nodes is considered. The energy consumption Psum required to execute all tasks on this processor is:

[0066]

[0067] Increasing the energy consideration of edge-aware nodes makes the power of priority local scheduling more inclined to processors with abundant energy and excellent performance, and also makes the power of priority offloading scheduling more inclined to processors with scarce energy and weak performance.

[0068] Preferably, step (3) includes:

[0069] (3.1) If the number of tasks that can be scheduled on this node is less than Tsf1 of the priority decision comprehensive ranking, then the fast self-decision method for intelligent test tasks of the edge sensing node is to... Scheduling decisions are made sequentially for Tsf1; j During scheduling, if a task belongs to the first or second tier of priority local execution, it will be placed on the local processor for execution. Furthermore, while satisfying processor resource constraints and execution time constraints, the task fitness (Tfit) will be prioritized. j The largest processor is placed; under the current scheduling decision, for task Tsf1 j If the fastest local processor has already accepted other scheduled tasks, making it unable to meet its resource or execution time constraints, then the task will be scheduled to the local processor with the second-lowest task fitness value until the constraints are met; if Tsf1 j If a task belongs to the first tier of priority offloading execution, it will be preferentially placed on an edge computing node for execution. Furthermore, given the constraints of processor resources and execution time, the task fitness (Tfit) will be preferentially selected. j The task is placed on the processor with the highest fitness value. This scheduling decision also satisfies the processor's resource constraints and execution time constraints. If not, the task is scheduled to the local processor with the second lowest fitness value until the constraints are met.

[0070] (3.2) If the number of schedulable tasks on the current node is greater than Tsf1 obtained by comprehensive priority decision sorting, all tasks in Tsf1 shall be preferentially scheduled according to the above strategy first; then the tasks in Tsf2 are scheduled sequentially, and this part of tasks are placed on the processor with the highest task fitness Tfit. If the current decision does not satisfy the resource constraint of the processor, the task is placed on the processor with the second highest task fitness until the task is successfully scheduled.

[0071] Preferably, in said step (3), after each task is scheduled, the scheduling strategy is written into the decision matrix to reduce the number of unknown decisions to be optimized in the decision matrix. As the number of unknown decisions to be solved decreases, the solving time generally shows a trend of significant decrease; fast self-scheduling of tasks is performed on edge-aware nodes.

[0072] Preferably, in said step (4), the objective function of comprehensive optimized scheduling is a multi-objective optimization problem for maximum battery life and minimum power consumption:

[0073] where a+b=1 and a<<b, wherein Enp is the total power consumption of the test system, Eed=[Eed1,...,Eed N is the battery energy of each edge device, a and b respectively represent the optimization weights of maximum battery life and minimum power consumption; the optimization constraints include: the actual completion time of all tasks shall satisfy the deadline constraint of the tasks; the sum of the maximum execution time of all to-be-executed tasks on each processor shall be less than the scheduling period; the number of resources used for task scheduling shall be less than the actual resource upper limit.

[0074] The above description is only a preferred embodiment of the present invention, and does not limit the present invention in any form. Any simple modification, equivalent change and modification made to the above embodiments in accordance with the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A joint optimization scheduling method with edge scheduling degrees of freedom as its core, characterized in that: It includes the following steps: (1) Edge sensing node EDa sorts the intelligent test tasks on its own node using a task priority scheduling and sorting method; (2) Based on the two optimization objectives that intelligent testing systems usually focus on, namely minimizing overall power consumption and maximizing overall battery life, the corresponding edge scheduling degree of freedom parameter α was designed; (3) Based on the completion of task priority scheduling comprehensive sorting and edge scheduling freedom, the intelligent test tasks on the edge sensing nodes are rapidly self-scheduled. (4) After each task is scheduled, the scheduling strategy is written into the decision matrix to reduce the number of unknown decisions to be optimized in the decision matrix; (5) After the local task scheduling is completed, the local task decision and the undecided tasks to be scheduled are sent to the edge computing node for comprehensive optimization and scheduling. Step (1) includes the following sub-steps: (1.1) Edge-aware node EDa acquires local processor resources Pna and locally occurring tasks T. EDa According to the test system information, the edge sensing node cannot obtain the load status of the edge computing node or information about other edge sensing nodes; (1.2) Sort the intelligent test tasks that are prioritized for local execution to form a priority local execution queue; (1.3) Sort the smart test tasks that are prioritized for uninstallation and execution to form a priority uninstallation and execution queue; (1.4) Based on the sorting results of priority local execution and priority unload execution, the priority of intelligent test tasks is comprehensively sorted.

2. The joint optimization scheduling method based on edge scheduling degrees of freedom as described in claim 1, characterized in that: The sorting method in step (1.2) is as follows: (1.2.1) For intelligent testing tasks with privacy and data security requirements, local execution is given priority, and these tasks are prioritized as the first tier of task ranking, represented as: Tlf1=[Tlf11,Tlf12,...],Tlf1∈T EDa ; (1.2.2) For tasks whose execution time uploaded to the edge computing node cannot meet the minimum execution time requirement Tfsmin, they are prioritized for local execution and are placed in the second tier of task ranking. This part of the tasks is represented as: Tlf2=[Tlf21,Tlf22,...],Tlf2∈T EDa ; (1.2.3) For intelligent test tasks suitable for the local processor, local execution is given priority, and these tasks are placed in the third tier of task ranking. The task fitness (Tfit) is used to evaluate the suitability of the task for execution on different processors. This parameter is calculated as follows: Calculate the total energy consumption Te required to execute each task on different processors. i，j Including task T j In processor P i Energy consumption during execution, and communication consumption when tasks are offloaded to other nodes; For Te i，j The values ​​in each column are reordered in descending order, and the sorted index is assigned to the task fitness Tfit. At this point, the task fitness Tfit... i，j Indicates task T j The suitability for execution on different processors; the larger the number, the better it is suitable for execution on that processor. Different processors accelerate tasks to varying degrees; task adaptability takes into account the differences in acceleration across different processors. The new task fitness is represented as: Where Pmaxr represents the maximum possible difference ratio of the task across different processors, and the task fitness Tfit is at this point. i,j It can provide a more refined representation of how well a task is suited for execution on different processors; According to Tfit i,j The difference ΔTfit between the maximum fitness of processors on local nodes and edge computing nodes. j Arranged in descending order, the larger the difference, the more suitable it is for priority local scheduling; The sorted tasks are represented as Tlf3 = [Tlf31, Tlf32, ...]; The final priority local execution queue is Tlf = [Tlf1, Tlf2, Tlf3].

3. The joint optimization scheduling method based on edge scheduling degrees of freedom as described in claim 2, characterized in that: The sorting method in step (1.3) is as follows: For tasks whose execution time on local nodes cannot meet the minimum execution time requirement Tfsmin, they are preferentially offloaded to edge computing nodes for execution and placed in the first tier of task sorting. These tasks are represented as Tof1 = [Tof11, Tof12, ...]; For intelligent testing tasks suitable for edge computing nodes, they should be prioritized for offloading and execution, and placed in the second tier of task sorting, according to Tfit. i，j The difference ΔTfit between the maximum fitness of processors on edge computing nodes and local nodes. j Arranged in descending order, the larger the difference, the more suitable it is for priority edge computing node scheduling; The tasks are sorted according to their task fitness, and those with higher task fitness are executed on the edge computing nodes first. This part of the tasks is represented as Tof2 = [Tof21, Tof22, ...]; The final priority local unloading queue is formed as Tof = [Tof1, Tof2].

4. The joint optimization scheduling method based on edge scheduling degrees of freedom as described in claim 3, characterized in that: In step (1.4): First, sort the tasks that must be executed locally or must be unloaded. Prioritize the first and second tiers of the locally executed sorting queue and prioritize the first tier of the unloaded execution sorting queue. There is no order between locally executed and unloaded tasks. This is represented as: Tsf1 = [Tlf1, Tlf2, Tof1]. Based on the priority of local execution ΔTfit j Perform a unified sort, ΔTfit j The larger the absolute value, the greater the difference between the local and edge processors, and the better the task scheduling effect; according to |ΔTfit j Sort the remaining tasks in descending order, and the sorting result is Tsf2; The final task priority scheduling comprehensive sorting queue is: Tsf = [Tsf1, Tsf2].

5. The joint optimization scheduling method based on edge scheduling degrees of freedom as described in claim 4, characterized in that: Step (2) includes: (2.1) To achieve the optimization goal of minimizing overall power consumption, those with higher energy efficiency ratios are given more priority for local execution. First, the total power consumption Ptp of different tasks executed on different processors is calculated at the edge computing node. i，j ; Then calculate a certain processor P. i Energy consumption Psum required to execute all tasks: The final edge processor priority scheduling factor Ppr is expressed as: When Ppr ≥ 0, a larger value indicates higher energy consumption for executing all tasks, making those tasks more suitable for priority offloading and scheduling. When Ppr ≤ 0, a smaller value indicates lower energy consumption for executing all tasks, making those tasks more suitable for priority local execution. A larger |Ppr| value indicates greater suitability for self-scheduling on the edge sensing node. The maximum priority scheduling coefficient Ppr for the processors on the edge sensing node is selected. EDa The edge scheduling degree of freedom parameter α of the edge sensing node is expressed as: Where β is a scheduling coefficient that controls the task scheduling power of the edge computing node to control the edge sensing node. In practical applications, it can be adjusted according to the actual situation. The larger the parameter is, the more tasks the edge sensing node can schedule, and the faster the overall optimization scheduling speed will be, but the final optimization result may be worse. (2.2) For scheduling scenarios with optimal overall battery life, the energy consumption of edge sensing nodes is considered. The energy consumption Psum required to execute all tasks on this processor is: Increasing the energy consideration of edge-aware nodes makes the power of priority local scheduling more inclined to processors with abundant energy and excellent performance, and also makes the power of priority offloading scheduling more inclined to processors with scarce energy and weak performance.

6. The joint optimization scheduling method based on edge scheduling degrees of freedom as described in claim 5, characterized in that: Step (3) includes: (3.1) If the number of tasks that can be scheduled on this node is less than Tsf1 of the priority decision comprehensive ranking, then the fast self-decision method for intelligent test tasks of the edge sensing node is to... Scheduling decisions are made sequentially for Tsf1; j During scheduling, if a task belongs to the first or second tier of priority local execution, it will be placed on the local processor for execution. Furthermore, while satisfying processor resource constraints and execution time constraints, the task fitness (Tfit) will be prioritized. j The largest processor is placed; under the current scheduling decision, for task Tsf1 j If the fastest local processor has already accepted other scheduled tasks, making it unable to meet its resource or execution time constraints, then the task will be scheduled to the local processor with the second-lowest task fitness value until the constraints are met; if Tsf1 j If a task belongs to the first tier of priority offloading execution, it will be preferentially placed on an edge computing node for execution. Furthermore, given the constraints of processor resources and execution time, the task fitness (Tfit) will be preferentially selected. j The task is placed on the processor with the highest fitness value. This scheduling decision also satisfies the processor's resource constraints and execution time constraints. If not, the task is scheduled to the local processor with the second lowest fitness value until the constraints are met. (3.2) If the number of tasks that can be scheduled on this node is greater than the number of tasks in the priority decision comprehensive ranking Tsf1, then all tasks in Tsf1 are first scheduled according to the above strategy; then the tasks in Tsf2 are scheduled in turn, and these tasks are placed on the processor with the highest task fitness Tfit. If the current decision does not meet the resource constraints of the processor, then it is placed on the processor with the second highest task fitness, until the task is successfully scheduled.

7. The joint optimization scheduling method based on edge scheduling degrees of freedom as described in claim 6, characterized in that: In step (3), after each task is scheduled, the scheduling strategy is written into the decision matrix to reduce the number of unknown decisions to be optimized in the decision matrix. As the number of unknown decisions to be solved decreases, the solution time will show a significant downward trend. Perform rapid self-scheduling of tasks on edge-aware nodes.

8. The joint optimization scheduling method based on edge scheduling degrees of freedom as described in claim 7, characterized in that: In step (4), the objective function of the comprehensive optimization scheduling is a multi-objective optimization problem of maximizing battery life and minimizing power consumption: Where a + b = 1, and a << b, where Enp is the total power consumption of the test system, and Eed = [Eed1, ..., Eed2]. N [Battery energy for each edge device, where a and b represent the optimization weights for longest battery life and minimum power consumption, respectively;] The optimization constraints include: the actual completion time of all tasks should meet the task deadline constraints; the sum of the maximum execution times of all tasks to be executed on each processor should be less than the scheduling cycle; and the number of resources used for task scheduling should be less than the actual resource limit.

Images