CN-121996433-A - Intelligent distribution method and device for edge computing tasks and edge equipment

CN121996433ACN 121996433 ACN121996433 ACN 121996433ACN-121996433-A

Abstract

The invention provides an intelligent distribution method, a device and edge equipment for edge calculation tasks, wherein each model obtained by front debugging through small sample learning can determine the corresponding distribution type in various alternative sub types, in the excavation link of description elements, description elements corresponding to main information in the excavated task description elements can be enhanced through different types of attribute description elements learned in the front debugging link through a target optimization model, the processed task description elements highlight data in the main information, the expression effect of the main data is improved, noise disturbance of invalid information is reduced, the accuracy of task type identification is improved, and accurate task distribution is facilitated.

Inventors

WANG XIBIN
LI XIAOKANG
ZHANG XINGXING

Assignees

贵州理工学院
深圳市索川应用技术有限公司

Dates

Publication Date: 20260508
Application Date: 20260409

Claims (10)

1. An intelligent distribution method for edge computing tasks is characterized by comprising the following steps: Acquiring tasks to be allocated; Acquiring various reference tasks which are established in advance corresponding to various reference sub-types; Loading the tasks to be allocated into a target task description element mining model which is subjected to pre-debugging to obtain corresponding initial task description elements to be allocated, and loading the various reference tasks into the target task description element mining model to obtain corresponding initial reference task description elements; Based on a pre-debugging target optimization model, based on various attribute description elements obtained by learning different kinds of task data samples during pre-debugging, main task information strengthening operation is respectively carried out on the initial task description elements to be allocated and various initial reference task description elements to obtain corresponding target task description elements to be allocated and various target reference task description elements; Based on a pre-debugged target distribution model, determining a task type corresponding to the task to be distributed based on a commonality measurement result between the target task description element to be distributed and each target reference task description element; and distributing the task to be distributed to an edge computing node corresponding to the task type based on the task type.
2. The method of claim 1, wherein performing a main task information enhancement operation on the initial task description element to be allocated based on each kind of attribute description element obtained by learning different kinds of task data samples at the time of pre-debugging, comprises: Based on various kinds of attribute description elements obtained by learning different kinds of task data samples during front-end debugging, respectively determining the commonality condition between the initial task description elements to be allocated and the various kinds of attribute description elements; based on the commonality, determining a pairing influence coefficient between the initial task description element to be allocated and each kind of attribute description element; And determining the strengthening intensity of each kind of attribute description element for each distribution vector in the initial task description element to be allocated based on the pairing influence coefficient, and completing the main task information strengthening operation of the initial task description element to be allocated based on each strengthening intensity.
3. The method according to claim 2, wherein the determining the commonality between the initial task description element to be allocated and each kind of attribute description element based on each kind of attribute description element obtained by learning on different kinds of task data samples at the time of pre-debugging includes: Acquiring various kinds of attribute description elements obtained by learning different kinds of task data samples during front-end debugging, performing morphological transformation on the various kinds of attribute description elements to obtain attribute description arrays corresponding to the various kinds of attribute description elements, and performing morphological transformation on the initial task description elements to be allocated to obtain corresponding task description arrays; And calculating an array multiplication result between the attribute description array and the task description array to obtain a common array indicating the common situation between the initial task description elements to be allocated and the attribute description elements of each kind.
4. A method according to claim 3, wherein the determining, based on the commonality, a pairing influence coefficient between the initial task description element to be allocated and each kind of attribute description element, and determining, by the pairing influence coefficient, strength of reinforcement of each kind of attribute description element for each distribution vector in the initial task description element to be allocated, includes: Carrying out standardization operation on the commonality coefficient of each element in the commonality array, and obtaining an influence coefficient array indicating pairing influence coefficients between the initial task description elements to be allocated and the attribute description elements of each kind through the commonality coefficient of each element after the standardization operation; and determining the strengthening intensity of each element description value in the initial task description elements to be allocated by each kind of attribute description elements based on the result of the array multiplication between the description array formed by each kind of attribute description elements and the influence coefficient array.
5. The method of claim 2, wherein the performing a primary task information enhancement operation on the initial task description element to be assigned based on the respective enhancement strengths comprises: Performing description value accumulation operation on the reinforcement intensity of each element and the description value of the corresponding element in the initial task description element to be allocated to obtain respective corresponding accumulated description value of each element; and performing feedforward operation on the accumulated description values corresponding to each element to complete strengthening of main task information in the initial task description elements to be allocated.
6. The method of any of claims 1-5, further comprising the step of pre-debugging a task-allocation algorithm that encompasses a task-descriptive element mining model, an optimization model, and an allocation model, comprising: And repeating iterative pre-debugging for the task allocation algorithm for a plurality of times based on a target training library until a preset debugging cut-off condition is met, and performing the following steps in one debugging process: Repeatedly debugging a task allocation algorithm to be debugged for a preset number of times through a task data sample group determined in a sub-debugging development set to obtain a task allocation algorithm to be verified, and determining allocation accuracy of the task allocation algorithm to be verified through verification task data determined in the sub-debugging verification set, wherein the sub-debugging development set and the sub-debugging verification set belong to the target training library; And selecting a target task allocation algorithm meeting the set selection requirement from all the task allocation algorithms to be verified based on the allocation accuracy determined by the task allocation algorithms to be verified in different test rounds.
7. The method of claim 6, wherein determining task data samples from a subset of the debug development set comprises: Respectively extracting j task data samples as a class of comparison samples for i subtask types which are determined in advance in the subtask development set, and respectively extracting u disjoint task data samples as samples to be distributed; Based on the determined various control samples and various to-be-allocated samples, i multiplied by u task data sample groups are established, wherein each task data sample group comprises one to-be-allocated sample corresponding to one sub-task type, and i, j and u are natural numbers greater than or equal to 1.
8. The method of claim 7, wherein the performing the repeated debugging for the preset number of times on the task allocation algorithm to be debugged through the task data sample group determined in the sub-debugging development set to obtain the task allocation algorithm to be verified comprises: based on the established task data sample group, repeatedly debugging a task allocation algorithm to be debugged for a preset number of times, and optimizing algorithm parameters of the task allocation algorithm through errors obtained through calculation during each debugging, wherein during one repeated debugging, the method comprises the following steps: Loading the obtained debugging sample pair into a task allocation algorithm to be debugged, obtaining a predicted task type determined by the sample to be allocated in the corresponding debugging sample pair, and determining an error through an error between the predicted task type and the corresponding prior allocation type; the task allocation algorithm to be loaded to the task to be debugged by the obtained one debug sample pair obtains a predicted task type determined corresponding to the task to be debugged in the debug sample pair, and the method comprises the following steps: Loading a to-be-allocated sample and various reference samples included in a debugging sample pair into a task description element mining model to be debugged to obtain to-be-allocated task data sample description elements and various reference task data sample description elements; Based on an optimization model to be debugged, based on category attribute description elements built for each category learning, main task information strengthening operation is respectively carried out on the task data sample description elements to be distributed and each comparison task data sample description element to obtain target task data sample description elements to be distributed and each target comparison task data sample description element; And determining the task type corresponding to the task data sample to be distributed based on a distribution model to be debugged and based on a commonality measurement result between the target task data sample description element to be distributed and each target comparison task data sample description element.
9. An intelligent distribution device for edge computing tasks, which is characterized by comprising: The target task acquisition module is used for acquiring tasks to be distributed; the reference task acquisition module is used for acquiring various reference tasks which are established in advance corresponding to various reference sub-types; The description element mining module is used for loading the tasks to be distributed into a target task description element mining model which is subjected to pre-debugging to obtain corresponding initial task description elements to be distributed, and loading the various reference tasks into the target task description element mining model to obtain corresponding initial reference task description elements; the element information strengthening module is used for carrying out main task information strengthening operation on the initial task description elements to be allocated and the various initial reference task description elements respectively based on the target optimization model which is subjected to pre-debugging and on various attribute description elements obtained by learning different types of task data samples during pre-debugging, so as to obtain corresponding target task description elements to be allocated and various target reference task description elements; the task type determining module is used for determining the task type corresponding to the task to be distributed based on a target distribution model which is subjected to pre-debugging and based on a commonality measurement result between the target task description element to be distributed and each target reference task description element; and the task allocation module is used for allocating the task to be allocated to the edge computing node corresponding to the task type based on the task type.
10. An edge device comprising a memory and a processor, the memory storing a computer program executable on the processor, wherein the processor performs the steps of the method of any one of claims 1 to 8 when the program is executed.

Description

Intelligent distribution method and device for edge computing tasks and edge equipment Technical Field The present application relates to the field of, but not limited to, data processing technologies, and in particular, to an intelligent allocation method and apparatus for edge computing tasks, and an edge device. Background With the rapid development of information technology, task allocation algorithms play a vital role in numerous fields, such as cloud computing resource scheduling, workflow allocation of intelligent manufacturing systems, and big data processing task partitioning. The algorithms can reasonably distribute the tasks to corresponding processing units according to the characteristics of the tasks, the capability of processing the resources and the real-time state of the system, thereby realizing the efficient utilization of the resources and the optimized operation of the system. However, when facing complex and variable task environments and processing resources, the conventional task allocation algorithm often has the problems of low allocation precision, poor self-adaptive capacity and the like. These problems lead to the adverse consequences of underutilization of resources, inefficient task execution, and even reduced system performance. In order to solve the above problems, researchers have continuously explored and tried new task allocation algorithms and debugging methods in recent years. Among them, a task allocation algorithm based on machine learning is gradually attracting attention. The algorithm can automatically adjust algorithm parameter values and optimize task allocation strategies by learning and mining key information in the task data sample, so that the accuracy and the adaptability of task allocation are improved. However, existing machine learning-based task allocation algorithms still present some challenges in the debugging process. For example, the problems of how to effectively select the tuning sample pairs, how to accurately extract task description elements, how to strengthen main task information, how to determine task types, and the like all need further research and solution. In addition, with the increasing complexity of task environments and processing resources, higher demands are also placed on the performance and stability of task allocation algorithms. Disclosure of Invention Accordingly, the embodiment of the application at least provides an intelligent distribution method and device for edge computing tasks and edge equipment. The technical scheme of the embodiment of the application is realized as follows: In one aspect, an embodiment of the present application provides an intelligent allocation method for edge computing tasks, including: Acquiring tasks to be allocated; Acquiring various reference tasks which are established in advance corresponding to various reference sub-types; Loading the tasks to be allocated into a target task description element mining model which is subjected to pre-debugging to obtain corresponding initial task description elements to be allocated, and loading the various reference tasks into the target task description element mining model to obtain corresponding initial reference task description elements; Based on a pre-debugging target optimization model, based on various attribute description elements obtained by learning different kinds of task data samples during pre-debugging, main task information strengthening operation is respectively carried out on the initial task description elements to be allocated and various initial reference task description elements to obtain corresponding target task description elements to be allocated and various target reference task description elements; Based on a pre-debugged target distribution model, determining a task type corresponding to the task to be distributed based on a commonality measurement result between the target task description element to be distributed and each target reference task description element; and distributing the task to be distributed to an edge computing node corresponding to the task type based on the task type. Optionally, based on each kind of attribute description elements obtained by learning different kinds of task data samples during pre-debugging, performing main task information enhancement operation on the initial task description elements to be allocated, including: Based on various kinds of attribute description elements obtained by learning different kinds of task data samples during front-end debugging, respectively determining the commonality condition between the initial task description elements to be allocated and the various kinds of attribute description elements; based on the commonality, determining a pairing influence coefficient between the initial task description element to be allocated and each kind of attribute description element; And determining the strengthening intensity of each kind of attribute description element