An edge computing task intelligent allocation method and device, and an edge device

Abstract

The application provides an edge computing task intelligent distribution method and device and an edge device, each model obtained through pre-debugging of small sample learning can determine a corresponding distribution type in various alternative subtypes, in the mining link of description elements, through different kinds of attribute description elements learned in the pre-debugging link by the target optimization model, the main information corresponding to the description elements in the mined task description elements can be strengthened, the processed task description elements highlight the data in the main information, improve the expression effect of the main data, reduce the noise disturbance of invalid information, improve the accuracy of task type identification, and facilitate accurate task distribution.

Description

Technical Field

[0001] This application relates to, but is not limited to, the field of data processing technology, and in particular to a method, apparatus, and edge device for intelligent allocation of edge computing tasks. Background Technology

[0002] With the rapid development of information technology, task allocation algorithms play a crucial role in many fields, such as cloud computing resource scheduling, workflow allocation in intelligent manufacturing systems, and task partitioning in big data processing. These algorithms can rationally allocate tasks to appropriate processing units based on task characteristics, processing resource capabilities, and the real-time status of the system, thereby achieving efficient resource utilization and optimized system operation.

[0003] However, traditional task allocation algorithms often suffer from low allocation accuracy and poor adaptability when faced with complex and ever-changing task environments and processing resources. These problems lead to adverse consequences such as insufficient resource utilization, low task execution efficiency, and even system performance degradation.

[0004] To address these issues, researchers have been exploring and experimenting with new task allocation algorithms and debugging methods in recent years. Among them, a machine learning-based task allocation algorithm has gradually attracted widespread attention. This type of algorithm learns and mines key information from task data samples, automatically adjusting algorithm parameters and optimizing task allocation strategies, thereby improving the accuracy and adaptability of task allocation.

[0005] However, existing machine learning-based task allocation algorithms still face some challenges during the debugging process. For example, how to effectively select debugging sample pairs, how to accurately extract task description elements, how to enhance key task information, and how to determine task types all require further research and solutions. Furthermore, with the increasing complexity of task environments and processing resources, higher demands are being placed on the performance and stability of task allocation algorithms. Summary of the Invention

[0006] In view of this, embodiments of this application provide at least one method, apparatus, and edge device for intelligent allocation of edge computing tasks.

[0007] The technical solution of this application embodiment is implemented as follows:

[0008] On the one hand, embodiments of this application provide a method for intelligent allocation of edge computing tasks, including:

[0009] Get tasks to be assigned;

[0010] Obtain various reference tasks that have been pre-established for various reference subtypes;

[0011] The task to be assigned is loaded into the pre-debugged target task description element mining model to obtain the corresponding initial task description elements to be assigned, and the various reference tasks are loaded into the target task description element mining model to obtain the corresponding initial reference task description elements.

[0012] Based on the pre-debugging target optimization model, and based on the attribute description elements of various types obtained from learning different types of task data samples during pre-debugging, the main task information enhancement operations are performed on the initial task description elements to be assigned and various initial reference task description elements to obtain the corresponding target task description elements to be assigned and various target reference task description elements.

[0013] Based on the pre-tested target allocation model, and based on the common measurement results between the target task description elements and each target reference task description element, the task type corresponding to the task to be allocated is determined.

[0014] Based on the task type, the task to be assigned is distributed to the edge computing node corresponding to the task type.

[0015] Optionally, based on the attribute description elements of various types obtained from learning different types of task data samples during pre-debugging, the initial task description elements to be assigned are subjected to main task information enhancement operations, including:

[0016] Based on the attribute description elements of each type obtained from learning different types of task data samples during pre-debugging, the commonalities between the initial task description elements to be assigned and the attribute description elements of each type are determined respectively.

[0017] Based on the commonalities mentioned above, the pairing influence coefficients between the initial task description elements to be assigned and the attribute description elements of each category are determined.

[0018] Based on the pairing influence coefficient, the reinforcement intensity of each type of attribute description element for each distribution vector in the initial task description element to be assigned is determined, and based on each reinforcement intensity, the main task information reinforcement operation of the initial task description element to be assigned is completed.

[0019] Optionally, the step of determining the commonalities between the initial task description elements to be assigned and the attribute description elements of each type, based on the attribute description elements of different types obtained during pre-debugging learning from different types of task data samples, includes:

[0020] The system acquires attribute description elements of various types obtained from learning different types of task data samples during pre-debugging, performs morphological transformation on each type of attribute description element to obtain the attribute description array corresponding to each type of attribute description element, and performs morphological transformation on the initial task description elements to be assigned to obtain the corresponding task description array.

[0021] Calculate the array multiplication result between the attribute description array and the task description array to obtain a common array indicating the commonalities between the initial task description elements to be assigned and the attribute description elements of each type.

[0022] Optionally, determining the pairing influence coefficient between the initial task description elements to be assigned and each type of attribute description element based on the commonalities, and determining the reinforcement strength of each type of attribute description element for each distribution vector in the initial task description elements to be assigned through the pairing influence coefficient, includes:

[0023] The common coefficients of each element in the commonity array are standardized, and the common coefficients of each element after standardization are used to obtain an influence coefficient array that indicates the pairing influence coefficients between the initial task description elements to be assigned and the description elements of each type of attribute.

[0024] Based on the array multiplication result between the description array composed of various attribute description elements and the influence coefficient array, the reinforcement strength of each attribute description element for each element description value in the initial task to be assigned is determined.

[0025] Optionally, the step of enhancing the main task information of the initial task description elements based on various enhancement intensities includes:

[0026] The enhancement intensity of each element is added to the description value of the corresponding element in the initial task description elements to obtain the accumulated description value of each element.

[0027] The accumulated description value corresponding to each element is fed forward to enhance the main task information in the initial task description elements to be assigned.

[0028] Optionally, it also includes a step of pre-debugging the task allocation algorithm that covers the task description element mining model, optimization model, and allocation model, including:

[0029] Based on the target training library, the task allocation algorithm is repeatedly iterated and pre-tested until a preset debugging cutoff condition is met. During each debugging process, the following steps are performed:

[0030] By repeatedly debugging the task allocation algorithm to be debugged a preset number of times using the task data sample group determined in the sub-debug development set, a task allocation algorithm to be verified is obtained. The allocation accuracy of the task allocation algorithm to be verified is determined by the verification task data determined in the sub-debug verification set. The sub-debug development set and the sub-debug verification set belong to the target training library.

[0031] Based on the allocation accuracy determined by the proposed verification task allocation algorithm in different debugging rounds, a target task allocation algorithm that meets the set selection requirements is selected from each proposed verification task allocation algorithm.

[0032] Optionally, task data samples are determined from the sub-debug development set, including:

[0033] In the sub-debugging development set, for each of the i pre-determined sub-task types, j task data samples are extracted as a type of control sample, and u disjoint task data samples are extracted as proposed assignment samples.

[0034] Based on the identified various control samples and various proposed assignment samples, establish i×u task data sample groups, where each task data sample group includes a proposed assignment sample corresponding to a subtask type, and i, j, and u are all natural numbers greater than or equal to 1.

[0035] Optionally, the step of repeatedly debugging the task allocation algorithm to be debugged a preset number of times by determining the task data sample group in the sub-debug development set to obtain the task allocation algorithm to be verified includes:

[0036] Based on the established task data sample group, the task allocation algorithm to be debugged is repeatedly debugged a preset number of times, and the algorithm parameters of the task allocation algorithm are optimized by calculating the error obtained in each debugging session; wherein, in one repeated debugging session, the following steps are performed:

[0037] The obtained debugging sample pair is loaded into the task allocation algorithm to be debugged, and the predicted task type determined by the proposed allocation sample in the debugging sample pair is obtained. The error is determined by the error between the predicted task type and the corresponding prior allocation type.

[0038] The step of loading a pair of debug samples into the task allocation algorithm to be debugged, and obtaining the predicted task type determined by the corresponding allocation sample in the debug sample pair, includes:

[0039] The proposed assignment samples and various control samples included in a debugging sample pair are loaded into the task description element mining model to be debugged, so as to obtain the data sample description elements of the proposed assignment task and the data sample description elements of each control task.

[0040] Based on the optimization model to be debugged, and based on the category attribute description elements established by learning each category, the main task information enhancement operation is performed on the description elements of the data sample to be assigned and the description elements of each reference task data sample, to obtain the description elements of the target data sample to be assigned and the description elements of each target reference task data sample.

[0041] Based on the allocation model to be debugged, and based on the commonality measurement results between the descriptive elements of the target task data sample and the descriptive elements of each target comparison task data sample, the task type corresponding to the target task data sample is determined.

[0042] On the other hand, an edge computing task intelligent allocation device is provided, including:

[0043] The target task acquisition module is used to acquire tasks to be assigned.

[0044] The reference task acquisition module is used to acquire various reference tasks that have been pre-established for various reference subtypes;

[0045] The description element mining module is used to load the task to be assigned into the pre-debugged target task description element mining model to obtain the corresponding initial task description elements, and to load the various reference tasks into the target task description element mining model to obtain the corresponding initial reference task description elements.

[0046] The element information enhancement module is used to perform main task information enhancement operations on the initial task description elements to be assigned and various initial reference task description elements based on the target optimization model that has been pre-tested and the attribute description elements of various types obtained by learning from different types of task data samples during pre-testing, so as to obtain the corresponding target task description elements to be assigned and various target reference task description elements.

[0047] The task type determination module is used to determine the task type corresponding to the task to be assigned based on the commonality measurement results between the description elements of the task to be assigned to the target and the description elements of each target reference task, based on the pre-debugged target allocation model.

[0048] The task allocation module is used to allocate the task to be allocated to the edge computing node corresponding to the task type based on the task type.

[0049] Thirdly, an edge device is provided, including a memory and a processor, the memory storing a computer program that can run on the processor, the processor executing the program to implement the steps in the methods described above.

[0050] Beneficial Effects: In this embodiment, after obtaining the task to be assigned, various reference tasks corresponding to different reference subtypes are obtained in advance. The task to be assigned is loaded into the pre-tested target task description element mining model to obtain the corresponding initial task description elements, and various reference tasks are loaded into the target task description element mining model to obtain the corresponding initial reference task description elements. Next, based on the pre-tested target optimization model and the various attribute description elements obtained from learning different types of task data samples during pre-testing, the initial task description elements and various initial reference task description elements are subjected to main task information enhancement operations to obtain the corresponding target task description elements and various target reference task description elements. Then, based on the pre-tested target allocation model and the commonality measurement results between the target task description elements and each target reference task description element, the task type corresponding to the task to be assigned is determined. In this way, the models obtained through small-sample learning and pre-tuning can determine the corresponding allocation type among various candidate subtypes. In the mining of descriptive elements, the different types of attribute descriptive elements learned by the target optimization model in the pre-tuning stage can strengthen the descriptive elements corresponding to the main information in the mined task descriptive elements, making the processed task descriptive elements highlight the data in the main information, improving the expression effect of the main data, reducing the noise disturbance of invalid information, improving the accuracy of task type identification, and facilitating accurate task allocation.

[0051] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and are not intended to limit the technical solutions of this application. Attached Figure Description

[0052] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the specification, serve to explain the technical solutions of this application.

[0053] Figure 1 This is a schematic diagram illustrating the implementation process of an edge computing task intelligent allocation method provided in an embodiment of this application.

[0054] Figure 2 This is a schematic diagram of the composition structure of an edge computing task intelligent allocation device provided in an embodiment of this application.

[0055] Figure 3 This is a schematic diagram of the hardware entity of an edge device provided in an embodiment of this application. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application are further described in detail below with reference to the accompanying drawings and embodiments. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0057] In the following description, references to "some embodiments" describe a subset of all possible embodiments; however, it is understood that "some embodiments" may be the same or different subsets of all possible embodiments and may be combined with each other without conflict. The terms "first / second / third" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first / second / third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing this application only and is not intended to limit this application.

[0058] This application provides an intelligent task allocation method for edge computing, which can be executed by the processor of an edge device. The type of edge device is not limited.

[0059] Figure 1 This is a schematic diagram illustrating the implementation process of an edge computing task intelligent allocation method provided in an embodiment of this application, as shown below. Figure 1 As shown, the method includes the following steps:

[0060] Step S1: Obtain tasks to be assigned;

[0061] Step S2: Obtain various reference tasks that have been pre-established for various reference subtypes;

[0062] Step S3: Load the task to be assigned into the pre-debugged target task description element mining model to obtain the corresponding initial task description elements to be assigned, and load the various reference tasks into the target task description element mining model to obtain the corresponding initial reference task description elements.

[0063] Step S4: Based on the pre-debugging target optimization model, and based on the attribute description elements of various types obtained from learning different types of task data samples during pre-debugging, perform main task information enhancement operations on the initial task description elements to be assigned and various initial reference task description elements respectively to obtain the corresponding target task description elements to be assigned and various target reference task description elements.

[0064] Step S5: Based on the pre-tested target allocation model, and based on the commonality measurement results between the target task description elements and each target reference task description element, determine the task type corresponding to the task to be allocated;

[0065] Step S6: Based on the task type, assign the task to be assigned to the edge computing node corresponding to the task type.

[0066] Edge computing is a distributed computing architecture that moves computation and data storage to the edge of the network, near devices or end users, to reduce data transmission latency and improve response speed. For example, a smart speaker in a smart home, upon receiving a user's voice command, doesn't need to send the command to a remote cloud server for processing; instead, it performs the computation on a gateway device at the edge of the home network and returns the result, thus achieving a rapid response. Task allocation refers to the process of assigning specific computational or processing tasks to appropriate computing resources or nodes. For example, in a network of multiple computers, when a user requests video editing, the system allocates the video editing task to the most suitable computer based on the load and processing power of each computer.

[0067] In this embodiment, the reference task is a predefined task template with typical characteristics, used to provide a reference or comparison benchmark for tasks to be assigned. Task description elements refer to key information that characterizes task features or requirements, such as task size, required computing resources, and real-time requirements. For example, in an online game application, a player's game task may consume a large amount of computing resources and network bandwidth, while requiring real-time response. These requirements constitute the task description elements of that task. The target task description element mining model is a machine learning model used to extract key task description elements from a task.

[0068] A goal optimization model is a machine learning model used to optimize or enhance task description elements based on task characteristics and requirements, highlighting the key information of the task. Commonality measurement refers to the process or method of measuring the similarity or commonalities between two or more tasks, which can be implemented using distance metrics.

[0069] A target allocation model is a machine learning model used to assign tasks to the most suitable computing resources or nodes based on the characteristics and requirements of the task. For example, in a cloud computing environment consisting of multiple servers, a target allocation model might assign a newly arriving computing task to the most suitable server for processing based on factors such as the load, processing power, and network conditions of each server.

[0070] The edge computing task intelligent allocation method provided in this application begins with acquiring tasks to be allocated. This means that the system needs to receive or retrieve the tasks that need to be allocated. These tasks may be requests from different applications or services and need to be assigned to appropriate nodes in the edge computing environment for processing.

[0071] Next, various reference tasks corresponding to different reference subtypes are obtained. These reference tasks can be understood as predefined task templates with different characteristics. They are used as a reference in the subsequent task allocation process. Then, the tasks to be allocated and the reference tasks are loaded into a pre-debugged target task description element mining model. The role of this model is to extract key feature information from the tasks, i.e., task description elements. These elements can be the task size, required computing resources, real-time requirements, etc. By loading the tasks to be allocated and the reference tasks into this model, their initial task description elements can be obtained. After obtaining the initial task description elements, based on a pre-debugged target optimization model, the main task information enhancement operation is performed on these elements. The role of this optimization model is to obtain attribute description elements of various types by learning different types of task data samples. Then, it uses these attribute description elements to enhance the initial task description elements, highlighting the main information of the task, and obtaining the corresponding target task description elements to be allocated and target reference task description elements.

[0072] Next, based on a pre-tested target assignment model, the task type corresponding to the task to be assigned is determined according to the commonality measurement results between the description elements of the target task to be assigned and the description elements of each target reference task. The role of this assignment model is to find the most matching task type by comparing the similarity or commonalities between the task to be assigned and the reference tasks. This process can be viewed as a classification or clustering process, dividing the task to be assigned into the most appropriate task category.

[0073] Finally, based on the determined task type, the method assigns the task to be allocated to the edge computing node corresponding to that task type. Specifically, in the edge computing environment, multiple mapping relationships between task types and edge computing nodes are pre-established. For example, tasks with high real-time requirements are mapped to edge nodes with low network latency and high processing power, while data-intensive tasks are mapped to edge nodes with abundant storage resources and sufficient local bandwidth. After identifying the task type of the task to be allocated through the aforementioned steps, the system selects one or more target nodes corresponding to the task type from multiple available edge computing nodes according to the mapping relationship. When selecting target nodes, in addition to the basic matching rules corresponding to the task type, the real-time status information of each candidate node can be further combined, including but not limited to the node's current load, remaining computing power, available memory, communication latency, and energy consumption level, so as to select the optimal node to execute the task while satisfying the task type constraints. After determining the target node, the system sends the task data, execution parameters, and related dependent resources of the task to be allocated to the target node and triggers the node to execute. By using the above method, task type is taken as the core basis for allocation decision-making. This ensures that different types of tasks can be matched with the most suitable processing nodes, and can also dynamically adapt to the real-time changes in node status in the edge computing environment, thereby improving the overall efficiency of task processing, resource utilization and service quality.

[0074] It's important to note that this technical solution involves the application of multiple machine learning models, including a target task description feature mining model, a target optimization model, and a target assignment model. These models all require training and learning during the pre-launch debugging phase to acquire the ability to identify and process different task types and attributes. In practical applications, these models can be implemented using various algorithms or neural networks, such as decision trees, support vector machines, and deep learning. The specific choice depends on the characteristics and requirements of the task, as well as the system's resources and limitations.

[0075] In one implementation, step S4 involves strengthening the initial task description elements to be assigned by performing a main task information enhancement operation based on the various attribute description elements obtained from learning different types of task data samples during pre-debugging. Specifically, this may include: Step S41: Determining the commonalities between the initial task description elements to be assigned and each type of attribute description element based on the various attribute description elements obtained from learning different types of task data samples during pre-debugging; Step S42: Determining the pairing influence coefficient between the initial task description elements to be assigned and each type of attribute description element based on the commonalities; Step S43: Determining the enhancement strength of each type of attribute description element for each distribution vector in the initial task description elements to be assigned based on the pairing influence coefficient, and completing the main task information enhancement operation for the initial task description elements to be assigned based on each enhancement strength.

[0076] Based on this, step S4 is further refined, describing in detail how to perform main task information enhancement operations on the initial task description elements based on the attribute description elements of different types of task data samples learned during pre-debugging.

[0077] First, in step S41, based on the various attribute description elements learned during the pre-debugging phase, the commonalities between the initial task description elements to be assigned and these attribute description elements are evaluated. Commonalities can be understood as similarities or correlations. For example, if an initial task description element indicates that the task requires a large amount of computing resources and real-time response, then it has a high degree of commonality with a task category whose attribute description element is "computationally intensive and latency-sensitive".

[0078] Next, in step S42, based on the commonalities determined in step S41, a paired influence coefficient is assigned to each category of attribute description element. This coefficient can be understood as a weight, reflecting the importance or degree of influence of the attribute description element on the initial task description elements to be assigned. For example, for a task that requires a large amount of computing resources, the "computation-intensive" attribute description element that is highly related to it may be assigned a higher paired influence coefficient.

[0079] Then, in step S43, the pairing influence coefficients determined in step S42 are used to determine the reinforcement strength of each type of attribute description element on each distribution vector in the initial task description elements to be assigned. Here, the distribution vector can be understood as different dimensions or features of the task description elements. The reinforcement strength determines which aspects of the initial task description elements each attribute description element reinforces or enhances. For example, for a task requiring real-time response, the associated "latency-sensitive" attribute description element might reinforce its description of response time.

[0080] Finally, after enhancing the main task information of the initial task description elements, the computer obtains a more accurate and comprehensive task description that better reflects the essential requirements and characteristics of the task. This will aid the subsequent task allocation process, enabling tasks to be more accurately assigned to the most suitable processing nodes, thereby improving the efficiency and performance of the entire system.

[0081] Step S41, based on the attribute description elements of each type obtained from learning different types of task data samples during pre-debugging, determines the commonalities between the initial task description elements to be assigned and the attribute description elements of each type, which may specifically include:

[0082] Step S411: Obtain the attribute description elements of each type obtained from learning different types of task data samples during pre-debugging, and perform morphological transformation on the attribute description elements of each type to obtain the attribute description array corresponding to each type of attribute description element, and perform morphological transformation on the initial task description elements to be assigned to obtain the corresponding task description array.

[0083] Step S412: Calculate the array multiplication result between the attribute description array and the task description array to obtain a commonality array indicating the commonalities between the initial task description elements to be assigned and the attribute description elements of each type.

[0084] Step S41 is further refined, detailing how to determine the commonalities between the initial task description elements to be assigned and the attribute description elements of each category. First, in step S411, the attribute description elements of each category learned during the pre-debugging phase are acquired. These attribute description elements are obtained by analyzing different types of task data samples; they describe the characteristics of the task in some form (such as text, numerical values, etc.). To facilitate subsequent calculations and comparisons, these attribute description elements undergo morphological transformation, that is, they are converted into a form more suitable for calculation, such as a matrix or array. Similarly, the initial task description elements to be assigned also undergo morphological transformation to obtain a corresponding task description array. The specific method of morphological transformation depends on the nature of the attribute description elements and task description elements, as well as the needs of subsequent calculations. For example, if the attribute description elements are in text form, the morphological transformation may include operations such as text segmentation and vectorization to convert them into numerical arrays or matrices. If the attribute description elements are already numerical, the morphological transformation may include operations such as normalization and standardization to eliminate dimensional differences between different features.

[0085] Next, in step S412, the result of the array multiplication between the attribute description array and the task description array is calculated. This array multiplication is not a simple element-wise multiplication, but a calculation method that reflects the similarity or correlation between the two arrays. The specific calculation method depends on the properties of the arrays and the computational requirements. For example, dot product, cosine similarity, etc., can be used.

[0086] A commonality array is obtained by multiplying the attribute description array and the task description array. Each element in this commonality array reflects the commonalities between the initial task description elements to be assigned and the corresponding category attribute description elements. The higher the value of the commonality array, the stronger the commonalities between the initial task description elements to be assigned and the category attribute description elements, meaning they are more similar or related.

[0087] In summary, by using morphological transformation and array multiplication, the commonalities between the initial task description elements and the attribute description elements of each category were quantified. This facilitates the subsequent task allocation process, enabling tasks to be more accurately assigned to processing nodes that match their characteristics, thereby improving the efficiency and performance of the entire system.

[0088] As one implementation, in steps S42 and S43, based on the commonalities, the pairing influence coefficient between the initial task description elements to be assigned and each type of attribute description element is determined, and the reinforcement strength of each type of attribute description element for each distribution vector in the initial task description elements to be assigned is determined through the pairing influence coefficient. Specifically, this may include:

[0089] Step S42a: Perform a standardization operation on the common coefficient of each element in the commonity array, and obtain an influence coefficient array that indicates the pairing influence coefficient between the initial task description elements to be assigned and the description elements of each type of attribute through the common coefficient of each element after the standardization operation.

[0090] Step S42b: Based on the array multiplication result between the description array composed of the description elements of each type of attribute and the influence coefficient array, determine the reinforcement strength of each type of attribute description element for each element description value in the initial task to be assigned description element.

[0091] Steps S42 and S43 are further refined, detailing how to determine the pairing influence coefficients between the initial task description elements and the various category attribute description elements based on commonalities, and how to determine the reinforcement strength of each category attribute description element on each distribution vector in the initial task description elements using the pairing influence coefficients. First, in step S42a, each element in the commonalities array is standardized. The purpose of standardization is to convert the commonalities coefficients (i.e., similarity values) into values ​​within a standard range for subsequent calculation and comparison. Standardization can employ various methods, such as min-max standardization or Z-score standardization. After standardization, an influence coefficient array is obtained, where each element indicates the pairing influence coefficient between the initial task description elements and the corresponding category attribute description elements.

[0092] The pairwise influence coefficient can be understood as a weight, reflecting the importance or degree of influence of the attribute description element on the initial task description element to be assigned. For example, if an attribute description element is very similar to the initial task description element to be assigned, then their pairwise influence coefficient will be high, indicating that the attribute description element has a greater influence on the initial task description element to be assigned. Next, in step S42b, the influence coefficient array obtained in step S42a, and the description array composed of each type of attribute description element, are used to calculate the reinforcement strength of each type of attribute description element on each element description value in the initial task description element to be assigned. This calculation can be performed by array multiplication, that is, multiplying the influence coefficient array by the description array.

[0093] Reinforcement intensity can be understood as the degree to which various attribute descriptive elements strengthen the distribution vectors in the initial task description elements. The calculation of reinforcement intensity considers the pairwise influence coefficient, thus reflecting the different influences of different attribute descriptive elements on the initial task description elements. Through the calculation of reinforcement intensity, the distribution vectors in the initial task description elements will be strengthened or adjusted accordingly, thereby more accurately reflecting the essential requirements and characteristics of the task.

[0094] In summary, through standardization operations and array multiplication calculations, the pairing influence coefficients between the initial task description elements and the attribute description elements of each category were determined. Furthermore, the reinforcement strength of each attribute description element on each distribution vector in the initial task description elements was calculated. This facilitates the subsequent task allocation process, enabling tasks to be more accurately assigned to processing nodes that match their characteristics, thereby improving the efficiency and performance of the entire system.

[0095] As one implementation method, in step S43, based on each enhancement intensity, the main task information enhancement operation of the initial task description elements to be assigned is completed, which may specifically include:

[0096] Step S431: Add the enhancement strength of each element to the description value of the corresponding element in the initial task description elements to obtain the accumulated description value of each element.

[0097] Step S432: Perform a feedforward operation on the accumulated description value corresponding to each element to strengthen the main task information in the initial task description elements to be assigned.

[0098] Step S43 is further refined, detailing how to enhance the main task information of the initial task description elements based on various enhancement intensities. First, in step S431, the enhancement intensity of each element is accumulated with the description value of the corresponding element in the initial task description elements. Here, the description value can be a numerical representation of each feature or attribute in the task description element. The purpose of the accumulation operation is to apply the enhancement intensity to the corresponding description value, thereby enhancing the main task information in the initial task description elements. For example, suppose the initial task description elements include features such as task computation requirements, memory requirements, and data transmission requirements, and the enhancement intensity corresponding to each feature has been calculated. The computer will add the enhancement intensity of each feature to its corresponding description value, thus obtaining the accumulated description value. In this way, features that are more important for task allocation will be more significantly enhanced.

[0099] Next, in step S432, the accumulated description values ​​corresponding to each element obtained in step S431 are subjected to a feedforward operation. A feedforward operation typically refers to directly passing the input data to the next layer processing unit without any feedback or loop connections. Here, the feedforward operation can be understood as updating the initial task description elements to be assigned with the accumulated description values, thereby strengthening their main task information. Based on this, the main task information in the initial task description elements to be assigned can be targeted and strengthened according to various strengthening intensities. This helps improve the accuracy and completeness of the task description, making the subsequent task allocation process more efficient and accurate. Furthermore, this strengthening method can be flexibly adjusted and optimized according to actual needs to adapt to different task allocation scenarios.

[0100] The method provided in this application embodiment further includes a step of pre-debugging the task allocation algorithm covering the task description element mining model, optimization model, and allocation model. Specifically, it may include the following steps:

[0101] Step S10: Based on the target training library, perform repeated iterative pre-debugging of the task allocation algorithm until the preset debugging cutoff condition is met. During each debugging process, the following steps are performed:

[0102] Step S20: By repeatedly debugging the task allocation algorithm to be debugged a preset number of times using the task data sample group determined in the sub-debug development set, a task allocation algorithm to be verified is obtained, and the allocation accuracy of the task allocation algorithm to be verified is determined by the verification task data determined in the sub-debug verification set, wherein the sub-debug development set and the sub-debug verification set belong to the target training library.

[0103] Step S30: Based on the allocation accuracy determined for the proposed verification task allocation algorithm in different debugging rounds, select the target task allocation algorithm that meets the set selection requirements from each proposed verification task allocation algorithm.

[0104] The method provided in this application includes a pre-debugging step of a task allocation algorithm that covers a task description element mining model, an optimization model, and an allocation model. This series of steps aims to ensure that the task allocation algorithm meets preset performance requirements before practical application.

[0105] First, in step S10, the task allocation algorithm is iteratively debugged multiple times based on the target training library. This process continues until preset debugging cutoff conditions are met, such as reaching a preset number of iterations, algorithm performance convergence, or meeting specific accuracy requirements. During each debugging process, a series of operations are performed to verify and optimize the algorithm's performance.

[0106] Next, in step S20, the task allocation algorithm to be debugged is repeatedly debugged a predetermined number of times using the task data sample set determined in the sub-debug development set. This process aims to verify the algorithm's performance under different conditions using actual task data samples. Through repeated debugging, a task allocation algorithm to be verified is obtained, and the allocation accuracy of the algorithm is determined using the verification task data determined in the sub-debug verification set. Here, both the sub-debug development set and the sub-debug verification set are part of the target training library, which are used for algorithm development and verification, respectively.

[0107] Finally, in step S30, based on the allocation accuracy determined for the proposed task allocation algorithms in different debugging rounds, a target task allocation algorithm that meets the set selection requirements is selected from each proposed task allocation algorithm. This process aims to select the algorithm with the best performance from multiple candidate algorithms as the final target task allocation algorithm. The set selection requirements may include multiple aspects such as the algorithm's allocation accuracy, computational complexity, and real-time requirements.

[0108] A series of pre-training (or debugging) steps are used to ensure that the task allocation algorithm meets the preset performance requirements before practical application. By repeatedly iterating and debugging, validating with actual task data examples, and selecting the optimal algorithm from multiple candidate algorithms, the accuracy and reliability of the task allocation algorithm can be effectively improved, thereby ensuring that the subsequent task allocation process can be carried out more efficiently and accurately.

[0109] In step S20, the task data samples are determined from the sub-debug development set, including:

[0110] Step S21: In the sub-debug development set, for each of the i pre-determined sub-task types, extract j task data samples as a type of control sample, and extract u non-overlapping (i.e. non-repeating) task data samples as proposed assignment samples.

[0111] Step S22: Based on the determined various comparison samples and various proposed assignment samples, establish i×u task data sample groups, wherein each task data sample group includes a proposed assignment sample corresponding to a subtask type, where i, j and u are natural numbers greater than or equal to 1.

[0112] Step S20 is further refined, detailing how task data samples are determined from the sub-debug development set. First, in step S21, operations are performed on the pre-determined i sub-task types within the sub-debug development set. For each sub-task type, j task data samples are extracted as a set of control samples. These control samples will be used for subsequent performance comparison and verification. Simultaneously, for each sub-task type, the computer also extracts u disjoint (i.e., non-repeating) task data samples as proposed allocation samples. These proposed allocation samples will be used to simulate the actual task allocation process.

[0113] For example, suppose a sub-debug development set contains 1000 task data samples, and 5 sub-task types are predetermined (i.e., i=5). For each sub-task type, the computer might select 10 task data samples as control samples (i.e., j=10) and 20 disjoint task data samples as proposed assignment samples (i.e., u=20). In this way, each sub-task type will have its own set of control samples and its own set of proposed assignment samples.

[0114] Next, in step S22, task data sample groups are established based on the various comparison samples and proposed allocation samples determined in step S21. Specifically, i×u task data sample groups are established, where each task data sample group includes a proposed allocation sample corresponding to a subtask type. These task data sample groups will be used for subsequent task allocation algorithm debugging and verification.

[0115] Continuing with the example above, for the 5 sub-task types and 20 proposed assignment examples for each sub-task type, we create 5 × 20 = 100 task data example groups. Each task data example group will contain one proposed assignment example corresponding to one sub-task type. In this way, during the subsequent task allocation algorithm debugging process, the computer can use these task data example groups to simulate actual task allocation scenarios and evaluate the allocation accuracy of the algorithm.

[0116] This allows for targeted selection of task data samples from the sub-debugging development set for algorithm debugging and verification. This helps ensure that the task allocation algorithm demonstrates good performance across various sub-task types before practical application. Furthermore, using disjoint task data samples as proposed allocation examples more realistically simulates the actual task allocation process, thereby improving the effectiveness and reliability of algorithm debugging.

[0117] Based on this, in step S20, the task allocation algorithm to be debugged is repeatedly debugged a preset number of times using the task data sample group determined in the sub-debug development set to obtain the task allocation algorithm to be verified, including:

[0118] Step S22a: Based on the established task data sample group, the task allocation algorithm to be debugged is repeatedly debugged a preset number of times, and the algorithm parameters of the task allocation algorithm are optimized by calculating the error obtained in each debugging session; wherein, in one repeated debugging session, the following steps are performed:

[0119] Step S22b: Load the obtained debugging sample pair into the task allocation algorithm to be debugged, obtain the predicted task type determined by the proposed allocation sample in the debugging sample pair, and determine the error by the error between the predicted task type and the corresponding prior allocation type.

[0120] Step S20 further describes in detail how to repeatedly debug the task allocation algorithm to be debugged using task data sample groups, and optimize the algorithm parameters.

[0121] First, in step S22a, based on the established task data sample group, the task allocation algorithm to be debugged is repeatedly debugged a predetermined number of times. The purpose of this process is to optimize the algorithm's performance through multiple trials and adjustments. During each debugging session, the obtained error is calculated, and the algorithm parameters of the task allocation algorithm are optimized based on these errors. Algorithm parameters are key factors affecting the algorithm's performance and output; by adjusting these parameters, the algorithm can better adapt to different task allocation scenarios.

[0122] To illustrate this process more concretely, consider an example where the task assignment algorithm is a machine learning algorithm, such as a Support Vector Machine (SVM) or a neural network. During each debugging session, a subset of task data samples is used as training data. This data is fed into the algorithm, and the error is calculated by comparing the algorithm's output with the actual task assignment results (i.e., the prior assignment type or training labels). Then, an optimization algorithm (such as gradient descent) is used to adjust the algorithm's parameters to minimize the calculated error. This process is repeated until a preset number of debugging attempts is reached or a specific stopping condition is met.

[0123] Next, in step S22b, a series of operations are performed during each iteration of debugging. First, a debugging sample pair is obtained, which includes a proposed assignment sample and its corresponding prior assignment type (i.e., training label). Then, this debugging sample pair is loaded into the task assignment algorithm to be debugged. The algorithm generates a predicted task type based on the input proposed assignment sample. Finally, the error is determined by comparing the error between the predicted task type and the corresponding prior assignment type.

[0124] This error calculation process is crucial for optimizing algorithm performance. By comparing the predicted results with the actual results, the computer can understand where the algorithm needs improvement and adjust its parameters accordingly. With increasing iterations, the algorithm's performance gradually improves, and the prediction error gradually decreases.

[0125] The above describes a method for repeatedly debugging and optimizing a task allocation algorithm using a set of task data samples. By continuously adjusting the algorithm's parameters and calculating the error, the algorithm's performance and accuracy can be gradually improved, resulting in a more reliable and efficient task allocation algorithm.

[0126] The step of loading a pair of debug samples into the task allocation algorithm to be debugged, and obtaining the predicted task type determined by the corresponding allocation sample in the debug sample pair, includes:

[0127] Step S22b1: Load the proposed assignment samples and various control samples included in a debugging sample pair into the task description element mining model to be debugged, and obtain the task data sample description elements to be assigned and the description elements of each control task data sample.

[0128] Step S22b2: Based on the optimization model to be debugged, and based on the category attribute description elements established by learning each category, perform main task information enhancement operations on the description elements of the data sample to be assigned and the description elements of each reference task data sample to obtain the description elements of the target data sample to be assigned and the description elements of each target reference task data sample.

[0129] Step S22b3: Based on the allocation model to be debugged, and based on the commonality measurement results between the target task data sample description elements and the target comparison task data sample description elements, determine the task type corresponding to the target task data sample.

[0130] First, in step S22b1, the proposed assignment examples and various control examples contained in a debugging sample pair are loaded into the task description element mining model to be debugged. The task description element mining model is a machine learning model whose purpose is to extract key task description elements from task data examples, which are crucial for subsequent task allocation. By processing the proposed assignment examples and control examples, the model can generate description elements for the proposed task data examples and description elements for each control task data example. These description elements are abstract representations of task characteristics and requirements, helping the algorithm to more accurately understand the task content. Next, in step S22b2, the extracted task description elements are further processed using the optimization model to be debugged. The goal of the optimization model is to strengthen the main task information in the task data example description elements based on the category attribute description elements learned from each classification category. This process helps to highlight the key features of the task and improve the accuracy of task allocation. By performing main task information strengthening operations on the description elements for the proposed task data examples and description elements for each control task data example, the computer obtains the target proposed task data example description elements and description elements for each target control task data example. These target description elements focus more on the core information of the task, which helps to improve the accuracy of subsequent task allocation.

[0131] Finally, in step S22b3, the task type corresponding to the proposed task data sample is determined using the allocation model to be debugged. The role of the allocation model is to assign tasks to the most suitable processing unit or resource based on extracted and enhanced task description elements. In this step, the task type is determined based on the commonality measurement results between the description elements of the target proposed task data sample and the description elements of each target reference task data sample. The commonality measurement results reflect the similarity and differences between different task data samples, helping the algorithm to accurately determine the task type to which the proposed task data sample belongs.

[0132] In summary, this paper details the steps involved in loading debug sample pairs into the task allocation algorithm and obtaining the corresponding predicted task types. These steps include extracting task description elements using a task description element mining model, enhancing key task information using an optimization model, and determining the task type using an allocation model. Through these operations, the computer can more accurately understand and allocate tasks, improving the efficiency and accuracy of task processing.

[0133] Based on the foregoing embodiments, this application provides an edge computing task intelligent allocation device. The units and modules included in the device can be implemented by a processor in a computer device; of course, they can also be implemented by specific logic circuits. In the implementation process, the processor can be a central processing unit (CPU), a microprocessor unit (MPU), a digital signal processor (DSP), or a field programmable gate array (FPGA), etc.

[0134] Figure 2 This is a schematic diagram of the composition structure of an edge computing task intelligent allocation device provided in an embodiment of this application, as shown below. Figure 2 As shown, the edge computing task intelligent allocation device 200 includes:

[0135] The target task acquisition module 210 is used to acquire tasks to be assigned.

[0136] Reference task acquisition module 220 is used to acquire various reference tasks that have been pre-established for various reference subtypes;

[0137] The description element mining module 230 is used to load the task to be assigned into the pre-debugged target task description element mining model to obtain the corresponding initial task description elements to be assigned, and to load the various reference tasks into the target task description element mining model to obtain the corresponding initial reference task description elements.

[0138] The element information enhancement module 240 is used to perform main task information enhancement operations on the initial task description elements to be assigned and various initial reference task description elements based on the target optimization model that has been pre-tested and on the attribute description elements of various types obtained by learning from different types of task data samples during pre-testing, so as to obtain the corresponding target task description elements to be assigned and various target reference task description elements.

[0139] The task type determination module 250 is used to determine the task type corresponding to the task to be assigned based on the commonality measurement results between the target task description elements and each target reference task description elements, based on the pre-debugged target allocation model.

[0140] The task allocation module 260 is used to allocate the task to be allocated to the edge computing node corresponding to the task type based on the task type.

[0141] The descriptions of the apparatus embodiments above are similar to those of the method embodiments above, and have similar beneficial effects. In some embodiments, the functions or modules included in the apparatus provided in this application can be used to perform the methods described in the method embodiments above. For technical details not disclosed in the apparatus embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0142] It should be noted that, in the embodiments of this application, if the above-described intelligent allocation method for edge computing tasks is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of this application, or the part that contributes to the related technology, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, mobile hard drives, read-only memory (ROM), magnetic disks, or optical disks. Thus, the embodiments of this application are not limited to any specific hardware, software, or firmware, or any combination of hardware, software, and firmware.

[0143] This application provides an edge device including a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the program, it implements some or all of the steps in the above-described method.

[0144] This application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements some or all of the steps in the above-described method. The computer-readable storage medium can be transient or non-transient.

[0145] This application provides a computer program including computer-readable code, wherein when the computer-readable code is executed in a computer device, a processor in the computer device performs some or all of the steps in the above-described method.

[0146] This application provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program. When the computer program is read and executed by a computer, it implements some or all of the steps in the above-described method. This computer program product can be implemented specifically through hardware, software, or a combination thereof. In some embodiments, the computer program product is specifically embodied as a computer storage medium; in other embodiments, the computer program product is specifically embodied as a software product, such as a software development kit (SDK), etc.

[0147] It should be noted that the descriptions of the various embodiments above tend to emphasize the differences between them, while their similarities or commonalities can be referred to interchangeably. The descriptions of the above embodiments of the device, storage medium, computer program, and computer program product are similar to the descriptions of the above method embodiments and have similar beneficial effects. For technical details not disclosed in the embodiments of the device, storage medium, computer program, and computer program product of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0148] Figure 3 This is a schematic diagram of the hardware entity of an edge device provided in an embodiment of this application, such as... Figure 3 As shown, the hardware entity of the edge device 1000 includes a processor 1001 and a memory 1002, wherein the memory 1002 stores a computer program that can run on the processor 1001, and the processor 1001 executes the program to implement the steps in the method of any of the above embodiments.

[0149] The memory 1002 stores computer programs that can run on the processor. The memory 1002 is configured to store instructions and applications that can be executed by the processor 1001. It can also cache data to be processed or already processed (e.g., image data, audio data, voice communication data, and video communication data) in the processor 1001 and the various modules in the edge device 1000. It can be implemented by flash memory or random access memory (RAM).

[0150] The processor 1001 executes the program to implement the steps of the edge computing task intelligent allocation method described above. The processor 1001 typically controls the overall operation of the edge device 1000.

[0151] This application provides a computer storage medium that stores one or more programs that can be executed by one or more processors to implement the steps of the edge computing task intelligent allocation method as described in any of the above embodiments.

[0152] It should be noted that the descriptions of the storage medium and device embodiments above are similar to those of the method embodiments above, and have similar beneficial effects. For technical details not disclosed in the storage medium and device embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding. The processor described above can be at least one of the following: Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), Central Processing Unit (CPU), Controller, Microcontroller, and Microprocessor. It is understood that the electronic device implementing the above processor function can also be other types, and this application does not specifically limit the specific types.

[0153] The aforementioned computer storage media / memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc, or compact disc read-only memory (CD-ROM), etc.; or it can be various terminals that include one or any combination of the above-mentioned memories, such as mobile phones, computers, tablet devices, personal digital assistants, etc.

[0154] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of this application, the sequence number of the above-described steps / processes does not imply the order of execution; the execution order of each step / process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. The sequence numbers of the above-described embodiments are merely descriptive and do not represent the superiority or inferiority of the embodiments. It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. In the absence of further restrictions, an element defined by the phrase "including one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0155] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0156] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0157] In addition, each functional unit in the various embodiments of this application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0158] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, read-only memory (ROM), magnetic disks, or optical disks.

[0159] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence or the part that contributes to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, magnetic disks, or optical disks.

[0160] The above description is merely an embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for intelligent allocation of edge computing tasks, characterized in that, include: Get tasks to be assigned; Obtain various reference tasks that have been pre-established for various reference subtypes; The task to be assigned is loaded into the pre-debugged target task description element mining model to obtain the corresponding initial task description elements to be assigned, and the various reference tasks are loaded into the target task description element mining model to obtain the corresponding initial reference task description elements. Based on the pre-debugging target optimization model, and based on the attribute description elements of various types obtained from learning different types of task data samples during pre-debugging, the main task information enhancement operations are performed on the initial task description elements to be assigned and various initial reference task description elements to obtain the corresponding target task description elements to be assigned and various target reference task description elements. Based on the pre-tested target allocation model, and based on the common measurement results between the target task description elements and each target reference task description element, the task type corresponding to the task to be allocated is determined. Based on the task type, the task to be assigned is allocated to the edge computing node corresponding to the task type; Specifically, based on the attribute description elements of various types obtained from learning different types of task data samples during pre-debugging, the initial task description elements to be assigned undergo a main task information enhancement operation, including: Based on the attribute description elements of each type obtained from learning different types of task data samples during pre-debugging, the commonalities between the initial task description elements to be assigned and the attribute description elements of each type are determined respectively. Based on the commonalities mentioned above, the pairing influence coefficients between the initial task description elements to be assigned and the attribute description elements of each category are determined. Based on the pairing influence coefficient, the reinforcement intensity of each type of attribute description element for each distribution vector in the initial task description element to be assigned is determined, and based on each reinforcement intensity, the main task information reinforcement operation of the initial task description element to be assigned is completed. The process of determining the commonalities between the initial task description elements to be assigned and the attribute description elements of each category, based on the attribute description elements of different types obtained during pre-debugging, includes: The system acquires attribute description elements of various types obtained from learning different types of task data samples during pre-debugging, performs morphological transformation on each type of attribute description element to obtain the attribute description array corresponding to each type of attribute description element, and performs morphological transformation on the initial task description elements to be assigned to obtain the corresponding task description array. Calculate the array multiplication result between the attribute description array and the task description array to obtain a common array indicating the commonalities between the initial task description elements to be assigned and the attribute description elements of each type.

2. The method as described in claim 1, characterized in that, The step of determining the pairing influence coefficient between the initial task description elements to be assigned and each type of attribute description element based on the commonalities, and determining the reinforcement strength of each type of attribute description element for each distribution vector in the initial task description elements to be assigned through the pairing influence coefficient, includes: The common coefficients of each element in the commonity array are standardized, and the common coefficients of each element after standardization are used to obtain an influence coefficient array that indicates the pairing influence coefficients between the initial task description elements to be assigned and the description elements of each type of attribute. Based on the array multiplication result between the description array composed of various attribute description elements and the influence coefficient array, the reinforcement strength of each attribute description element for each element description value in the initial task to be assigned is determined.

3. The method as described in claim 1, characterized in that, The step of enhancing the main task information of the initial task description elements based on various enhancement intensities includes: The enhancement intensity of each element is added to the description value of the corresponding element in the initial task description elements to obtain the accumulated description value of each element. The accumulated description value corresponding to each element is subjected to a feedforward operation to enhance the main task information in the initial task description element to be assigned. The feedforward operation refers to updating the accumulated description value to the initial task description element to be assigned.

4. The method according to any one of claims 1-3, characterized in that, It also includes a pre-debugging step for the task allocation algorithm, which covers the task description element mining model, optimization model, and allocation model, including: Based on the target training library, the task allocation algorithm is repeatedly iterated and pre-tested until a preset debugging cutoff condition is met. During each debugging process, the following steps are performed: By repeatedly debugging the task allocation algorithm to be debugged a preset number of times using the task data sample group determined in the sub-debug development set, a task allocation algorithm to be verified is obtained. The allocation accuracy of the task allocation algorithm to be verified is determined by the verification task data determined in the sub-debug verification set. The sub-debug development set and the sub-debug verification set belong to the target training library. Based on the allocation accuracy determined by the proposed verification task allocation algorithm in different debugging rounds, a target task allocation algorithm that meets the set selection requirements is selected from each proposed verification task allocation algorithm.

5. The method as described in claim 4, characterized in that, Task data samples were determined from the sub-debug development set, including: In the sub-debugging development set, for each of the i pre-determined sub-task types, j task data samples are extracted as a type of control sample, and u disjoint task data samples are extracted as proposed assignment samples. Based on the identified various control samples and various proposed assignment samples, establish i×u task data sample groups, where each task data sample group includes a proposed assignment sample corresponding to a subtask type, and i, j, and u are all natural numbers greater than or equal to 1.

6. The method as described in claim 5, characterized in that, The process involves repeatedly debugging the task allocation algorithm to be debugged a predetermined number of times using a set of task data sample groups determined in the sub-debug development set, to obtain the task allocation algorithm to be verified. This includes: Based on the established task data sample group, the task allocation algorithm to be debugged is repeatedly debugged a preset number of times, and the algorithm parameters of the task allocation algorithm are optimized by calculating the error obtained in each debugging session; wherein, in one repeated debugging session, the following steps are performed: The obtained debugging sample pair is loaded into the task allocation algorithm to be debugged, and the predicted task type determined by the proposed allocation sample in the debugging sample pair is obtained. The error is determined by the error between the predicted task type and the corresponding prior allocation type. The step of loading a pair of debug samples into the task allocation algorithm to be debugged, and obtaining the predicted task type determined by the corresponding allocation sample in the debug sample pair, includes: The proposed assignment samples and various control samples included in a debugging sample pair are loaded into the task description element mining model to be debugged, so as to obtain the data sample description elements of the proposed assignment task and the data sample description elements of each control task. Based on the optimization model to be debugged, and based on the category attribute description elements established by learning each category, the main task information enhancement operation is performed on the description elements of the data sample to be assigned and the description elements of each reference task data sample, to obtain the description elements of the target data sample to be assigned and the description elements of each target reference task data sample. Based on the allocation model to be debugged, and based on the commonality measurement results between the descriptive elements of the target task data sample and the descriptive elements of each target comparison task data sample, the task type corresponding to the target task data sample is determined.

7. An intelligent task allocation device for edge computing, characterized in that, include: The target task acquisition module is used to acquire tasks to be assigned. The reference task acquisition module is used to acquire various reference tasks that have been pre-established for various reference subtypes; The description element mining module is used to load the task to be assigned into the pre-debugged target task description element mining model to obtain the corresponding initial task description elements, and to load the various reference tasks into the target task description element mining model to obtain the corresponding initial reference task description elements. The element information enhancement module is used to perform main task information enhancement operations on the initial task description elements to be assigned and various initial reference task description elements based on the target optimization model that has been pre-tested and the attribute description elements of various types obtained by learning from different types of task data samples during pre-testing, so as to obtain the corresponding target task description elements to be assigned and various target reference task description elements. The task type determination module is used to determine the task type corresponding to the task to be assigned based on the commonality measurement results between the description elements of the task to be assigned to the target and the description elements of each target reference task, based on the pre-debugged target allocation model. The task allocation module is used to allocate the task to be assigned to the edge computing node corresponding to the task type based on the task type. Specifically, based on the attribute description elements of various types obtained from learning different types of task data samples during pre-debugging, the initial task description elements to be assigned undergo a main task information enhancement operation, including: Based on the attribute description elements of each type obtained from learning different types of task data samples during pre-debugging, the commonalities between the initial task description elements to be assigned and the attribute description elements of each type are determined respectively. Based on the commonalities mentioned above, the pairing influence coefficients between the initial task description elements to be assigned and the attribute description elements of each category are determined. Based on the pairing influence coefficient, the reinforcement intensity of each type of attribute description element for each distribution vector in the initial task description element to be assigned is determined, and based on each reinforcement intensity, the main task information reinforcement operation of the initial task description element to be assigned is completed. The process of determining the commonalities between the initial task description elements to be assigned and the attribute description elements of each category, based on the attribute description elements of different types obtained during pre-debugging, includes: The system acquires attribute description elements of various types obtained from learning different types of task data samples during pre-debugging, performs morphological transformation on each type of attribute description element to obtain the attribute description array corresponding to each type of attribute description element, and performs morphological transformation on the initial task description elements to be assigned to obtain the corresponding task description array. Calculate the array multiplication result between the attribute description array and the task description array to obtain a common array indicating the commonalities between the initial task description elements to be assigned and the attribute description elements of each type.

8. An edge device comprising a memory and a processor, the memory storing a computer program executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method according to any one of claims 1 to 6.

Images