Edge device computing power resource scheduling system and method based on artificial intelligence

Abstract

The invention belongs to the technical field of computing power resource scheduling, provides an edge device computing power resource scheduling system and method based on artificial intelligence, and aims to solve the problem of scheduling failure caused by mismatching of edge device computing power release prediction and a real state. Comprising the following steps: extracting AI predicted computing power release time and real release time from historical scheduling failure data in a past period, calculating deviation, outputting a deviation distribution set through DBSCAN clustering by taking timestamps and deviation as characteristics, analyzing relevance between the deviation and the corresponding timestamps, constructing a time-computing power deviation coupling model, and calculating the time-computing power deviation. And outputting a current time range deviation based on the model, comparing the current time range deviation with a real deviation, calculating compensation time to optimize AI prediction when a scheduling risk exists, and generating a scheduling scheme in combination with a task demand and a real-time state of equipment.

Description

Technical Field

[0001] This invention belongs to the field of computing power resource scheduling technology, specifically, a method for scheduling computing power resources of edge devices based on artificial intelligence. Background Technology

[0002] As edge computing becomes the core carrier of computing power deployment due to its advantages of low latency and high bandwidth, artificial intelligence (AI) technology, with its dynamic perception and accurate prediction capabilities, has become the core driver of computing power resource scheduling for edge devices and a key support for the efficient operation of current edge computing systems.

[0003] AI-driven edge computing scheduling relies on a closed-loop process of state awareness, prediction and decision-making, and resource allocation: The scheduling system builds a data foundation by periodically reporting the computing power status of edge devices, trains a prediction model with the help of historical task execution data, accurately estimates the release time of computing power resources after the task is completed, and then allocates new tasks based on the determination that computing power has been released; the smooth operation of this process depends on the core premise that the actual computing power release status of the device is completely matched with the AI ​​prediction results. However, edge devices may experience situations where computing power resources are not fully released due to hardware characteristics, software mechanisms, or environmental disturbances. The AI ​​scheduling system, based on historical data predictions or preliminary status reports from devices, may misjudge that computing power has been fully released and allocate new tasks. Ultimately, the new tasks may become stuck, crash, or have resources preempted due to insufficient actual resources on the target device, leading to scheduling failure. This problem directly restricts the reliability of edge computing systems and has become a key bottleneck for the implementation of artificial intelligence (AI) scheduling technology.

[0004] To this end, the present invention provides an edge device computing resource scheduling system and method based on artificial intelligence. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0006] The technical solution adopted by this invention to solve its technical problem is: One objective of this invention is to provide a method for scheduling computing resources for edge devices based on artificial intelligence, comprising: S10: Obtain the predicted edge device computing power release time and the actual edge computing power release time from historical computing power scheduling data, calculate the edge device computing power release time deviation, perform cluster analysis on the edge device computing power release time deviation, and output the edge device computing power release time deviation distribution set; S20: Based on the distribution set of edge device computing power release time deviation, conduct correlation analysis on the edge device computing power release time deviation and the timestamps corresponding to the edge device computing power release time deviation, and establish a time-computing power deviation coupling model based on the correlation analysis results; S30: Based on the time-computing power deviation coupling model, combined with the current time range, output the edge device computing power release time deviation corresponding to the current time range, compare the edge device computing power release time deviation corresponding to the current time range with the actual edge device computing power release time deviation, optimize artificial intelligence based on the comparison results, and schedule edge device computing power resources based on the optimized artificial intelligence.

[0007] As a further improvement, the specific process for calculating the time deviation of computing power release from the edge device is as follows: Extract the AI-predicted edge device computing power release time from historical computing power scheduling data, and the actual time when the edge device completes the computing power release. Record this as the actual edge computing power release time. Subtract the predicted edge device computing power release time from the actual edge computing power release time. That is, the edge device computing power release time deviation is equal to the predicted edge device computing power release time minus the actual edge computing power release time.

[0008] As a further improvement, the specific process of the output edge device computing power release time deviation distribution set is as follows: Based on the calculated edge device computing power release time deviation and the timestamp corresponding to the edge device computing power release time deviation, the DBSCAN density clustering algorithm is selected. The timestamp of the edge device computing power release time deviation is used as the time dimension feature and the edge device computing power release time deviation is used as the core feature. Clustering parameters are set: neighborhood radius and minimum number of points to obtain multiple deviation distribution clusters. All deviation distribution clusters are integrated to output the edge device computing power release time deviation distribution set.

[0009] As a further improvement, the specific process of performing correlation analysis on the edge device computing power release time deviation and the timestamps corresponding to the edge device computing power release time deviation is as follows: Based on the distribution set of edge device computing power release time deviation, obtain the deviation distribution cluster corresponding to any edge device computing power release time deviation distribution in the distribution set, calculate the time range corresponding to the deviation distribution cluster, obtain the edge device computing power release time deviation value in the deviation distribution cluster, use the timestamp value in the time range corresponding to the edge device computing power release time deviation distribution as the independent variable, use the specific deviation value corresponding to each timestamp as the dependent variable, and calculate the Pearson correlation coefficient between the independent variable and the dependent variable. Preset correlation coefficient threshold; If the calculated correlation coefficient is greater than or equal to the correlation coefficient threshold, it indicates that there is a linear relationship between the independent variable and the dependent variable. If the calculated correlation coefficient is less than the correlation coefficient threshold, it indicates that there is no linear relationship between the independent variable and the dependent variable.

[0010] As a further improvement, the specific process of establishing the time-computing power deviation coupling model based on the correlation analysis results is as follows: If the correlation analysis shows that there is a linear correlation, then a time-computing power deviation coupling model is constructed by using a univariate linear regression algorithm to linearly fit the timestamp values ​​in the time range corresponding to the distribution of computing power release time deviation of edge devices and the specific deviation values ​​corresponding to each timestamp. If the correlation analysis shows that there is no linear correlation, then the timestamp values ​​in the time range corresponding to the distribution of computing power release time deviation of edge devices and the specific deviation values ​​corresponding to each timestamp are linearly fitted to construct a time-computing power deviation coupling model.

[0011] As a further improvement, the specific process of outputting the edge device computing power release time deviation corresponding to the current time range based on the time-computing power deviation coupling model and the current time range is as follows: Obtain the deviation distribution cluster corresponding to the edge device computing power release time deviation distribution within the current time range, calculate the time range corresponding to the deviation distribution cluster, and the time-computing power deviation coupling model. Input the time range to which the current time range belongs into the time-computing power deviation coupling model, and output the edge device computing power release time deviation corresponding to the current time range.

[0012] As a further improvement, the specific process of comparing the edge device computing power release time deviation corresponding to the current time range with the actual edge device computing power release time deviation is as follows: Obtain the predicted edge device computing power release time from the AI ​​output within the current time range, as well as the actual edge computing power release time within the current time range, and calculate the deviation of the actual edge device computing power release time. Based on the time-computing power deviation coupling model, the time deviation of edge device computing power release is obtained within the current time range; If the deviation of the edge device computing power release time corresponding to the current time range is greater than or equal to the actual edge device computing power release time deviation, it is determined that there is a risk of edge computing power resource scheduling.

[0013] As a further improvement, the specific process of optimizing artificial intelligence based on the comparison results is as follows: If it is determined that there is a risk of edge computing power scheduling, the compensation computing power release time deviation is obtained by subtracting the actual edge device computing power release time deviation from the edge device computing power release time deviation corresponding to the current time range. Obtain all the compensated computing power release time deviations in the deviation distribution cluster corresponding to the edge device computing power release time deviation distribution within the current time range, and take the maximum value as the compensation time; By adding a compensation time to the edge device computing power release time predicted by artificial intelligence, an optimized computing power release time prediction value is obtained, which in turn optimizes the artificial intelligence.

[0014] As a further improvement, the specific process of scheduling edge device computing resources based on optimized artificial intelligence is as follows: The core requirements of the task to be scheduled are obtained, and real-time status data of all candidate edge devices are collected. The distribution cluster of edge device computing power release time deviation to which the current time range belongs is determined. The optimized artificial intelligence prediction model is called. Based on the compensation time of the time deviation distribution cluster and combined with the core requirements of the task to be scheduled, the optimized computing power release completion time of the target edge device is output. Using the optimized computing power release completion time as the time node, and combining the real-time status data of candidate devices, devices that can release sufficient resources to undertake tasks at the time node are selected, and an edge device computing power resource scheduling scheme is generated.

[0015] The second objective of this invention is to provide an edge device computing resource scheduling system based on artificial intelligence, comprising: Deviation Data Processing and Distribution Mining Module: Obtains the predicted edge device computing power release time and the actual edge computing power release time from historical computing power scheduling data, calculates the edge device computing power release time deviation, performs cluster analysis on the edge device computing power release time deviation, and outputs the edge device computing power release time deviation distribution set; Time-Computing Power Deviation Correlation Analysis and Model Building Module: Based on the distribution set of edge device computing power release time deviation, this module performs correlation analysis on the edge device computing power release time deviation and the timestamps corresponding to the edge device computing power release time deviation, and establishes a time-computing power deviation coupling model based on the correlation analysis results. Edge computing resource scheduling and execution module: Based on the time-computing power deviation coupling model, combined with the current time range, it outputs the edge device computing power release time deviation corresponding to the current time range, compares the edge device computing power release time deviation corresponding to the current time range with the actual edge device computing power release time deviation, optimizes artificial intelligence based on the comparison results, and schedules edge device computing power resources based on the optimized artificial intelligence.

[0016] The beneficial effects of this invention are as follows: 1. By employing historical scheduling failure data mining and DBSCAN clustering technology, the time distribution characteristics of computing power release deviation are accurately captured, providing a targeted basis for subsequent prediction and optimization, and improving the pertinence of scheduling strategies; through Pearson correlation coefficient analysis and the construction of a time-computing power deviation coupling model, the correlation between deviation and timestamp is accurately quantified, significantly improving the scientificity and accuracy of computing power release time prediction.

[0017] 2. A dynamic optimization mechanism for compensation time is introduced to correct AI prediction logic in real time for scheduling risk scenarios, effectively avoiding task lag and crashes caused by prediction deviations and reducing scheduling failure rate; relying on lightweight data processing and model operation on the edge side, no new hardware is required, adapting to the low computing power requirements of edge devices, significantly reducing system deployment and operation and maintenance costs, and making it highly practical; a closed-loop technical system of data mining-model building-prediction optimization-scheduling execution-data feedback is constructed, which can dynamically respond to changes in deviation patterns, continuously iterate and optimize, and ensure the long-term efficient and reliable operation of the edge computing system. Attached Figure Description

[0018] The invention will now be further described with reference to the accompanying drawings.

[0019] Figure 1 This is a flowchart illustrating the steps of the edge device computing resource scheduling method based on artificial intelligence according to the present invention. Figure 2 This is a system module diagram of the edge device computing power resource scheduling system based on artificial intelligence, as described in this invention. Detailed Implementation

[0020] To make the technical means, creative features, objectives, and effects of this invention easily understood, this invention provides an edge device computing power resource scheduling system and method based on artificial intelligence. The system aims to solve the scheduling failure problem caused by the mismatch between the predicted and actual state of edge device computing power release. The method includes: extracting the AI-predicted computing power release time and the actual release time from historical scheduling failure data within past periods; calculating the deviation and then using timestamps and deviations as features to output a deviation distribution set through DBSCAN clustering; analyzing the correlation between the deviation and the corresponding timestamps; constructing a time-computing power deviation coupling model; outputting the deviation within the current time range based on the model; comparing it with the actual deviation; calculating compensation time to optimize AI prediction when scheduling risks exist; and then generating a scheduling scheme by combining task requirements and the real-time status of the device. The invention will be further described below with reference to specific implementation methods. Example

[0021] like Figure 1 As shown in the embodiment of the present invention, the method for scheduling computing resources of edge devices for artificial intelligence includes: S10: Obtain the predicted edge device computing power release time and the actual edge computing power release time from historical computing power scheduling data, calculate the edge device computing power release time deviation, perform cluster analysis on the edge device computing power release time deviation, and output the edge device computing power release time deviation distribution set; In S10, the specific process of obtaining the predicted edge device computing power release time and the actual edge computing power release time from historical computing power scheduling data, and calculating the edge device computing power release time deviation is as follows: Extract the AI-predicted edge device computing power release time from historical computing power scheduling data, and the actual time when the edge device completes the computing power release. Record this as the actual edge computing power release time. The difference between the predicted edge device computing power release time and the actual edge computing power release time can be used to obtain the edge device computing power release time deviation. For example, the edge device computing power release time deviation = predicted edge device computing power release time - actual edge computing power release time; The predicted edge device computing power release time is obtained by extracting it from the historical scheduling decision log of the AI ​​computing power scheduling system, associating the corresponding scheduling record through the unique task identifier (such as task ID), and reading the computing power release time prediction value field in the log. This field is an estimated value calculated by the AI ​​scheduling algorithm before the task is issued, combined with the task complexity (model parameters, data volume) and the historical load data of the device. It is synchronously written to the local log database of the scheduling system along with the scheduling instruction. The method for obtaining the actual release time of edge computing power is as follows: it is extracted from the logs of the lightweight resource monitoring process deployed on the edge device. The device-side monitoring process will detect core resource indicators such as CPU / GPU utilization, memory / video memory utilization, and related process status in real time. When all indicators drop to the preset idle threshold (such as CPU≤10%, GPU≤5%, no task-related process residue), the monitoring process automatically records the current system timestamp as the actual release time of edge computing power. It should be noted that the historical computing power scheduling data involved in this step specifically refers to the dataset corresponding to computing power scheduling failure cases in the past period. This dataset contains the AI-predicted edge device computing power release time for each failure case, as well as the actual edge computing power release time. In S10, the specific process of performing cluster analysis on the edge device computing power release time deviation and outputting the distribution set of edge device computing power release time deviation is as follows: Based on the calculated edge device computing power release time deviation and the timestamp corresponding to the edge device computing power release time deviation, the DBSCAN density clustering algorithm is selected. The timestamp of the edge device computing power release time deviation is used as the time dimension feature and the edge device computing power release time deviation is used as the core feature. The clustering parameters are set as follows: neighborhood radius (determined based on the timestamp and the edge device computing power release time deviation value, such as 1 timestamp + 50ms deviation) and minimum number of points (such as a single cluster containing at least 5 valid records). Multiple deviation distribution clusters with significant feature differences are obtained. All deviation distribution clusters are integrated to output the edge device computing power release time deviation distribution set. It should be understood that this step, through cluster analysis, can effectively uncover the temporal distribution characteristics of the deviation in the computing power release time of edge devices, specifically defining which time periods are high-incidence periods where deviations occur in concentrated form and which time periods are low-incidence periods where deviations occur in dispersed form.

[0022] S20: Based on the distribution set of edge device computing power release time deviation, conduct correlation analysis on the edge device computing power release time deviation and the timestamps corresponding to the edge device computing power release time deviation, and establish a time-computing power deviation coupling model based on the correlation analysis results; In S20, the specific process of performing correlation analysis on the edge device computing power release time deviation and the timestamps corresponding to the edge device computing power release time deviation based on the edge device computing power release time deviation distribution set is as follows: Based on the edge device computing power release time deviation distribution set, obtain the deviation distribution cluster corresponding to any edge device computing power release time deviation distribution in the edge device computing power release time deviation distribution set, calculate the time range corresponding to the deviation distribution cluster, and obtain the edge device computing power release time deviation value in the deviation distribution cluster (the specific deviation value is calculated by S10). Using the timestamp value in the time range corresponding to the edge device computing power release time deviation distribution as the independent variable and the specific deviation value corresponding to each timestamp as the dependent variable, calculate the Pearson correlation coefficient between the independent variable and the dependent variable. Preset correlation coefficient threshold (e.g., 0.7, which can be dynamically configured according to the real-time requirements of edge scheduling scenarios); If the calculated correlation coefficient is greater than or equal to the correlation coefficient threshold, it indicates that there is a linear relationship between the independent variable and the dependent variable. If the calculated correlation coefficient is less than the correlation coefficient threshold, it indicates that there is no linear relationship between the independent variable and the dependent variable. In S20, based on the correlation analysis results, the specific process of establishing the time-computing power deviation coupling model is as follows: If the correlation analysis shows that there is a linear correlation, then a time-computing power deviation coupling model is constructed by using a univariate linear regression algorithm to linearly fit the timestamp values ​​in the time range corresponding to the distribution of computing power release time deviation of edge devices and the specific deviation values ​​corresponding to each timestamp. If the correlation analysis shows that there is no linear correlation, then the timestamp values ​​in the time range corresponding to the distribution of computing power release time deviation of edge devices and the specific deviation values ​​corresponding to each timestamp are linearly fitted to construct a time-computing power deviation coupling model. S30: Based on the time-computing power deviation coupling model, combined with the current time range, output the edge device computing power release time deviation corresponding to the current time range, compare the edge device computing power release time deviation corresponding to the current time range with the actual edge device computing power release time deviation, optimize artificial intelligence based on the comparison results, and schedule edge device computing power resources based on the optimized artificial intelligence; In S30, based on the time-computing power deviation coupling model and combined with the current time range, the specific process of outputting the edge device computing power release time deviation corresponding to the current time range is as follows: Obtain the deviation distribution cluster corresponding to the edge device computing power release time deviation distribution of the current time range, calculate the time range corresponding to the deviation distribution cluster, and the time-computing power deviation coupling model. Input the time range to which the current time range belongs into the time-computing power deviation coupling model, and output the edge device computing power release time deviation corresponding to the current time range. In S30, the specific process of comparing the edge device computing power release time deviation corresponding to the current time range with the actual edge device computing power release time deviation is as follows: Obtain the predicted edge device computing power release time from the AI ​​output within the current time range, as well as the actual edge computing power release time within the current time range, and calculate the deviation of the actual edge device computing power release time. Based on the time-computing power deviation coupling model, the time deviation of edge device computing power release is obtained within the current time range; If the deviation of the edge device computing power release time corresponding to the current time range is greater than or equal to the actual edge device computing power release time deviation, it indicates that the edge device computing power release time predicted by artificial intelligence is inaccurate, and it is determined that there is a risk of edge computing power resource scheduling. If the deviation of the edge device computing power release time corresponding to the current time range is less than the actual deviation of the edge device computing power release time, it means that although the edge device computing power release time predicted by artificial intelligence is inaccurate, computing power resource scheduling can still be performed. In S30, the specific process of optimizing artificial intelligence based on the comparison results is as follows: If the comparison result indicates that there is a risk of edge computing power scheduling, the compensation computing power release time deviation is obtained by subtracting the actual edge device computing power release time deviation from the edge device computing power release time deviation corresponding to the current time range. Obtain all the compensated computing power release time deviations in the deviation distribution cluster corresponding to the edge device computing power release time deviation distribution within the current time range, and take the maximum value as the compensation time; By adding a compensation time to the original AI-predicted edge device computing power release time, an optimized computing power release time prediction value is obtained, which in turn optimizes the AI. In S30, the specific process of scheduling edge device computing resources based on optimized artificial intelligence is as follows: The core requirements of the tasks to be scheduled (including task complexity, latency threshold, and computing resource requirements) are obtained, and real-time status data of all candidate edge devices (current CPU / GPU utilization, remaining memory / video memory, and execution progress of running tasks) are collected to form a basic data pool for scheduling decisions. Determine the edge device computing power release time deviation distribution cluster to which the current time range belongs, and call the optimized artificial intelligence prediction model: based on the compensation time of the time deviation distribution cluster, combined with the core requirements of the task to be scheduled, output the optimized computing power release completion time of the target edge device (i.e. the result of the original prediction time plus the compensation time). Using the optimized computing power release completion time as the time node, and combining the real-time status data of candidate devices, devices that can release sufficient resources to undertake tasks at this time node (optimized computing power release completion time) are selected, and an edge device computing power resource scheduling scheme is generated. Example

[0023] like Figure 2 As shown in the specific implementation process of the embodiment, the present invention provides an edge device computing resource scheduling system based on artificial intelligence, including: Deviation Data Processing and Distribution Mining Module: Obtains the predicted edge device computing power release time and the actual edge computing power release time from historical computing power scheduling data, calculates the edge device computing power release time deviation, performs cluster analysis on the edge device computing power release time deviation, and outputs the edge device computing power release time deviation distribution set; Time-Computing Power Deviation Correlation Analysis and Model Building Module: Based on the distribution set of edge device computing power release time deviation, this module performs correlation analysis on the edge device computing power release time deviation and the timestamps corresponding to the edge device computing power release time deviation, and establishes a time-computing power deviation coupling model based on the correlation analysis results. Edge computing resource scheduling and execution module: Based on the time-computing power deviation coupling model, combined with the current time range, it outputs the edge device computing power release time deviation corresponding to the current time range, compares the edge device computing power release time deviation corresponding to the current time range with the actual edge device computing power release time deviation, optimizes artificial intelligence based on the comparison results, and schedules edge device computing power resources based on the optimized artificial intelligence.

[0024] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An artificial intelligence-based edge device computing resource scheduling method, characterized in that: Comprise: S10: Obtain the predicted edge device computing power release time and the edge computing power real release time from the historical computing power scheduling data, calculate the edge device computing power release time deviation, and perform clustering analysis on the edge device computing power release time deviation, and output the edge device computing power release time deviation distribution set; S20: Based on the edge device computing power release time deviation distribution set, the edge device computing power release time deviation and the time stamp corresponding to the edge device computing power release time deviation are analyzed, and a time-computing power deviation coupling model is established according to the correlation analysis result; S30: Based on the time-computing power deviation coupling model, the current time range is combined to output the edge device computing power release time deviation corresponding to the current time range, the edge device computing power release time deviation corresponding to the current time range is compared with the real edge device computing power release time deviation, and the artificial intelligence is optimized based on the comparison result. Based on the optimized artificial intelligence, the edge device computing power resource is scheduled.

2. The artificial intelligence-based edge device computing resource scheduling method of claim 1, wherein: The specific process of calculating the edge device computing power release time deviation is: Extract the artificial intelligence AI predicted edge device computing power release time and the actual edge device computing power release real time from the historical computing power scheduling data, and record the actual edge device computing power release real time as the edge computing power real release time. The difference between the predicted edge device computing power release time and the edge computing power real release time is the edge device computing power release time deviation. 3.The AI-based edge device computing resource scheduling method of claim 1, wherein: The specific process of outputting the edge device computing power release time deviation distribution set is: Based on the calculated edge device computing power release time deviation and the time stamp corresponding to the edge device computing power release time deviation, the DBSCAN density clustering algorithm is selected, the time stamp of the edge device computing power release time deviation is taken as the time dimension feature, and the edge device computing power release time deviation is taken as the core feature. Set the clustering parameters: neighborhood radius, minimum point number, get multiple deviation distribution clusters, integrate all deviation distribution clusters, and output the edge device computing power release time deviation distribution set.

4. The artificial intelligence-based edge device computing resource scheduling method of claim 1, wherein: The specific process of analyzing the correlation between the edge device computing power release time deviation and the time stamp corresponding to the edge device computing power release time deviation is: Based on the edge device computing power release time deviation distribution set, obtain the deviation distribution cluster corresponding to any edge device computing power release time deviation distribution in the edge device computing power release time deviation distribution set, calculate the time range corresponding to the deviation distribution cluster, and obtain the edge device computing power release time deviation value in the deviation distribution cluster. Take the time stamp value in the time range corresponding to the edge device computing power release time deviation distribution as the independent variable, and take the specific deviation value corresponding to each time stamp as the dependent variable. Calculate the Pearson correlation coefficient between the independent variable and the dependent variable; A preset correlation coefficient threshold is set; If the calculated correlation coefficient is greater than or equal to the correlation coefficient threshold, it indicates that there is a linear correlation between the independent variable and the dependent variable; If the calculated correlation coefficient is less than the correlation coefficient threshold, it indicates that there is no linear correlation between the independent variable and the dependent variable.

5. The artificial intelligence-based edge device computing resource scheduling method of claim 1, wherein: The specific process of establishing a time-computing power deviation coupling model according to the correlation analysis result is: If the correlation analysis is that there is a linear correlation, a linear fitting is performed on the timestamp values in the time range corresponding to the edge device computing power release time deviation distribution and the specific deviation values corresponding to each timestamp by using a linear regression algorithm, and a time-computing power deviation coupling model is constructed. If the correlation analysis is that there is no linear correlation, a linear fitting is performed on the timestamp values in the time range corresponding to the edge device computing power release time deviation distribution and the specific deviation values corresponding to each timestamp, and a time-computing power deviation coupling model is constructed. 6.The AI-based edge device computing resource scheduling method of claim 1, wherein: The specific process of outputting the specific edge device computing power release time deviation corresponding to the current time range based on the time-computing power deviation coupling model is: Obtain the deviation distribution cluster corresponding to the edge device computing power release time deviation distribution to which the current time range belongs, calculate the time range corresponding to the deviation distribution cluster, and input the time range to which the current time range belongs into the time-computing power deviation coupling model to output the edge device computing power release time deviation corresponding to the current time range.

7. The artificial intelligence-based edge device algorithm resource scheduling method of claim 1, wherein: The specific process of comparing the edge device computing power release time deviation corresponding to the current time range with the real edge device computing power release time deviation is: Obtain the predicted edge device computing power release time output by the artificial intelligence in the current time range and the real edge computing power release time in the current time range, and calculate the real edge device computing power release time deviation; Obtain the edge device computing power release time deviation corresponding to the current time range based on the time-computing power deviation coupling model; If the edge device computing power release time deviation corresponding to the current time range is greater than or equal to the real edge device computing power release time deviation, it is determined that there is an edge computing power resource scheduling risk. 8.The AI-based edge device computing resource scheduling method of claim 1, wherein: The specific process of optimizing the artificial intelligence based on the comparison result is: If it is determined that there is an edge computing power scheduling risk, the compensation computing power release time deviation is obtained by subtracting the real edge device computing power release time deviation from the edge device computing power release time deviation corresponding to the current time range; Obtain all compensation computing power release time deviations in the deviation distribution cluster corresponding to the edge device computing power release time deviation distribution to which the current time range belongs, and take the maximum value as the compensation time; Increase the compensation time based on the predicted edge device computing power release time of the artificial intelligence to obtain the optimized computing power release time prediction value, and then optimize the artificial intelligence. 9.The AI-based edge device computing resource scheduling method of claim 1, wherein: The specific process of scheduling the edge device computing power resource based on the optimized artificial intelligence is: Obtain the core demand of the task to be scheduled, simultaneously collect the real-time state data of all candidate edge devices, determine the edge device computing power release time deviation distribution cluster to which the current time range belongs, call the optimized artificial intelligence prediction model, and output the optimized computing power release completion time of the target edge device based on the compensation time of the time deviation distribution cluster and the core demand of the task to be scheduled. Take the optimized computing power release completion time as the time node, combine the real-time state data of the candidate devices, and select the devices that can release enough resources to undertake the task at the time node to generate an edge device computing power resource scheduling scheme.

10. An artificial intelligence-based edge device computing resource scheduling system for performing any of the methods of claims 1-9. It comprises: The bias data processing and distribution mining module obtains predicted edge device computing power release time and real edge computing power release time from historical computing power scheduling data, calculates edge device computing power release time bias, performs clustering analysis on the edge device computing power release time bias, and outputs an edge device computing power release time bias distribution set; The time-computing power bias correlation analysis and model construction module performs correlation analysis on the edge device computing power release time bias and the time stamp corresponding to the edge device computing power release time bias based on the edge device computing power release time bias distribution set, and establishes a time-computing power bias coupling model according to the correlation analysis result; The edge computing power resource scheduling execution module outputs the edge device computing power release time bias corresponding to the current time range based on the time-computing power bias coupling model and the current time range, compares the edge device computing power release time bias corresponding to the current time range with the real edge device computing power release time bias, optimizes the artificial intelligence based on the comparison result, and schedules the edge device computing power resource based on the optimized artificial intelligence.

Images