Grinding and cutting production line computing resource elastic scaling method based on edge-cloud cooperation

Abstract

This application relates to the field of cloud-edge collaborative resource scheduling technology, specifically to a method for elastic scaling and scheduling of computing resources in a grinding and cutting production line based on edge-cloud collaboration. The method includes: acquiring the edge resources of each computing node on an edge computing server; statistically analyzing the task queue information of each grinding and cutting robot in the production line, including urgent insertion tasks and original processing tasks; acquiring the node load of each computing node on the edge computing server; calculating the task latency and waiting urgency; and thus obtaining the processing latency of the task queue of each computing node. An objective function is constructed using the node load and processing latency of each computing node, and an optimization algorithm is used to elastically scale and schedule the computing resources of the grinding and cutting production line based on edge-cloud collaboration. This application can improve resource scheduling efficiency and optimize computing resource allocation.

Description

Technical Field

[0001] This application relates to the field of cloud-edge collaborative resource scheduling technology, specifically to a method for elastic scaling scheduling of computing resources in grinding and cutting production lines based on edge-cloud collaboration. Background Technology

[0002] Edge-cloud collaboration optimizes data processing, intelligent services, and resource scheduling through the division of labor and cooperation among cloud, edge, and terminal devices. With the development of intelligent technologies and edge-cloud computing, smart manufacturing plants are introducing grinding robot clusters to achieve collaborative control of mixed-flow production. By dynamically scheduling the grinding robot clusters used for mixed-flow production and combining edge-cloud collaborative computing technology, mixed-flow production can be completed efficiently. However, the limitations of computing power resources on the grinding production line will directly affect the task completion efficiency of the grinding robots.

[0003] In mixed-flow production in factories, there are often sudden emergency order insertion tasks, which cause the workload and complexity of edge grinding lines and cloud computing processing to change dynamically and unpredictably. The traditional computing resource scheduling methods designed for fixed workflow tasks and independent tasks of grinding robots are difficult to adapt to this dynamic edge-cloud collaborative computing scenario. They ignore the urgency of emergency order insertion tasks, resulting in insufficient allocation of computing resources to the grinding line. This leads to poor timeliness and accuracy in the optimization of grinding line processing parameters. Therefore, it is necessary to elastically scale the computing resources of the grinding line according to the dynamic characteristics of edge cloud computing tasks, optimize the allocation of computing resources, and improve the efficiency and quality of mixed-flow production in the grinding line. Summary of the Invention

[0004] To address the aforementioned technical issues, this application provides a method for elastic scaling and scheduling of computing resources in grinding and cutting production lines based on edge-cloud collaboration, thereby resolving the existing problems.

[0005] The edge-cloud collaborative method for elastically scaling and scheduling computing resources in grinding and cutting production lines in this application adopts the following technical solution: One embodiment of this application provides a method for elastically scaling and scheduling computing resources in a grinding and cutting production line based on edge-cloud collaboration, including the following steps: Obtain edge resources of each computing node of the edge computing power server, and collect the task queue information of each grinding robot in the grinding production line, including emergency order insertion tasks and original processing tasks. Also collect the product types of completed grinding tasks and the actual processing time and standard processing time of the corresponding product types. Each grinding robot corresponds to one computing node in the edge computing power server. Based on the fluctuation of resource utilization of each computing node of the edge computing server and the proportion of emergency order insertion tasks, the node load of each computing node of the edge computing server is obtained. The deviation between the actual processing time of the grinding and cutting tasks completed by each computing node and the standard processing time is analyzed, and the task latency is calculated. Based on the waiting time of emergency order insertion tasks and the waiting time of the original processing tasks of each computing node, the waiting urgency is calculated. Combined with the task latency, the processing latency of the task queue of each computing node is obtained. The objective function is constructed by considering the node load and processing latency of each computing node, and the computing resources of the edge-cloud collaborative grinding and cutting production line are elastically scaled and scheduled using optimization algorithms.

[0006] Preferably, the edge resources of each computing node within a preset time period, including memory utilization, CPU utilization, and GPU utilization, are arranged in time sequence to obtain three time-series data sequences. The average of the permutation entropy of the three time-series data sequences is used as the resource utilization fluctuation of each computing node.

[0007] Preferably, before obtaining the node load, the total number of tasks and the number of urgent tasks in the task queue of each computing power node are counted, and the ratio of the number of urgent tasks to the total number of tasks is calculated as the node task urgency of each computing power node.

[0008] Preferably, the node load of each computing node is the result of multiplying the resource utilization fluctuation and the node task urgency and then normalizing them.

[0009] Preferably, for each computing node, the deviation between the actual processing time and the standard processing time of the j-th grinding task completed by the i-th computing node is calculated. : ,in, Let represent the actual processing time and standard processing time of the j-th grinding task completed by the i-th computing node, respectively. The average deviation of all completed grinding tasks of the i-th computing node is taken as the task delay of the i-th computing node.

[0010] Preferably, the sum of the standard processing times of all grinding and cutting tasks before the kth emergency insertion task of each computing power node is used as the waiting time of the kth emergency insertion task. The sum of the waiting times of all emergency insertion tasks is used as the first waiting time of the emergency insertion task of each computing power node. Correspondingly, the second waiting time of the original processing task of each computing power node is calculated.

[0011] Preferably, the weighted sum of the first and second waiting times of each computing node of the edge computing server is used as the waiting urgency of each computing node.

[0012] Preferably, the formula for obtaining the processing latency of each computing node's task queue is: In the formula, This represents the processing latency of the task queue of the i-th computing node in the edge computing server. These represent the task latency and waiting urgency of the i-th computing node, respectively. This indicates the maximum value normalization method.

[0013] Preferably, if there is only one grinding task in the task queue of the computing power node, the waiting urgency is 0. When the maximum processing latency of all computing power nodes is 0, the processing latency is set to 0.

[0014] Preferably, the formula for obtaining the objective function is: In the formula, This represents minimizing the objective function. This represents the total number of computing nodes in the edge computing server. This represents the proportion of computing resources allocated to the i-th computing node, satisfying... and , This is a preset lower bound constant inherent to the underlying resources. Indicates the weighting coefficient. This represents the node load of the i-th computing node in the edge computing server.

[0015] This application has at least the following beneficial effects: This application addresses the problem that traditional computing resource scheduling methods designed for fixed workflow tasks and independent tasks of grinding robots are difficult to adapt to dynamic edge-cloud collaborative computing scenarios with sudden increases in urgent order insertion tasks, leading to unreasonable allocation of computing resources in the grinding production line. This application fully analyzes the load level of each computing node of the edge computing server and the latency of task queue processing, as well as the characteristics of computing node load and waiting urgency, and uses them to construct the objective function for elastic scaling scheduling of computing resources in the grinding production line, solves the computing resource scheduling instruction set of each computing node, and improves the efficiency and accuracy of elastic scaling scheduling of computing resources in the grinding production line based on edge-cloud collaboration. To address the issue of insufficient computing resources in grinding and cutting production lines that rely solely on edge computing servers, leading to overloaded computing nodes and low computational efficiency, this application combines edge-cloud collaboration technology. It performs elastic scaling and scheduling analysis of edge computing server resources on a cloud computing server, distributing computing resource scheduling instructions to the edge computing server to allocate reasonable computing resources to each computing node. This improves the resource utilization and computational efficiency of the computing nodes, thereby enhancing the production efficiency and quality of the grinding and cutting production line. Attached Figure Description

[0016] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 The flowchart illustrates the steps of the edge-cloud collaborative method for elastic scaling and scheduling of computing resources in grinding and cutting production lines provided in this application. Detailed Implementation

[0018] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the edge-cloud collaborative method for elastic scaling and scheduling of computing resources in grinding and cutting production lines proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, terms such as “comprising,” “including,” or any other variations thereof are intended to cover a non-exclusive inclusion, such that a circuit structure, article, or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase “comprising one…” does not exclude the presence of other identical elements in the article or device that includes said element. Furthermore, the term “and / or” as used herein includes any and all combinations of one or more of the associated listed items. 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 pertains.

[0020] The following description, in conjunction with the accompanying drawings, details the specific scheme of the edge-cloud collaborative method for elastic scaling and scheduling of computing resources in grinding and cutting production lines provided in this application.

[0021] This application provides an embodiment of a method for elastically scaling and scheduling computing resources in a grinding and cutting production line based on edge-cloud collaboration. For details, please refer to [link to relevant documentation]. Figure 1 This includes the following steps: Step 1: Obtain the edge resources of each computing node of the edge computing server, and collect the task queue information of each grinding robot in the grinding production line, including emergency order insertion tasks and original processing tasks. Also, collect the product types of completed grinding tasks and the actual processing time and standard processing time of the corresponding product types. Each grinding robot corresponds to one computing node in the edge computing server.

[0022] To reduce the computational load on edge computing servers and improve their computational efficiency, a large amount of production data from the grinding and cutting production line, as well as the edge resources of each computing node on the edge computing server, are uploaded to a cloud computing server for computational resource scheduling and analysis. The edge resources refer to memory utilization, CPU utilization, and GPU utilization.

[0023] Specifically, the MES (Manufacturing Execution System) used for mixed-flow production in the factory uploads the edge resources of each computing node of the edge computing server and the grinding production line production data of the edge computing server to the cloud computing server via the OPC UA protocol. The grinding production line production data includes information on completed grinding tasks and task queue information for each grinding robot in the grinding production line. The task queue information includes emergency order insertion tasks and original processing tasks. Each grinding robot corresponds to a computing node in the edge computing server. The grinding task information includes the product type processed by each grinding task, the actual processing time for that product type, and its standard processing time. The data sampling frequency is 20Hz. The MES system and the OPC UA protocol are well-known technologies, and their specific implementation will not be described in detail.

[0024] Step 2: Based on the fluctuation of resource utilization of each computing node of the edge computing server and the proportion of emergency order insertion tasks, obtain the node load of each computing node of the edge computing server. Analyze the deviation between the actual processing time of the completed grinding and cutting tasks of each computing node and the standard processing time, calculate the task latency, calculate the waiting urgency based on the waiting time of emergency order insertion tasks and the waiting time of the original processing tasks of each computing node, and obtain the processing latency of the task queue of each computing node in combination with the task latency.

[0025] S1: Obtain the node load of each computing node on the edge computing server based on the fluctuation of resource utilization of each computing node and the proportion of emergency order insertion tasks.

[0026] Timely and precise scheduling of computing resources for edge computing servers can maintain the stability of edge resources and computational efficiency of each computing node. For example, GPU utilization is highest when it is maintained in the range of 40%-60%. Conversely, if computing resources cannot be scheduled in a timely and precise manner, it may lead to overload of edge computing servers when processing different tasks on the grinding and cutting production line or at different stages of the same task, resulting in unstable fluctuations in the edge resources of its computing nodes.

[0027] To analyze the fluctuation characteristics of edge resources of each computing node at different times, the cloud computing server accesses the edge resources (memory utilization, CPU utilization, and GPU utilization) of each computing node of the edge computing server within 1 hour. The memory utilization, CPU utilization, and GPU utilization of each computing node within a preset time period are arranged in time sequence to obtain three time-series data sequences. The average of the permutation entropy of the three time-series data sequences is used as the resource utilization fluctuation of each computing node.

[0028] Preferably, in this embodiment, taking the i-th computing node as an example, the memory utilization, CPU utilization, and GPU utilization of the i-th computing node within one hour are obtained. These are then arranged in chronological order to form time-series data sequences of memory utilization, CPU utilization, and GPU utilization, respectively. The mean of the permutation entropy of the time-series data sequences of memory utilization, CPU utilization, and GPU utilization of the i-th computing node is calculated and denoted as the resource utilization volatility of the i-th computing node. This reflects the degree of fluctuation in the edge resource utilization of each computing node of the edge computing server. The greater the resource utilization volatility, the greater the fluctuation in edge resource utilization, indicating that the computing resource scheduling of the edge computing server is unreasonable. The smaller the resource utilization volatility, the more reasonable the computing resource scheduling of the edge computing server. Permutation entropy is a well-known technology, and its specific implementation will not be described in detail.

[0029] When scheduling computing resources, it is necessary to consider not only the historical edge resource occupancy of each computing node of the edge computing server, but also the urgency of the task queue to be processed in the future. When a large number of urgent insert tasks are suddenly added to the grinding and cutting production line, the urgency of the task queue will increase, thereby increasing the load on the computing nodes, and computing resources need to be scheduled first.

[0030] To analyze the urgency of the task queues of each computing node, the cloud computing server accesses the task queue information of each computing node on the edge computing server, and calculates the total number of tasks and the number of urgent tasks in the task queues. The ratio of the number of urgent tasks to the total number of tasks is then used as the node task urgency, reflecting the urgency of the task queues of each computing node on the edge computing server. A higher node task urgency indicates that the computing node has more urgent tasks among its pending tasks, increasing the node's load. It should be noted that in this embodiment, to avoid the denominator being zero during the ratio calculation, a very small positive number is added to the denominator. This does not affect the result and prevents calculation crashes caused by a zero denominator. In this embodiment, the very small positive number is 0.001.

[0031] Resource utilization volatility and node task urgency are usually closely related to the resource allocation of computing nodes on edge computing servers. The greater the resource utilization volatility, the more unreasonable the resource scheduling of computing nodes is, and the more overloaded the computing nodes are. The greater the task urgency, the more likely it is to increase the load on subsequent computing nodes. The completion efficiency of urgent tasks is related to the resource scheduling of computing power.

[0032] Based on the above analysis, the node load of the i-th computing node of the edge computing server is calculated to characterize the load level of each computing node of the edge computing server: In the formula, This represents the node load of the i-th computing node in the edge computing server. Let represent the resource utilization volatility and task urgency of the i-th computing node, respectively. This represents the error coefficient, with a value ranging from 0 to 1, without special restrictions. It is calculated considering situations where the edge resource utilization of some computing nodes remains unchanged or there are no urgent tasks in the task queue. or The value is equal to 0, therefore in this embodiment The value is set to 1 to avoid one parameter being 0 and the calculation of the other parameter being meaningless. This indicates the maximum value normalization method, which normalizes the current computing node based on the maximum node load of all computing nodes in all edge computing servers before normalization processing.

[0033] It should be noted that all parameters are raw physical quantities without physical units; they represent only mathematical calculations, and both sides of the equation are dimensionless. In existing data analysis methods, a fusion of multiple features is typically used to evaluate a specific feature. In this embodiment, to assess the load level of a computing node, the fluctuation of historical edge resources and the urgency of the upcoming task queues are analyzed to calculate resource utilization fluctuation. and node task urgency , The load level of a computing node is usually positively correlated with the load level of the computing node. Therefore, the node load level of a computing node is obtained by multiplying and fusing the two feature parameters.

[0034] S2: Analyze the deviation between the actual processing time of each computing node's completed grinding and cutting task and the standard processing time, calculate the task latency, calculate the waiting urgency based on the waiting time of each computing node's emergency insertion task and the waiting time of the original processing task, and obtain the processing latency of each computing node's task queue in combination with the task latency.

[0035] Under normal circumstances, if the computing nodes are overloaded, the edge resources will be under high load, which will reduce the data transmission and computing efficiency between the edge and cloud, directly affecting the production efficiency of the grinding and cutting production line, and causing the completion time of each grinding and cutting task to deviate from the standard processing time.

[0036] To analyze the deviation of the actual processing time of each grinding and cutting task on each computing node from the standard processing time, the cloud computing server retrieves information on the grinding and cutting tasks completed by each computing node on the edge computing server within one hour (the product type processed by each grinding and cutting task, the actual processing time of that product type, and its standard processing time). Taking the j-th grinding and cutting task as an example, based on the principle of rate of change, the deviation between the actual processing time of the j-th grinding and cutting task completed by the i-th computing node and the standard processing time is calculated. The deviation ,in, These represent the actual processing time and the standard processing time of the j-th grinding and cutting task completed by the i-th computing node, respectively.

[0037] It should be noted that the standard processing time for each grinding and cutting task in this embodiment is also the shortest processing time. Even with the most efficient scheduling of computing resources by the computing nodes, the actual processing time of the grinding and cutting task may be equal to or exceed the standard processing time. Furthermore, the average deviation between the actual processing time and the standard processing time of all completed grinding tasks on the i-th computing node is calculated and denoted as the task latency. This reflects the degree to which the actual processing time of each computing node on the edge computing server deviates from the standard processing time when completing a grinding task. A larger task latency indicates that the edge resource scheduling of the computing node is unreasonable, resulting in reduced grinding production efficiency and extended actual processing time.

[0038] To reduce the waiting time for urgent order insertion tasks on each computing node, the scheduling of computing resources should be optimized to ensure that grinding and cutting tasks preceding urgent order insertion tasks in the task queue are completed as early as possible (i.e., completed according to standard processing time). To analyze the waiting time of urgent order insertion tasks in the task queue, the cloud computing server accesses the task queue information of each computing node on the edge computing server. Taking the k-th urgent order insertion task in the task queue as an example, based on the product type and standard processing time of the grinding and cutting tasks preceding the k-th urgent order insertion task, the sum of the standard processing times of all grinding and cutting tasks preceding the k-th urgent order insertion task is calculated and recorded as the waiting time of the k-th urgent order insertion task. Then, the sum of the waiting times of all urgent order insertion tasks is calculated and recorded as the first waiting time of the ith computing node's urgent order insertion task, reflecting the waiting time of urgent order insertion tasks in the task queue of each computing node on the edge computing server. Using the same calculation method as the first waiting time, the second waiting time of the original task processed by the i-th computing node is calculated, which reflects the waiting time of non-urgent insert tasks in the task queue of each computing node of the edge computing server.

[0039] The longer the first and second waiting times are, the longer the tasks in the task queue of the computing power node need to wait, and the higher the urgency of the tasks waiting to be processed. In particular, the presence of urgent tasks in the task queue will exacerbate the urgency of the tasks waiting to be processed.

[0040] To analyze the urgency of the task queues of computing nodes, the weighted sum of the first and second waiting times of the i-th computing node on the edge computing server is calculated and denoted as the urgency of the i-th computing node. This reflects the urgency of all grinding tasks waiting in the task queues of each computing node on the edge computing server. A higher urgency indicates a longer waiting time for urgent tasks or a larger total number of tasks to be processed in the task queue of that computing node, resulting in a heavier load on the node. In this case, it is necessary to optimize the computing resource scheduling method to allocate computing resources reasonably and reduce the latency of the task queues.

[0041] The sum of the weighting coefficients for the first and second waiting times is 1. Since urgent tasks have higher time urgency and processing priority than non-urgent tasks, the weighting coefficient for the first waiting time should be larger. Assuming... Let be the weighting coefficients for the first waiting time and the second waiting time, respectively. and ,in The value range is 0-1. In this embodiment, The values ​​are 0.6 and 0.4 respectively.

[0042] Since the waiting urgency is calculated based on the standard processing time of each grinding task in the task queue, but in actual processing, the processing time of each grinding task may be too long due to unreasonable allocation of computing resources, that is, deviating from the standard processing time, this embodiment corrects the waiting urgency based on the task latency of the grinding tasks that the computing node has already completed, so that the latency of processing tasks in the task queue is closer to the actual processing scenario.

[0043] Based on the above analysis, the processing latency of the task queue of the i-th computing node of the edge computing server is calculated to characterize the latency of task processing in each computing node's task queue of the edge computing server: In the formula, This represents the processing latency of the task queue of the i-th computing node in the edge computing server. These represent the task latency and waiting urgency of the i-th computing node, respectively. This indicates a maximum value normalization method, which normalizes the processing latency of the current computing node's task queue based on the maximum processing latency of all computing node task queues across all edge computing servers. It's important to note that if each computing node's task queue contains only one grinding task, the urgency level is 0. When normalizing, the denominator may be 0. Therefore, if the maximum value of the processing delay is 0, the processing delay... Set it to 0.

[0044] In this formula, both sides of the equal sign are dimensionless. Based on prior knowledge of coefficient correction, if a parameter is too large or too small, it is usually corrected by subtracting or adding a correction parameter from its original coefficient of 1. In this embodiment, the waiting urgency is calculated based on the standard processing time, which may be smaller than the actual processing time. Therefore, the waiting urgency parameter needs to be corrected by using the task delay as the correction parameter. This makes the delay of processing tasks in the task queue closer to the actual processing scenario. After correction, the maximum value is normalized to eliminate the influence of dimensions and obtain the processing delay.

[0045] Step 3: Construct an objective function based on the node load and processing latency of each computing node, and use optimization algorithms to elastically scale and schedule the computing resources of the edge-cloud collaborative grinding and cutting production line.

[0046] When edge computing servers perform elastic scaling and scheduling of computing resources for each computing node, it is necessary to minimize the load on each computing node and shorten the completion latency of emergency order insertion tasks and original processing tasks (shorten the waiting processing latency) in order to improve the computing efficiency and accuracy of edge computing servers, thereby improving the efficiency of mixed-flow production of grinding robots in the grinding production line and ensuring that emergency order insertion tasks are completed on time and efficiently.

[0047] Based on the above analysis, an objective function for the elastic scaling and scheduling of computing resources in the grinding and cutting production line based on edge-cloud collaboration is constructed by considering the node load and processing latency of each computing node. In the formula, This represents minimizing the objective function. This represents the total number of computing nodes in the edge computing server. This represents the proportion of computing resources allocated to the i-th computing node, satisfying... and The objective function is adjusted by assigning power to each computing node. The ratio minimizes system load and latency. This is a preset lower limit constant inherent to the underlying resources, used to maintain the basic operation of the computing node's operating system and the background process connection of the OPC UA communication protocol. In this embodiment, the value is set to 0.05 to avoid the computing node having no computing resources and being insufficient to support subsequent tasks and resource scheduling data processing. This represents the weighting coefficient, with values ​​ranging from 0 to 1. This embodiment The value is 0.5.

[0048] Furthermore, a genetic algorithm is used to solve the objective function. The optimal solution is obtained by setting the initial parameters of the genetic algorithm without special restrictions, which can be set by the implementer. In this embodiment, the population size is set to 100, the crossover probability is set to 0.5, the mutation probability is set to 0.1, and the number of iterations is set to 200. After the iteration is completed, the solution that satisfies the objective function is obtained. The minimum optimal solution is obtained by extracting the overall proportional variable corresponding to the i-th computing node in the optimal solution. Then, the historical averages of CPU utilization, GPU utilization, and memory utilization of the computing node over the past hour are retrieved. Based on these historical averages, the basic proportion of resources consumed by CPU, GPU, and memory is calculated. Furthermore, using these basic proportions, the overall proportion variable is adjusted. Weighted decomposition is performed to obtain the corresponding CPU allocation ratio, GPU allocation ratio, and memory allocation ratio. In this embodiment, the overall ratio variable is... This directly serves as the unified allocation benchmark ratio for the computing power node relative to the total CPU, GPU, and memory of the edge computing power server; that is, the CPU allocation ratio, GPU allocation ratio, and memory allocation ratio are all equal to... Similarly, the CPU allocation ratio, GPU allocation ratio, and memory allocation ratio corresponding to all computing nodes are obtained and combined into a computing resource scheduling instruction set. Genetic algorithms are a well-known technique, and the process of obtaining the optimal solution will not be described in detail in this embodiment.

[0049] Subsequently, the cloud computing server transmits the set of computing resource scheduling instructions for each computing node, contained in the optimal solution, to the factory's MES system via the OPCUA protocol. The MES system then issues computing resource scheduling instructions to the edge computing server on the grinding and cutting production line. Finally, the edge computing server allocates appropriate computing resources (memory, GPU memory, and CPU resources) to each computing node according to the computing resource scheduling instructions, realizing elastic scaling and scheduling of computing resources on the grinding and cutting production line based on edge-cloud collaboration.

[0050] It is understood that references to "one embodiment" or "some embodiments" in this specification mean that one or more embodiments of this application include the specific features, structures, or characteristics described in connection with that embodiment. Therefore, the appearance of phrases such as "in one embodiment," "in some embodiments," "in other embodiments," or "in still other embodiments" in different parts of this specification does not necessarily refer to the same embodiment, but rather means "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0051] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous. Moreover, the sequence numbers of the steps in the embodiments do not imply a specific order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments in this specification.

[0052] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for elastic scaling and scheduling of computing resources in a grinding and cutting production line based on edge-cloud collaboration, characterized in that, Includes the following steps: Obtain edge resources of each computing node of the edge computing power server, and collect the task queue information of each grinding robot in the grinding production line, including emergency order insertion tasks and original processing tasks. Also collect the product types of completed grinding tasks and the actual processing time and standard processing time of the corresponding product types. Each grinding robot corresponds to one computing node in the edge computing power server. Based on the fluctuation of resource utilization of each computing node of the edge computing server and the proportion of emergency order insertion tasks, the node load of each computing node of the edge computing server is obtained. The deviation between the actual processing time of the grinding and cutting tasks completed by each computing node and the standard processing time is analyzed, and the task latency is calculated. Based on the waiting time of emergency order insertion tasks and the waiting time of the original processing tasks of each computing node, the waiting urgency is calculated. Combined with the task latency, the processing latency of the task queue of each computing node is obtained. The objective function is constructed by considering the node load and processing latency of each computing node, and the computing resources of the edge-cloud collaborative grinding and cutting production line are elastically scaled and scheduled using optimization algorithms.

2. The method for elastic scaling and scheduling of computing resources in a grinding and cutting production line based on edge-cloud collaboration as described in claim 1, characterized in that, The edge resources of each computing node within a preset time period, including memory utilization, CPU utilization, and GPU utilization, are arranged in time sequence to obtain three time-series data sequences. The average of the permutation entropy of the three time-series data sequences is used as the resource utilization fluctuation of each computing node.

3. The method for elastic scaling and scheduling of computing resources in a grinding and cutting production line based on edge-cloud collaboration as described in claim 1, characterized in that, Before obtaining the node load, the total number of tasks and the number of urgent tasks in the task queue of each computing power node are counted, and the ratio of the number of urgent tasks to the total number of tasks is calculated as the node task urgency of each computing power node.

4. The method for elastic scaling and scheduling of computing resources in a grinding and cutting production line based on edge-cloud collaboration as described in claim 3, characterized in that, The node load of each computing node is the result of multiplying the resource utilization fluctuation and the node task urgency and then normalizing them.

5. The method for elastic scaling and scheduling of computing resources in a grinding and cutting production line based on edge-cloud collaboration as described in claim 1, characterized in that, For each computing node, calculate the deviation between the actual processing time and the standard processing time of the j-th grinding task completed by the i-th computing node. : ,in, Let represent the actual processing time and standard processing time of the j-th grinding task completed by the i-th computing node, respectively. The average deviation of all completed grinding tasks of the i-th computing node is taken as the task delay of the i-th computing node.

6. The method for elastic scaling and scheduling of computing resources in a grinding and cutting production line based on edge-cloud collaboration as described in claim 1, characterized in that, The sum of the standard processing times of all grinding and cutting tasks before the kth emergency insertion task on each computing node is used as the waiting time for the kth emergency insertion task. The sum of the waiting times of all emergency insertion tasks is used as the first waiting time for the emergency insertion task on each computing node. Correspondingly, the second waiting time of the original processing task on each computing node is calculated.

7. The method for elastic scaling and scheduling of computing resources in a grinding and cutting production line based on edge-cloud collaboration as described in claim 6, characterized in that, The weighted sum of the first and second waiting times of each computing node on the edge computing server is used as the waiting urgency of each computing node.

8. The method for elastic scaling and scheduling of computing resources in a grinding and cutting production line based on edge-cloud collaboration as described in claim 1, characterized in that, The formula for obtaining the processing latency of each computing node's task queue is as follows: In the formula, This represents the processing latency of the task queue of the i-th computing node in the edge computing server. These represent the task latency and waiting urgency of the i-th computing node, respectively. This indicates the maximum value normalization method.

9. The method for elastic scaling and scheduling of computing resources in a grinding and cutting production line based on edge-cloud collaboration as described in claim 8, characterized in that, If there is only one grinding task in the task queue of a computing node, the waiting urgency is 0. When the maximum processing latency of all computing nodes is 0, the processing latency is set to 0.

10. The method for elastic scaling and scheduling of computing resources in a grinding and cutting production line based on edge-cloud collaboration as described in claim 1, characterized in that, The formula for obtaining the objective function is: In the formula, This represents minimizing the objective function. This represents the total number of computing nodes in the edge computing server. This represents the proportion of computing resources allocated to the i-th computing node, satisfying... and , This is a preset lower bound constant inherent to the underlying resources. Indicates the weighting coefficient. This represents the node load of the i-th computing node in the edge computing server.

Images