Task decomposition optimization method and device for end-side computing power network and storage medium

Abstract

The present application relates to the technical field of mobile edge computing, and particularly relates to a task decomposition optimization method and device for an end-side computing power network and a storage medium. The task decomposition optimization method comprises: obtaining decomposition parameters required for task decomposition in an end-side computing power network scenario; preprocessing and correcting a grey wolf coding matrix in an improved grey wolf optimization algorithm according to the decomposition parameters and task decomposition constraints to obtain a grey wolf coding matrix that meets the task decomposition constraints; decomposing the task using the improved grey wolf optimization algorithm to obtain a task decomposition strategy set; calculating the time delay and energy consumption of local and nearby idle terminal devices and edge servers for each strategy in the task decomposition strategy set; calculating the total cost of task decomposition according to the time delay and energy consumption of each strategy in the task decomposition strategy; and outputting a task decomposition subset with the minimum total cost of task decomposition as the optimal task decomposition strategy. The present application effectively shortens the task processing time delay, fully utilizes limited computing nodes, and activates the utilization of end-side device resources.

Description

Technical Field

[0001] This invention relates to the field of mobile edge computing technology, and in particular to a method, apparatus and storage medium for task decomposition and optimization for edge computing networks. Background Technology

[0002] With the development of the Internet of Things (IoT) and artificial intelligence (AI), the intelligence level of terminal devices is constantly increasing, and their numbers are enormous and their types are diverse. However, due to strict device size limitations and production cost considerations, terminal devices typically have batteries with limited capacity and energy-efficient / low-performance processors. Therefore, limited device lifespan and low computing power cannot support the increasing number of new applications that require sustainable and high-performance computing. Mobile cloud computing is one of the effective means to solve the problem of limited terminal device resources. Although cloud computing has powerful computing capabilities, its distance from the task terminal results in significant latency when there is a large amount of data transmission.

[0003] With the improvement of edge computing power, there is insufficient utilization of computing power supply methods between the edge and cloud. The development of edge computing and chip technologies has enabled edge networks to support a large number of computing tasks, resulting in a significant surplus of computing power. However, compared to the resource-rich cloud centers, the resource reserves of edge nodes and mobile terminals are still insufficient to meet the resource demands of applications. Furthermore, the constraints and heterogeneity of computing resources on edge devices mean that traditional binary offloading, which offloads the entire task to other nodes with computing capabilities, may lead to a high task failure rate. Therefore, task decomposition is a more resource-efficient and energy-saving approach. In task decomposition, users can break down tasks containing large amounts of data into a small number of subtasks with appropriate resource requirements to support distributed deployment. These subtasks can then be deployed to different computing nodes for parallel execution. Compared to the traditional approach of treating tasks as an indivisible whole, where tasks can only be deployed on a limited number of computing nodes, task decomposition divides massive computing tasks into subtasks of varying sizes. These decomposed parallel subtasks can be distributed and deployed across multiple terminal computing nodes, overcoming the bottleneck of limited resource capabilities of a single device, enabling flexible deployment of services, and providing a better user experience.

[0004] While task decomposition alleviates the bottlenecks of binary unloading and proportional unloading, existing decomposition methods have certain drawbacks:

[0005] 1) Defining the computation order of subtasks as serial, the latency caused by serial processing increases linearly with the number of subtasks, resulting in unacceptable latency in low-latency and ultra-low-latency applications.

[0006] 2) The task is decomposed according to its function, and the resulting granularity depends on the original size of the functional subtask.

[0007] 3) Relying on a single computing model and not considering the limitations of server computing resources will prevent the full utilization of computing resources. Summary of the Invention

[0008] The main objective of this invention is to provide a task decomposition and optimization method, apparatus, and storage medium for edge computing networks, aiming to shorten task processing latency, make full use of limited computing nodes, and revitalize the utilization of edge device resources.

[0009] To achieve the above objectives, the first aspect of this invention proposes a task decomposition and optimization method for edge computing networks, applied to edge computing network scenarios, wherein the edge computing network scenario includes edge servers and edge terminals, and the edge terminals are divided into local and nearby idle terminal devices, comprising the following steps:

[0010] Obtain the decomposition parameters required for task decomposition in edge computing power network scenarios;

[0011] Based on the decomposition parameters and task decomposition constraints, the gray wolf encoding matrix in the improved gray wolf optimization algorithm is preprocessed and corrected to obtain a gray wolf encoding matrix that conforms to the task decomposition constraints; the improved gray wolf optimization algorithm is a gray wolf optimization algorithm that incorporates the whale algorithm and the Levy flight algorithm.

[0012] The task is decomposed using the preprocessed and corrected improved Grey Wolf optimization algorithm to obtain a set of task decomposition strategies;

[0013] Calculate the latency and energy consumption of each strategy in the task decomposition strategy set for local and nearby idle terminal devices and edge servers;

[0014] Calculate the total cost of task decomposition based on the latency and energy consumption of each strategy in the task decomposition strategy set;

[0015] The subset of tasks that minimizes the total cost of task decomposition is taken as the optimal task decomposition strategy.

[0016] Optionally, this includes establishing a network model for edge computing power network scenarios;

[0017] The edge computing network scenario includes edge servers and multiple edge terminal devices. The multiple edge terminal devices are connected through D2D communication, and the edge server and the edge terminal devices are connected through cellular communication.

[0018] Multiple end-side terminal devices simultaneously generate a large number of tasks that need to be decomposed. The set of such end-side terminal devices under the base station coverage is defined as M = {m1, m2, ..., m}. o The set of computing power nodes for nearby idle terminals is defined as Help = {h1, h2, ..., h}. x};

[0019] Define the transmission rate at which users transmit computing tasks to edge servers via cellular links;

[0020]

[0021] Where W is defined as the bandwidth allocated to the user by the system. This indicates the transmission power from the user to the edge server. This represents the straight-line distance between the user and the edge server. τ represents the channel gain between the task terminal and the edge server. 2 Indicates noise power;

[0022] And define the transmission rate at which users transmit computation task data to nearby idle terminals via a D2D communication link;

[0023]

[0024] in, This indicates the transmission power from the user to a nearby idle terminal. This represents the straight-line distance between the user and a nearby available terminal. This indicates the channel gain between the task terminal and nearby idle terminals.

[0025] Optionally, this includes establishing a decomposition model for the edge computing power network scenario:

[0026] The set of tasks to be decomposed is defined as Task = {T1, T2, ..., T} n Each task contains three attributes: T i ={data i ,c i ,t max,i};data i c represents the amount of data required for task i. i This represents the computational density of task i, i.e., the number of CPU cycles required to compute 1 bit of data; t max,i This represents the maximum allowed latency to complete computation task i. Additionally, the computing power of the edge server is f. MEC The computing resources allocated to task i are f. mec,i ;

[0027] The task is defined as decomposable and can be distributed across edge servers, surrounding idle computing nodes, and the local machine. Each task can be partially or completely sent to any other computing device. Therefore, it is necessary to decompose the task and determine the components and locations of the decomposition. The decision set for all task decompositions can be represented as X = {a1, a2, ..., a...} n}, in and These represent the proportions of task i deployed on local, nearby idle computing nodes, and edge servers, respectively. If the decision rate of a task is 0, it means that the task will not be deployed to the corresponding computing device. When the decision rate of a task is equal to 1, it means that task i is fully executed on the corresponding computing device.

[0028] Optionally, it includes establishing a computing node computing model for the edge computing network scenario, wherein the computing node computing model includes a local computing model, a nearby idle terminal computing model, and an edge server computing model;

[0029] The local computation model is as follows:

[0030]

[0031] Among them, t local,i For local calculation latency, e local,i For local calculation of energy consumption, f i This indicates the amount of computing resources allocated to task i by the device. This represents the amount of task data allocated for local computation, where k is the energy consumption coefficient.

[0032] The calculation model for nearby idle terminals is as follows:

[0033]

[0034] Among them, t help,i For nearby idle terminals, For task transmission time, For task return time, f is the processing time of task i on a nearby idle terminal j; j This represents the computing power provided by nearby idle terminal j. This represents the transmission rate at which the user transmits data from computation task i to a nearby idle terminal j via the D2D communication link. This represents the amount of data to be calculated for task i on a nearby idle terminal, and ρ is the return factor, which represents the ratio between the output and input of the transmitted data.

[0035] The energy consumption of nearby idle terminals is calculated as follows:

[0036]

[0037] Among them, e help,i Calculate energy consumption for nearby idle terminals. The energy consumption for transmitting task i to a nearby idle terminal. The energy consumption for nearby idle terminals to transmit task i back to the end-side device. P represents the energy consumption of task i being processed by a nearby idle terminal. helpThis indicates the device power of the idle terminal j in the attachment. It is a terminal task device m i Upload power;

[0038] The edge server computing model is as follows:

[0039]

[0040] Among them, t mec,i For edge server latency, For data transmission delay, For data return latency, f represents the processing latency of the task on the edge server. mec,i The share of computing resources allocated to the edge server for the current task. This indicates the transmission rate at which the user transmits data for computing task i to the edge server via the cellular link. This indicates the amount of data for task i to be calculated on the edge server;

[0041] The energy consumption model for edge server computing is as follows:

[0042]

[0043] Among them, e mec,i Edge server computing power consumption The energy consumption for transmitting task i to the edge server; The energy consumption for the edge server to transmit task i back to the end device. P represents the energy consumption of task i processed on the edge server. mec This indicates the device power of the edge server.

[0044] Optionally, this includes establishing an optimization model for edge computing power network scenarios:

[0045] Total task latency is:

[0046]

[0047] Among them, t sum For the total task delay, t local,i For local calculation of latency, t help,i For the latency of nearby idle terminals, t mec,i For edge server latency;

[0048] Total energy consumption is:

[0049]

[0050] Among them, e sum For total energy consumption, e local,i For local calculation of energy consumption, ehelp,i Calculate energy consumption for nearby idle terminals, e mec,i Calculate energy consumption for edge servers;

[0051] Design a system utility function to evaluate the efficiency of task decomposition and deployment. The system utility function is as follows:

[0052] Q = λ × t sum +μ×e sum Formula (11);

[0053] Where Q represents the total cost of the system, λ+μ=1, and λ and μ represent the proportions of time delay and energy consumption in the system utility function, respectively;

[0054] The optimized model is:

[0055]

[0056] The task decomposition constraints are:

[0057]

[0058] Optionally, the decomposition parameters include the size S of the gray wolf population, the maximum number of iterations T, the number of nearby idle terminals x, the number of computation tasks n, and information about the computing power devices;

[0059] The gray wolf coding matrix in the preprocessed and corrected improved gray wolf optimization algorithm includes:

[0060] The encoding matrix for each gray wolf Where A∈R n×4 The first three columns These represent the task decomposition ratios where task i is assigned to local devices, nearby idle terminals, and edge servers, respectively; f mec,i This represents the share of computing resources allocated by the edge server to the current task i;

[0061] The constraint encoding matrix A The sum of the three is 1;

[0062] The sum of the values ​​in the fourth column of the constraint coding matrix A does not exceed 1; and the value is proportional to the proportion of task decomposition processed by the edge server.

[0063] The number of all non-zero values ​​in the second column of the constraint coding matrix A satisfies the number of nearby idle terminals;

[0064] Convert the constrained encoding matrix A of each gray wolf into a single row and store it sequentially into encoding matrix B, where B∈R C ×(n×4) C represents the size of the gray wolf population.

[0065] Optionally, decomposing the task using the preprocessed and corrected improved Grey Wolf optimization algorithm includes the following steps:

[0066] Calculate the fitness value of each gray wolf, and select the three with the lowest fitness values ​​as the alpha wolf X. α ;X β ;X δ ;

[0067] Generate a random probability factor p;

[0068] When the random probability factor p is less than 0.5, the position of the gray wolf is updated using formula (15);

[0069]

[0070] The parameter d is a constant with values ​​in the range [0,2], and u and v follow normal distributions u~N(0,σ) and v~N(0,σ), respectively.

[0071]

[0072] in, This is the globally optimal solution. Let b represent the position vector of the gray wolf, b be the constant coefficient of the spiral equation, and l be a random number between [-1, 1].

[0073] When the random probability factor p is greater than or equal to 0.5, the position of the gray wolf is updated using formulas (16) to (19).

[0074]

[0075] Where n1 and n2 represent vectors with values ​​between [0,1], m = 2 - 2t / T, t represents the current iteration number, T represents the maximum iteration number, and m decreases linearly from 2 to 0 as the iteration number increases;

[0076]

[0077] Among them, a new position update formula is designed by incorporating dynamic weights:

[0078]

[0079] Compare the current positions of the Grey Wolves And calculated using formula (20) The fitness value is retained as the position variable after this update.

[0080]

[0081] Determine if the maximum number of iterations has been reached. If it has, output the result. Otherwise, return to the gray wolf coding matrix in the preprocessed and improved gray wolf optimization algorithm.

[0082] The second aspect of this invention discloses a task decomposition and optimization apparatus for edge-side computing power networks, applied to the task decomposition and optimization method for edge-side computing power networks described in the first aspect of this invention. The apparatus includes:

[0083] The acquisition module is used to acquire the decomposition parameters required for task decomposition in edge computing power network scenarios.

[0084] The preprocessing and correction module is used to preprocess and correct the gray wolf encoding matrix in the improved gray wolf optimization algorithm according to the decomposition parameters and task decomposition constraints, so as to obtain the gray wolf encoding matrix that meets the task decomposition constraints; the improved gray wolf optimization algorithm is a gray wolf optimization algorithm that incorporates the whale algorithm and the Levy flight algorithm.

[0085] The decomposition module is used to decompose the task using the preprocessed and corrected improved Grey Wolf optimization algorithm to obtain a set of task decomposition strategies.

[0086] The latency and energy consumption calculation module is used to calculate the latency and energy consumption of local and nearby idle terminal devices and edge servers for each strategy in the task decomposition strategy set.

[0087] The total cost calculation module is used to calculate the total cost of task decomposition based on the latency and energy consumption of each strategy in the task decomposition strategy set.

[0088] The output module is used to output the subset of tasks that minimizes the total cost of task decomposition as the optimal task decomposition strategy.

[0089] The third aspect of the present invention discloses a task decomposition and optimization apparatus for edge computing networks, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described in the first aspect of the present invention.

[0090] The fourth aspect of the present invention discloses a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first aspect of the present invention.

[0091] The technical solution provided by this invention may include the following beneficial effects:

[0092] This invention uses task computation latency, computation energy consumption, and mobile edge server computing resources as constraints to calculate the computation latency and energy consumption of tasks on local, nearby idle terminal devices, and edge servers, respectively. It designs a system utility function to evaluate the decomposition strategy of the computation task, transforming the decomposition problem into a system utility optimization problem under constraints. Furthermore, it effectively explores and approximates the optimal solution of the task decomposition problem through an improved Grey Wolf optimization algorithm, thereby improving the overall performance and resource utilization efficiency.

[0093] This invention not only considers how to decompose tasks (the task decomposition problem), but also needs to decide where to process the tasks (deployment decision). Furthermore, this invention fully considers the multiple heterogeneous computing nodes in the edge device, treating them as an important part of the subtask offloading and deployment, thus making full use of limited computing nodes and revitalizing the utilization of edge device resources.

[0094] It is worth noting that in the application scenarios of the gray wolf optimization algorithm, the initialization process of the gray wolf population is often accompanied by numerical randomness. While this randomness provides a wide exploration space for the search process, it may also cause the generated initial gray wolf encoding matrix A to not satisfy the predetermined constraints. In view of this problem, this invention preprocesses and corrects the gray wolf encoding matrix in the improved gray wolf optimization algorithm according to the decomposition parameters and task decomposition constraints, thereby constraining the gray wolf encoding matrix to obtain a gray wolf encoding matrix that conforms to the task decomposition constraints, ensuring that it meets the needs of practical applications.

[0095] By modifying the gray wolf encoding matrix, its uniqueness and efficiency are enhanced, making the decomposition process more closely aligned with the actual needs of the problem. This allows for more effective decomposition of complex problems into manageable subproblems, enabling customized decomposition. It also allows for faster location of key problem components, reducing unnecessary computational overhead and improving solution efficiency. Furthermore, the optimization and adjustment of the initial encoding results, based on a deep understanding of the characteristics of task decomposition optimization problems, ensures that the initial solution is closer to the global optimum, thereby accelerating the convergence process. Precise initialization reduces the number of iterations required to reach the optimal solution, improving the overall solution speed. Attached Figure Description

[0096] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 the present invention.

[0097] Figure 1 This is a flowchart illustrating the task decomposition and optimization method for edge computing networks according to the present invention.

[0098] Figure 2This is a schematic diagram of the edge computing network scenario of the present invention;

[0099] Figure 3 This is a flowchart illustrating the improved gray wolf optimization algorithm of this invention.

[0100] Figure 4 This is a system schematic diagram of the task decomposition and optimization device for edge-side computing power networks according to the present invention;

[0101] Figure 5 This is a schematic diagram of the task decomposition and optimization device for edge computing networks according to the present invention. Detailed Implementation

[0102] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0103] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0104] In this invention, unless otherwise explicitly specified and limited, the terms "connection," "fixed," etc., should be interpreted broadly. For example, "fixed" can mean a fixed connection, a detachable connection, or an integral part; it can mean a mechanical connection or an electrical connection; it can mean a direct connection or an indirect connection through an intermediate medium; it can mean the internal communication of two components or the interaction between two components, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0105] Furthermore, in this invention, descriptions involving "first," "second," etc., are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the word "and / or" throughout the text means including three parallel solutions; taking "A and / or B" as an example, it includes solution A, solution B, or a solution that simultaneously satisfies A and B. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0106] The following is combined with Figure 1 This invention describes a task decomposition and optimization method for edge computing networks, applicable to edge computing network scenarios. The edge computing network scenario includes edge servers and edge terminals, and the edge terminals are divided into local and nearby idle terminal devices. Multiple edge terminal devices are connected through D2D communication, and the base station and edge terminal devices are connected through cellular communication.

[0107] The method includes the following steps:

[0108] Obtain the decomposition parameters required for task decomposition in the edge computing power network scenario; the decomposition parameters include the size of the gray wolf population, the maximum number of iterations, the number of nearby idle terminals, the number of tasks, and computing power device information;

[0109] Based on the decomposition parameters and task decomposition constraints, the gray wolf encoding matrix in the improved gray wolf optimization algorithm is preprocessed and corrected to obtain a gray wolf encoding matrix that conforms to the task decomposition constraints; the improved gray wolf optimization algorithm is a gray wolf optimization algorithm that incorporates the whale algorithm and the Levy flight algorithm; the task decomposition constraints include the proportion of tasks deployed to local, nearby idle terminals and edge servers, which together constitute the entire computing task; the computing resources allocated by the edge computing processor to each subtask do not exceed its own resource limit and the total resources allocated to all tasks do not exceed its own system limit; the actual completion time of the task must not exceed the maximum allowable latency for completing the task;

[0110] The task is decomposed using the preprocessed and corrected improved Grey Wolf optimization algorithm to obtain a set of task decomposition strategies;

[0111] The latency and energy consumption of each strategy in the set of computational task decomposition strategies are measured on local and nearby idle terminal devices and edge servers.

[0112] Calculate the total cost of task decomposition based on the latency and energy consumption of each strategy in the task decomposition strategy set;

[0113] The subset of tasks that minimizes the total cost of task decomposition is taken as the optimal task decomposition strategy.

[0114] This invention uses task computation latency, computation energy consumption, and mobile edge server computing resources as constraints to calculate the computation latency and energy consumption of tasks on local, nearby idle terminal devices, and edge servers, respectively. It designs a system utility function to evaluate the decomposition strategy of the computation task, transforming the decomposition problem into a system utility optimization problem under constraints. Furthermore, it effectively explores and approximates the optimal solution of the task decomposition problem through an improved Grey Wolf optimization algorithm, thereby improving the overall performance and resource utilization efficiency.

[0115] This invention not only considers how to decompose tasks (the task decomposition problem), but also needs to decide where to process the tasks (deployment decision). Furthermore, this invention fully considers the multiple heterogeneous computing nodes in the edge device, treating them as an important part of the subtask offloading and deployment, thus making full use of limited computing nodes and revitalizing the utilization of edge device resources.

[0116] It is worth noting that in the application scenarios of the gray wolf optimization algorithm, the initialization process of the gray wolf population is often accompanied by numerical randomness. While this randomness provides a wide exploration space for the search process, it may also cause the generated initial gray wolf encoding matrix A to not satisfy the predetermined constraints. In view of this problem, this invention preprocesses and corrects the gray wolf encoding matrix in the improved gray wolf optimization algorithm according to the decomposition parameters and task decomposition constraints, thereby constraining the gray wolf encoding matrix to obtain a gray wolf encoding matrix that conforms to the task decomposition constraints, ensuring that it meets the needs of practical applications.

[0117] By modifying the gray wolf encoding matrix, its uniqueness and efficiency are enhanced, making the decomposition process more closely aligned with the actual needs of the problem. This allows for more effective decomposition of complex problems into manageable subproblems, enabling customized decomposition. It also allows for faster location of key problem components, reducing unnecessary computational overhead and improving solution efficiency. Furthermore, the optimization and adjustment of the initial encoding results, based on a deep understanding of the characteristics of task decomposition optimization problems, ensures that the initial solution is closer to the global optimum, thereby accelerating the convergence process. Precise initialization reduces the number of iterations required to reach the optimal solution, improving the overall solution speed.

[0118] Specifically, such as Figure 2The invention also includes establishing a network model for the edge computing network scenario shown in the diagram. The edge computing network scenario includes a base station and multiple edge terminal devices. The multiple edge terminal devices are connected through D2D communication, and the base station and the edge terminal devices are connected through cellular communication. Both cellular communication and D2D communication in the edge computing network scenario are implemented using multi-user orthogonal frequency division multiple access technology. Since the wireless communication channels in the system are orthogonally separated, mutual interference between channels can be effectively mitigated.

[0119] Multiple end-side terminal devices simultaneously generate a large number of tasks that need to be decomposed. The set of such end-side terminal devices under the base station coverage is defined as M = {m1, m2, ..., m}. o The set of computing power nodes for nearby idle terminals is defined as Help = {h1, h2, ..., h}. x};

[0120] Define the transmission rate at which users transmit computing tasks to edge servers via cellular links;

[0121]

[0122] Where W is defined as the bandwidth allocated to the user by the system. This indicates the transmission power from the user to the edge server. This represents the straight-line distance between the user and the edge server. τ represents the channel gain between the task terminal and the edge server. 2 Indicates noise power;

[0123] And define the transmission rate at which users transmit computation task data to nearby idle terminals via a D2D communication link;

[0124]

[0125] in, This indicates the transmission power from the user to a nearby idle terminal. This represents the straight-line distance between the user and a nearby available terminal. This indicates the channel gain between the task terminal and nearby idle terminals.

[0126] This invention also includes establishing a decomposition model for edge computing power network scenarios:

[0127] The set of tasks to be decomposed is defined as Task = {T1, T2, ..., T} n Each task contains three attributes: T i ={data i ,c i ,t max,i};data ic represents the amount of data required for task i. i This represents the computational density of task i, i.e., the number of CPU cycles required to compute 1 bit of data; t max,i This represents the maximum allowed latency to complete computation task i. Additionally, the computing power of the edge server is f. MEC The computing resources allocated to task i are f. mec,i ;

[0128] The task is defined as decomposable and can be distributed across edge servers, nearby idle terminals, and local machines. Each task can be partially or completely sent to any other computing device. Therefore, it is necessary to decompose the task and determine the components and locations of the decomposition. The decision set for all task decompositions can be represented as X = {a1, a2, ..., a...} n}, in and These represent the proportions of task i deployed on local, nearby idle computing nodes, and edge servers, respectively. If the decision rate of a task is 0, it means that the task will not be deployed to the corresponding computing device. When the decision rate of a task is equal to 1, it means that task i is fully executed on the corresponding computing device.

[0129] The present invention also includes a computing node computing model for establishing a computing power network scenario on the edge side, the computing node computing model including a local computing model, a nearby idle terminal computing model and an edge server computing model;

[0130] The local computation model is as follows:

[0131] The computational load and time required to compute task i on the local smart device depend on the device's own computing power. Task i does not require data transmission during local computation, therefore there is no transmission delay, only computational delay. The local computational latency and energy consumption for task i can be expressed as:

[0132]

[0133] Among them, t local,i For local calculation latency, e local,i For local calculation of energy consumption, f i This indicates the amount of computing resources allocated to task i by the device. This represents the amount of task data allocated for local computing, and k is the energy consumption coefficient, the value of which is related to the chip architecture of the computing device and the computer's power management strategy.

[0134] The local latency and energy consumption of each strategy in the task decomposition strategy set can be calculated using formulas (3) and (4);

[0135] The calculation model for nearby idle terminals is as follows:

[0136] When a user transmits computation task i to a nearby idle terminal j with abundant resources via a D2D link for execution, the completion delay of computation task i on the nearby idle terminal is the sum of the task transmission time, the task return time, and the processing time of task i on the nearby idle terminal j, that is:

[0137]

[0138] Among them, t help,i For nearby idle terminals, For task transmission time, For task return time, f is the processing time of task i on a nearby idle terminal j; j This represents the computing power provided by nearby idle terminal j. This represents the transmission rate at which the user transmits data from computation task i to a nearby idle terminal j via the D2D communication link. This represents the amount of data to be calculated for task i on a nearby idle terminal. ρ is the return factor, which represents the ratio between the output and input of the transmitted data; it is generally a constant in the range of [0.01, 0.3].

[0139] The energy consumption of nearby idle terminals is calculated as follows:

[0140]

[0141] Among them, e help,i Calculate energy consumption for nearby idle terminals. The energy consumption for transmitting task i to a nearby idle terminal. The energy consumption for nearby idle terminals to transmit task i back to the end-side device. P represents the energy consumption of task i being processed by a nearby idle terminal. help This indicates the device power of the idle terminal j in the attachment. It is a terminal task device m i Upload power;

[0142] The latency and energy consumption of nearby idle terminal devices for each strategy in the task decomposition strategy set can be calculated using formulas (5) and (6).

[0143] The edge server computing model is as follows:

[0144] When computation task i transmits some data to the edge server (i.e., MEC base station) for execution, the completion latency of the computation task is divided into three parts: data transmission latency, data return latency, and task processing latency on the edge server. The latency at the edge server is defined as:

[0145]

[0146] Among them, t mec,i For edge server latency, For data transmission delay, For data return latency, f represents the processing latency of the task on the edge server. mec,i The share of computing resources allocated to the edge server for the current task. This indicates the transmission rate at which the user transmits data for computing task i to the edge server via the cellular link. This indicates the amount of data for task i to be calculated on the edge server;

[0147] The energy consumption model for edge server computing is as follows:

[0148]

[0149] Among them, e mec,i Edge server computing power consumption The energy consumption for transmitting task i to the edge server; The energy consumption for the edge server to transmit task i back to the end device. P represents the energy consumption of task i processed on the edge server. mec This indicates the device power of the edge server.

[0150] The latency and energy consumption of the edge server for each strategy in the task decomposition strategy set can be calculated using formulas (7) and (8);

[0151] This invention also includes establishing an optimization model for edge computing power network scenarios:

[0152] Total task latency is:

[0153]

[0154] Among them, t sum For the total task delay, t local,i For local calculation of latency, t help,i For the latency of nearby idle terminals, t mec,i For edge server latency;

[0155] Total energy consumption is:

[0156]

[0157] Among them, e sum For total energy consumption, e local,i For local calculation of energy consumption, e help,i Calculate energy consumption for nearby idle terminals, e mec,iCalculate energy consumption for edge servers;

[0158] Design a system utility function to evaluate the efficiency of task decomposition and deployment. The system utility function is as follows:

[0159] Q = λ × t sum +μ×e sum Formula (11);

[0160] Where Q represents the total cost of the system, λ+μ=1, and λ and μ represent the proportions of latency and energy consumption in the system utility function, respectively; these can be set according to the service requirements and status of the task. The total cost of task decomposition is calculated by calculating the latency and energy consumption of each strategy in the task decomposition strategy set using formula (11).

[0161] The optimized model is:

[0162]

[0163] The task decomposition constraints are:

[0164]

[0165] Specifically, the gray wolf coding matrix in the preprocessed and improved gray wolf optimization algorithm includes:

[0166] The encoding matrix for each gray wolf Where A∈R n×4 The first three columns These represent the proportions of tasks allocated to local devices, nearby idle terminals, and edge servers, respectively; f mec,i This indicates the share of computing resources allocated to the edge server for the current task;

[0167] The constraint encoding matrix A The sum of the three is 1;

[0168] The sum of the values ​​in the fourth column of the constraint coding matrix A does not exceed 1; and the value is proportional to the proportion of task decomposition processed by the edge server.

[0169] The number of all non-zero values ​​in the second column of the constraint coding matrix A satisfies the number of nearby idle terminals;

[0170] Convert the constrained encoding matrix A of each gray wolf into a single row and store it sequentially into encoding matrix B, where B∈R C ×(n×4) S represents the size of the gray wolf population. The encoding matrix B is the gray wolf encoding matrix that meets the task decomposition constraints.

[0171] In applications of the gray wolf optimization algorithm, the initialization process of the gray wolf population is often accompanied by numerical randomness. While this randomness provides a wide exploration space for the search process, it may also cause the generated initial gray wolf encoding matrix A to fail to meet the predetermined constraints. This step was designed to address this issue by constraining the gray wolf encoding matrix to ensure it meets the requirements of practical applications.

[0172] Specifically, each task contains four parameters to be optimized, namely: and f mec,i Assuming there are n tasks to be decomposed, the encoding matrix A of a gray wolf is: (The first three columns) represent the proportions of task i allocated to local devices, surrounding idle computing power terminals, and edge servers, respectively; f mec,i (Column 4) represents the share of computing resources allocated by the edge server to the current task i. To facilitate subsequent calculations and optimizations, the encoding matrix A for each gray wolf is converted into a single row and stored sequentially in encoding matrix B, thus forming the encoding matrix for the entire gray wolf pack. The sum of the first three columns of matrix A must be 1, reflecting that the total task decomposition ratio must satisfy the requirement of a single task, ensuring task integrity. Simultaneously, the sum of the values ​​in the fourth column should not exceed 1, and the value should be proportional to the task decomposition ratio processed by the edge server, ensuring that the share of computing resources allocated to the edge server is within a reasonable range, guaranteeing the limited availability of edge server resources. Furthermore, in the second column of matrix A, the number of all non-zero values ​​must satisfy the preset number of nearby idle terminal devices, ensuring that the algorithm can fully consider and rationally utilize existing idle computing resources when allocating tasks.

[0173] However, due to the randomness of gray wolf population initialization, the above constraints are often difficult to satisfy directly. Therefore, we modify the gray wolf encoding matrix. Based on the proposed decomposition model, we make necessary adjustments to the encoding results to ensure that the final encoding matrix B strictly satisfies all the given constraints.

[0174] By modifying the gray wolf encoding matrix, its uniqueness and efficiency are enhanced, making the decomposition process more closely aligned with the actual needs of the problem. This allows for more effective decomposition of complex problems into manageable subproblems, enabling customized decomposition. It also allows for faster location of key problem components, reducing unnecessary computational overhead and improving solution efficiency. Furthermore, optimizing the initial encoding results, based on a deep understanding of the problem's characteristics, ensures that the initial solution is closer to the global optimum, thereby accelerating the convergence process. Precise initialization reduces the number of iterations required to reach the optimal solution, improving overall solution speed.

[0175] More specifically, the pseudocode description of the preprocessing and correction of the gray wolf coding matrix is ​​shown in Table 1 below. Rows 5-9, in the second column of matrix A, require that the number of all non-zero values ​​satisfy a preset number of nearby idle terminal devices, ensuring that the algorithm can fully consider and rationally utilize existing idle computing resources when allocating tasks. Rows 10-14 constrain the sum of the first three columns of matrix A to be 1, reflecting that the total task decomposition ratio must satisfy a single task, ensuring task integrity. Rows 16-20 ensure that the sum of the values ​​in the fourth column should not exceed 1 and that the value is proportional to the task decomposition ratio processed by the edge server, ensuring that the share of computing resources allocated to the MEC server is within a reasonable range, guaranteeing the limited resources of the edge server.

[0176] Table 1: Pseudocode description of preprocessing and correcting the gray wolf coding matrix

[0177]

[0178]

[0179] Specifically, the task decomposition using the preprocessed and corrected improved Gray Wolf optimization algorithm includes the following steps:

[0180] Calculate the fitness value of each gray wolf, and select the three with the lowest fitness values ​​as the alpha wolf X. α ;X β ;x δ ;

[0181] Generate a random probability factor p;

[0182] When the random probability factor p is less than 0.5, the position of the gray wolf is updated using formula (15);

[0183]

[0184] The parameter d is a constant with a value of [0,2], and in this embodiment it is 1.5. u and v follow normal distributions u~N(0,σ) and v~N(0,σ), respectively.

[0185]

[0186] in, This is the globally optimal solution. Let b represent the position vector of the gray wolf, b is the constant coefficient of the spiral equation, usually b is 1, and l is a random number between [-1, 1].

[0187] When the random probability factor p is greater than or equal to 0.5, the position of the gray wolf is updated using formulas (16) to (19).

[0188]

[0189] Where n1 and n2 represent vectors with values ​​between [0,1], m = 2 - 2t / T, t represents the current iteration number, T represents the maximum iteration number, and m decreases linearly from 2 to 0 as the iteration number increases;

[0190]

[0191] Among them, a new position update formula is designed by incorporating dynamic weights:

[0192]

[0193] Compare the current positions of the Grey Wolves And calculated using formula (20) The fitness value is retained as the position variable after this update.

[0194]

[0195] Determine if the maximum number of iterations we set has been reached. If so, output the result; otherwise, return to the gray wolf encoding matrix in the preprocessed and improved gray wolf optimization algorithm.

[0196] The pseudocode description of the improved gray wolf optimization algorithm is shown in Table 2 below:

[0197] Table 2: Pseudocode description of the improved gray wolf optimization algorithm

[0198]

[0199] A second aspect of the present invention discloses a task decomposition and optimization apparatus 400 for edge computing networks, applied to the aforementioned task decomposition and optimization method for edge computing networks, the apparatus comprising:

[0200] The acquisition module 401 is used to acquire the decomposition parameters required for task decomposition in the edge computing power network scenario;

[0201] The preprocessing and correction module 402 is used to preprocess and correct the gray wolf encoding matrix in the improved gray wolf optimization algorithm according to the decomposition parameters and task decomposition constraints, so as to obtain the gray wolf encoding matrix that meets the task decomposition constraints; the improved gray wolf optimization algorithm is a gray wolf optimization algorithm that incorporates the whale algorithm and the Levy flight algorithm.

[0202] The decomposition module 403 is used to decompose the task using the preprocessed and corrected improved Grey Wolf optimization algorithm to obtain a set of task decomposition strategies.

[0203] The latency and energy consumption calculation module 404 is used to calculate the latency and energy consumption of local and nearby idle terminal devices and edge servers for each strategy in the task decomposition strategy set.

[0204] The total cost calculation module 405 is used to calculate the total cost of task decomposition based on the latency and energy consumption of each strategy in the task decomposition strategy set.

[0205] Output module 406 is used to output the subset of tasks with the minimum total cost of task decomposition as the optimal task decomposition strategy.

[0206] This invention uses task computation latency, computation energy consumption, and mobile edge server computing resources as constraints to calculate the computation latency and energy consumption of tasks on local, nearby idle terminal devices, and edge servers, respectively. It designs a system utility function to evaluate the decomposition strategy of the computation task, transforming the decomposition problem into a system utility optimization problem under constraints. Furthermore, it effectively explores and approximates the optimal solution of the task decomposition problem through an improved Grey Wolf optimization algorithm, thereby improving the overall performance and resource utilization efficiency.

[0207] This invention not only considers how to decompose tasks (the task decomposition problem), but also needs to decide where to process the tasks (deployment decision). Furthermore, this invention fully considers the multiple heterogeneous computing nodes in the edge device, treating them as an important part of the subtask offloading and deployment, thus making full use of limited computing nodes and revitalizing the utilization of edge device resources.

[0208] It is worth noting that in the application scenarios of the gray wolf optimization algorithm, the initialization process of the gray wolf population is often accompanied by numerical randomness. While this randomness provides a wide exploration space for the search process, it may also cause the generated initial gray wolf encoding matrix A to not satisfy the predetermined constraints. In view of this problem, this invention preprocesses and corrects the gray wolf encoding matrix in the improved gray wolf optimization algorithm according to the decomposition parameters and task decomposition constraints, thereby constraining the gray wolf encoding matrix to obtain a gray wolf encoding matrix that conforms to the task decomposition constraints, ensuring that it meets the needs of practical applications.

[0209] By modifying the gray wolf encoding matrix, its uniqueness and efficiency are enhanced, making the decomposition process more closely aligned with the actual needs of the problem. This allows for more effective decomposition of complex problems into manageable subproblems, enabling customized decomposition. It also allows for faster location of key problem components, reducing unnecessary computational overhead and improving solution efficiency. Furthermore, the optimization and adjustment of the initial encoding results, based on a deep understanding of the characteristics of task decomposition optimization problems, ensures that the initial solution is closer to the global optimum, thereby accelerating the convergence process. Precise initialization reduces the number of iterations required to reach the optimal solution, improving the overall solution speed.

[0210] like Figure 5 As shown, the third aspect of this invention discloses a task decomposition and optimization device 500 for edge computing networks, including a processor 501 and a memory 502. The processor 501 and the memory 502 are connected, for example, via a bus 503. Further, the industrial IoT device intelligent identification device 500 may also include a transceiver 504. It should be noted that in practical applications, the transceiver 504 is not limited to one, and the structure of the industrial IoT device intelligent identification device 500 does not constitute a limitation on the embodiments of this application. The processor 501 is used in the embodiments of this application to implement... Figure 4 The diagram shows an acquisition module 401, a preprocessing and correction module 402, a decomposition module 403, a latency and energy consumption calculation module 404, a total cost calculation module 405, and an output module 406. The processor 501 can be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. The processor 501 can also be a combination that implements computational functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0211] Bus 503 may include a path for transmitting information between the aforementioned components. Bus 503 may be a PCI bus or an EISA bus, etc. Bus 503 may be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0212] The memory 502 may be a ROM or other type of static storage device capable of storing static information and instructions, RAM or other type of dynamic storage device capable of storing information and instructions, or it may be an EEPROM, CD-ROM or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.

[0213] The memory 502 stores the application code that executes the scheme of this application, and its execution is controlled by the processor 501. The processor 501 executes the application code stored in the memory 502 to implement... Figure 1 The actions of the task decomposition and optimization method for edge computing networks provided in the illustrated embodiment.

[0214] The fourth aspect of the present invention discloses a computer-readable storage medium storing a computer program that, when executed by a processor, implements the task decomposition and optimization method for edge computing networks described in the first aspect of the present invention.

[0215] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. All equivalent structural transformations made under the concept of the present invention using the contents of the present invention specification and drawings, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A task decomposition and optimization method for edge computing networks, applied to edge computing network scenarios, wherein the edge computing network scenario includes edge servers and edge terminals, and the edge terminals are divided into local and nearby idle terminal devices, characterized in that, Includes the following steps: Obtain the decomposition parameters required for task decomposition in edge computing power network scenarios; Based on the decomposition parameters and task decomposition constraints, the gray wolf encoding matrix in the improved gray wolf optimization algorithm is preprocessed and corrected to obtain a gray wolf encoding matrix that conforms to the task decomposition constraints; the improved gray wolf optimization algorithm is a gray wolf optimization algorithm that incorporates the whale algorithm and the Levy flight algorithm. The task is decomposed using an improved gray wolf optimization algorithm, resulting in a set of task decomposition strategies. Calculate the latency and energy consumption of each strategy in the task decomposition strategy set for local and nearby idle terminal devices and edge servers; Calculate the total cost of task decomposition based on the latency and energy consumption of each strategy in the task decomposition strategy set; The subset of tasks that minimizes the total cost of task decomposition is taken as the optimal task decomposition strategy. The decomposition parameters include the size of the gray wolf population S, the maximum number of iterations T, the number of nearby idle terminals x, the number of computation tasks n, and information about the computing power devices. The gray wolf coding matrix in the preprocessed and corrected improved gray wolf optimization algorithm includes: The encoding matrix for each gray wolf Where A∈R n×4 The first three columns These represent the task decomposition ratios where task i is assigned to local devices, nearby idle terminals, and edge servers, respectively; f mec,i This represents the share of computing resources allocated by the edge server to the current task i; The constraint encoding matrix A The sum of the three is 1; The sum of the values ​​in the fourth column of the constraint coding matrix A does not exceed 1; and the value is proportional to the proportion of task decomposition processed by the edge server. The number of all non-zero values ​​in the second column of the constraint coding matrix A satisfies the number of nearby idle terminals; Convert the constrained encoding matrix A of each gray wolf into a single row and store it sequentially into encoding matrix B.

2. The task decomposition and optimization method for edge-side computing power networks according to claim 1, characterized in that: This includes establishing a network model for edge computing power network scenarios; The edge computing network scenario includes edge servers and multiple edge terminal devices. The multiple edge terminal devices are connected through D2D communication, and the edge server and the edge terminal devices are connected through cellular communication. Multiple end-side terminal devices simultaneously generate a large number of tasks that need to be decomposed. The set of such end-side terminal devices under the base station coverage is defined as M = {m1, m2, ..., m}. o The set of computing power nodes for nearby idle terminals is defined as Help = {h1, h2, ..., h}. x }; Define the transmission rate at which users transmit computing tasks to edge servers via cellular links; Where W is defined as the bandwidth allocated to the user by the system. This indicates the transmission power from the user to the edge server. This represents the straight-line distance between the user and the edge server. τ represents the channel gain between the task terminal and the edge server. 2 Indicates noise power; And define the transmission rate at which users transmit computation task data to nearby idle terminals via a D2D communication link; in, This indicates the transmission power from the user to a nearby idle terminal. This represents the straight-line distance between the user and a nearby available terminal. This indicates the channel gain between the task terminal and nearby idle terminals.

3. The task decomposition and optimization method for edge-side computing power networks according to claim 2, characterized in that: This includes establishing a decomposition model for edge computing power network scenarios: The set of tasks to be decomposed is defined as Task = {T1, T2, ..., T} n Each task contains three attributes: T i ={data i ,c i ,t max,i };data i c represents the amount of data required for task i. i This represents the computational density of task i, i.e., the number of CPU cycles required to compute 1 bit of data; t max,i This represents the maximum allowed latency to complete computation task i. Additionally, the computing power of the edge server is f. MEc The computing resources allocated to task i are f. mec,i ; The task is defined as decomposable and can be distributed across edge servers, surrounding idle computing nodes, and the local machine. Each task can be partially or completely sent to any other computing device. Therefore, it is necessary to decompose the task and determine the components and locations of the decomposition. The decision set for all task decompositions can be represented as X = {a1, a2, ..., a...} n }, If the decision rate of a task is 0, it means that the task will not be deployed to the corresponding computing device; when the decision rate of a task is equal to 1, it means that task i will perform the computation entirely on the corresponding computing device.

4. The task decomposition and optimization method for edge-side computing power networks according to claim 3, characterized in that: This includes establishing a computing node computing model for edge computing network scenarios, which includes a local computing model, a nearby idle terminal computing model, and an edge server computing model. The local computation model is as follows: Among them, t local,i For local calculation latency, e local,i For local calculation of energy consumption, f i This indicates the amount of computing resources allocated to task i by the device. This represents the amount of task data allocated for local computation, where k is the energy consumption coefficient. The calculation model for nearby idle terminals is as follows: Among them, t help,i For nearby idle terminals, For task transmission time, For task return time, f is the processing time of task i on a nearby idle terminal j; j This represents the computing power provided by nearby idle terminal j. This represents the amount of data to be calculated for task i on a nearby idle terminal, and ρ is the return factor, which represents the ratio between the output and input of the transmitted data. The energy consumption of nearby idle terminals is calculated as follows: Among them, e help,i Calculate energy consumption for nearby idle terminals. The energy consumption for transmitting task i to a nearby idle terminal. The energy consumption for nearby idle terminals to transmit task i back to the end-side device. P represents the energy consumption of task i being processed by a nearby idle terminal. help P represents the device power of the idle terminal j in the attachment. i up It is a terminal task device m i Upload power; The edge server computing model is as follows: Among them, t mec,i For edge server latency, For data transmission delay, For data return latency, For task processing latency on edge servers, This indicates the amount of data for task i to be calculated on the edge server; The energy consumption model for edge server computing is as follows: Among them, e mec,i Calculate energy consumption for edge servers. The energy consumption for transmitting task i to the edge server; The energy consumption for the edge server to transmit task i back to the end device. P represents the energy consumption of task i processed on the edge server. mec This indicates the device power of the edge server.

5. The task decomposition and optimization method for edge-side computing power networks according to claim 3, characterized in that: This includes establishing an optimization model for edge computing network scenarios: Total task latency is: Among them, t sum For the total task delay, t local,i For local calculation of latency, t help,i For the latency of nearby idle terminals, t mec,i For edge server latency; Total energy consumption is: Among them, e sum For total energy consumption, e local,i For local calculation of energy consumption, e help,i Calculate energy consumption for nearby idle terminals, e mec,i Calculate energy consumption for edge servers; Design a system utility function to evaluate the efficiency of task decomposition and deployment. The system utility function is as follows: Q=λ×t sum +μ×e sum ; Where Q represents the total cost of the system, λ+μ=1, and λ and μ represent the proportions of time delay and energy consumption in the system utility function, respectively; The optimized model is: The task decomposition constraints are:

6. The task decomposition and optimization method for edge-side computing power networks according to claim 1, characterized in that, The task decomposition using the preprocessed and corrected improved Grey Wolf optimization algorithm includes the following steps: Calculate the fitness value of each gray wolf, and select the three with the lowest fitness values ​​as the alpha wolf X. α ;X β ;X δ ; Generate a random probability factor p; When the random probability factor p is less than 0.5, the position of the gray wolf is updated using the following formula; The parameter d is a constant with values ​​in the range [0,2], and u and v follow normal distributions u~N(0,σ) and v~M(0,σ), respectively. in, This is the globally optimal solution. Let b represent the position vector of the gray wolf, b be the constant coefficient of the spiral equation, and l be a random number between [-1, 1]. When the random probability factor p is greater than or equal to 0.5, the position of the gray wolf is updated using the following formula; Where n1 and n2 represent vectors with values ​​between [0,1], m = 2 - 2t / T, t represents the current iteration number, T represents the maximum iteration number, and m decreases linearly from 2 to 0 as the iteration number increases; Among them, a new position update formula is designed by incorporating dynamic weights: Compare the current positions of the Grey Wolves and The fitness value is retained as the position variable after this update. Determine if the maximum number of iterations has been reached. If it has, output the result. Otherwise, return to the gray wolf coding matrix in the preprocessed and improved gray wolf optimization algorithm.

7. A task decomposition and optimization device for edge-side computing power networks, characterized in that: The apparatus used in the task decomposition and optimization method for edge-side computing networks as described in any one of claims 1-6 includes: The acquisition module is used to acquire the decomposition parameters required for task decomposition in edge computing power network scenarios. The preprocessing and correction module is used to preprocess and correct the gray wolf encoding matrix in the improved gray wolf optimization algorithm according to the decomposition parameters and task decomposition constraints, so as to obtain the gray wolf encoding matrix that meets the task decomposition constraints; the improved gray wolf optimization algorithm is a gray wolf optimization algorithm that incorporates the whale algorithm and the Levy flight algorithm. The decomposition module is used to decompose tasks using an improved gray wolf optimization algorithm to obtain a set of task decomposition strategies. The latency and energy consumption calculation module is used to calculate the latency and energy consumption of local and nearby idle terminal devices and edge servers for each strategy in the task decomposition strategy set. The total cost calculation module is used to calculate the total cost of task decomposition based on the latency and energy consumption of each strategy in the task decomposition strategy set. The output module is used to output the subset of tasks that minimizes the total cost of task decomposition as the optimal task decomposition strategy.

8. A task decomposition and optimization device for edge-side computing power networks, characterized in that: It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method described in any one of claims 1-6.

Images