A computing offloading method based on improved coral reef algorithm in cloud-edge collaborative computing

Abstract

The application discloses a cloud edge collaborative computing computing offloading method based on an improved coral reef algorithm, which comprises the following steps: firstly, modeling a task set and variables, an edge server set and variables, and a cloud server set and variables; then, determining a task offloading variable and a corresponding offloading matrix model of the task and the server; then, determining a system total time delay and a load balancing index model; then, converting a multi-objective optimization problem of the system total time delay and the load balancing index into a single-objective optimization problem by using a linear weighting method, and determining an optimization target of the task offloading; finally, solving an optimal task offloading strategy based on minimized system time delay and load balancing level by using the improved coral reef algorithm. The method avoids the system load imbalance phenomenon, can reduce the multi-objective optimization problem of the task offloading to the single-objective optimization problem, and solves the iteration stagnation and local optimal problem of the standard coral reef algorithm by using the coral reef optimization algorithm of the hybrid genetic algorithm, so that the global optimal solution is obtained.

Description

Technical Field

[0001] This invention belongs to the field of cloud computing and edge computing technology, and in particular relates to a computation offloading method based on an improved coral reef algorithm in cloud-edge collaborative computing. Background Technology

[0002] In recent years, the Internet of Things (IoT) has been one of the major revolutions in information and communication technologies. IoT extends internet connectivity from traditional smart devices to sectors like transportation and industrial manufacturing. By 2023, global IoT connections will account for half of all connected devices worldwide, with a dramatic increase in the number and scale of network services and applications. Concentrating computing and storage resources in large data centers will be insufficient to meet the data and communication requirements of IoT. Due to the long distances between cloud computing centers and IoT devices, massive data transmission leads to severe latency issues, resulting in low Quality of Service (QoS). Therefore, cloud computing cannot meet the demands of low-latency, high-QoS applications, and issues such as user privacy and data security cannot be guaranteed. This has led to the development of the edge computing paradigm, which deploys cloud services at the network edge to overcome the network congestion and transmission latency problems of traditional cloud computing.

[0003] Due to the rapid development of the Internet of Things (IoT) and 5G communication in recent years, and the complementary advantages of cloud computing and edge computing, researchers have been conducting research in the field of cloud-edge collaborative design. Offloading tasks with lower latency requirements to edge servers and tasks requiring large computing resources to cloud servers is the basic collaborative approach of cloud-edge computing. Current research mostly focuses on latency or system load; studies combining latency and load balancing have not yielded ideal results. Therefore, how to establish a cloud-edge collaborative system model and obtain a reasonable task scheduling strategy while balancing the overall system load is an urgent problem to be solved in the field of cloud-edge collaborative computing. Summary of the Invention

[0004] The purpose of this invention is to address the problem that existing task offloading methods in cloud-edge collaborative computing models do not comprehensively consider the load imbalance caused by the resource gap between cloud servers and edge servers, as well as the task scheduling problem of multiple tasks and multiple servers. This invention provides a task offloading method based on an improved Coral Reef algorithm. This method comprehensively considers task completion time and the overall system load balancing level, establishes a task scheduling model under the cloud-edge collaborative system, and uses the improved Coral Reef algorithm to solve the task offloading strategy, ultimately obtaining the optimal offloading strategy and the optimal system load balancing.

[0005] To achieve the objectives of this invention, a computational offloading method based on an improved coral reef algorithm in cloud-edge collaborative computing is disclosed, specifically including the following steps:

[0006] Step 1: Model the task set and task variables;

[0007] Step 2: Model the edge server set and edge server variables, and the cloud server set and cloud server variables;

[0008] Step 3: Determine the task unloading variables and the corresponding unloading matrix model between tasks and servers;

[0009] Step 4: Model the task completion latency and server load indicators, and thereby determine the system total latency and load balancing indicator models, laying the model foundation for the establishment of subsequent optimization objectives;

[0010] Step 5: Transform the multi-objective optimization problem of total system latency and load balancing metrics into a single-objective optimization problem using a linear weighting method to determine the optimization objective of task offloading;

[0011] Step 6: Based on the task offloading strategy of minimizing system latency and load balancing level, the number of parents in the Coral Reef algorithm during the genetic crossover process is increased by using a three-parent hybridization method, thereby increasing the search range of the solution; then, the solution of the Coral Reef algorithm is mutated by using a Gaussian random distribution function to improve the neighborhood search capability of the algorithm in the later stage.

[0012] Further, step 1 specifically involves: within a task scheduling cycle, the task queue contains a set of n independent tasks to be processed: Task = {T1, T2, T3, ... T}. n Each independent task in the set of independent tasks is represented by a quintuple T. i ={num i ,data i ,cal i mem i ,sto i} indicates that each element in it is a task variable, specifically the task number num. i The amount of data transmitted by the task i The CPU computing power required for the task i The memory resources required by the task (mem) i The task requires hard disk resources. i .

[0013] Further, step 2 specifically involves: the entire cloud-edge collaborative computing network comprises an edge server set consisting of q edge computing nodes and a cloud server set consisting of p cloud computing nodes; each edge node in the edge server set is represented by a quintuple E. j ={edge j etra j ,ecal j ,emem j ,esto j} represents the edge server variable, where each element is an edge server variable. j Number the edge nodes, etra j The ecal represents the data transmission rate between edge nodes and terminal devices. j Emem represents the computing power of the edge nodes. j esto represents the memory resources of edge nodes. j The disk resources of the edge node are represented; each cloud computing node in the cloud server set is also represented by a quintuple C. k ={cloud k ,ctra k ,ccal k ,cmem k ,csto k Each element in this variable is a cloud server variable, cloud. k Indicates the cloud node number, ctra k ccal represents the data transfer rate between the terminal and the cloud node. k cmem represents the computing power of a cloud node. k csto represents the memory resources of a cloud node. k This represents the hard disk resources of the cloud node.

[0014] Furthermore, step 3 specifically involves: if each task can be computed on an edge node or a cloud node, then each task has m = p + q allocation options; define the task offloading variable as x. i =[a i,1 ,a i,2 ,...,a i,m ] T Combining the unloading variables for each task, we obtain the corresponding unloading matrix between the task and the server node as A = {a i,j}, j∈{1,2,...,q} corresponds to edge nodes, j∈{q+1,q+2,...,q+p} corresponds to the corresponding cloud nodes, a i,j =1 indicates that task i is assigned to node j for execution.

[0015] Furthermore, step 4 specifically includes:

[0016] Step 4-1: If task i is executed on edge node j, then the completion delay of task i, which is offloaded to edge server j for processing, is... for

[0017]

[0018] Step 4-2, λ1, λ2, λ3 are weights, and the load R on edge server j is... j for

[0019]

[0020] Step 4-3: The load balancing metric LBe (Load Balance of edge servers) for the edge server cluster is...

[0021]

[0022] Step 4-4: Task i is executed on cloud node k. The distance between the cloud computing center and the edge device is L, and the propagation speed in the optical fiber is R. What is the completion delay of task i, which is offloaded to cloud node k for processing? for

[0023]

[0024] Steps 4-5: Load balancing on cloud server k k for

[0025]

[0026] Steps 4-6: The load balancing metric LBc (Load Balance of cloud servers) for the cloud server cluster is...

[0027]

[0028] Steps 4-7: For the described cloud-edge collaborative computing task offloading model, the task completion delay t of task i under cloud-edge collaborative computing is... i for

[0029]

[0030] In the formula This represents the value of a when j∈{1,2,...,q} in step 3. i,j The value, Let a represent the value of a when j∈{q+1,q+2,...,q+p} in step 3. i,j value

[0031] The total latency (makespan) for all tasks in the entire task queue is

[0032]

[0033] Steps 4-9: The overall load balancing index (LBEC) of the system is...

[0034] LBEC = m1LBe + m2LBc, m1 + m2 = 1

[0035] m1 and m2 represent weights.

[0036] Further, step 5 specifically involves satisfying the following constraints during the task unloading process: (1) a task can only be assigned to a server node for execution; (2) there is a correspondence between tasks and nodes; (3) the sum of α and β is 1, and the sum of m1 and m2 is 1; (4) the resources required by the task are not greater than the total resources of the corresponding computing node to which it is assigned. Under the condition of satisfying the above constraints, find an unloading strategy among all unloading strategies that satisfies the optimization objective P1: minE(x)=αmakespan+βLBEC, where α and β represent weights, and the formula is as follows.

[0037] P1:minE(x)=αmakespan+βLBEC

[0038]

[0039] C2:a i,j ={0,1}

[0040] C3: α+β=1,m1+m2=1

[0041] C4:mem ij ≤mem j CPU ij ≤cpu j ,sto ij ≤sto j

[0042] This established the optimization objective and constraints.

[0043] Furthermore, step 6 specifically includes:

[0044] Step 6-1: Initialize the N×M grid Λ, which represents the size of the coral reef, the relevant parameters of the improved coral reef algorithm, and the initial population S0. The algorithm parameters include the asexual reproduction probability P. a The probability of selection P b Probability of death P d and the constraint algebra k;

[0045] Step 6-2: Calculate the coral reef population S generated after the t-th iteration using the fitness value formula f(x) = αmakespan + βLBEC. t Fitness value for each coral polyp;

[0046] Step 6-3, according to P b Three parents are selected. Parents already selected within k iterations will not be selected again. Coral larvae, i.e., new candidate solutions, are generated through multiple crossover operations of the three parents.

[0047] Step 6-4, for coral polyps Gaussian mutations were performed to obtain new coral larvae. A random function that follows a Gaussian distribution with mean μ and variance σ. 2 Gaussian probability density function f G as follows

[0048]

[0049] Let μ = 0, σ = σ max -(σ max -σ min )·t / t max σ decreases as the number of iterations increases;

[0050] Step 6-5: Discard the solutions from Step 6-3 that have not been selected as parents in k iterations;

[0051] Step 6-6: Settle the coral larvae generated in Steps 6-3 and 6-4 on the coral reef; if the target coral square is empty, settle in that square; if the square is occupied, the new candidate solution will compete with the existing solution, the solution with the lower fitness value will be left in the square, and the old solution will be discarded; if the new candidate solution fails to compete, try to find other squares, and if it still has not settled after k attempts, discard the solution;

[0052] Steps 6-7: Sort the coral polyps that have settled on the reef according to their fitness values, and then sort them according to the probability of asexual reproduction P. a Select individual coral polyps for replication, and then place the resulting larvae according to steps 6-6 to obtain population S. t ';

[0053] Step 6-8: Determine if the loop termination condition is met. If it is met, output the optimal solution and proceed to step 6-10; otherwise, proceed to step 6-9, the elimination stage.

[0054] Steps 6-9: The population S obtained in step 6-6 t The corals in the image are sorted by fitness value, and then based on mortality probability P. d Eliminating highly adaptable coral polyps results in a new population S t+1 Return to step 6-3 to proceed to the next iteration;

[0055] Steps 6-10: Based on the task allocation strategy obtained in Step 6-8, schedule each task to be executed on the corresponding server.

[0056] Compared with the prior art, the significant advancements of this invention are: 1) This invention models the load balancing levels of edge servers and cloud servers separately, avoiding system load imbalance caused by excessive resource differences between cloud servers and edge servers; 2) This invention reduces the multi-objective optimization problem of task offloading in cloud-edge collaborative computing scenarios to a single-objective optimization problem; 3) This invention designs a coral reef optimization algorithm with hybrid genetic algorithm to solve the task offloading strategy, avoiding the iteration stagnation and local optimum problems caused by insufficient population diversity in the later stages of local iteration of the standard coral reef algorithm, thereby obtaining the global optimum solution.

[0057] To more clearly illustrate the functional characteristics and structural parameters of the present invention, further explanation is provided below in conjunction with the accompanying drawings and specific embodiments. Attached Figure Description

[0058] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0059] Figure 1 This is a model diagram of a cloud-edge collaborative computing system based on an improved coral reef algorithm for computation offloading in cloud-edge collaborative computing.

[0060] Figure 2 A flowchart of the Coral Reef Algorithm, which is a computational offloading method based on the improved Coral Reef Algorithm in cloud-edge collaborative computing;

[0061] Figure 3 This is a schematic diagram of a three-parent hybridization method for computational offloading based on an improved coral reef algorithm in cloud-edge collaborative computing. Detailed Implementation

[0062] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] The task scheduling method based on the improved coral reef algorithm proposed in this invention, such as... Figure 1As shown, in a cloud-edge collaborative computing system, edge devices generate multiple tasks, and edge servers and cloud servers can provide computing services to users to process these tasks. Tasks can be processed on either the cloud server or the edge server, but only one type of server can be selected. Local devices are directly connected to the base station wirelessly, while the base station and server are connected via fiber optic cable. This deployment method minimizes the number of hops for task data transmission to the server, thus reducing communication latency.

[0064] Step 1: Within a task scheduling cycle, the task queue contains a set of n independent tasks to be processed: Task = {T1, T2, T3, ..., T}. n Each independent task in the set of independent tasks is represented by a quintuple T. i ={num i ,data i ,cal i mem i ,sto i} indicates that each element in it is a task variable, specifically the task number num. i The amount of data transmitted by the task i The CPU computing power required for the task i The memory resources required by the task (mem) i The task requires hard disk resources. i .

[0065] Step 2: The entire cloud-edge collaborative computing network consists of an edge server set composed of q edge computing nodes and a cloud server set composed of p cloud computing nodes; each edge node in the edge server set is represented by a quintuple E. j ={edge j etra j ,ecal j ,emem j ,esto j} represents the edge server variable, where each element is an edge server variable. j Number the edge nodes, etra j The ecal represents the data transmission rate between edge nodes and terminal devices. j Emem represents the computing power of the edge nodes. j esto represents the memory resources of edge nodes. j The disk resources of the edge node are represented; each cloud computing node in the cloud server set is also represented by a quintuple C. k ={cloud k ,ctra k ,ccal k,cmem k ,csto k Each element in this variable is a cloud server variable, cloud. k Indicates the cloud node number, ctra k ccal represents the data transfer rate between the terminal and the cloud node. k cmem represents the computing power of a cloud node. k csto represents the memory resources of a cloud node. k This represents the hard disk resources of the cloud node.

[0066] Step 3: Each task can be computed on either an edge node or a cloud node, so each task has m = p + q allocation options; define the task offloading variable as x. i =[a i,1 ,a i,2 ,...,a i,m ] T Combining the unloading variables for each task, we obtain the corresponding unloading matrix between the task and the server node as A = {a i,j}, j∈{1,2,...,q} corresponds to edge nodes, j∈{q+1,q+2,...,q+p} corresponds to the corresponding cloud nodes, a i,j =1 indicates that task i is assigned to node j for execution.

[0067] Step 4: Model system latency and load balancing metrics, specifically including the following steps:

[0068] Step 4-1: If task i is executed on edge node j, then the completion delay of task i, which is offloaded to edge server j for processing, is... for

[0069]

[0070] Step 4-2, λ1, λ2, λ3 are weights, and the load R on edge server j is... j for

[0071]

[0072] Step 4-3: The load balancing metric LBe for the edge server cluster is...

[0073]

[0074] Step 4-4: Task i is executed on cloud node k. The distance between the cloud computing center and the edge device is L, and the propagation speed in the optical fiber is R. What is the completion delay of task i, which is offloaded to cloud node k for processing? for

[0075]

[0076] Steps 4-5: Load balancing on cloud server k k for

[0077]

[0078] Steps 4-6: The load balancing metric LBc of the cloud server cluster is...

[0079]

[0080] Steps 4-7: For the described cloud-edge collaborative computing task offloading model, the task completion delay t of task i under cloud-edge collaborative computing is... i for

[0081]

[0082] In the formula This represents the value of a when j∈{1,2,...,q} in step 3. i,j The value, Let a represent the value of a when j∈{q+1,q+2,...,q+p} in step 3. i,j The value;

[0083] The total latency (makespan) for all tasks in the entire task queue is

[0084]

[0085] Steps 4-9: The overall load balancing index (LBEC) of the system is...

[0086] LBEC=m1LBe+m2LBc,m1+m2=1(9)

[0087] Step 5, the optimization objectives for unloading the modeling task, specifically include the following steps:

[0088] During the task unloading process, the following constraints must be met: (1) a task can only be assigned to a server node for execution; (2) there is a correspondence between tasks and nodes; (3) the sum of α and β is 1, and the sum of m1 and m2 is 1; (4) the resources required by a task are not greater than the total resources of the corresponding computing node to which it is assigned. Under the above constraints, find an unloading strategy among all unloading strategies that satisfies the optimization objective P1: minE(x)=αmakespan+βLBEC, where α and β represent weights, and the formula is as follows.

[0089] P1:minE(x)=αmakespan+βLBEC (10)

[0090]

[0091] C2:a i,j ={0,1}

[0092] C3: α+β=1,m1+m2=1

[0093] C4:mem ij ≤mem j CPU ij ≤cpu j ,sto ij ≤sto j

[0094] like Figure 2 As shown, step 6, the task offloading strategy based on minimizing system latency and load balancing, specifically includes the following steps:

[0095] Step 6-1: Initialize the N×M grid Λ, which represents the size of the coral reef, the relevant parameters of the improved coral reef algorithm, and the initial population S0. The algorithm parameters include the asexual reproduction probability P. a The probability of selection P b Probability of death P d and the constraint algebra k;

[0096] Step 6-2: For the coral reef population S generated after the t-th iteration... t The fitness value of each coral polyp is calculated according to formula (10);

[0097] like Figure 3 As shown, step 6-3, according to P b Select three parents (parents already selected within k steps will not be selected again), and generate coral larvae (new candidate solutions) through multiple crossover operations of the three parents;

[0098] Step 6-4, for coral polyps Gaussian mutations were performed to obtain new coral larvae. A random function that follows a Gaussian distribution with mean μ and variance σ. 2 Gaussian probability density function f G ()as follows

[0099]

[0100] Let μ = 0, σ = σ max -(σ max -σ min )·t / t max σ decreases as the number of iterations increases.

[0101] Step 6-5: Discard the solutions from Step 6-3 that have not been selected as parents within k steps;

[0102] Step 6-6: Settle the coral larvae generated in Steps 6-3 and 6-4 on the coral reef; if the target coral square is empty, settle in that square; if the square is occupied, the new candidate solution will compete with the existing solution, the solution with the lower fitness value will be left in the square, and the old solution will be discarded; if the new candidate solution fails to compete, try to find other squares, and if it still has not settled after k attempts, discard the solution;

[0103] Steps 6-7: Sort the coral polyps that have settled on the reef according to their fitness values, and then sort them according to the probability of asexual reproduction P. a Select individual coral polyps for replication, and then place the resulting larvae according to steps 6-6 to obtain population S. t ';

[0104] Step 6-8: Determine if the loop termination condition is met. If it is met, output the optimal solution and proceed to step 6-10; otherwise, proceed to step 6-9, the elimination stage.

[0105] Steps 6-9: The population S obtained in step 6-6 t The corals in the image are sorted by fitness value, and then ranked according to mortality probability P. d Eliminating highly adaptable coral polyps results in a new population S t+1 Return to step 6-3 to proceed to the next iteration;

[0106] Steps 6-10: Based on the task allocation strategy obtained in Step 6-8, schedule each task to be executed on the corresponding server.

[0107] Unless otherwise specified, the raw materials and equipment used in this invention are all commonly used in the field; unless otherwise specified, the methods used in this invention are all conventional methods in the field.

[0108] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0109] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A computational offloading method based on an improved coral reef algorithm in cloud-edge collaborative computing, characterized in that, Specifically, the following steps are included: Step 1: Model the task set and task variables; Step 2: Model the edge server set and edge server variables, and the cloud server set and cloud server variables; Step 3: Determine the task unloading variables and the corresponding unloading matrix model between tasks and servers; Step 4: Model the task completion latency and server load indicators, and thereby determine the system total latency and load balancing indicator models, laying the model foundation for the establishment of subsequent optimization objectives; Step 5: Transform the multi-objective optimization problem of total system latency and load balancing metrics into a single-objective optimization problem using a linear weighting method to determine the optimization objective of task offloading; Step 6: Based on the task offloading strategy of minimizing system latency and load balancing level, the number of parents in the Coral Reef algorithm during the genetic crossover process is increased by using a three-parent hybridization method, thereby increasing the search range of the solution; then, the solution of the Coral Reef algorithm is mutated by using a Gaussian random distribution function to improve the neighborhood search capability of the algorithm in the later stage.

2. The computational offloading method based on the improved coral reef algorithm in cloud-edge collaborative computing according to claim 1, characterized in that, Step 1 specifically involves: within a task scheduling cycle, the task queue contains... A set of independent tasks to be processed Each independent task in the set of independent tasks consists of a quintuple. This indicates that each element is a task variable, specifically a task number. Task data transmission volume CPU computing power required for the task The memory resources required by the task The hard disk resources required for the task .

3. The computational offloading method based on the improved coral reef algorithm in cloud-edge collaborative computing according to claim 2, characterized in that, Step 2 specifically involves: Within the entire cloud-edge collaborative computing network... An edge server set consisting of edge computing nodes and A cloud server set consisting of cloud computing nodes; each edge node in the edge server set consists of a quintuple. This indicates that each element is an edge server variable. Number the edge nodes. This indicates the data transmission rate between the edge node and the terminal device. This represents the computing power of the edge nodes. This represents the memory resources of the edge nodes. The disk resources of the edge node are represented; each cloud computing node in the cloud server set is also represented by a 5-tuple. Each element in this context is a cloud server variable. Indicates the cloud node number. This indicates the data transmission rate between the terminal and the cloud node. This indicates the computing power of a cloud node. This represents the memory resources of the cloud node. This represents the hard disk resources of the cloud node.

4. The computational offloading method based on the improved coral reef algorithm in cloud-edge collaborative computing according to claim 3, characterized in that, Step 3 specifically involves: Since each task can be computed on either an edge node or a cloud node, each task has m = p + q allocation options; the task unloading variable is defined as... By combining the unloading variables for each task, we obtain the corresponding unloading matrix between the task and the server node. , Corresponding edge nodes, Corresponding to the corresponding cloud nodes, This indicates that task i is assigned to node j for execution.

5. A computational offloading method based on an improved coral reef algorithm in cloud-edge collaborative computing according to claim 4, characterized in that, Step 4 specifically includes: Step 4-1: If task i is executed on edge node j, then the completion delay of task i, which is offloaded to edge server j for processing, is... for Step 4-2 The load on edge server j is the weight. for Step 4-3: The load balancing metric LBe for the edge server cluster is... Step 4-4: Task i is executed on cloud node k. The distance between the cloud computing center and the edge device is... The propagation speed in the optical fiber is The completion delay of task i, which is unloaded to cloud node k for processing, is... for Steps 4-5: Load balancing on cloud server k for Steps 4-6: The load balancing metric LBc of the cloud server cluster is... Steps 4-7: For the described cloud-edge collaborative computing task offloading model, the task completion delay of task i under cloud-edge collaborative computing. for In the formula In step 3 Time corresponding The value, In step 3 Time corresponding The value; Total latency of all tasks in the entire task queue for Steps 4-9: Overall load balancing metrics of the system for , Indicates the weight.

6. A computational offloading method based on an improved coral reef algorithm in cloud-edge collaborative computing according to claim 5, characterized in that, Step 5 specifically involves: during the task unloading process, the following restrictions must be met: a task can only be assigned to one server node for execution; and the task-node correspondence must be established. and The sum of is 1. and The sum of all resources required by the task is 1; the total resources required by each task are no greater than the total resources of the corresponding computing nodes allocated to it; under the above constraints, find one unloading strategy among all unloading strategies that satisfies the optimization objective. , and The weight is represented by the following formula: This established the optimization objective and constraints.

7. A computational offloading method based on an improved coral reef algorithm in cloud-edge collaborative computing according to claim 6, characterized in that, Step 6 specifically includes: Step 6-1, Initialization Grid That is, the size of the coral reef, the relevant parameters of the improved coral reef algorithm, and the initial population. The algorithm parameters include the probability of asexual reproduction. Choice probability Probability of death and the constraint algebra k; Step 6-2: Use the fitness value formula Calculate the coral reef population generated after the t-th iteration. Fitness value for each coral polyp; Step 6-3, according to Three parents are selected. Parents already selected within k iterations will not be selected again. Coral larvae, i.e., new candidate solutions, are generated through multiple crossover operations of the three parents. Step 6-4, for coral polyps Gaussian mutations were performed to obtain new coral larvae. , For a random function that follows a Gaussian distribution, the mean is... The variance is Gaussian probability density function as follows Pick , , It decreases as the number of iterations increases; Step 6-5: Discard the solutions from Step 6-3 that have not been selected as parents in k iterations; Step 6-6: Settle the coral larvae generated in Steps 6-3 and 6-4 on the coral reef; if the target coral square is empty, settle in that square; if the square is occupied, the new candidate solution will compete with the existing solution, the solution with the lower fitness value will remain in the square, and the old solution will be discarded; if the new candidate solution fails to compete, try to find other squares, and so on. If the solution still fails to settle after the first attempt, then discard the solution. Steps 6-7: Sort the coral polyps that have settled on the reef according to their fitness values ​​and their asexual reproduction probability. Select individual coral polyps for replication, and then place the resulting larvae according to steps 6-6 to obtain a population. ; Step 6-8: Determine if the loop termination condition is met. If it is met, output the optimal solution and proceed to step 6-10; otherwise, proceed to step 6-9, the elimination stage. Steps 6-9: The population obtained in step 6-6 The coral polyps were sorted by fitness value, and then further sorted by mortality probability. New populations are formed by eliminating highly adaptable coral polyps. Return to step 6-3 to proceed to the next iteration; Steps 6-10: Based on the task allocation strategy obtained in Step 6-8, schedule each task to be executed on the corresponding server.

Images