CN118585341B - A cloud computing resource scheduling method, system, device and medium based on enhanced Cougar optimization algorithm - Google Patents

Abstract

The invention provides a cloud computing resource scheduling method, a system, equipment and a medium based on an enhanced American lion optimization algorithm; the method comprises the following steps: s1, acquiring a cloud task byte size vector, a computing resource vector, a load cost vector and a resource bandwidth vector which are possessed by a virtual machine; then establishing the following objective function which minimizes the total cost; s2, optimizing an objective function by using an enhanced American lion optimization algorithm to obtain an optimal individual; s3, mapping the optimal individuals obtained in the S2 to a binary scheduling matrix, and obtaining a task scheduling strategy through the scheduling matrix so as to perform task scheduling; the invention can reduce the response time of the system and improve the solution precision so as to improve the performance when dealing with the problem of large-scale task scheduling. The rationality of cloud computing task scheduling is further improved, the resource utilization rate can be maximized, and the problems of resource waste or service performance reduction in cloud computing task scheduling are solved.

Description

A cloud computing resource scheduling method, system, device and medium based on enhanced Puma optimization algorithm

Technical Field

The present invention relates to the technical field of task scheduling methods under cloud computing, and in particular to a cloud computing resource scheduling method, system, device and medium based on an enhanced Puma optimization algorithm.

Background Art

Cloud computing technology has been on the rise since the beginning of the 21st century and has gradually become an important support for modern society. It provides computing resources, storage resources, application programs and other services to users through the Internet through virtualization technology, distributed data storage technology, large-scale data management technology and other technical means, realizing the decoupling of IT and business, so that IT resources can be supplied on demand. Cloud computing optimization scheduling is used to reduce the waste rate of resources and prevent the waste of virtual machine computing resources, storage resources and resource bandwidth resources under large-scale task scheduling.

At present, many task scheduling problems are generally solved using heuristic and meta-heuristic algorithms. Heuristic algorithms are search and optimization algorithms based on experience and intuition, which are usually used to solve complex combinatorial optimization problems, such as simulated annealing, genetic algorithm, and ant colony optimization. However, these algorithms often have difficulty in handling large-scale task scheduling problems, resulting in excessive computer response time. On the other hand, existing meta-heuristic algorithms use random exploration, such as the whale optimization algorithm (WOA), the gray wolf optimization algorithm (GWO), and the differential evolution algorithm (DE) to handle scheduling problems, making them have a higher response speed. However, although the existing meta-heuristic algorithms have their advantages in handling large-scale task scheduling problems, most meta-heuristic algorithms have problems such as long exploration time and insufficient optimization performance when handling large-scale task scheduling problems. Therefore, the existing cloud computing task scheduling methods based on meta-heuristic algorithms need to be further improved.

Summary of the invention

The purpose of the present invention is to solve the technical problems existing in large-scale task scheduling, and in particular to introduce a cloud computing resource scheduling method, system, device and medium based on an enhanced Puma optimization algorithm.

In order to achieve the above-mentioned object of the present invention, the present invention provides a cloud computing resource scheduling method based on an enhanced Puma optimization algorithm, comprising the following steps:

S1. Obtain the cloud task byte size vector and the computing resource vector, load cost vector, and resource bandwidth vector of the virtual machine, and then establish the following objective function that minimizes the total cost:

Fun＝min{ω ₁ F ₁ +ω ₂ F ₂ +ω ₃ F ₃ }(1)

In formula (1), ω ₁ , ω ₂ and ω ₃ represent the weights of functions F ₁ , F ₂ and F ₃ respectively, which are used to adjust the center of gravity of the objective function when optimizing different aspects;

F ₁ represents the time cost function;

F ₂ represents the load cost function;

F ₃ represents the resource bandwidth cost function;

_F1 represents the time cost function and is calculated as follows:

In formula (2), T represents the number of tasks, K represents the number of computing resources (virtual machines), ui _,j represents that the ^ith task is calculated on the ^jth virtual machine, and it is stipulated that each task can only be calculated on one virtual machine, that is, for each i∈[1,T], we get _En,j represents the computing resource vector of the ^jth virtual machine, Et _,i represents the computing resource vector required by the i ^-th task, and _F2 represents the load cost function, which is calculated by the following formula:

In formula (3), T represents the number of tasks, K represents the number of computing resources (virtual machines), ui _,j represents that the ^ith task is calculated on the ^jth virtual machine, and it is stipulated that each task can only be calculated on one virtual machine, that is, for each i∈[1,T], we get Sn _,j represents the load vector of the ^jth virtual machine, St _,i represents the load vector required by the ^ith task, and _F3 represents the resource bandwidth cost function, which is calculated by the following formula:

In formula (4), T represents the number of tasks, K represents the number of computing resources (virtual machines), ui _,j represents that the ^ith task is calculated on the ^jth virtual machine, and it is stipulated that each task can only be calculated on one virtual machine, that is, for each i∈[1,T], we get E _t,i represents the computing resource vector required for the ^ith task, C _n,j represents the resource bandwidth vector of the j ^th virtual machine, C _t,i represents the resource bandwidth vector required for the ^ith task, En _,j represents the computing resource vector of the j ^th virtual machine, and P represents a parameter related to the price.

S2. Use the enhanced Puma optimization algorithm to optimize the objective function and obtain the optimal individual:

S21. Initialize the original population using formula (5) to obtain the initialized original population Sol:

Sol _i = lb + rand·(ub-lb) (5)

In formula (5), Sol _i represents the position information of the i-th individual in the population Sol, ub represents the upper bound vector of the solution space, lb represents the lower bound vector of the solution space, and rand represents the generation of a random number between [0,1];

S22, initially Iter=1, by judging the relationship between Iter and 4, if Iter<4, then execute S23, otherwise jump to execute S28; wherein Iter represents the current number of iterations;

S23, updating individual positions through the exploration phase;

S24, updating individual positions through the development phase;

S25. Merge populations Sol, Sol ^* and Sol ^** , and take the first Nc individuals, where Sol represents the original population after initialization, Sol ^* represents the new population updated through the exploration phase, and Sol ^** represents the new population updated through the development phase;

S26. Calculate the optimal individual fitness of populations Sol ^* and Sol ^** and At the same time, Iter = Iter + 1;

S27, calculate the score of the exploration phase at the Iter iteration and development stage scores

S28. If Iter≤MaxIter, execute S29, otherwise execute step S33; wherein MaxIter represents the maximum number of iterations, and MaxIter≥4;

S29, updating the individual positions through the exploration phase, where the individuals here are the first Nc individuals in step S25;

S30, updating the individual positions through the development phase, where the individuals here are the first Nc individuals in step S25;

S31, yes and Make updates;

S32, Iter=Iter+1, return to S28;

S33, selecting the individual with the smallest objective function value in the population Sol ^* at the last iteration as the scheduling method;

Furthermore, the specific steps of S23 are:

S231. Calculate the new position of the i- ^th individual in the j- ^th dimension by the following formula:

In formula (6), ub represents the upper bound vector of the solution space, lb represents the lower bound vector of the solution space, N is a random number, and N = 2 rand-1, rand is a random number between 0 and 1, Dim represents the dimension of the problem, Dif _a,b is the difference between two different random solutions Sol _a,N and Sol _b,N (Sol _a,N and Sol _b,N are specifically two individuals randomly drawn from the population Sol), and Gap ₁ is calculated by the following formula:

Gap ₁ = Dif _a,b -Dif _c,d (7)

In formula (7), Dif _c,d represents the difference between two random solutions Sol _c,N and Sol d,N that are different from Sol _a,N or Sol _{b,N (} Sol _c,N and Sol _d,N are specifically two individuals randomly drawn from the population Sol), and Gap ₂ is given by the following formula:

Gap ₂ = Dif _c,d - Dif _e,f (8)

In formula (8), Dife _,f is the difference between two random solutions Sol _e,N and Sol f, _N that are different from Sol _a,N , Sol _b,N , Sol _c,N or Sol _d,N (Sol _e,N and Sol _f,N are specifically two individuals randomly drawn from the population Sol); the new population Sol ^* is obtained by the following formula, Sol ^* = {Sol _inew }, and the solution formula for Sol _inew is as follows:

In formula (9), Sol _inew represents the i ^th individual in the new population Sol ^* , represents the new position of the i ^th individual of the new population Sol ^* on the j ^th dimension, Sol _i represents the i ^th individual of the original population after initialization, represents the fitness value of the i ^th individual newly generated by formula (6), and Sol _icost represents the fitness value of the i ^th individual of the initialized original population Sol.

Furthermore, the specific steps of S24 are:

S241. Calculate the new position of the i- ^th individual in the j- ^th dimension by the following formula:

In formula (10), represents the ^ith individual in the new population Sol ^** , Sol _inew represents the ith individual in the population Sol ^* , Sol _r1 and Sol _r2 are two random different solutions in the population Sol ^* , ε and L ^* are parameters set before the optimization starts, δ represents a random number of 0 or 1, Sol _male represents the optimal individual in the population Sol ^* , exp represents an exponential function with e as the base, mean represents the mean value function, rand represents a random number between 0 and 1, Sol _all represents all individuals in the population Sol ^* , Nc represents the number of individuals in the population Sol ^* , randn represents a random number that conforms to the normal distribution in the problem dimension, and Sol _nn is calculated by the following formula:

Sol _nn =(Q ₁ ·P·Sol _inew )+Q ₂ ·(1-P)·Sol _male (11)

In formula (11), P is a random number, Sol _inew represents the i-th individual in the population Sol ^* , Sol _male represents the best individual in the population Sol ^* , and Q ₁ is calculated by the following formula:

In formula (12), Iter represents the current number of iterations, MaxIter represents the total number of iterations, exp represents the exponential function with e as the base, randn represents the random number that conforms to the normal distribution in the problem dimension, _{and Q2} is calculated by the following formula:

Q ₂ = randn·(randn) ² ·cos((2·rand)·randn) (13)

In formula (13), cos represents the cosine function, randn represents a random number that conforms to the normal distribution in the problem dimension, and rand represents a random number between 0 and 1;

Furthermore, the specific steps of S25 are:

S251. Merge the three populations and take the first Nc individuals using the following formula:

Sol _new = [Sol Sol ^* Sol ^** ] (14)

In formula (14), Sol represents the original population after initialization, Sol ^* represents the new population updated through the exploration phase, and Sol ^** represents the new population updated through the development phase, which is then sorted using the following formula:

[～,index]＝sort(Sol _newcost ) (15)

In formula (15), [～, index] represents the representation after Sol _new is sorted in ascending order, sort represents the ascending function, Sol _newcost represents the fitness matrix of the population Sol _new , and index represents the subscript after Sol _newcost is sorted in ascending order. The new population is obtained by the following formula:

Sol _new ＝Sol _new (index(1:Nc)) (16)

In formula (16), index represents the subscript after Sol _newcost is sorted in ascending order, Nc represents the number of populations; Sol _new (index(1:Nc)) means taking the first Nc individuals of the population Sol _new ;

Furthermore, the specific steps of S26 are:

S261. Calculate the optimal fitness individuals of populations Sol ^* and Sol ^** by the following formula: and

Formula (17), It represents the minimum function, Sol ^* .cost and Sol ^** .cost represent the fitness vectors of populations Sol ^* and Sol ^** respectively, and Iter = Iter + 1;

Furthermore, the specific steps of S27 are:

S271, update the score of the exploration phase at the Iter iteration by the following formula and development stage scores

In formula (18), PF ₁ and PF ₂ are parameters set before optimization and are initially set to PF ₁ =PF ₂ =0.5. f1 _Exploration and f2 _Exploration are calculated by the following formulas:

In formula (19), Gap _time , Gap ¹ _time , Gap ² _time and Gap ³ _time are four numbers with initial values of 1, PF ₁ and PF ₂ are parameters set before optimization and are initially set to PF ₁ = PF ₂ = 0.5, and Calculated by the following formula:

In formula (20), is the fitness value of the best individual in the initialized original population Sol, and They represent the optimal fitness values of the population Sol ^* during the first three iterations. f1 _Exploitation and f2 _Exploitation are calculated by the following formulas:

In formula (21), PF ₁ and PF ₂ are parameters set before optimization and are initialized to PF ₁ =PF ₂ =0.5, Gap _time , Gap ¹ _time , Gap ² _time and Gap ³ _time are four numbers with initial values of 1, and Calculated by the following formula:

In formula (22), and They represent the optimal fitness values of the population Sol ^** during the first three iterations; It is the fitness value of the best individual in the initialized original population Sol;

Furthermore, the specific steps of S31 are:

S311, through the following formula and To update:

In formula (23), and Respectively represent the scores of the exploration and development phases during the Iter ^th iteration, and Calculated by the following formula:

In formula (24), is the rounding symbol; and Respectively represent the scores of the exploration and development phases at the Iter ^th iteration, and Calculated by the following formula:

In formula (25), Kc is calculated by the following formula:

In formula (26), {|cost _old -cost _new |} ^exploit and {|cost _old -cost _new |} ^explore represent the absolute value of the difference between the fitness value before and after the update in the development and exploration phases, respectively. min represents the minimum value function. and Calculated by the following equation:

In formula (27), and They represent the optimal fitness before and after the development phase respectively. and They represent the optimal fitness before the exploration phase and the optimal fitness after the exploration phase respectively. represents the number of times the individual is updated through formula (10), represents the number of individuals updated in the exploration phase selected by formula (9), PF ₁ is the parameter set before optimization and is initially set to PF ₁ = 0.5, and Calculated by the following formula:

In formula (28), and are the best fitness values of the current solution in the development phase and exploration phase before the Iter ^th update, respectively. and are the fitness values of the best solutions in the (Iter+1) ^th development phase and exploration phase, respectively. and are the fitness values of the best solutions in the (Iter+2) ^th development phase and exploration phase, respectively. and are the best fitness values of the current solution in the development phase and exploration phase after the Iter ^th update, respectively. and are the fitness values of the best solutions in the (Iter+1) ^th development phase and exploration phase, respectively. and are the fitness values of the best solutions in the development phase and exploration phase after the (Iter+2) ^th selection, respectively. It is the number of individuals updated in the development phase from the Iter ^th iteration to the (Iter+1) ^th iteration in the iteration process. It is the number of individuals updated in the exploration phase from the Iter ^th iteration to the (Iter+1) ^th iteration during the iteration process. is the number of individuals updated in the development phase from the (Iter+1) ^th iteration to the (Iter+2) ^th iteration in the iteration process. is the number of individuals updated through the exploration phase from the (Iter+1) ^th iteration to the (Iter+2) ^th iteration in the iterative process. is the number of individuals updated in the development phase from the (Iter+2) ^th iteration to the (Iter+3) ^th iteration in the iteration process. is the number of individuals selected from the exploration phase from the (Iter+2) ^th iteration to the (Iter+3) ^th iteration in the iterative process after being updated in the exploration phase. PF ₂ is the parameter set before optimization and is initially set to PF ₂ = 0.5. and Calculated by the following formula:

In formula (29), PF ₃ is the parameter set before optimization and is initially set to PF ₃ = 0.3;

S3, mapping the optimal individuals obtained in S2 to a binary scheduling matrix to perform task scheduling;

The binary scheduling matrix is represented as:

Among them, matrix U is the scheduling matrix;

u _i,j indicates that the ^ith task runs on the ^jth computing resource, and it is stipulated that each task can only be calculated on one virtual machine, that is, for each i∈[1,T], we get The first row of the scheduling matrix represents the first task, and ^{the jth} column of the first row has a value of 1 (a row has at least one value 1); this means that the first task runs on the ^jth computing resource.

Furthermore, the initialization of the population Sol also includes: updating the initialized population using a Sinusoidal chaotic map:

In formula (30), α is the control parameter, and the initial α=2.3;

sin represents the sine function;

Sol _i represents the i-th individual in the initialized original population Sol;

Represents the individual after being updated by the Sinusoidal chaotic map.

By adding the Sinusoidal chaotic map initialization method to the algorithm, on the one hand, the diversity of the population is improved, thereby reducing the average total cost of all strategies in the initial task scheduling strategy. On the other hand, the search speed for the optimal solution is accelerated, thereby improving the efficiency of obtaining the task scheduling strategy.

Furthermore, in the development stage, the individual positions are updated and an intermediate point sample learning strategy is introduced to improve the development stage, which is specifically the following formula:

In formula (31), Represents the new individual updated in the original development stage; the specific expression is formula (10) in step S241.

Sol _mid represents the individual generated by the midpoint sample learning strategy and is represented by the following formula:

In formula (32), represents the R _mth individual in the updated population Sol ^** after it is sorted in ascending order by its fitness value, Temp represents the updated population Sol ^** after it is sorted in ascending order by its fitness value, and R _m is calculated by the following formula:

In formula (33), i represents the i-th individual, represents the upward rounding function, and rand represents a random number between the interval [0,1];

Since the median is not affected by extreme values and can also judge the quality of a batch of data, the middle point carries relevant information about the optimal solution during the task scheduling process. By adding the middle point sample learning strategy during the algorithm development phase, the convergence speed of finding the optimal solution in the task scheduling process is accelerated, thereby reducing the response time of cloud computing task scheduling.

Furthermore, the optimal individual is mapped to a binary scheduling matrix by using a random S-T transfer function to convert the optimal individual into a binary matrix for task scheduling, which is specifically expressed as the following formula:

In formula (34), abs represents the absolute value function, tan represents the tangent function, Sol _male represents the optimal individual in the iterative process, and Matrix _new represents the matrix after transformation by the random ST transfer function;

A single or traditional transfer function will use a lot of conversion time, resulting in a waste of computing resources during task scheduling. Using a random S-T transfer function, the time for converting the optimal solution to the optimal scheduling method during task scheduling can be reduced through the above steps, thereby reducing the response time of the computer system during scheduling.

The present invention also proposes a cloud task scheduling system based on a multi-strategy Cougar optimization algorithm.

Furthermore, the system comprises:

Data collection module: obtains the cloud task byte size vector and the computing resource vector, load cost vector and resource bandwidth vector of the virtual machine;

Task scheduling module: The objective function of minimizing the total cost is optimized by using the enhanced Puma optimization algorithm to obtain the optimal individual. The optimal individual is then mapped to a binary scheduling matrix, and the task scheduling strategy is obtained through the scheduling matrix to perform task scheduling.

The binary scheduling matrix is represented as:

Among them, matrix U is the scheduling matrix;

u _i,j indicates that the ^ith task runs on the ^jth computing resource, and it is stipulated that each task can only be calculated on one virtual machine, that is, for each i∈[1,T], we get The first row of the scheduling matrix represents the first task, and ^{the jth} column of the first row has a value of 1 (a row has at least one value 1); this means that the first task runs on the ^jth computing resource.

The objective function of minimizing the total cost is:

Fun＝min{ω ₁ F ₁ +ω ₂ F ₂ +ω ₃ F ₃ }(1)

In formula (1), ω ₁ , ω ₂ and ω ₃ represent the weights of functions F ₁ , F ₂ and F ₃ respectively, which are used to adjust the center of gravity of the objective function when optimizing different aspects;

F ₁ represents the time cost function;

F ₂ represents the load cost function;

F ₃ represents the resource bandwidth cost function;

The objective function of minimizing the total cost is optimized by the enhanced Cougar optimization algorithm to obtain the optimal individual, including:

S21. Initialize the original population using formula (5) to obtain the initialized original population Sol:

Sol _i = lb + rand·(ub-lb) (5)

In formula (5), Sol _i represents the location information of the i-th individual in the population Sol;

ub represents the upper bound vector of the solution space;

lb represents the lower bound vector of the solution space;

rand means generating a random number between [0,1];

S22, initially Iter=1, by judging the relationship between Iter and 4, if Iter<4, then execute S23, otherwise jump to execute S28; wherein Iter represents the current number of iterations;

S23, updating individual positions through the exploration phase;

S24, updating individual positions through the development phase;

S25. Merge the populations Sol, Sol ^* and Sol ^** , and take the first Nc individuals; where Sol represents the original population after initialization, Sol ^* represents the new population updated through the exploration phase, and Sol ^** represents the new population updated through the development phase;

S26. Calculate the optimal individual fitness of populations Sol ^* and Sol ^** and At the same time, Iter = Iter + 1;

S27, calculate the score of the exploration phase at the Iter iteration and development stage scores

S28. If Iter≤MaxIter, execute S29, otherwise execute step S33; wherein MaxIter represents the maximum number of iterations, and MaxIter≥4;

S29, updating the individual positions through the exploration phase, where the individuals here are the first Nc individuals in step S25;

S30, updating the individual positions through the development phase, where the individuals here are the first Nc individuals in step S25;

S31, yes and Make updates;

S32, Iter=Iter+1, return to S28;

S33. Select the individual with the smallest objective function value in the population Sol ^* at the last iteration as the scheduling method.

Furthermore, the initialization of the population Sol also includes: updating the initialized population using a Sinusoidal chaotic map:

In formula (30), α is the control parameter, and the initial α=2.3;

sin represents the sine function;

Sol _i represents the i-th individual in the initialized original population Sol;

Represents the individual after being updated by the Sinusoidal chaotic map.

Furthermore, in the development stage, the individual positions are updated and an intermediate point sample learning strategy is introduced to improve the development stage, which is specifically the following formula:

In formula (31), Sol _midnew represents the new individual updated in the development phase after the introduction of the midpoint sample learning strategy;

Represents the new individual updated in the original development stage; the specific expression is formula (10) in step S241.

Sol _mid represents the individual generated by the midpoint sample learning strategy and is represented by the following formula:

In formula (32), Temp represents the updated population Sol ^** after being sorted in ascending order of its fitness value, and _Rm is calculated by the following formula:

In formula (33), i represents the i-th individual, It represents the rounding function, and rand represents a random number between the interval [0,1].

Furthermore, the optimal individual is mapped to a binary scheduling matrix by using a random S-T transfer function to convert the optimal individual into a binary matrix for task scheduling, which is specifically expressed as the following formula:

In formula (34), abs represents the absolute value function, tan represents the tangent function, Sol _male represents the optimal individual in the iterative process, and Matrix _new represents the matrix after transformation by the random ST transfer function.

The present invention also proposes a computer device, wherein when the processor executes the program stored in the memory, a cloud task scheduling method based on the multi-strategy Cougar optimization algorithm is implemented.

The present invention also proposes a readable storage medium storing a program, which, when executed by a processor, implements a cloud task scheduling method based on a multi-strategy Cougar optimization algorithm.

In summary, due to the adoption of the above technical solutions, the task scheduling method proposed in the present invention obtains a scheduling strategy by introducing an advanced optimization algorithm, which significantly reduces the time cost of task scheduling. This optimization not only reduces the response time of the computer system, but also enhances the search range of the understanding space by adding an intermediate point sample learning strategy in the development stage, thereby improving the accuracy of the optimal solution. This method is particularly suitable for large-scale task scheduling problems, and can effectively improve the rationality of cloud computing task scheduling, maximize resource utilization, and effectively reduce resource waste and business performance degradation. At the same time, by optimizing the task execution process and resource allocation, this method can also improve hardware computing efficiency and hardware processing speed, thereby improving system performance as a whole.

Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS

The above and/or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

FIG1 is a schematic diagram of a task scheduling process based on the enhanced Puma optimization algorithm SMRSTPO of the present invention.

FIG. 2 is a comparison diagram of task scheduling based on the enhanced Puma optimization algorithm SMRSTPO of the present invention and other optimization algorithms.

DETAILED DESCRIPTION

Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

With the rise of cloud computing, the scale of tasks has also continued to rise. Existing task scheduling methods often face the problem of falling into suboptimal task scheduling. The fundamental reason is that the algorithm's exploration and utilization performance is weak, and it is impossible to quickly and effectively find the optimal task scheduling method. The selected task scheduling method is inefficient, resulting in problems such as wasting resources when the system performs large-scale task scheduling.

In order to alleviate these problems, this paper proposes a cloud computing resource scheduling method SMRSTPO based on the enhanced Puma optimization algorithm, as shown in Figure 1, which specifically includes the following steps:

S1. Obtain the cloud task (Task) byte size vector and the computing resource vector, load cost vector and resource bandwidth vector of the virtual machine (VM); then establish the following objective function that minimizes the total cost;

S2, using the enhanced Puma optimization algorithm to optimize the objective function and obtain the optimal individual;

S3, mapping the optimal individual (optimal solution) obtained in S2 to a binary scheduling matrix, obtaining the task scheduling strategy through the scheduling matrix, and thus performing task scheduling;

The binary scheduling matrix is represented as:

Among them, matrix U is the scheduling matrix;

u _i,j indicates that the ^ith task runs on the ^jth computing resource, and it is stipulated that each task can only be calculated on one virtual machine, that is, for each i∈[1,T], we get The first row of the scheduling matrix represents the first task. If the value of ^{the jth} column in the first row is 1 (there is at least one value 1 in a row), it means that the first task runs on the ^jth computing resource. For example, if only the value of u _1,2 in the first row of the scheduling matrix U is 1, it means that the first task is assigned to run on the second computing resource. If the values of u _2,2 and u _2,5 in the second row of the scheduling matrix U are 1, it means that the second task is assigned to run on the second computing resource and the fifth computing resource. Therefore, the specific task scheduling strategy can be obtained from the u _i,j value of the scheduling matrix U, thereby realizing task scheduling.

Furthermore, the objective function of minimizing the total cost established in step S1 is as follows:

Fun＝min{ω ₁ F ₁ +ω ₂ F ₂ +ω ₃ F ₃ }(1)

In formula (1), ω ₁ , ω ₂ and ω ₃ represent the weights of functions F ₁ , F ₂ and F ₃ respectively, which are used to adjust the center of gravity of the objective function when optimizing different aspects. F ₁ represents the time cost function, F ₂ represents the load cost function, and F ₃ represents the resource bandwidth cost function.

_F1 represents the time cost function and is calculated as follows:

In formula (2), T represents the number of tasks, K represents the number of computing resources (virtual machines), ui _,j represents that the ^ith task is calculated on the ^jth virtual machine, and it is stipulated that each task can only be calculated on one virtual machine, that is, for each i∈[1,T], we get _En,j represents the computing resource vector of the ^jth virtual machine, Et _,i represents the computing resource vector required by the i ^-th task, and _F2 represents the load cost function, which is calculated by the following formula:

In formula (3), T represents the number of tasks, K represents the number of computing resources (virtual machines), ui _,j represents that the ^ith task is calculated on the ^jth virtual machine, and it is stipulated that each task can only be calculated on one virtual machine, that is, for each i∈[1,T], we get Sn _,j represents the load vector of the ^jth virtual machine, St _,i represents the load vector required by the ^ith task, and _F3 represents the resource bandwidth cost function, which is calculated by the following formula:

In formula (4), T represents the number of tasks, K represents the number of computing resources (virtual machines), ui _,j represents that the ^ith task is calculated on the ^jth virtual machine, and it is stipulated that each task can only be calculated on one virtual machine, that is, for each i∈[1,T], we get E _t,i represents the computing resource vector required for the ^ith task, C _n,j represents the resource bandwidth vector of the j ^th virtual machine, C _t,i represents the resource bandwidth vector required for the ^ith task, En _,j represents the computing resource vector of the j ^th virtual machine, and P represents a parameter related to the price.

Furthermore, the S2 optimizes the cloud task scheduling method using the enhanced Puma optimization algorithm, and the specific steps of obtaining the optimal task scheduling strategy are:

S21. Initialize the original population using formula (5) to obtain the initialized original population Sol:

Sol _i = lb + rand·(ub-lb) (5)

In formula (5), Sol _i represents the location information of the i-th individual in the population Sol;

ub represents the upper bound vector of the solution space;

lb represents the lower bound vector of the solution space;

rand means generating a random number between [0,1];

S22, initially Iter=1, by judging the relationship between Iter and 4, if Iter<4, then execute S23, otherwise jump to execute S28; wherein Iter represents the current number of iterations;

S23, updating individual positions through the exploration phase;

S24, updating individual positions through the development phase;

S25. Merge the populations Sol, Sol ^* and Sol ^** , and take the first Nc individuals; where Sol represents the original population after initialization, Sol ^* represents the new population updated through the exploration phase, and Sol ^** represents the new population updated through the development phase;

S26. Calculate the optimal individual fitness of populations Sol ^* and Sol ^** and At the same time, Iter = Iter + 1;

S27, calculate the score of the exploration phase at the Iter iteration and development stage scores

S28. If Iter≤MaxIter, execute S29, otherwise execute step S33; wherein MaxIter represents the maximum number of iterations, and MaxIter≥4;

S29, updating the individual positions through the exploration phase, where the individuals here are the first Nc individuals in step S25;

S30, updating the individual positions through the development phase, where the individuals here are the first Nc individuals in step S25;

S31, yes and Make updates;

S32, Iter=Iter+1, return to S28;

S33. Select the individual with the smallest objective function value in the population Sol ^* at the last iteration as the scheduling method.

Furthermore, the specific steps of S23 are:

S231. Calculate the new position of the i- ^th individual in the j- ^th dimension by the following formula:

In formula (6), ub represents the upper bound vector of the solution space, lb represents the lower bound vector of the solution space, N is a random number, and N = 2 rand-1, rand is a random number between 0 and 1, Dim represents the dimension of the problem, Dif _a,b is the difference between two different random solutions Sol _a,N and Sol _b,N (Sol _a,N and Sol _b,N are specifically two individuals randomly drawn from the population Sol), and Gap ₁ is calculated by the following formula:

Gap ₁ = Dif _a,b -Dif _c,d (7)

In formula (7), Dif _c,d represents the difference between two random solutions Sol _c,N and Sol d,N that are different from Sol _a,N or Sol _{b,N (} Sol _c,N and Sol _d,N are specifically two individuals randomly drawn from the population Sol), and Gap ₂ is given by the following formula:

Gap ₂ = Dif _c,d - Dif _e,f (8)

In formula (8), Dife _,f is the difference between two random solutions Sol _e,N and Sol f, _N that are different from Sol _a,N , Sol _b,N , Sol _c,N or Sol _d,N (Sol _e,N and Sol _f,N are specifically two individuals randomly drawn from the population Sol); the new population Sol ^* is obtained by the following formula, Sol ^* = {Sol _inew }, and the solution formula for Sol _inew is as follows:

In formula (9), Sol _inew represents the i ^th individual in the new population Sol ^* , represents the new position of the i ^th individual of the new population Sol ^* on the j ^th dimension, Sol _i represents the i ^th individual of the original population after initialization, represents the fitness value of the i ^th individual newly generated by formula (6), and Sol _icost represents the fitness value of the i ^th individual of the initialized original population Sol.

Furthermore, the specific steps of S24 are:

S241. Calculate the new position of the i- ^th individual in the j- ^th dimension by the following formula:

In formula (10), represents the ^ith individual in the new population Sol ^** , Sol _inew represents the ith individual in the population Sol ^* , Sol _r1 and Sol _r2 are two random different solutions in the population Sol ^* , ε and L ^* are parameters set before the optimization starts, δ represents a random number of 0 or 1, Sol _male represents the optimal individual in the population Sol ^* , exp represents an exponential function with e as the base, mean represents the mean value function, rand represents a random number between 0 and 1, Sol _all represents all individuals in the population Sol ^* , Nc represents the number of individuals in the population Sol ^* , randn represents a random number that conforms to the normal distribution in the problem dimension, and Sol _nn is calculated by the following formula:

Sol _nn =(Q ₁ ·P·Sol _inew )+Q ₂ ·(1-P)·Sol _male (11)

In formula (11), P is a random number, Sol _inew represents the i-th individual in the population Sol ^* , Sol _male represents the best individual in the population Sol ^* , and Q ₁ is calculated by the following formula:

In formula (12), Iter represents the current number of iterations, MaxIter represents the total number of iterations, exp represents the exponential function with e as the base, randn represents the random number that conforms to the normal distribution in the problem dimension, _{and Q2} is calculated by the following formula:

Q ₂ = randn·(randn) ² ·cos((2·rand)·randn) (13)

In formula (13), cos represents the cosine function, randn represents a random number that conforms to the normal distribution in the problem dimension, and rand represents a random number between 0 and 1;

Furthermore, the specific steps of S25 are:

S251. Merge the three populations and take the first Nc individuals using the following formula:

Sol _new = [Sol Sol ^* Sol ^** ] (14)

In formula (14), Sol represents the original population after initialization, Sol ^* represents the new population updated through the exploration phase, and Sol ^** represents the new population updated through the development phase, which is then sorted using the following formula:

[～,index]＝sort(Sol _newcost ) (15)

In formula (15), [～, index] represents the representation after Sol _new is sorted in ascending order, sort represents the ascending function, Sol _newcost represents the fitness matrix of the population Sol _new , and index represents the subscript after Sol _newcost is sorted in ascending order. The new population is obtained by the following formula:

Sol _new ＝Sol _new (index(1:Nc)) (16)

In formula (16), index represents the subscript after Sol _newcost is sorted in ascending order, Nc represents the number of populations; Sol _new (index(1:Nc)) means taking the first Nc individuals of the population Sol _new ;

Furthermore, the specific steps of S26 are:

S261. Calculate the optimal fitness individuals of populations Sol ^* and Sol ^** by the following formula: and

Formula (17), It represents the minimum function, Sol ^* .cost and Sol ^** .cost represent the fitness vectors of populations Sol ^* and Sol ^** respectively, and Iter = Iter + 1;

Furthermore, the specific steps of S27 are:

S271. Update the score of the exploration phase at the Iter iteration using the following formula: and development stage scores

In formula (18), PF ₁ and PF ₂ are parameters set before optimization and are initially set to PF ₁ =PF ₂ =0.5. f1 _Exploration and f2 _Exploration are calculated by the following formulas:

In formula (19), Gap _time , Gap ¹ _time , Gap ² _time and Gap ³ _time are four numbers with initial values of 1, PF ₁ and PF ₂ are parameters set before optimization and are initially set to PF ₁ = PF ₂ = 0.5, and Calculated by the following formula:

In formula (20), is the fitness value of the best individual in the initialized original population Sol, and They represent the optimal fitness values of the population Sol ^* during the first three iterations. f1 _Exploitation and f2 _Exploitation are calculated by the following formulas:

In formula (21), PF ₁ and PF ₂ are parameters set before optimization and are initialized to PF ₁ =PF ₂ =0.5, Gap _time , Gap ¹ _time , Gap ² _time and Gap ³ _time are four numbers with initial values of 1, and Calculated by the following formula:

In formula (22), and They represent the optimal fitness values of the population Sol ^** during the first three iterations; It is the fitness value of the best individual in the initialized original population Sol;

Furthermore, the specific steps of S31 are:

S311, through the following formula and To update:

In formula (23), and Respectively represent the scores of the exploration and development phases during the Iter ^th iteration, and Calculated by the following formula:

In formula (24), is the rounding symbol; and Respectively represent the scores of the exploration and development phases at the Iter ^th iteration, and Calculated by the following formula:

In formula (25), Kc is calculated by the following formula:

In formula (26), {|cost _old -cost _new |} ^exploit and {|cost _old -cost _new |} ^explore represent the absolute value of the difference between the fitness value before and after the update in the development and exploration phases, respectively. min represents the minimum value function. and Calculated by the following equation:

In formula (27), and They represent the optimal fitness before and after the development phase respectively. and They represent the optimal fitness before the exploration phase and the optimal fitness after the exploration phase respectively. represents the number of times the individual is updated through formula (10), represents the number of individuals updated in the exploration phase selected by formula (9), PF ₁ is the parameter set before optimization and is initially set to PF ₁ = 0.5, and Calculated by the following formula:

In formula (28), and are the best fitness values of the current solution in the development phase and exploration phase before the Iter ^th update, respectively. and are the fitness values of the best solutions in the (Iter+1) ^th development phase and exploration phase, respectively. and are the fitness values of the best solutions in the (Iter+2) ^th development phase and exploration phase, respectively. and are the best fitness values of the current solution in the development phase and exploration phase after the Iter ^th update, respectively. and are the fitness values of the best solutions in the (Iter+1) ^th development phase and exploration phase, respectively. and are the fitness values of the best solutions in the development phase and exploration phase after the (Iter+2) ^th selection, respectively. It is the number of individuals updated in the development phase from the Iter ^th iteration to the (Iter+1) ^th iteration in the iteration process. It is the number of individuals updated in the exploration phase from the Iter ^th iteration to the (Iter+1) ^th iteration during the iteration process. is the number of individuals updated in the development phase from the (Iter+1) ^th iteration to the (Iter+2) ^th iteration in the iteration process. is the number of individuals updated through the exploration phase from the (Iter+1) ^th iteration to the (Iter+2) ^th iteration in the iterative process. is the number of individuals updated in the development phase from the (Iter+2) ^th iteration to the (Iter+3) ^th iteration in the iteration process. is the number of individuals selected from the exploration phase from the (Iter+2) ^th iteration to the (Iter+3) ^th iteration in the iterative process after being updated in the exploration phase. PF ₂ is the parameter set before optimization and is initially set to PF ₂ = 0.5. and Calculated by the following formula:

In formula (29), PF ₃ is the parameter set before optimization and is initially set to PF ₃ = 0.3;

Furthermore, the Sinusoidal chaotic map is introduced to update the initialized original population, which is expressed as the following formula:

In formula (30), α is the control parameter, the initial α = 2.3, sin represents the sine function, Sol _i represents the i-th individual in the initialized original population Sol, Represents the individual after updating through the Sinusoidal chaotic mapping;

By adding the Sinusoidal chaotic map initialization method to the algorithm, on the one hand, the diversity of the population is improved, thereby reducing the average total cost of all strategies in the initial task scheduling strategy. On the other hand, the search speed for the optimal solution is accelerated, thereby improving the efficiency of obtaining the task scheduling strategy.

Furthermore, the intermediate point sample learning strategy is introduced to improve the development stage, which is specifically the following formula:

In formula (31), represents the new individual updated in the original development stage, and Sol _mid represents the individual generated by the midpoint sample learning strategy and is represented by the following formula:

In formula (32), Temp represents the updated population Sol ^** after being sorted in ascending order of its fitness value, and _Rm is calculated by the following formula:

In formula (33), i represents the i-th individual, represents the upward rounding function, and rand represents a random number between the interval [0,1];

Since the median is not affected by extreme values and can also judge the quality of a batch of data, the middle point carries relevant information about the optimal solution during the task scheduling process. By adding the middle point sample learning strategy during the algorithm development phase, the convergence speed of finding the optimal solution in the task scheduling process is accelerated, thereby reducing the response time of cloud computing task scheduling.

Furthermore, the optimal individual (optimal solution) is mapped to a binary scheduling matrix by using a random S-T transfer function to convert the optimal individual into a binary matrix for task scheduling, which is specifically expressed as the following formula:

In formula (34), abs represents the absolute value function, tan represents the tangent function, Sol _male represents the optimal individual in the iterative process, and Matrix _new represents the matrix after transformation by the random ST transfer function;

In addition to the binary scheduling matrix obtained by adopting the proposed random S-T transfer function, general S-shaped transfer function, T-shaped transfer function, V-shaped transfer function, etc. can also be used. However, a single or traditional transfer function will use a lot of conversion time, resulting in a waste of computing resources during task scheduling. Using a random S-T transfer function, the time for converting the optimal solution into the optimal scheduling method during task scheduling can be reduced through the above steps, thereby reducing the response time of the computer system during scheduling.

The pseudo code of the SMRSTPO algorithm is as follows:

In order to further illustrate the technical effect of SMRSTPO, simulation experiments are conducted. We analyze the performance of the proposed SMRSTPO in solving the cloud resource scheduling problem and compare it with the ant colony optimization algorithm (ACO), particle swarm optimization algorithm (PSO), whale optimization algorithm (WOA), puma optimization algorithm (PO) and improved whale optimization algorithm (IWC). The description of the computing tasks and virtual machine parameters is shown in Table 1. Load cost, time cost (i.e., resource bandwidth cost), price cost and total cost are used as performance analysis indicators. The experimental conditions are set to a population size of 50 and a maximum number of iterations of 100.

We tested the ability of SMRSTPO to handle task scheduling with 100 computing tasks. Figure 2 shows the simulation results. It can be clearly seen from Figure 2(a) that the total cost of SMRSTPO is reduced by more than 20% compared with PSO, WOA, and ACO, which shows that SMRSTPO has strong optimization ability. In addition, compared with the powerful IWC, the total cost of SMRSTPO is also reduced by about 1%, which further shows that SMRSTPO has a more accurate solution. As can be seen from Figure 2(b), it is always better than ACO, PSO, PO, and WOA in price cost, and after the 60th iteration, SMRSTPO is comparable to IWC in price cost. As can be seen from Figure 2(c), although SMRSTPO is slightly inferior to WOA in time optimization, it also has strong optimization performance compared with other algorithms. As can be seen from Figure 2(d), in terms of load performance, SMRSTPO is much lower than ACO, PSO, WOA, and PO, indicating that SMRSTPO has strong search ability.

Table 1. Virtual machine and computing task parameters

parameter Virtual Machine Scope Scope of Tasks Computational resource vector En [200,500] [10,50] Load cost vector Sn [100,500] [50,100] Resource bandwidth vector Cn [100,250] [20,50]

Claims (8)

1. The cloud computing resource scheduling method based on the enhanced American lion optimization algorithm is characterized by comprising the following steps of:

s1, acquiring a cloud task byte size vector, a computing resource vector, a load cost vector and a resource bandwidth vector of a virtual machine, and then establishing the following objective function for minimizing total cost:

Fun＝min{ω₁F₁+ω₂F₂+ω₃F₃}(1)

In formula (1), ω ₁,ω₂ and ω ₃ represent weights of the functions F ₁,F₂ and F ₃, respectively;

F ₁ represents a time cost function;

f ₂ represents a load cost function;

F ₃ represents a resource bandwidth cost function;

S2, optimizing an objective function by using an enhanced American lion optimization algorithm to obtain an optimal individual:

S21, initializing an original population through a formula (5) to obtain an initialized original population Sol:

Sol_i＝lb+rand·(ub-lb)(5)

In the formula (5), sol _i represents the position information of the ith individual in the population Sol;

ub represents the upper bound vector of the solution space;

lb represents the lower bound vector of the solution space;

rand means generating a random number between 0, 1;

s22, the initial Iter=1, if Iter is less than 4, S23 is executed by judging the relation between Iter and 4, otherwise, S28 is executed in a jumping manner; wherein Iter represents the current iteration number;

s23, updating the individual position through an exploration stage;

S24, updating the individual position through a development stage;

S25, combining the population Sol, sol ^* and Sol ^**, and taking the first Nc individuals; where Sol represents the original population after initialization, sol ^* represents the new population after updating by the exploration phase, sol ^** represents the new population after updating by the development phase;

s26, calculating the optimal individual fitness of the populations Sol ^* and Sol ^** AndMeanwhile, iter=iter+1;

s27, respectively calculating scores of exploration phases in the Iter iteration And score of development stage

S28, executing S29 if Iter is not more than MaxIter, otherwise executing step S33; wherein MaxIter represents the maximum iteration number and MaxIter is not less than 4;

s29, updating the individual position through the exploration phase;

s30, updating the individual position through a development stage;

S31, pair AndUpdating;

S32, iter=iter+1, returning to S28;

S33, selecting an individual with the smallest objective function value in the population Sol ^* at the last iteration as a scheduling method;

S3, mapping the optimal individuals obtained in the S2 to a binary scheduling matrix, and obtaining a task scheduling strategy through the scheduling matrix so as to perform task scheduling;

The binary scheduling matrix is expressed as:

Wherein the matrix U is a scheduling matrix;

T, K are the number of rows and columns of the matrix U respectively;

u _i,j denotes that the i ^th task runs on the j ^th computing resource and specifies that each task can only be computed on one virtual machine;

The optimal individual is mapped to a binary scheduling matrix, and the optimal individual is changed into the binary matrix by adopting a random S-T-shaped transfer function so as to perform task scheduling, wherein the optimal individual is expressed by the following formula:

In equation (34), abs represents an absolute function, tan represents a tangent function, sol _male represents an optimal individual in the iterative process, and Matrix _new represents a Matrix after conversion by a random S-T transfer function.

2. The cloud computing resource scheduling method based on the enhanced american lion optimization algorithm as claimed in claim 1, wherein said initializing the original population further comprises: and updating the initialized population by adopting Sinusoidal chaotic mapping:

in the formula (30), alpha is a control parameter;

sin represents a sine function;

Sol _i represents the ith individual in the initialized original population Sol;

Representing the individual after updating by Sinusoidal chaotic map.

3. The cloud computing resource scheduling method based on the enhanced american lion optimization algorithm according to claim 1, wherein updating the individual location in the development phase introduces a mid-point sample learning strategy to improve the development phase, specifically the following formula:

In the formula (31), sol _midnew represents a new individual updated in the development stage after the introduction of the intermediate point sample learning strategy;

representing new individuals updated at the original development stage;

Sol _mid represents the individual generated by the mid-point sample learning strategy and is represented by the following equation:

In the formula (32), the amino acid sequence of the compound, Representing the R _m th individual in the population after the updated population Sol ^** is sorted in ascending order of fitness value, R _m is calculated by the following formula:

in the formula (33), i represents the ith individual, Representing an upward rounding function, rand represents a random number between intervals 0, 1.

4. A cloud computing resource scheduling system based on an enhanced american lion optimization algorithm, the system comprising:

And a data acquisition module: acquiring a cloud task byte size vector, a computing resource vector, a load cost vector and a resource bandwidth vector which are provided by a virtual machine;

Task scheduling module: optimizing the established objective function of the minimum total cost through an enhanced American lion optimization algorithm to obtain an optimal individual; mapping the optimal individuals to a binary scheduling matrix, and obtaining a task scheduling strategy through the scheduling matrix so as to perform task scheduling;

The binary scheduling matrix is expressed as:

Wherein the matrix U is a scheduling matrix;

T, K are the number of rows and columns of the matrix U respectively;

u _i,j denotes that the i ^th task runs on the j ^th computing resource and specifies that each task can only be computed on one virtual machine, the objective function to minimize the total cost is:

Fun＝min{ω₁F₁+ω₂F₂+ω₃F₃}(1)

In formula (1), ω ₁,ω₂ and ω ₃ represent weights of the functions F ₁,F₂ and F ₃, respectively;

F ₁ represents a time cost function;

f ₂ represents a load cost function;

F ₃ represents a resource bandwidth cost function;

The optimization of the objective function of the minimum total cost established by the enhanced American lion optimization algorithm to obtain the optimal individual comprises the following steps:

S21, initializing an original population through a formula (5) to obtain an initialized original population Sol:

Sol_i＝lb+rand·(ub-lb)(5)

in the formula (5), sol _i represents the position information of the ith individual in the population Sol, ub represents the upper bound vector of the solution space, lb represents the lower bound vector of the solution space, and rand represents the generation of a random number between [0,1 ];

s22, the initial Iter=1, if Iter is less than 4, S23 is executed by judging the relation between Iter and 4, otherwise, S28 is executed in a jumping manner; wherein Iter represents the current iteration number;

s23, updating the individual position through an exploration stage;

S24, updating the individual position through a development stage;

S25, merging the population Sol, sol ^* and Sol ^**, and taking the first Nc individuals, wherein Sol represents an initialized original population, sol ^* represents a new population updated by an exploration phase, and Sol ^** represents a new population updated by a development phase;

s26, calculating the optimal individual fitness of the populations Sol ^* and Sol ^** AndMeanwhile, iter=iter+1;

s27, respectively calculating scores of exploration phases in the Iter iteration And score of development stage

S28, executing S29 if Iter is not more than MaxIter, otherwise executing step S33; wherein MaxIter represents the maximum iteration number and MaxIter is not less than 4;

s29, updating the individual position through the exploration phase;

s30, updating the individual position through a development stage;

S31, pair AndUpdating;

S32, iter=iter+1, returning to S28;

S33, selecting an individual with the smallest objective function value in the population Sol ^* at the last iteration as a scheduling method;

The optimal individual is mapped to a binary scheduling matrix, and the optimal individual is changed into the binary matrix by adopting a random S-T-shaped transfer function so as to perform task scheduling, wherein the optimal individual is expressed by the following formula:

In equation (34), abs represents an absolute function, tan represents a tangent function, sol _male represents an optimal individual in the iterative process, and Matrix _new represents a Matrix after conversion by a random S-T transfer function.

5. The cloud computing resource scheduling system based on the enhanced american lion optimization algorithm of claim 4, wherein said initializing the original population further comprises: and updating the initialized population by adopting Sinusoidal chaotic mapping:

in the formula (30), alpha is a control parameter;

sin represents a sine function;

Sol _i represents the ith individual in the initialized original population Sol;

Sol _i ^new represents the individual after updating by Sinusoidal chaotic map.

6. The cloud computing resource scheduling system based on the enhanced american lion optimization algorithm as claimed in claim 4, wherein updating the individual location during the development phase introduces a mid-point sample learning strategy to improve the development phase, specifically the following formula:

In the formula (31), Representing new individuals updated at the original development stage;

Sol _mid represents the individual generated by the mid-point sample learning strategy and is represented by the following equation:

In equation (32), temp represents the population after the updated population Sol ^** is sorted in ascending order of its fitness value, and R _m is calculated by the following equation:

in the formula (33), i represents the ith individual, Representing an upward rounding function, rand represents a random number between intervals 0, 1.

7. A computer device comprising a processor and a memory for storing a program executable by the processor, wherein the processor implements the method of claim 1 when executing the program stored in the memory.

8. A readable storage medium storing a program, which when executed by a processor, implements the method of claim 1.