Cloud-based Flexible Job Scheduling Method for Resources Based on Improved Chimp Optimization Algorithm

Abstract

The present invention discloses a flexible job scheduling method for cloud-based resources based on an improved chimp optimization algorithm, and the method comprises the following steps: (1) adjusting the convergence factor of the chimp algorithm and setting its position update strategy according to the chimp algorithm; (2) designing a cloud-based resource scheduling scheme according to the chimp algorithm obtained in step (1) to provide data reference. The present invention can solve the flexible job shop scheduling problem. The present invention provides a new initialization method combining GS, LS and random search, which improves the quality of the initial solution of the population and speeds up the convergence rate of the genetic algorithm. By comparing with the test results of other genetic algorithms in the prior art, the calculation results are further improved, and at the same time, the calculation time is shortened to a certain extent, verifying the feasibility and effectiveness of the proposed initialization method.

Description

Technical Field

[0001] The present invention belongs to the technical field of cloud-based resource flexible job scheduling, and relates to a cloud-based resource flexible job scheduling method based on an improved chimp optimization algorithm. Background Art

[0002] As a typical representative of the integration of informatization and industrialization, however, there are still some problems to be further studied in the application of the relevant theories and researches of cloud manufacturing in scheduling. Based on the relevant researches at home and abroad, this application explores the process-oriented cloud-based resource scheduling problem in the cloud manufacturing environment in combination with the production process. The flexible job shop scheduling problem of cloud-based resource modeling is an extension based on the traditional shop scheduling problem. For the traditional shop scheduling problem, the processing procedures of each workpiece, the corresponding machines and processing times of each procedure are pre-determined. However, for the flexible shop resource scheduling problem of cloud-based, the procedures involved in each workpiece can be processed on multiple machines, and the processing times of the selected machines under this premise are different. Compared with the traditional shop scheduling method, the cloud-based resource scheduling problem increases the flexibility of scheduling and relatively conforms to the actual situation of actual production. Therefore, the cloud-based resource scheduling problem is a problem that needs to be solved urgently at present. Summary of the Invention

[0003] The technical problem to be solved by the present invention is: to provide a cloud-based resource flexible job scheduling method based on an improved chimp optimization algorithm to solve the technical problems existing in the prior art.

[0004] The technical solution adopted by the present invention is: a cloud-based resource flexible job scheduling method based on an improved chimp optimization algorithm, and the method includes the following steps:

[0005] (1) Adjust the convergence factor of the chimp algorithm and set its position update strategy according to the chimp algorithm;

[0006] (2) Design a cloud-based resource scheduling plan according to the chimp algorithm obtained in step (1) to provide data reference.

[0007] The adjustment of the convergence factor and the position update strategy method are as follows:

[0008] S1-1, the initial chimp algorithm; the standard ChOA algorithm divides the chimpanzee population into four types: attackers, obstacles, drivers, and chasers. Among them, the attackers are the leaders of the population, and the other three types of chimpanzees assist in hunting, and their social status decreases in turn. The mathematical models of chimpanzees expelling and chasing prey are as follows:

[0009] d = |C·Xprey(t)-m·Xchimp(t)|

[0010] Xchimp(t + 1) = Xprey(t) - a·d

[0011] Where: t represents the current iteration number; Xprey is the prey position vector; Xchimp is the current chimpanzee position vector; a, m, C are coefficient vectors, and their calculation formulas are as follows:

[0012] a = 2f·r 1 -f

[0013] m = Chaotic value

[0014] C = 2·r 2

[0015] Where: r1 and r2 are random vectors in the range of [0, 1] respectively; f is the convergence factor, and its value decreases non-linearly from 2.5 to 0 as the number of iterations increases; a is a random vector that determines the distance between the chimpanzee and the prey, and its value is a random number between [-f, f]; m is the chaotic mapping vector, representing the influence of the sexual motivation of the chimpanzee during the hunting process; C is the control coefficient for the chimpanzee to expel and chase the prey, and its value is a random number between [0, 2]; after the population is initialized, four optimal solutions are selected in turn as the positions of the attacker, barrier, driver, and chaser. The positions of the other chimpanzees in the population are updated around the positions of the following four chimpanzees, and its mathematical model is described as follows:

[0016] X 1 = Xattacker - a1·|C1·Xattacker - m1·X|

[0017] X 2 = Xbarrier - a2·|C2·Xbarrier - m2·X|

[0018] X 3 = Xchaser - a3·|C3·Xchaser - m3·X|

[0019] X 4 = Xdriver - a4·|C4·Xdriver - m4·X|

[0020] X (t+1) = (X1 + X2 + X3 + X4) / 4

[0021] S1 - 2, adjust the genetic factor of the chimpanzee algorithm to accelerate the convergence speed of the algorithm:

[0022]

[0023] Among them, t is the current iteration number, Max iter is the maximum number of iterations, a initial and afinal They are the initial value and the final value of a, taking values of 2 and 0 respectively;

[0024] S1-3. At the same time, in order to better balance the global search and local development processes of the algorithm, a new adaptive step strategy is set:

[0025]

[0026] S1-4. Set a dynamic proportional weight based on the weight of the guiding position vector to enable the chimpanzee algorithm to perform efficient optimization;

[0027] The distance weights between the current chimpanzee individual and attacker, barrier, chaser, and driver

[0028]

[0029]

[0030]

[0031]

[0032] Combined with the previous adaptive position update strategy, the final position update method can be expressed as

[0033]

[0034] The detailed method in step (3) is as follows:

[0035] S3-1. Consider three performance indicators at the same time: minimizing the makespan, minimizing the load of the machine with the maximum load, and minimizing the total load on all machines;

[0036] The makespan C M

[0037] minC M = min(max(C K )) 1 ≤ k ≤ m

[0038] In the formula, Ck is the completion time of machine M K ;

[0039] The load of the machine with the maximum load W M

[0040] minW M = min(max(W K )) 1 ≤ k ≤ m

[0041] In the formula, W K is the machine M KWorkload;

[0042] Total load of all machines

[0043]

[0044] S3-2. The encoding of the LOV rule is more suitable for the chimpanzee algorithm to solve the FJSP problem. Design the data structure for the scheduling scheme according to the LOV rule:

[0045] S3-3. Use the SDChOA to decode the data structure in step S3-2 to obtain a complete scheduling scheme.

[0046] Advantages of the present invention: Compared with the prior art, the present invention can effectively combine customer groups, scientifically and reasonably arrange the processing sequences of each workpiece and each process, realize the full utilization of processing resources, and improve the workpiece processing efficiency. By following the principle of comprehensive operation balance, apply the standard model to plan the processing sequence of workpieces; finally, the following effects can be achieved:

[0047] 1) Establish a data analysis method around key links, analyze and study the data of key links, and establish a distribution service standard model;

[0048] 2) The configuration of manufacturing (machines, personnel) resources is reasonable, the load of machines is reasonably increased, and the processing tasks of each machine are balanced;

[0049] 3) Establish a task scheduling decision system, and form a process specification for regular data collection, analysis, application, and improvement, etc. Description of the drawings

[0050] Figure 1 is the process structure diagram of the present invention;

[0051] Figure 2 shows the curve graph of the linear decreasing strategy of parameter a and the proposed non-linear decreasing strategy. Detailed implementation manners

[0052] The present invention will be further introduced below in conjunction with specific embodiments.

[0053] Embodiment 1: As Figure 1-2 shown, a cloud-based resource flexible job scheduling method based on an improved chimpanzee optimization algorithm includes the following steps:

[0054] (1) Adjust the convergence factor of the chimpanzee algorithm and set its position update strategy; the methods for adjusting the convergence factor and the position update strategy are as follows:

[0055] S1-1, Initial chimpanzee algorithm; The standard ChOA algorithm divides the chimpanzee population into four types: attackers, blockers, drivers, and chasers. Among them, the attacker is the leader of the population, and the other three types of chimpanzees assist in hunting, with their social status decreasing in turn. The mathematical models for chimpanzees to expel and chase prey are as follows:

[0056] d = |C·Xprey(t) - m·Xchimp(t)|

[0057] Xchimp(t + 1) = Xprey(t) - a·d

[0058] In the formula: t represents the current iteration number; Xprey is the prey position vector; Xchimp is the current chimpanzee position vector; a, m, and C are coefficient vectors, and their calculation formulas are as follows:

[0059] a = 2f·r 1 -f

[0060] m = Chaotic value

[0061] C = 2·r 2

[0062] In the formula: r1 and r2 are random vectors in [0, 1] respectively; f is the convergence factor, and its value decreases non-linearly from 2.5 to 0 as the number of iterations increases; a is a random vector that determines the distance between the chimpanzee and the prey, and its value is a random number between [-f, f]; m is the chaotic mapping vector, representing the influence of the sexual motivation of chimpanzees during the hunting process; C is the control coefficient for chimpanzees to expel and chase prey, and its value is a random number between [0, 2]. After the population is initialized, the four optimal solutions are selected in turn as the positions of the attacker, blocker, driver, and chaser. The positions of other chimpanzees in the population are updated around the positions of the following four types of chimpanzees, and its mathematical model is described as follows:

[0063] X 1 = Xattacker - a1·|C1·Xattacker - m1·X|

[0064] X 2 = Xbarrier - a2·|C2·Xbarrier - m2·X|

[0065] X 3 = Xchaser - a3·|C3·Xchaser - m3·X|

[0066] X 4 = Xdriver - a4·|C4·Xdriver - m4·X|

[0067] X (t+1)=(X1 + X2 + X3 + X4) / 4

[0068] S1 - 2. Adjust the genetic factors of the chimpanzee algorithm to accelerate the convergence speed of the algorithm:

[0069]

[0070] where t is the current iteration number, and Max iter is the maximum iteration number, and a initial and a final are the initial value and the final value of a, taking 2 and 0 respectively; Figure 1 shows the curve comparison between the linear decreasing strategy of parameter a and the proposed non - linear decreasing strategy;

[0071] From Figure 2 it can be clearly seen that compared with the original linear decreasing strategy, this non - linear transition parameter focuses more on local exploitation in more iterations. In the middle and late stages of iteration, the value of the proposed non - linear parameter is smaller, indicating that compared with global exploration, it helps to conduct local exploitation for a long time (about 62% of the maximum iteration number). This figure also shows that during the search process, the proposed non - linear parameter strategy is beneficial to global exploration only in about 38% of the iterations.

[0072] S1 - 3. In ChOA, the initialization of four groups of solutions, attacker, barrier, chaser, and driver, will be recorded and retained until an individual with a better fitness value appears in the iteration process to replace them. That is to say, if in the t - th generation, no solution better than the recorded one appears in the population, the current population still updates its position towards these four chimpanzees. However, when all four of them fall into local optima, it is difficult for the entire population to find a better solution. It can be understood that when the decision - maker of the chimpanzee group misjudges the location where the prey appears, then all the encircling actions of the chimpanzees will be ineffective, and it is difficult for them to find the prey in the wrong place.

[0073] The present invention proposes a new definition method for barrier, chaser, and driver to strengthen the role of the current - generation optimal individual, thereby enhancing the global search ability of the algorithm. In the implementation of the algorithm, like ChOA, however, barrier, chaser, and driver are defined as local variables, which are the chimpanzee individuals with the best, second - best, and third - best fitness values in the t - th generation. At the same time, in order to better balance the global search and local exploitation processes of the algorithm, a new adaptive step - moving strategy is proposed, and its mathematical expression is:

[0074]

[0075] S1-4. The disadvantage of the position update formula in ChOA is that X 1 and X 2 and X 3 and X 4 The method of taking the average cannot highlight the importance among the four. Therefore, two new proportional weight strategies are proposed in the existing technology, namely the weighted average strategy and the proportional weight strategy based on fitness value, and experimental verification is carried out: someone proposed a dynamic proportional weight based on the weight of the guiding position vector; another person analyzed and experimentally studied different weight strategies, and theoretically proved the reason why the dynamic weight strategy can optimize efficiently.

[0076] The distance weights between the current chimpanzee individual and attacker, barrier, chaser, and driver are as follows:

[0077]

[0078]

[0079]

[0080]

[0081] In practical applications, the denominator of the above formulas is likely to be 0. Therefore, a very small constant ε needs to be added, with a value of 10 -16 . They are modified to

[0082]

[0083]

[0084]

[0085]

[0086] Set a dynamic proportional weight based on the weight of the guiding position vector to enable the chimpanzee algorithm to optimize efficiently: Combining the previous adaptive position update strategy, the final position update method can be expressed as

[0087]

[0088] (2) Compare the new chimpanzee algorithm obtained in step (1) with other optimization algorithms using test functions;

[0089] The other optimization algorithms are shown in the following table.

[0090] Table 1 Benchmark test functions

[0091]

[0092] Table 1 shows the basic information of 10 benchmark functions, including 5 unimodal functions F1 - F5, 3 non - linear multimodal functions F6 - F8, and 2 multimodal functions F9 - F10 with fixed dimensions. Table 2 shows the parameter settings of each comparison algorithm;

[0093] (3) Design a cloud - based resource scheduling scheme according to the chimpanzee algorithm obtained in step (1) to provide data reference; the detailed method in step (3) is as follows:

[0094] S3 - 1, considering three performance metrics simultaneously: minimizing the makespan, minimizing the load of the machine with the maximum load, and minimizing the total load on all machines;

[0095] Makespan C M

[0096] minC M = min(max(C K )) 1 ≤ k ≤ m

[0097] where Ck is the completion time of machine M K ;

[0098] Load of the machine with the maximum load W M

[0099] minW M = min(max(W K )) 1 ≤ k ≤ m

[0100] where W K is the workload of machine M K ;

[0101] Total load on all machines

[0102]

[0103] S3 - 2, the LOV - rule - based encoding is more suitable for the chimpanzee algorithm to solve the FJSP problem. Design the data structure for the design scheme according to the LOV rule:

[0104] S3 - 3, use SDChOA to decode the data structure to obtain a complete scheduling scheme.

[0105] The present invention fully verifies the effectiveness and superiority of SDChOA. The present invention compares SDChOA with the Salp Swarm Algorithm (SSA), Grey Wolf Optimizer (GWO), basic Chimp Optimization Algorithm (ChOA), and Chimp Optimization Algorithm with Lévy flight improvement (ChOA_Levi). The population size N = 30, the spatial dimension dim = 10 / 30 / 50, and the number of iterations tmax = 1000. It can be seen from Table 2 that SDChOA has better optimization effects than other algorithms on most functions.

[0106] Table 2 Algorithm test results

[0107]

[0108] Specific application example: Problem description and model construction: The description of the flexible job shop scheduling problem is as follows: n jobs {J1,..., Jn} need to be processed on m machines {M1,..., Mm}. Each job contains one or more operations, and the operation sequence is pre-determined. Each operation can be processed on multiple different processing machines, and the processing time of the operation varies with the different processing machines. The scheduling objective is to select the most suitable machine for each operation, determine the best processing sequence and start time of each job operation on each machine, so that certain performance indicators of the entire system reach the optimal. Therefore, the flexible job shop scheduling problem contains two sub-problems: determining the processing machines of each job and determining the processing order on each machine. The scheduling where each operation can be processed on any one of the selectable processing machines is called a fully flexible job shop scheduling; conversely, the scheduling where each operation can only be processed on some of the selectable processing machines is called a partially flexible job shop scheduling, as shown in Table 3.

[0109] Table 3 An example of a partially flexible job shop scheduling problem

[0110]

[0111] Note: J 1 represents job 1, O 12 represents the second operation of the first job, and so on.

[0112] In addition, the following constraint conditions need to be satisfied during the processing.

[0113] (1) Only one job can be processed on the same machine at the same time.

[0114] (2) The same operation of the same job can only be processed by one machine at the same time.

[0115] (3) Each operation of each job cannot be interrupted once it starts processing.

[0116] (4) Different workpieces have the same priority.

[0117] (5) There is no precedence constraint between the processes of different workpieces, while there is a precedence constraint between the processes of the same workpiece.

[0118] (6) All workpieces can be processed at time zero.

[0119] Considering three performance indicators simultaneously: minimizing the makespan, minimizing the load of the most loaded machine, and minimizing the total load on all machines, the objective functions of these three performance indicators are as follows respectively.

[0120] (1) Makespan C M

[0121] min C M =min(max(C k )) 1 ≤ k ≤ m

[0122] In the formula, C k is the completion time of machine M k .

[0123] (2) Load of the most loaded machine W M

[0124] min W M =min(max(W k )) 1 ≤ k ≤ m

[0125] In the formula, W k is the workload of machine M k .

[0126] (3) Total load of all machines WT

[0127]

[0128] Table 1 is the processing machine and processing schedule of a flexible job shop scheduling problem including 2 workpieces and 5 machines. Among them, “—” indicates that this process cannot select the corresponding machine for processing. The problem listed in Table 2 is a partial flexible job shop scheduling problem. If all the “—” in Table 1 correspond to processing times, it means that each process of each workpiece can select all machines for processing, which is a fully flexible job shop scheduling problem.

[0129] By adopting the method of the present invention, the flexible job shop scheduling problem can be solved. A new initialization method combining GS, LS and random search provided by the present invention improves the quality of the initial solution of the population and accelerates the convergence speed of the genetic algorithm. By comparing with the test results of other genetic algorithms in the prior art, the calculation results are further improved, and at the same time, the calculation time is shortened to a certain extent, verifying the feasibility and effectiveness of the proposed initialization method.

[0130] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A flexible job scheduling method for cloud-based resources based on an improved chimp optimization algorithm, characterized in that: The method includes the following steps: (1) Adjust the convergence factor of the chimp algorithm and set its position update strategy; (2) Design a cloud-based resource scheduling scheme according to the chimp algorithm obtained in step (1) to provide data reference; The adjustment of the convergence factor and the position update strategy method are as follows: S1-1, the initial chimp algorithm; the standard ChOA algorithm divides the chimpanzee population into four types: attackers, blockers, drivers, and chasers. Among them, the attacker is the leader of the population, and the other three types of chimpanzees assist in hunting, and their social status decreases in turn. The mathematical models of chimpanzees expelling and chasing prey are as follows: d = |C·Xprey(t)) - mXchimp(t) Xchimp(t + 1) = Xprey(t) - a·d In the formula: t represents the current iteration number; Xprey is the prey position vector; Xchimp is the current chimpanzee position vector; a, m, C are coefficient vectors, and the calculation formulas are as follows: a = 2f·r1f m = Chaotic value C2r2 In the formula: r1 and r2 are random vectors between [0, 1]; f is the convergence factor, and its value non-linearly decreases from 2.5 to 0 as the number of iterations increases; a is a random vector that determines the distance between the chimpanzee and the prey, and its value is a random number between [-f, f], m is the chaotic mapping vector, representing the influence of sexual motivation of chimpanzees during hunting; C is the control coefficient for chimpanzees to expel and chase prey, and its value is a random number between [0, 2]. After the population is initialized, the four optimal solutions are selected in turn as the positions of the attacker, blocker, driver, and chaser. The positions of other chimpanzees in the population are updated around the positions of the following four chimpanzees, and its mathematical model is described as follows: X = Xattacker - a1|C1·Xattacker - m1·X| X2 = Xbarrier - a2·|C2·Xbarrier - m2·X] X3 = Xchaser - a3|C3·Xchaser - m3·X| X4 = Xdriver - a4·|C4·Xdriver - m4·X] X(t + 1) = (X1 + X2 + X3 + X4) / 4 S1-2, adjust the genetic factor of the chimp algorithm: Among them, t is the current iteration number, Maxiter is the maximum iteration number, ainitial and afinal are the initial value and the final value of a respectively, and the values are 2 and 0 respectively; S1-3, set a new adaptive moving step strategy: S1-4, set a dynamic proportional weight based on the weight of the guiding position vector; The distance weights between the current chimpanzee individual and the attacker, barrier, chaser, and driver Combined with the previous adaptive position update strategy, the final position update method is expressed as The detailed method in step (2) is as follows: S2-1, considering three performance metrics simultaneously: minimizing the makespan, minimizing the load of the machine with the maximum load, and minimizing the total load on all machines; The maximum completion time C M minC M = min(max(C K )) 1≤k≤m where Ck is the completion time of machine M K ; Maximum load Machine load W M minW M = min(max(W K )) for 1 ≤ k ≤ m where W K is the workload of machine M K ; Total load on all machines S2-2, designing the data structure for the scheduling scheme according to the LOV rule: S2-3, decoding the data structure in step S3-2 using SDChOA to obtain a complete scheduling scheme.

Images