Industrial control software public test resource scheduling method based on adaptive genetic algorithm

Abstract

The invention discloses an industrial control software public test resource scheduling method based on a self-adaptive genetic algorithm. The method comprises the steps of task scheduling modeling, resource constraint condition definition, task dependency relationship construction and dynamic scheduling strategy making. By introducing a self-adaptive mechanism, the crossover rate and the mutation rate in the genetic algorithm are adjusted to adapt to the search requirements of different stages, so that the optimization efficiency and the diversity of solutions are improved. According to the method, a public test resource scheduling optimization process specially aiming at industrial control software characteristics is designed, and the complex scheduling problems of multiple tasks, multiple constraints and limited resources can be effectively solved. Compared with a traditional static scheduling strategy, the method has the advantages that the test period is remarkably shortened, the resource utilization rate and the task completion rate are increased, the method is particularly suitable for industrial control software public test scenes with large-scale and multi-skill requirements and complex task dependence, and the method has good application prospects and popularization value.

Description

Technical Field

[0001] The present invention relates to the field of project scheduling optimization, and in particular to an industrial control software crowd-testing resource scheduling method based on an adaptive genetic algorithm. Background Art

[0002] Existing industrial control software crowd-testing resource scheduling methods (MS-RCPSP) generally suffer from low scheduling efficiency, insufficient resource utilization, and difficulty adapting to complex and changing scenarios. Traditional methods are mostly based on static planning or heuristic strategies, lacking sufficient global search and local optimization capabilities, resulting in scheduling results easily falling into local optimality in large-scale tasks and multi-skill resource environments. Genetic algorithms (GAs), as a global optimization method, have strong search capabilities, but fixed parameter settings often lead to too fast early convergence or slow late convergence, affecting scheduling results. In summary, it is urgent to design a new method to solve the problem of industrial control software crowd-testing resource scheduling. At the same time, an optimization strategy based on an adaptive genetic algorithm is introduced to effectively solve the MS-RCPSP problem, thereby improving the overall performance and efficiency of industrial control software crowd-testing resource scheduling. Summary of the Invention

[0003] The present invention provides an industrial control software crowd-testing resource scheduling method based on an adaptive genetic algorithm. The method aims to overcome the shortcomings of the existing crowd-testing resource scheduling process in complex resource allocation, skill matching, and task dependency processing by introducing an adaptive genetic evolution mechanism, thereby improving the global optimization capability and convergence performance of the scheduling scheme.

[0004] The technical solution adopted in the present invention is as follows:

[0005] A crowd-testing resource scheduling method based on multi-level random property particle swarm optimization includes the following steps:

[0006] Step 1: Build a multi-dimensional constrained optimization resource scheduling model. First, build a multi-dimensional resource scheduling optimization model that combines task dependencies, resource constraints, and skill matching requirements. The model uses a multi-objective optimization framework to simultaneously minimize project duration and maximize resource utilization efficiency. By introducing constraints such as resource availability, task execution sequence, and skill requirement matching, the scheduling solution not only meets basic operational requirements but also optimizes the time and resource allocation efficiency of the entire project process.

[0007] Step 2: Design a dynamic adjustment strategy for the adaptive genetic algorithm. In the early stages of the algorithm, a high crossover and mutation rate is used to achieve broad search space coverage, avoiding falling into local optimal solutions. In the later stages of the algorithm, these parameters are gradually reduced to focus on the refined optimization of local optimal solutions. By combining task characteristics with real-time fitness feedback, parameters can be intelligently adjusted at different scheduling stages to improve the algorithm's search efficiency and accuracy.

[0008] Step 3: Introduce multi-level constraint integration and objective function optimization. By integrating factors such as task execution time, resource consumption, and task dependencies, a unified fitness evaluation framework is constructed, and the scheduling scheme is verified and optimized with multiple constraints. Within this framework, the objective function not only focuses on a single optimization goal, but also balances conflicts between different goals to ensure that the final result can achieve optimal performance in multiple dimensions while satisfying constraints such as resource allocation, task sequence, and skill matching.

[0009] Step 4: Generate a scheduling plan and conduct multi-level performance verification. After the algorithm optimization is completed, a specific task scheduling plan is generated by decoding the optimal solution in the genetic algorithm, and the start time, end time, and resource allocation of each task are calculated. Next, a multi-level solution verification is performed to ensure that all constraints, such as resource constraints, task dependencies, skill matching, etc., are met. At the same time, by comparing with other scheduling methods (such as traditional heuristic methods or other optimization algorithms), the advantages of this method in solving the resource scheduling problem of crowd testing of industrial control software are systematically evaluated, and its applicability and superiority in complex scenarios are verified.

[0010] Step 1: Build a resource scheduling model with multi-dimensional constraint optimization. The specific steps are as follows:

[0011] Step 1.1: Define four important index sets: task index set, resource index set, time index set, and skill index set. The specific definitions are as follows:

[0012] Task index: T = {t1, t2, ...t n}, represents all tasks to be scheduled, where t i For task i.

[0013] Resource index: R = {r1, r2, ... r n}, indicating all available resources, where r j For resource j.

[0014] Time index: Γ={t0,t1,...t k}, indicating the time point of the project, the time interval is from t0 (start time) to t k (End Time).

[0015] Skill index: S = {s1,s2,...s p}, indicating all skills, where s p For skill p.

[0016] Step 1.2: Define the model input parameters, which include six parts: resource set R, task set T, skill plan S, task quantity N, resource quantity M, and skill quantity P. The resource set R includes all available resources; the task set T includes all tasks that need to be scheduled; the skill plan S includes all skills required to execute tasks; the task quantity N is the total number of tasks in the project; the resource quantity M is the total number of resources in the project; and the skill quantity P is the total number of skills in the project.

[0017] Resource collection: R = {R1, R2, ... R m}, indicating all available resources.

[0018] Task set: T = {T1, T2, ...T n}, indicating all tasks to be scheduled.

[0019] Skill plan: P = {P1, P2, ... P n}, indicating the skill plan required for each task.

[0020] Number of tasks: N = |T|, which is the total number of tasks.

[0021] Number of resources: M = |R|, that is, the total number of resources.

[0022] Number of skills: P = |S|, which is the total number of skills.

[0023] Step 1.3: Define the variables related to the scheduling task, which includes seven parts: task duration D(T i ), task start time S(T i ), Skill Requirements K(T i ,S j ), resource mastery level L(R i ,S j ), the set of pre-tasks P(T i );

[0024] Mission duration: D(T i ), indicating task T i The execution duration of

[0025] Task start time: S(T i ), indicating task T i The start time of execution.

[0026] Skill requirements: K(Ti ,S j ), indicating task T i Skill S j requirements.

[0027] Resource mastery level: L(T i ,S j ), indicating task T i Skill S j level of mastery.

[0028] Predecessor task set: P(T i ), indicating task T i The set of prerequisite tasks.

[0029] Step 1.4: Define the decision variables, which include two parts: the assignment variable x jr and execute variable y jrt :

[0030]

[0031] Assign variable x jr Is a binary decision variable used to indicate whether task T j Assigned to resource R r ; Execute variable y jrt Used to represent the resource R at time t r Whether to execute task T j ;

[0032] Step 1.5: Construct the objective function. The core of this step is to optimize the project's make span by constructing the objective function. The objective function is used to describe the maximum completion time of all tasks in the project, that is, the total duration. The specific definition is as follows:

[0033] Minimize max{s j +d j ,j∈J}

[0034] In the exploration and development phases, the task priority weights are adjusted according to the importance of the tasks, so that the optimization objectives are more adapted to the optimization needs of different phases:

[0035]

[0036] Where w1+w2=1, and w1 and w2 are adjusted dynamically according to the stage;

[0037] Step 1.6: Construct constraints. In the model, a series of constraints need to be introduced to ensure the feasibility of task scheduling and the rationality of resource allocation. These constraints include the following categories:

[0038] The task priority constraints are as follows:

[0039]

[0040] This constraint ensures that task T j Only in all its predecessor tasks T j After completion, start

[0041] The skill matching constraints are as follows:

[0042]

[0043] This constraint ensures that the allocated resource R r Must meet the task T j Required skills k Proficiency requirements,

[0044] The unique assignment constraints are as follows:

[0045]

[0046] This constraint stipulates that each task can only be assigned to one resource for execution.

[0047] Resource usage restrictions are as follows:

[0048]

[0049] This constraint ensures that each resource can only process one task at a time.

[0050] The variable value range constraints are as follows:

[0051]

[0052] These constraints define the range of values ​​that the variables can take, making the construction of the solution reasonable;

[0053] Step 1.7: Multi-level optimization strategy setup. To improve the optimization performance of the genetic algorithm, this method uses a multi-level optimization strategy, dividing the algorithm into an exploration phase and an exploitation phase. The goals of these two phases are to extensively search the solution space and refine the solution quality. The specific strategies are as follows:

[0054] Exploration phase: In the exploration phase, the goal of the genetic algorithm is to search the solution space extensively to avoid early convergence to a local optimal solution. To achieve this goal, a larger crossover probability and mutation probability are used to increase the diversity of the solution space. The operations in this phase include:

[0055] Crossover operation: Assume we have two parent individuals P1 and P2, whose gene sequences are G1 and G2 respectively. The crossover operation generates two offspring individuals C1 and C2 by exchanging genes at a certain position. The crossover operation formula is as follows:

[0056] C1=Crossover(P1,P2)C2=Crossover(P2,P1)

[0057] Among them, the crossover operation is based on a certain probability P c ,Right now:

[0058] P c =probability of crossover

[0059] Mutation operation: Mutation operation increases the diversity of solutions by randomly changing the genes of individuals and avoids falling into local optimal solutions. The mutation operation formula is as follows:

[0060] M(x)=x+Δx whereΔx~u(-δ,δ)

[0061] Where Δx represents the amount of variation, usually according to a certain probability P m Perform mutation

[0062] P m =probablity of mutation

[0063] Development phase: After the exploration phase is complete, the algorithm enters the development phase, focusing on fine-tuning the area near the current optimal solution. This phase gradually reduces the mutation probability and crossover probability, enhancing the algorithm's local search capabilities, thereby accelerating convergence to the global optimal solution. The specific operations are as follows:

[0064] Crossover operation: In the development stage, the crossover operation is more refined, and the crossover probability P' c is gradually reduced and is usually set to:

[0065] P' c =P c -α whereα∈[0,1]

[0066] The crossover method is still to select two parent individuals for exchange to generate offspring individuals.

[0067] Mutation operation: mutation probability P' m It is also gradually reduced during the development phase, and mutation operations become more cautious to avoid excessive perturbation of the current optimal solution. The mutation probability is set to:

[0068] P' m =P m -β whereβ∈[0,1]

[0069] Fitness evaluation and selection: During the development phase, the fitness of individuals is evaluated to determine which individuals will be selected for the next generation. Assuming the fitness function is f(x), individuals are selected for reproduction based on the value of the fitness function. Selection operations usually use strategies such as roulette wheel selection, bidding selection, or tournament selection to ensure that individuals with higher fitness are more likely to enter the next generation. The selection probability P select (x) can be expressed as:

[0070]

[0071] Among them, x i represents individual i, f(x i ) represents the fitness of the individual, and N is the size of the population.

[0072] Phase transition criteria: The transition between the exploration phase and the development phase can be performed dynamically based on the following criteria:

[0073] Iterations: When a certain number of iterations I is reached max When , it automatically switches to the development stage. At this time, the number of iterations I explore Requirements:

[0074]

[0075] Fitness change rate: If the fitness value does not improve significantly after multiple iterations, the system will determine that the search has reached saturation and automatically enter the development phase.

[0076] Dynamic adjustment mechanism: By analyzing the changing trend of fitness distribution, the conversion timing is flexibly adjusted according to the performance of the particle swarm to ensure the adequacy of the search in the exploration phase.

[0077] Step 2: Introduce an adaptive inertia weight adjustment strategy. In the optimization process of genetic algorithms, crossover and mutation operations are key factors affecting search efficiency and solution quality. This method aims to balance the contradiction between global search and local search by introducing adaptive crossover probability and mutation probability adjustment strategies, in order to improve the algorithm's performance in complex scheduling problems. The specific steps are as follows:

[0078] Step 2.1, Adaptive crossover probability adjustment: In the early stage of the genetic algorithm, the crossover operation is used to enhance the diversity of the population. To this end, a higher crossover probability P is used in the early stage. c , to ensure that the solution space is widely covered during the exploration process. As the algorithm iterates, the crossover probability gradually decreases, causing the population to converge to the optimal solution area. The adjustment formula for the crossover probability is as follows:

[0079]

[0080] Among them, P c(t) is the crossover probability of the current iteration t; P c0 is the initial crossover probability; T is the maximum number of iterations; α is the parameter that controls the crossover probability deceleration rate;

[0081] Step 2.2, adaptive mutation probability adjustment: The role of mutation operation in genetic algorithm is to increase the diversity of solutions and avoid the dilemma of local optimal solution. In order to improve the efficiency of mutation operation, this method adopts a higher mutation probability P in the early stage of the algorithm. m , as the iteration progresses, the mutation probability is gradually reduced so that fine-tuning can be performed when approaching the optimal solution. The decreasing formula for the mutation probability is as follows:

[0082]

[0083] Among them, P m (t) is the mutation probability at the current iteration t; P m0 is the initial mutation probability; T is the maximum number of iterations; β is the parameter that controls the crossover probability deceleration rate;

[0084] Step 2.3, fitness detection and adaptive adjustment mechanism: In order to further improve the convergence and stability of the genetic algorithm, an adaptive mechanism is used to monitor and dynamically adjust the crossover and mutation probabilities in real time. By analyzing the changing trend of the fitness function, when the improvement of the fitness function decreases or the number of iterations reaches a certain number, the fitness change range Δf(t) is used to determine whether the crossover and mutation probabilities need to be adjusted:

[0085] Δf(t)=f(t)-f(t-1)

[0086] If Δf(t) is less than the preset threshold ε, it means that the optimization of the solution is close to saturation. At this time, the intensity of the crossover and mutation operations can be appropriately reduced to improve the accuracy of the local search and enhance the stability of the algorithm.

[0087] Step 3: Implement adaptive genetic algorithm optimization within a multi-level optimization framework. The exploration phase randomly initializes individuals in the population to extensively search for potential high-quality scheduling solutions. The development phase focuses on the area surrounding the current excellent solution, performing localized fine-tuning and gradually improving the quality of the scheduling solution. The synergy between the exploration and development phases effectively enhances the algorithm's global search capabilities and local convergence performance. The specific steps are as follows:

[0088] Step 3.1, population initialization: Randomly generate the initial population, and each individual uses a genetic code to represent the task scheduling order, which is defined as follows:

[0089] Chromosome j =RandomPermutation(N),j=1,2...P

[0090] Among them, N is the number of tasks, P is the population size; at the same time, the initial crossover probability P is set c0 and the initial mutation probability P m0 ;

[0091] Step 3.2: Genetic operation execution: selection operation is performed according to the fitness function, and the parents are determined by roulette or tournament method; crossover and mutation operations are performed with dynamic probability P c (t), P m (t), the specific adjustment formula is as follows:

[0092]

[0093] Among them, P c0 、P m0 is the initial probability, t is the current generation, T is the maximum number of iterations, α and β are control parameters;

[0094] Step 3.3, fitness evaluation and winner selection: through fitness function

[0095]

[0096] Evaluate individual quality and select a new generation based on fitness. If there is an individual with better fitness than the current global optimal solution, update the global optimal solution and gradually guide the population to converge to the optimal scheduling solution.

[0097] Step 4: Generate and verify the scheduling plan. By decoding the chromosome gene sequence, the task scheduling order is determined, the start and end times of each task are calculated, and the scheduling plan is verified to meet multiple constraints such as resource allocation, task dependencies, and skill matching. The optimization effect is evaluated by comparing with traditional methods to verify its superiority and applicability in complex resource-constrained environments. The specific steps are as follows:

[0098] Step 4.1, generate a scheduling plan: In this stage, the execution order of tasks is determined according to the order of genes in individual chromosomes; the start time S of each task is calculated step by step according to the order i and completion time C i , the formula is as follows:

[0099]

[0100] C i =S i +D i

[0101] Among them, P i is the set of all predecessor tasks of task i, D iis the duration of task i, EarliestAvailable(i) represents the earliest time when task i can start if the resource availability condition is met;

[0102] Step 4.2, verify the constraints: resource constraint check: ensure that at any time point t, for each resource r, the following conditions are met:

[0103]

[0104] Where Γ(r,t) represents the set of tasks that are using resource r at time t, Cap r represents the maximum capacity of resource r;

[0105] Task pre-requisite constraint: For any task pair (i, j) with a dependency relationship, the following conditions must be met:

[0106] C i ≤S j

[0107] That is, task j can only start after all predecessor tasks i are completed;

[0108] Skill matching constraint: When resource r is assigned to task i, the skill level q of the resource r must be met. r,s Not less than the skill requirements for the task i,s ,Right now:

[0109]

[0110] Skills(i) represents the skill set associated with task i.

[0111] Advantages and beneficial effects of the present invention

[0112] The present invention proposes a resource scheduling method for crowd-testing of industrial control software based on an adaptive genetic algorithm. Aiming at the problems of resource heterogeneity and complex task dependencies in the crowd-testing process, an optimization algorithm including exploration and development phases is designed in combination with an adaptive genetic evolution strategy. This method achieves comprehensive optimization of resource allocation, skill matching, and task priority relationships by dynamically adjusting genetic parameters, effectively improving scheduling quality and shortening the total project duration (makespan). The present invention is applicable to industrial control software crowd-testing projects and other complex engineering scheduling scenarios. It can significantly enhance the stability and efficiency of project execution while improving resource utilization, providing effective technical support for improving productivity and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0113] Figure 1 This is a flow chart of a method for resource scheduling of industrial control software crowd testing based on an adaptive genetic algorithm according to the present invention;

[0114] Figure 2 It is a topological diagram of task dependencies and resource constraints in resource scheduling for crowd-testing of industrial control software. DETAILED DESCRIPTION

[0115] The present invention is described in detail below with reference to the accompanying drawings and examples.

[0116] A method for crowd-testing resource scheduling for industrial control software based on an adaptive genetic algorithm is proposed. Based on a crowd-testing resource scheduling problem with five task nodes and several dependencies (the structure is shown in Table 1), an adaptive genetic algorithm is constructed to solve the optimal scheduling solution for resources and tasks within a finite time unit. The specific implementation plan includes the following contents and steps:

[0117] Table 1 Structural information of a crowd-testing resource scheduling

[0118]

[0119]

[0120]

[0121] Step 1: Build a resource scheduling model with multi-dimensional constraint optimization. The specific steps are as follows:

[0122] Step 1.1: Define four important index sets: task index set, resource index set, time index set, and skill index set. The specific definitions are as follows:

[0123] Task index: T = {t1, t2, ...t n}, represents all tasks to be scheduled, where t i For task i.

[0124] Resource index: R = {r1, r2, ... r n}, indicating all available resources, where r j For resource j.

[0125] Time index: Γ={t0,t1,...t k}, indicating the time point of the project, the time interval is from t0 (start time) to t k (End Time).

[0126] Skill index: S = {s1,s2,...s p}, indicating all skills, where s p For skill p.

[0127] Step 1.2: Define the model input parameters, which include six parts: resource set R, task set T, skill plan S, task quantity N, resource quantity M, and skill quantity P. The resource set R includes all available resources; the task set T includes all tasks that need to be scheduled; the skill plan S includes all skills required to execute tasks; the task quantity N is the total number of tasks in the project; the resource quantity M is the total number of resources in the project; and the skill quantity P is the total number of skills in the project.

[0128] Resource collection: R = {R1, R2, ... R m}, indicating all available resources.

[0129] Task set: T = {T1, T2, ...T n}, indicating all tasks to be scheduled.

[0130] Skill plan: P = {P1, P2, ... P n}, indicating the skill plan required for each task.

[0131] Number of tasks: N = |T|, which is the total number of tasks.

[0132] Number of resources: M = |R|, that is, the total number of resources.

[0133] Number of skills: P = |S|, which is the total number of skills.

[0134] Step 1.3: Define the variables related to the scheduling task, which includes seven parts: task duration D(T i ), task start time S(T i ), Skill Requirements K(T i ,S j ), resource mastery level L(R i ,S j ), the set of pre-tasks P(T i );

[0135] Mission duration: D(T i ), indicating task T i The execution duration of

[0136] Task start time: S(T i ), indicating task T i The start time of execution.

[0137] Skill requirements: K(T i ,S j ), indicating task T i Skill S j requirements.

[0138] Resource mastery level: L(T i ,S j ), indicating task T i Skill S j level of mastery.

[0139] Predecessor task set: P(T i ), indicating task T i The set of prerequisite tasks.

[0140] Step 1.4: Define the decision variables, which consist of two parts: describing the allocation of tasks and resources and the execution status of tasks at specific points in time:

[0141] Assign decision variables: x ij ∈{0,1}, represents task T i Is it allocated to resource R? j .

[0142] Execution decision variable: y ijk ∈{0,1}, represents task T i Is it at time point T k Resource R j implement.

[0143]

[0144] Step 1.5: Construct an optimization objective function. The core of this step is to optimize the project's make span by constructing an objective function. The objective function is used to describe the maximum completion time of all tasks in the project, that is, the total duration. The specific definition is as follows:

[0145]

[0146] In the exploration and development phases, the task priority weights are adjusted according to the importance of the tasks, so that the optimization objectives are more adapted to the optimization needs of different phases:

[0147] Exploration phase: Task priority weight is set to the inverse of the task duration:

[0148]

[0149] Development phase: Task priority weights are set to the duration of the task:

[0150]

[0151] The optimization goal is:

[0152]

[0153] Step 1.6: Establish constraints. In the model, a series of constraints need to be introduced to ensure the feasibility of the task and the rational allocation of resources. These constraints include the following categories:

[0154] The task priority constraints are as follows:

[0155]

[0156] This constraint ensures that task T i Only in all its predecessor tasks T k After completion, start

[0157] The skill matching constraints are as follows:

[0158]

[0159] This constraint ensures that the allocated resource R j The skill mastery level can meet the task T i Skill S l The proficiency requirement is as follows:

[0160]

[0161] This constraint ensures that each task T i Can only be assigned to one resource R j ,

[0162] Resource usage restrictions are as follows:

[0163]

[0164] This constraint ensures that each resource R j At any point in time T k Only one task can be performed.

[0165] Step 1.7: Multi-level optimization strategy setup. To improve the optimization performance of the genetic algorithm, this method uses a multi-level optimization strategy, dividing the algorithm into an exploration phase and an exploitation phase. The goals of these two phases are to extensively search the solution space and refine the solution quality. The specific strategies are as follows:

[0166] Exploration phase: In the exploration phase, the goal of the genetic algorithm is to search the solution space extensively to avoid early convergence to a local optimal solution. To achieve this goal, a larger crossover probability and mutation probability are used to increase the diversity of the solution space. The operations in this phase include:

[0167] Crossover operation: Assume we have two parent individuals P1 and P2, whose gene sequences are G1 and G2 respectively. The crossover operation generates two offspring individuals C1 and C2 by exchanging genes at a certain position. The crossover operation formula is as follows:

[0168] C1=Crossover(P1,P2) C2=Crossover(P2,P1)

[0169] Among them, the crossover operation is based on a certain probability P c ,Right now:

[0170] P c =probability of crossover

[0171] Mutation operation: Mutation operation increases the diversity of solutions by randomly changing the genes of individuals and avoids falling into local optimal solutions. The mutation operation formula is as follows:

[0172] M(x)=x+Δx whereΔx~u(-δ,δ)

[0173] Where Δx represents the amount of variation, usually according to a certain probability P m Perform mutation

[0174] P m =probablity of mutation

[0175] Development phase: After the exploration phase is complete, the algorithm enters the development phase, focusing on fine-tuning the area near the current optimal solution. This phase gradually reduces the mutation probability and crossover probability, enhancing the algorithm's local search capabilities, thereby accelerating convergence to the global optimal solution. The specific operations are as follows:

[0176] Crossover operation: In the development stage, the crossover operation is more refined, and the crossover probability P' c is gradually reduced and is usually set to:

[0177] P' c =P c -α whereα∈[0,1]

[0178] The crossover method is still to select two parent individuals for exchange to generate offspring individuals.

[0179] Mutation operation: mutation probability P' m It is also gradually reduced during the development phase, and mutation operations become more cautious to avoid excessive perturbation of the current optimal solution. The mutation probability is set to:

[0180] P' m =P m -β whereβ∈[0,1]

[0181] Fitness evaluation and selection: During the development phase, the fitness of individuals is evaluated to determine which individuals will be selected for the next generation. Assuming the fitness function is f(x), individuals are selected for reproduction based on the value of the fitness function. Selection operations usually use strategies such as roulette wheel selection, bidding selection, or tournament selection to ensure that individuals with higher fitness are more likely to enter the next generation. The selection probability P select (x) can be expressed as:

[0182]

[0183] Among them, x i represents individual i, f(x i ) represents the fitness of the individual, and N is the size of the population.

[0184] Phase transition criteria: The transition between the exploration phase and the development phase can be performed dynamically based on the following criteria:

[0185] Iterations: When a certain number of iterations I is reached max When , it automatically switches to the development stage. At this time, the number of iterations I explore Requirements:

[0186]

[0187] Fitness change rate: If the fitness value does not improve significantly after multiple iterations, the system will determine that the search has reached saturation and automatically enter the development phase.

[0188] Dynamic adjustment mechanism: By analyzing the changing trend of fitness distribution, the conversion timing is flexibly adjusted according to the performance of the particle swarm to ensure the adequacy of the search in the exploration phase.

[0189] Step 2: Introduce an adaptive inertia weight adjustment strategy. In the optimization process of genetic algorithms, crossover and mutation operations are key factors affecting search efficiency and solution quality. This method aims to balance the contradiction between global search and local search by introducing adaptive crossover probability and mutation probability adjustment strategies, in order to improve the algorithm's performance in complex scheduling problems. The specific steps are as follows:

[0190] Step 2.1, Adaptive crossover probability adjustment: In the early stage of the genetic algorithm, the crossover operation is used to enhance the diversity of the population. To this end, a higher crossover probability P is used in the early stage. c , to ensure that the solution space is widely covered during the exploration process. As the algorithm iterates, the crossover probability gradually decreases, causing the population to converge to the optimal solution area. The adjustment formula for the crossover probability is as follows:

[0191]

[0192] Among them, P c(t) is the crossover probability of the current iteration t; P c0 is the initial crossover probability; T is the maximum number of iterations; α is the parameter that controls the crossover probability deceleration rate;

[0193] Step 2.2, adaptive mutation probability adjustment: The role of mutation operation in genetic algorithm is to increase the diversity of solutions and avoid the dilemma of local optimal solution. In order to improve the efficiency of mutation operation, this method adopts a higher mutation probability P in the early stage of the algorithm. m , as the iteration progresses, the mutation probability is gradually reduced so that fine-tuning can be performed when approaching the optimal solution. The decreasing formula for the mutation probability is as follows:

[0194]

[0195] Among them, P m (t) is the mutation probability at the current iteration t; P m0 is the initial mutation probability; T is the maximum number of iterations; β is the parameter that controls the crossover probability deceleration rate;

[0196] Step 2.3, fitness detection and adaptive adjustment mechanism: In order to further improve the convergence and stability of the genetic algorithm, an adaptive mechanism is used to monitor and dynamically adjust the crossover and mutation probabilities in real time. By analyzing the changing trend of the fitness function, when the improvement of the fitness function decreases or the number of iterations reaches a certain number, the fitness change range Δf(t) is used to determine whether the crossover and mutation probabilities need to be adjusted:

[0197] Δf(t)=f(t)-f(t-1)

[0198] If Δf(t) is less than the preset threshold ε, it means that the optimization of the solution is close to saturation. At this time, the intensity of the crossover and mutation operations can be appropriately reduced to improve the accuracy of the local search and enhance the stability of the algorithm.

[0199] Step 3: Implement adaptive genetic algorithm optimization within a multi-level optimization framework. The exploration phase randomly initializes individuals in the population to extensively search for potential high-quality scheduling solutions. The development phase focuses on the area surrounding the current excellent solution, performing localized fine-tuning and gradually improving the quality of the scheduling solution. The synergy between the exploration and development phases effectively enhances the algorithm's global search capabilities and local convergence performance. The specific steps are as follows:

[0200] Step 3.1, population initialization: Randomly generate the initial population, and each individual uses a genetic code to represent the task scheduling order, which is defined as follows:

[0201] Chromosome j =RandomPermutation(N),j=1,2...P

[0202] Among them, N is the number of tasks, P is the population size; at the same time, the initial crossover probability P is set c0 and the initial mutation probability P m0 ;

[0203] Step 3.2: Genetic operation execution: selection operation is performed according to the fitness function, and the parents are determined by roulette or tournament method; crossover and mutation operations are performed with dynamic probability P c (t), P m (t), the specific adjustment formula is as follows:

[0204]

[0205] Among them, P c0 、P m0 is the initial probability, t is the current generation, T is the maximum number of iterations, α and β are control parameters;

[0206] Step 3.3, fitness evaluation and winner selection: through fitness function

[0207]

[0208] Evaluate individual quality and select a new generation based on fitness. If there is an individual with better fitness than the current global optimal solution, update the global optimal solution and gradually guide the population to converge to the optimal scheduling solution.

[0209] Step 4: Generate and verify the scheduling plan. By decoding the chromosome gene sequence, the task scheduling order is determined, the start and end times of each task are calculated, and the scheduling plan is verified to meet multiple constraints such as resource allocation, task dependencies, and skill matching. The optimization effect is evaluated by comparing with traditional methods to verify its superiority and applicability in complex resource-constrained environments. The specific steps are as follows:

[0210] Step 4.1, generate a scheduling plan: In this stage, the execution order of tasks is determined according to the order of genes in individual chromosomes; the start time S of each task is calculated step by step according to the order i and completion time C i , the formula is as follows:

[0211]

[0212] C i =S i +D i

[0213] Among them, P i is the set of all predecessor tasks of task i, D iis the duration of task i, EarliestAvailable(i) represents the earliest time when task i can start if the resource availability condition is met;

[0214] Step 4.2, verify the constraints: resource constraint check: ensure that at any time point t, for each resource r, the following conditions are met:

[0215]

[0216] Where Γ(r,t) represents the set of tasks that are using resource r at time t, Cap r represents the maximum capacity of resource r;

[0217] Task pre-requisite constraint: For any task pair (i, j) with a dependency relationship, the following conditions must be met:

[0218] C i ≤S j

[0219] That is, task j can only start after all predecessor tasks i are completed;

[0220] Skill matching constraint: When resource r is assigned to task i, the skill level q of the resource r must be met. r,s Not less than the skill requirements for the task i,s ,Right now:

[0221]

[0222] Skills(i) represents the skill set associated with task i.

Claims

1. A method for resource scheduling of industrial control software crowd testing based on an adaptive genetic algorithm comprises the following steps: Step 1: Build a multi-dimensional constrained optimization resource scheduling model. First, build a multi-dimensional resource scheduling optimization model that combines task dependencies, resource constraints, and skill matching requirements. The model uses a multi-objective optimization framework to simultaneously minimize project duration and maximize resource utilization efficiency. By introducing constraints such as resource availability, task execution sequence, and skill requirement matching, the scheduling solution not only meets basic operational requirements but also optimizes the time and resource allocation efficiency of the entire project process. Step 2: Design a dynamic adjustment strategy for the adaptive genetic algorithm. In the early stages of the algorithm, a high crossover and mutation rate is used to achieve broad search space coverage, avoiding falling into local optimal solutions. In the later stages of the algorithm, these parameters are gradually reduced to focus on the refined optimization of local optimal solutions. By combining task characteristics with real-time fitness feedback, parameters can be intelligently adjusted at different scheduling stages to improve the algorithm's search efficiency and accuracy. Step 3: Introduce multi-level constraint integration and objective function optimization. By integrating factors such as task execution time, resource consumption, and task dependencies, a unified fitness evaluation framework is constructed, and the scheduling scheme is verified and optimized with multiple constraints. Within this framework, the objective function not only focuses on a single optimization goal, but also balances conflicts between different goals to ensure that the final result can achieve optimal performance in multiple dimensions while satisfying constraints such as resource allocation, task sequence, and skill matching. Step 4: Generate a scheduling plan and conduct multi-level performance verification. After the algorithm optimization is completed, a specific task scheduling plan is generated by decoding the optimal solution in the genetic algorithm, and the start time, end time, and resource allocation of each task are calculated. Next, a multi-level solution verification is performed to ensure that all constraints, such as resource constraints, task dependencies, skill matching, etc., are met. At the same time, by comparing with other scheduling methods (such as traditional heuristic methods or other optimization algorithms), the advantages of this method in solving the resource scheduling problem of crowd testing of industrial control software are systematically evaluated, and its applicability and superiority in complex scenarios are verified.

2. The method for resource scheduling of industrial control software crowd testing based on an adaptive genetic algorithm according to claim 1, characterized in that: The multi-dimensional constrained optimization resource scheduling model described in Step 1 comprehensively considers task dependencies, resource constraints, and skill matching, using a multi-objective optimization framework to minimize duration and maximize resource utilization. This model ensures the rationality and efficiency of the scheduling solution by integrating constraints such as task sequence and resource availability. Its advantages lie in improving resource utilization, shortening construction periods, and being able to flexibly cope with complex scheduling environments. It provides decision makers with optimized scheduling solutions and effectively improves the resource scheduling efficiency of industrial control software crowd testing. The specific steps are as follows: Step 1.1: Define four important index sets: task index set, resource index set, time index set, and skill index set. The specific definitions are as follows: Task index: T = {t1, t2, ...t n }, represents all tasks to be scheduled, where t i For task i. Resource index: R = {r1, r2, ... r n }, indicating all available resources, where r j For resource j. Time index: Γ={t0,t1,...t k }, indicating the time point of the project, the time interval is from t0 (start time) to t k (End Time). Skill index: S = {s1,s2,...s p }, indicating all skills, where s p For skill p. Step 1.2: Define the model input parameters, which include six parts: resource set R, task set T, skill plan S, task quantity N, resource quantity M, and skill quantity P. The resource set R includes all available resources; the task set T includes all tasks that need to be scheduled; the skill plan S includes all the skills required to execute tasks; and the task quantity N is the total number of tasks in the project. The number of resources M is the total number of resources in the project; the number of skills P is the total number of skills in the project; Resource collection: R = {R1, R2, ... R m }, indicating all available resources. Task set: T = {T1, T2, ...T n }, indicating all tasks to be scheduled. Skill plan: P = {P1, P2, ... P n }, indicating the skill plan required for each task. Number of tasks: N = |T|, which is the total number of tasks. Number of resources: M = |R|, that is, the total number of resources. Number of skills: P = |S|, which is the total number of skills. Step 1.3: Define the variables related to the scheduling task, which includes seven parts: task duration D(T i ), task start time S(T i ), Skill Requirements K(T i ,S j ), resource mastery level L(R i ,S j ), the set of pre-tasks P(T i ); Mission duration: D(T i ), indicating task T i The execution duration of Task start time: S(T i ), indicating task T i The start time of execution. Skill requirements: K(T i ,S j ), indicating task T i Skill S j requirements. Resource mastery level: L(T i ,S j ), indicating task T i Skill S j level of mastery. Predecessor task set: P(T i ), indicating task T i The set of prerequisite tasks. Step 1.4: Define the decision variables, which consist of two parts: describing the allocation of tasks and resources and the execution status of tasks at specific points in time: Assign decision variables: x ij ∈{0,1}, represents task T i Is it allocated to resource R? j . Execution decision variable: y ijk ∈{0,1}, represents task T i Is it at time point T k Resource R j implement. Step 1.5: Construct an optimization objective function. The core of this step is to optimize the project's make span by constructing an objective function. The objective function is used to describe the maximum completion time of all tasks in the project, that is, the total duration. The specific definition is as follows: In the exploration and development phases, the task priority weights are adjusted according to the importance of the tasks, so that the optimization objectives are more adapted to the optimization needs of different phases: Exploration phase: Task priority weight is set to the inverse of the task duration: Development phase: Task priority weights are set to the duration of the task: The optimization goal is: Step 1.6: Establish constraints. In the model, a series of constraints need to be introduced to ensure the feasibility of the task and the rational allocation of resources. These constraints include the following categories: The task priority constraints are as follows: This constraint ensures that task T i Only in all its predecessor tasks T k After completion, start The skill matching constraints are as follows: This constraint ensures that the allocated resource R j The skill mastery level can meet the task T i Skill S l Proficiency requirements, The unique assignment constraints are as follows: This constraint ensures that each task T i Can only be assigned to one resource R j , Resource usage restrictions are as follows: This constraint ensures that each resource R j At any point in time T k Only one task can be performed. Step 1.7: Multi-level optimization strategy setup. To improve the optimization performance of the genetic algorithm, this method uses a multi-level optimization strategy, dividing the algorithm into an exploration phase and an exploitation phase. The goals of these two phases are to extensively search the solution space and refine the solution quality. The specific strategies are as follows: Exploration phase: In the exploration phase, the goal of the genetic algorithm is to search the solution space extensively to avoid early convergence to a local optimal solution. To achieve this goal, a larger crossover probability and mutation probability are used to increase the diversity of the solution space. The operations in this phase include: Crossover operation: Assume we have two parent individuals P1 and P2, whose gene sequences are G1 and G2 respectively. The crossover operation generates two offspring individuals C1 and C2 by exchanging genes at a certain position. The crossover operation formula is as follows: C1=Crossover(P1,P2)C2=Crossover(P2,P1) Among them, the crossover operation is based on a certain probability P c ,Right now: P c =probability of crossover Mutation operation: Mutation operation increases the diversity of solutions by randomly changing the genes of individuals and avoids falling into local optimal solutions. The mutation operation formula is as follows: M(x)=x+Δx whereΔx~u(-δ,δ) Where Δx represents the amount of variation, usually according to a certain probability P m Perform mutation P m =probablity of mutation Development phase: After the exploration phase is complete, the algorithm enters the development phase, focusing on fine-tuning the area near the current optimal solution. This phase gradually reduces the mutation probability and crossover probability, enhancing the algorithm's local search capabilities, thereby accelerating convergence to the global optimal solution. The specific operations are as follows: Crossover operation: In the development stage, the crossover operation is more refined, and the crossover probability P c ' is gradually reduced and is usually set to: P' c =P c -αwhereα∈[0,1] The crossover method is still to select two parent individuals for exchange to generate offspring individuals. Mutation operation: mutation probability P' m It is also gradually reduced during the development phase, and mutation operations become more cautious to avoid excessive perturbation of the current optimal solution. The mutation probability is set to: P' m =P m -βwhereβ∈[0,1] Fitness evaluation and selection: During the development phase, the fitness of individuals is evaluated to determine which individuals will be selected for the next generation. Assuming the fitness function is f(x), individuals are selected for reproduction based on the value of the fitness function. Selection operations usually use strategies such as roulette wheel selection, bidding selection, or tournament selection to ensure that individuals with higher fitness are more likely to enter the next generation. The selection probability P select (x) can be expressed as: Among them, x i represents individual i, f(x i ) represents the fitness of the individual, and N is the size of the population. Phase transition criteria: The transition between the exploration phase and the development phase can be performed dynamically based on the following criteria: Iterations: When a certain number of iterations I is reached max When , it automatically switches to the development stage. At this time, the number of iterations I explore Requirements: Fitness change rate: If the fitness value does not improve significantly after multiple iterations, the system will determine that the search has reached saturation and automatically enter the development phase. Dynamic adjustment mechanism: By analyzing the changing trend of fitness distribution, the conversion timing is flexibly adjusted according to the performance of the particle swarm to ensure the adequacy of the search in the exploration phase.

3. The method for resource scheduling of industrial control software crowd testing based on an adaptive genetic algorithm according to claim 1, characterized in that: Step 2 introduces an adaptive inertia weight adjustment strategy. In the optimization process of genetic algorithms, crossover and mutation operations are key factors affecting search efficiency and solution quality. This method aims to balance the contradiction between global search and local search by introducing adaptive crossover and mutation probability adjustment strategies, in order to improve the algorithm's performance in complex scheduling problems. The specific steps are as follows: Step 2.1, Adaptive crossover probability adjustment: In the early stage of the genetic algorithm, the crossover operation is used to enhance the diversity of the population. To this end, a higher crossover probability P is used in the early stage. c , to ensure that the solution space is widely covered during the exploration process. As the algorithm iterates, the crossover probability gradually decreases, causing the population to converge to the optimal solution area. The adjustment formula for the crossover probability is as follows: Among them, P c (t) is the crossover probability of the current iteration t; P c0 is the initial crossover probability; T is the maximum number of iterations; α is the parameter that controls the crossover probability deceleration rate; Step 2.2, adaptive mutation probability adjustment: The role of mutation operation in genetic algorithm is to increase the diversity of solutions and avoid the dilemma of local optimal solution. In order to improve the efficiency of mutation operation, this method adopts a higher mutation probability P in the early stage of the algorithm. m , as the iteration progresses, the mutation probability is gradually reduced so that fine-tuning can be performed when approaching the optimal solution. The decreasing formula for the mutation probability is as follows: Among them, P m (t) is the mutation probability at the current iteration t; P m0 is the initial mutation probability; T is the maximum number of iterations; β is the parameter that controls the crossover probability deceleration rate; Step 2.3, fitness detection and adaptive adjustment mechanism: In order to further improve the convergence and stability of the genetic algorithm, an adaptive mechanism is used to monitor and dynamically adjust the crossover and mutation probabilities in real time. By analyzing the changing trend of the fitness function, when the improvement of the fitness function decreases or the number of iterations reaches a certain number, the fitness change range Δf(t) is used to determine whether the crossover and mutation probabilities need to be adjusted: Δf(t)=f(t)-f(t-1) If Δf(t) is less than the preset threshold ε, it means that the optimization of the solution is close to saturation. At this time, the intensity of the crossover and mutation operations can be appropriately reduced to improve the accuracy of the local search and enhance the stability of the algorithm.

4. The method for resource scheduling of industrial control software crowd testing based on an adaptive genetic algorithm according to claim 1, characterized in that: The third step is to implement adaptive genetic algorithm optimization within a multi-level optimization framework. The exploration phase randomly initializes individuals in the population to extensively search for potential high-quality scheduling solutions. The development phase focuses on the area surrounding the current excellent solution, performing localized fine-tuning and gradually improving the quality of the scheduling solution. The synergy between the exploration and development phases effectively enhances the algorithm's global search capabilities and local convergence performance. The specific steps are as follows: Step 3.1, population initialization: Randomly generate the initial population, and each individual uses a genetic code to represent the task scheduling order, which is defined as follows: Chromosome j =RandomPermutation(N),j=1,2...P Among them, N is the number of tasks, P is the population size; at the same time, the initial crossover probability P is set c0 and the initial mutation probability P m0 ; Step 3.2: Genetic operation execution: selection operation is performed according to the fitness function, and the parents are determined by roulette or tournament method; crossover and mutation operations are performed with dynamic probability P c (t), P m (t), the specific adjustment formula is as follows: Among them, P c0 、P m0 is the initial probability, t is the current generation, T is the maximum number of iterations, α and β are control parameters; Step 3.3, fitness evaluation and winner selection: through fitness function Evaluate individual quality and select a new generation based on fitness. If there is an individual with better fitness than the current global optimal solution, update the global optimal solution and gradually guide the population to converge to the optimal scheduling solution.

5. The method for resource scheduling of industrial control software crowd testing based on an adaptive genetic algorithm according to claim 1, characterized in that: Step 4: Generate and verify the scheduling plan. By decoding the chromosome gene sequence, the task scheduling order is determined, the start and end times of each task are calculated, and the scheduling plan is verified to meet multiple constraints such as resource allocation, task dependencies, and skill matching. The optimization effect is evaluated by comparing with traditional methods to verify its superiority and applicability in complex resource-constrained environments. The specific steps are as follows: Step 4.1, generate a scheduling plan: In this stage, the execution order of tasks is determined according to the order of genes in individual chromosomes; the start time S of each task is calculated step by step according to the order i and completion time C i , the formula is as follows: C i =S i +D i Among them, P i is the set of all predecessor tasks of task i, D i is the duration of task i, EarliestAvailable(i) represents the earliest time when task i can start if the resource availability condition is met; Step 4.2, verify the constraints: resource constraint check: ensure that at any time point t, for each resource r, the following conditions are met: Where Γ(r,t) represents the set of tasks that are using resource r at time t, Cap r represents the maximum capacity of resource r; Task pre-requisite constraint: For any task pair (i, j) with a dependency relationship, the following conditions must be met: C i ≤S j That is, task j can only start after all predecessor tasks i are completed; Skill matching constraint: When resource r is assigned to task i, the skill level q it possesses must be met. r,s Not less than the skill requirements for the task i,s ,Right now: Skills(i) represents the skill set associated with task i.

Images