An improved genetic algorithm unmanned cluster scheduling operation method and system

Abstract

The application discloses an improved genetic algorithm unmanned cluster scheduling operation method and system, and relates to unmanned aerial vehicles and intelligent optimization algorithms, which comprises the following steps: obtaining position information of each operation task and each unmanned aerial vehicle, and numbering any operation task and any unmanned aerial vehicle; randomly initializing a population and population algorithm parameters based on a genetic algorithm according to the position information and the number; establishing corresponding objective functions and fitness functions, and calculating individual fitness values; generating offspring individuals through a crossover operation for the parent individuals selected in the previous step, and performing a variable neighborhood search; performing optimization search on the offspring individuals subjected to the variable neighborhood search by using an improved large-scale neighborhood search operator, merging the obtained offspring population and the original population to obtain a new population; iterating until termination, and outputting a solution as an unmanned cluster scheduling distribution result. The application can quickly and efficiently generate an unmanned aerial vehicle scheduling scheme according to the operation task and the number of unmanned aerial vehicles.

Description

Technical Field

[0001] This application relates to the fields of unmanned aerial vehicles (UAVs) and intelligent optimization algorithms, and in particular to an improved genetic algorithm-based method and system for scheduling unmanned swarm operations. Background Technology

[0002] The scheduling optimization decision algorithm for unmanned aerial vehicle (UAV) swarms needs to be selected and adjusted according to the specific requirements and conditions of the task. In practical applications, combining real-time data and optimization strategies can further improve the UAV swarm scheduling algorithm to reduce operational costs and increase operational efficiency. Based on given optimization objectives and constraints, the UAV scheduling algorithm allocates available resources to the operational areas to be served. However, some scheduling algorithms may get trapped in local optima, leading to high scheduling costs. Summary of the Invention

[0003] This application provides an improved genetic algorithm-based unmanned swarm scheduling method and system, which uses the improved genetic algorithm in unmanned swarm scheduling operations to quickly and efficiently generate unmanned swarm scheduling schemes based on the task and the number of unmanned swarms.

[0004] This application provides an improved genetic algorithm-based unmanned cluster scheduling method for jobs, including: Obtain the total number of tasks to be completed (N), the total number of dispatch centers (M), and the total number of drones (K), as well as the location information of each task and each drone, and assign a number to each task and each drone. Based on the location information and the number, the population and population algorithm parameters are randomly initialized using a genetic algorithm. The population includes multiple individual solutions, each individual is a chromosome segment, and the chromosome contains the task number. The drone scheduling is transformed into a mathematical model to establish the corresponding objective function and fitness function, and the individual fitness value is calculated; For the parent individuals selected in the previous step, crossover operations are performed to generate child individuals, and a variable neighborhood search is conducted. An improved large-scale neighborhood search operator is used to perform an optimization search on the offspring after the variable neighborhood search. The resulting offspring population is then merged with the original population to obtain a new population. Iterate until termination, and output the solution as the result of unmanned cluster scheduling and allocation.

[0005] This application also proposes an improved unmanned cluster scheduling operation system based on a genetic algorithm, including a processor and a memory. The memory stores a computer program, which, when executed by the processor, implements the steps of the improved unmanned cluster scheduling operation method based on a genetic algorithm as described above.

[0006] This application utilizes an improved genetic algorithm for unmanned aerial vehicle (UAV) swarm scheduling, enabling rapid and efficient generation of UAV scheduling schemes based on the task and the number of UAVs. The improved genetic algorithm of this application features high efficiency and strong convergence, making it particularly suitable for scenarios requiring efficient UAV scheduling based on task allocation.

[0007] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0008] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a flowchart illustrating the improved unmanned cluster scheduling method for genetic algorithms in an embodiment of this application. Figure 2 This is a schematic diagram of the crossover operation in the improved unmanned cluster scheduling method of genetic algorithm according to an embodiment of this application; Figure 3 This is a schematic diagram of the optimization search for the improved unmanned cluster scheduling method of genetic algorithm in the embodiments of this application; Figure 4 This is a schematic diagram of the optimal scheduling scheme for the improved unmanned cluster scheduling method of the genetic algorithm in this application embodiment; Figure 5 This is a schematic diagram comparing the iterative curves of the improved unmanned cluster scheduling method of the genetic algorithm in the embodiments of this application. Detailed Implementation

[0009] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0010] This application provides an improved genetic algorithm-based unmanned cluster scheduling method for jobs, such as... Figure 1 As shown, it includes the following steps: In step S101, the total number of tasks to be done N, the total number of dispatch centers M, and the total number of drones K are obtained, along with the location information of each task and each drone, and each task and each drone is assigned a number.

[0011] In step S102, based on the location information and number, the population and population algorithm parameters are randomly initialized using a genetic algorithm. The population includes multiple individual solutions, each individual being a chromosome segment containing an integer sequence, which is the task sequence number. For example, in some examples, the population size NIND = 100.

[0012] In step S103, the drone scheduling is transformed into a mathematical model to establish corresponding objective functions and fitness functions, and individual fitness values ​​are calculated. In this specific example, the individual fitness value is used to evaluate the quality of an individual. Individuals can be further selected based on their fitness values, with superior individuals having a higher probability of being selected. Specifically, a roulette wheel betting method can be used.

[0013] In step S104, for the parent individuals selected in the previous step, a crossover operation is performed to generate child individuals, and a variable neighborhood search is performed.

[0014] In step S105, the improved large-scale neighborhood search operator is used to perform an optimization search on the offspring after the variable neighborhood search, and the resulting offspring population is merged with the original population to obtain a new population.

[0015] In step S106, the iteration continues until termination, and the output solution is used as the unmanned cluster scheduling and allocation result.

[0016] Obtaining the total number of pending tasks N, the total number of dispatch centers M, and the total number of drones K specifically includes: the task set F = { , ,…, },by Represents the i-th task, whose attributes are described as follows: ={Loc ,S }, where Loc and S Representing the work tasks Location and area.

[0017] The set of drones M = { , ,…, },by Representing the j-th drone, its attributes are described as follows: ={ , },in This indicates the average flight speed of the drone during the land transfer process. The average operating speed of the drone is indicated. In this embodiment of the application, the average flight speed and the average operating speed of each drone are the same.

[0018] An initial solution is constructed based on time window constraints and the daily workload constraints of the UAV. This initial solution may not necessarily satisfy both workload and time window constraints, which can reduce the search difficulty of the genetic algorithm to some extent. In some embodiments, the population and population algorithm parameters are randomly initialized based on location information and number using a genetic algorithm, including: The dispatch center penalizes drones that operate later than the latest task operation window, with a penalty factor w: in, Let be the start time of the mission for the k-th drone. For the first i The latest time to complete each task; Randomly select one task j∈{1,2,…,n} from all tasks, and initialize the number of drones to be used k=1; Generate a sequence Seq=[j,j+1,…,n,1,…,j-1], which is used to traverse the task, with the traversal index being... i The initial solution is generated when the process reaches n. Traverse the task Seq(i) according to the traversal sequence Seq, and add the UAV Seq(i) to the k-th path according to the time window and workload constraints.

[0019] In some embodiments, the process of converting drone scheduling into a mathematical model to establish corresponding objective functions and fitness functions, and calculating individual fitness values, includes: The objective function is defined as follows:

[0020] Where Z represents the total cost of drone scheduling operations, the total cost of minimizing the objective, the first term being the drone's operational cost, the second term being the distance cost, and the third term being the penalty cost. For drones Operating costs, Cost of drone flight distance, As decision variables, For drones from mission The distance to task j, K is the total number of drones, N is the total number of tasks, M is the total number of drone dispatch centers, α is the penalty weight for exceeding the operation capacity limit, and β is the penalty weight for exceeding the operation time window. This is a penalty item.

[0021] The fitness function is as follows: Calculate the fitness of each individual in the population ( i = 1, 2, …, NIND), where NIND is the population size.

[0022] In some embodiments, it further includes: Calculate the probability that each individual is inherited into the next generation population: Calculate the cumulative probability of each individual: <00001​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​

[0024] Variable neighborhood search is performed using swap, insertion, 2-Opt, and or-Opt neighborhood operators to introduce new gene variations. Specifically, variable neighborhood search is performed using swap, insertion, reverse end, and reverse j-neighborhood operators. The swap operator randomly selects the indices of two elements in the current solution and swaps their corresponding values. The insertion operator randomly selects a node in the current solution and inserts it into the new position. The reverse end operator randomly selects two different nodes i and j in the current solution, removes and reverses the content between i and j, and places it at the end of the entire sequence. The reverse j-neighborhood operator randomly selects two consecutive nodes i and j and inserts them in reverse order after the randomly selected node j.

[0025] In some embodiments, for the parent individuals selected in the previous step, generating child individuals through a crossover operation and performing a variable neighborhood search further includes: Set adaptive weighted neighborhood selection, set an initial weight for each neighborhood method, and set the range of weight variation, for example, set the weight to an integer between 1 and 5, that is, the range of variation is 1-5.

[0026] In the initial search, a neighborhood method is randomly selected. If the solution produced by the selected neighborhood method is better than the current best solution, the weight of that neighborhood is increased based on the range of variation, for example, by 1. Depending on the range of variation, the weight can be increased up to 5.

[0027] If no better solution is found, the weight of the neighborhood method is reduced based on the range of variation, for example, by 1, and the weight can be reduced to as low as 1 according to the range of variation.

[0028] After a search is completed, all neighborhoods are sorted by weight so that the neighborhood structure with the highest weight is selected in the next iteration. This design dynamically adjusts the weights of neighborhood structures based on the capabilities of the neighborhood search operator, ensuring that the search is performed using the best available neighborhood structure, thus accelerating the algorithm's convergence speed and improving its ability to find the optimal solution.

[0029] In some embodiments, such as Figure 3 As shown, the improved large-scale neighborhood search operator is used to perform further optimization search on the offspring after the variable neighborhood search, including: An improved large-scale neighborhood search operator ILNS is used to optimize the solution, where, First, remove all tasks that violate the time window. If no tasks violate the time window, select random tasks and remove a fixed number of highly correlated tasks, then reinsert them. The specific improved algorithm is as follows: The formula for calculating correlation is: in Indicates will The standardized value is between [0,1]. This represents the Euclidean distance between i and j. This indicates whether i and j are on the same path, i.e., whether they are operated by the same drone. A value of 0 indicates they are on the same path, and a value of 1 indicates they are not. Therefore... The larger the value, the greater the correlation between i and j.

[0030] The process iterates until the cost of the newly obtained route is compared with that of the unoptimized route. If the cost can be reduced, the new route is used.

[0031] This application proposes a specific example, employing an improved genetic algorithm in a simulation scheduling experiment. The population size NIND is set to 100, the maximum number of iterations MAXGEN is set to 100 generations, the generation gap GGAP is 0.9, the maximum crossover probability is 0.9, and the minimum crossover probability is 0.5. =10, =70, =2, all drones fly at a speed of 40km / h, and the maximum number of drones that can be scheduled is 25. Using an improved genetic algorithm in Matlab, the optimal scheduling scheme is derived on the test cases shown in Table 1. Figure 4 As shown, the total cost of the optimal scheduling scheme is 4258.18 yuan. Each scheduling route corresponds to a closed loop, requiring a total of 2 drones to be scheduled. The task sequence number and the drone scheduling center sequence number are 0-17. The starting point of each route is the central square dot, which is the drone scheduling center. The drone scheduling scheme is as follows: Scheduling route 1: 0->7->1->6->13->3->2->14->12->17->16->0; Drone scheduling route 2: 0->11->5->8->9->10->4->15->0. The scheduling scheme does not violate the time window route. When drones perform cross-plot operations, they prioritize serving the nearest plot while meeting the time window conditions. The task plot paths served by each drone are clear, and there are no detours or intersections between different paths, indicating that the optimized scheduling scheme generated by the algorithm has good rationality.

[0032] Table 1 Test Case Dataset To further verify the performance improvement of the improved genetic algorithm, a comparative experiment was conducted with two other scheduling algorithms under the same external conditions. Each algorithm was run three times, and the average results were taken. The resulting iterative curve comparison graphs are shown below. Figure 5 As shown. These include the Tabu Search (TS) algorithm and the Genetic Algorithm (GA).

[0033] Tabu search and genetic algorithms both have high initial costs, gradually converging after about 20 generations. In contrast, the algorithm presented in this application, GAVNSILNS, can optimize to around 15,000 from the first generation, stabilizing around the third generation and continuously searching for better solutions, ultimately obtaining the optimal solution 4258.18. This demonstrates stronger local search capability. Furthermore, GAVNSILNS converges faster than other algorithms with increasing iterations. This is because these algorithms lack effective local search strategies, resulting in small changes in the neighborhood and an inability to escape the current neighborhood, thus failing to find the optimal value within the neighborhood. Therefore, this application's method outperforms other algorithms in terms of both convergence speed and optimal value.

[0034] This application also proposes an improved unmanned cluster scheduling operation system based on a genetic algorithm, including a processor and a memory. The memory stores a computer program, which, when executed by the processor, implements the steps of the improved unmanned cluster scheduling operation method based on a genetic algorithm as described above.

[0035] It should be noted that, in the embodiments of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0036] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0037] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0038] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims. All of these forms are within the protection scope of this application.

Claims

1. An improved genetic algorithm-based unmanned cluster job scheduling method, characterized in that, It includes: Obtain the total number N of tasks to be executed, the total number M of scheduling centers, and the total number K of UAVs, as well as the location information of each task to be executed and each UAV, and number any task to be executed and any UAV. Based on the location information and numbers, randomly initialize the population and population algorithm parameters using a genetic algorithm. The population includes multiple individual solutions, each individual being a segment of chromosome, and the chromosome has task numbers. Convert the UAV scheduling into a mathematical model to establish the corresponding objective function and fitness function, and calculate the individual fitness value. For the parent individuals selected in the previous step, generate offspring individuals through crossover operations and perform variable neighborhood search. Use an improved large-scale neighborhood search operator to perform optimization search on the offspring after variable neighborhood search, and merge the obtained offspring population and the original population to obtain a new population. Iterate until termination, and output the solution as the scheduling and allocation result of the UAV cluster.

2. The unmanned cluster scheduling method for the improved genetic algorithm as described in claim 1, characterized in that, Based on the location information and numbers, randomly initialize the population and population algorithm parameters using a genetic algorithm, including: The scheduling center punishes the UAVs that are later than the latest task operation time window, and the penalty factor w is: in, Let be the start time of the mission for the k-th drone. For the first i The latest time to complete each task; Randomly select a task to be executed j ∈ {1, 2, …, n} from all tasks, and initialize the number of UAVs used k = 1. Generate a sequence Seq=[j,j+1,…,n,1,…,j-1], which is used to traverse the task, with the traversal index being... i The initial solution is generated when the process reaches n. Traverse the tasks Seq(i) according to the traversal sequence Seq, and add the UAV Seq(i) to the k-th path according to the time window and workload constraints.

3. The unmanned cluster scheduling method for the improved genetic algorithm as described in claim 1, characterized in that, Convert the UAV scheduling into a mathematical model to establish the corresponding objective function and fitness function, and calculate the individual fitness value, including: Establish the objective function as: Where Z represents the total cost of drone scheduling operations, and the objective is to minimize this total cost. For drones Operating costs, Cost of drone flight distance, As decision variables, For drones from mission The distance to task j, K is the total number of drones, N is the total number of tasks, M is the total number of drone dispatch centers, α is the penalty weight for exceeding the operation capacity limit, and β is the penalty weight for exceeding the operation time window. This is a penalty item; The fitness function is: Calculate the fitness of each individual in the population. ( i =1, 2, ..., NIND), where NIND is the population size.

4. The unmanned cluster scheduling method for the improved genetic algorithm as described in claim 3, characterized in that, It also includes: Calculate the probability that each individual is inherited into the next generation population: Calculate the cumulative probability of each individual: q[i] is the cumulative probability of chromosome x[i], i = 1, 2, …, n; Generate a uniformly distributed pseudo-random number r in the interval [0, 1]; If r < q[1], then select individual 1, otherwise, select individual k such that q[k - 1] < r ≤ q[k] holds; Repeat generating the pseudo-random number r and selecting individuals, a total of NIND times.

5. The unmanned cluster scheduling method for the improved genetic algorithm as described in claim 4, characterized in that, For the parent individuals selected in the previous step, generate offspring individuals through crossover operations and perform variable neighborhood search, including: Select parent 1 and parent 2 as crossover objects, select the crossover point positions i and j in parent 1 and parent 2, and simultaneously intercept the gene segment between the crossover points i and j in parent 1 and parent 2. Add the gene segment intercepted from parent 2 before parent 1, and add the gene segment intercepted from parent 1 before parent 2. Delete duplicate genes from front to back to obtain offspring 1 and offspring 2. Configure the adaptive crossover probability as: For crossover probability, For the maximum crossover probability, To minimize the crossover probability, Indicates the average fitness of the population. Indicates the maximum fitness of the population. This represents the higher fitness among the individuals to be crossed, and A is a constant obtained based on the characteristics of the Sigmoid curve. Use exchange, insertion, 2-Opt, or-Opt neighborhood operators to perform variable neighborhood search to introduce new gene changes.

6. The unmanned cluster scheduling method for the improved genetic algorithm as described in claim 5, characterized in that, For the parent individuals selected in the previous step, generating offspring individuals through crossover operations and performing variable neighborhood search also includes: Set the adaptive weight neighborhood selection, set the initial weight for each neighborhood method, and the initial weight is 1 for all, and set the change range of the weight. In the initial search, a neighborhood method is randomly selected. If the solution produced by the selected neighborhood method is better than the current best solution, the weight of that neighborhood is increased based on the range of variation. If no better solution is found, the weight of the neighborhood method is reduced based on the range of variation. After a search is completed, all neighborhoods are sorted by weight so that the neighborhood structure with higher weight is selected in the next loop search.

7. The unmanned cluster scheduling method for the improved genetic algorithm as described in claim 6, characterized in that, The optimization search for offspring after variable neighborhood search is performed using an improved large-scale neighborhood search operator, including: An improved large-scale neighborhood search operator ILNS is used to optimize the solution, where, First, remove all tasks that violate the time window. If there are no tasks that violate the time window, select random tasks and remove a fixed number of tasks that are highly related to them and reinsert them.

8. An unmanned cluster scheduling system based on an improved genetic algorithm, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program that, when executed by the processor, implements the steps of the unmanned cluster scheduling job method of the improved genetic algorithm as described in any one of claims 1 to 7.

Images