Multi-party multi-target unmanned aerial vehicle path planning method based on multi-department balance strategy

Abstract

The invention provides a multi-party multi-target unmanned aerial vehicle path planning method based on a multi-department balance strategy, and the method comprises the steps: 1, enabling a multi-party multi-target unmanned aerial vehicle path planning problem to be modeled into a multi-party multi-target continuous optimization problem, and enabling an efficiency department to pay attention to the height change of path scheduling, fuel consumption and other efficiency targets, the security department pays attention to security risks of people, objects and the like of path scheduling, and finally the decision-making department hopes to select an optimal path conforming to both departments at the same time; 2, solving a multi-party multi-target continuous optimization problem by adopting an evolution optimization algorithm based on a multi-department balance strategy, and in an individual selection stage, degrading adaptive values of inclined individuals; in the individual breeding stage, each individual is evolved towards the optimal solution of the worst department where the individual is located. According to the method, the unmanned aerial vehicle planning path meeting the requirements of the efficiency department and the safety department at the same time as much as possible is searched, the application scene is wide, and the method can adapt to the complex unmanned aerial vehicle path planning problem.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle path planning, and in particular to a multi-party and multi-objective unmanned aerial vehicle path planning method based on a multi-sector balance strategy. Background Art

[0002] UAV path planning has become a hot research area for many years, especially in single-objective and multi-objective optimization, where a large number of theoretical and technical achievements have been accumulated. Early path planning research mainly focused on single-objective optimization, focusing on how to optimize a certain aspect of the path, such as flight time, energy consumption, load, etc. However, with the diversification of UAV application scenarios, path planning problems have gradually become more complex and require comprehensive consideration among multiple objectives. Therefore, more and more research has begun to turn to multi-objective optimization, trying to find the optimal path solution by integrating the requirements of multiple objectives. These objectives may involve multiple aspects such as flight efficiency, path safety, environmental factors, and mission requirements. Researchers have proposed various methods to meet these challenges.

[0003] In recent years, with the continuous expansion of the demand for drone applications, some new progress has been made in the research on drone path planning. In particular, a new idea has been proposed in the prior art, which believes that the drone path planning problem involves multiple departments rather than a single department, mainly including the efficiency department and the safety department. The final optimal path should meet the needs of the efficiency department and the safety department as much as possible to achieve a balance between the two. Specifically, the efficiency department is concerned about the operating efficiency of the drone flight path, hoping to shorten the flight time and reduce energy consumption as much as possible to improve transportation efficiency; while the safety department is more concerned about the risks that drones may encounter during flight, especially the safety of the flight path. This problem can be modeled as a multi-party multi-objective optimization problem, which further deeply models the drone path planning in reality. This modeling method can better reflect the conflicts and trade-offs between the goals of multiple departments, thereby providing a more accurate solution for path planning in complex tasks.

[0004] However, the current multi-party and multi-objective path planning algorithms mostly use search methods based on a double multi-objective individual ranking mechanism to try to filter out the optimal solution from many solutions. These methods have effectively dealt with multi-party and multi-objective optimization problems to a certain extent, but they also have obvious shortcomings. Specifically, these algorithms usually search for the optimal solution for a certain department, but may not be optimal for the objectives of another department, which leads to a waste of certain computing resources and reduces the search accuracy. Therefore, the existing algorithms still have a lot of room for improvement in terms of effectiveness and accuracy, especially in multi-department and multi-objective path planning, which requires more efficient and accurate optimization strategies. Summary of the invention

[0005] Purpose of the invention: The technical problem to be solved by the present invention is to provide a multi-party and multi-objective UAV path planning method based on a multi-department balance strategy in view of the deficiencies of the prior art. The method can take into account both flight efficiency and safety while optimizing the flight path of the UAV, and solve the problem that the existing path planning technology cannot comprehensively consider the needs of the efficiency department and the safety department. By introducing a multi-department balance strategy, the method of the present invention can effectively balance multiple goals of multiple departments and search for the optimal path, thereby ensuring that the planned path meets the needs of the efficiency department to the greatest extent, while fully considering and guaranteeing the safety requirements of the safety department. Therefore, this method is particularly suitable for UAV task scheduling and path planning in complex environments.

[0006] The method of the present invention comprises the following steps: Step 1, model the multi-party and multi-objective UAV path planning problem as a multi-party and multi-objective continuous optimization problem; The multi-party multi-objective continuous optimization problem is expressed as: , , , in, It is a modeled multi-party multi-target UAV path search problem; given the search path Down, and They represent the target set that the efficiency department is concerned about and the target set that the safety department is concerned about, and T represents transposition; Concerned about efficiency department Common goals include: the total distance the drone flies along the planned path, fuel consumption, altitude change, and the total distance the drone passes through all hovering points; Of concern to security departments The first goal is usually manifested as the safety risk of drones, such as the safety risk of drones to pedestrians and vehicles, and the risk of property loss caused by drones; Step 2: Using an evolutionary optimization algorithm based on a multi-sector balance strategy to solve the multi-party and multi-objective continuous optimization problem.

[0007] Step 2 includes: generating an initial population based on the problem information, where each individual in the population is a planned path from the starting point to the end point; , lower limit On the basis of, the initial population is constructed according to the random number generator. The initial population contains N individuals, and the generation method is: , Where D is the dimension of the problem, It is a random number generator used to generate decimals in the range [0,1]. Represents the value of the jth dimension of the ith individual in the population; Based on the above, the target value of each individual in the population under the efficiency department and the safety department is calculated; Then, the population begins to evolve cyclically, which is divided into two steps: individual selection and individual reproduction. Individual selection and individual reproduction are continuously performed until the termination condition is met.

[0008] In step 2, the individual selection includes: Let the parent population be P and the child population be Q in each iteration, then define the merged population , using a multi-sector balance strategy for individual selection, the specific operations include: Step 2-11, for each individual in R , the individual is a planned drone path, and all target values ​​of the individual in all departments are obtained according to the target value calculation method, and the fast non-dominated sorting method is performed in turn to obtain the layer number vector L of the individual in each department: , in, is the set of goals that individuals care about in the efficiency sector The layer number value below, A collection of goals that individuals in the security sector are concerned about The layer number value below; Step 2-12, calculate the difference in the level number of each individual under the efficiency department and the safety department : , Step 2-13: Set the layer number difference to be greater than the set threshold Individuals with a value of less than or equal to the set threshold are defined as tilted individuals. Individuals with for: , in, Indicates the maximum value of the layer number difference of all individuals in the population. For the control parameters, set them as: , in, Indicates the number of evaluations currently consumed. Indicates the maximum number of evaluations; Step 2-14, after dividing the population into tilted individuals and non-tilted individuals, add the value X1 to all target values ​​of the tilted individuals, X1 is greater than all target values ​​of the individuals in the population, so that the tilted individuals move to the upper right in the target space, deteriorate the fitness values ​​of the tilted individuals in the efficiency sector and the safety sector, and make the fitness values ​​of the tilted individuals in the efficiency sector and the safety sector worse than the target values ​​of the non-tilted individuals; Step 2-15: After adjusting the target value, execute the multi-party multi-objective sorting algorithm to obtain the final layer number value of all individuals under the multi-party multi-objective problem. , and calculate the individual exclusion distance according to the layer number ; According to the final layer number value of each individual in the combined population R and exclusion distance To sort, The smaller the better. The bigger the better. The youngest goes first. The largest one is placed first, that is, sorted from good to bad; after sorting, select individuals as the next generation population, among which is the initial population size.

[0009] In step 2, the individual reproduction includes: Steps 2-21, respectively, according to Gaussian distribution and Cauchy distribution The scaling factor F and crossover probability CR required to generate the evolution of contemporary individuals are given by: , , in The range of is (0,1], the range of CR is [0,1]; when When it is less than or equal to zero, regenerate until the value The value of is in the feasible region (0,1], when When the value is greater than 1, The value is set to 1; for ,when When the value of is not within the feasible region [0,1], it is set to the nearest boundary value, that is, When it is less than 0, When the value is set to 0 and CR is greater than 1, The value is set to 1; Step 2-22, use the improved mutation strategy to mutate the individual, the formula is: , in, and Respectively represent The parent individual Dimension value and The variant Dimension value, represents the scaling factor, is a randomly selected individual in the population The value of the dimension, is a randomly selected individual from the population and the eliminated parent individuals in the The value on the dimension; after each population update is completed, the eliminated parent population individuals are selected and put together with the current population, and an individual is randomly selected as ; is the current parent individual Select the best solution from the worst square No. Dimension value, the specific settings are as follows: First, obtain the sequence number of the worst department based on the individual's layer number vector in multiple departments ; Secondly, select from the population the set of individuals that are the first level in the multi-party multi-objective sorting algorithm, that is, the optimal solution to the problem found by the algorithm in the current iteration; After that, select the individuals in the first layer from the individuals in the first layer. The best individuals under the department form a set ; Finally, from the collection Randomly select an individual as the current individual The best individual on the worst square Participate in differential variation; Step 2-23, after differential mutation, each variant With the parent individual Perform crossover operation to obtain offspring individuals , calculate the offspring individuals In the Values ​​on dimensions ; If the offspring If the random numbers in all dimensions are less than CR, a dimension is randomly selected. , so that the offspring individuals in this dimension The value on Equal to the variant individual in this dimension The value on : ; Step 2-24, after completing the mutation crossover operation, use the reflection mechanism to modify value, prevent The value exceeds the upper and lower limits of the problem.

[0010] Step 2-22, serial number The calculation formula is: , Among them, argmax is the function to obtain the sequence number of the maximum value. is the layer number vector of the individual.

[0011] Step 2-23, use the following formula to calculate the offspring individuals In the Values ​​on dimensions : , in, is the crossover probability, Represents a randomly generated decimal number in the range [0,1].

[0012] In step 2-24, the formula of the reflection mechanism is: , in, Indicates The first offspring The value on the dimension, It is the remainder method.

[0013] In step 2, after the termination condition is met, the individual with multi-party multi-objective layer number 1 is selected from the population, which is the optimal solution found by the algorithm.

[0014] The present invention also provides an electronic device, comprising a processor and a memory, wherein the memory stores program code, and when the program code is executed by the processor, the processor executes the steps of the described method.

[0015] The present invention also provides a storage medium storing a computer program or instruction. When the computer program or instruction is run on a computer, the steps of the method described are executed.

[0016] The method of the present invention can be applied to the following fields: Logistics and transportation: Used in drone delivery systems, it can optimize flight paths under multi-department and multi-objective conditions, taking into account both transportation efficiency and transportation safety. It is suitable for scenarios such as urban delivery, drone warehousing, and long-distance material transportation.

[0017] Environmental monitoring: When used in UAVs to perform environmental monitoring tasks, it can take into account the efficiency and safety of the flight path according to the requirements of different monitoring areas. It is particularly suitable for scenarios such as forest fire prevention, climate monitoring, and pollution emission monitoring.

[0018] Urban air travel: With the increasing development of air travel technology, the present invention can be applied to the path planning of new types of transportation such as urban air taxis and flying cars, optimize the safety and efficiency of flight routes, and improve the feasibility of urban air travel.

[0019] Agricultural monitoring: Used for path planning of agricultural drones, taking into account both flight efficiency and safety. It is suitable for tasks such as agricultural spraying and land monitoring, and can effectively improve the production efficiency and operational safety of agricultural operations.

[0020] Beneficial effects: Compared with the BPNNIA, BPHEIA and BPAIMA algorithms used in the prior art, the new scheme proposed by the present invention can search for a more efficient and safe UAV (Unmanned Aerial Vehicle) planning path. This scheme has a wide range of applicable scenarios and can be adapted to complex UAV path planning problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0022] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments, and the above and / or other advantages of the present invention will become more clear.

[0023] like Figure 1 As shown, an embodiment of the present invention provides a multi-party multi-objective UAV path planning method based on a multi-sector balance strategy, including: modeling the multi-party multi-objective UAV path planning problem into a multi-party multi-objective continuous optimization problem, the formula is: , , , in, This is the modeled multi-party multi-target UAV path search problem. Given a search path Down, and They represent the target sets that the efficiency department and the safety department are concerned about respectively, and T represents transposition. Concerned about efficiency department Common goals include: the total distance the drone flies along the planned path, fuel consumption, altitude change, and the total distance the drone passes through all hovering points. Of concern to security departments The first goal is usually manifested as the safety risk of drones, for example, the safety risk of drones to pedestrians and vehicles, and the risk of property loss caused by drones.

[0024] First, the initial population is generated based on the problem information. Each individual in the population is a planned path from the starting point to the end point. , lower limit On the basis of, the initial population is constructed according to the random number generator. The population contains N individuals, and the generation method is as follows: , Where D is the dimension of the problem, A random number generator that generates decimals in the range [0,1]. Represents the value of the jth dimension of the ith individual in the population. The target value of the path in the efficiency department and the safety department is calculated according to the above formula.

[0025] After that, the population begins to evolve in a cycle, which is divided into two steps: individual selection and individual reproduction. Individual selection and individual reproduction are continuously performed until the termination condition is met.

[0026] Individual selection process: Let the parent population at each iteration be , the offspring population is , how to get from the merged population Selecting the next generation population is the core step of the algorithm, so define the merged population The present invention adopts a multi-sector balance strategy for individual selection, and the specific operations are as follows: (1) For the combined population Each individual within , the individual is a planned drone path, obtains all its target values ​​under each department, and executes the Fast Non-dominated Sorting method in turn to obtain the layer number vector of the individual under each department : , in, is the set of goals that individuals care about in the efficiency sector The layer number value below, A collection of goals that individuals in the security sector are concerned about The layer number value describes the quality of the individual in the department. The larger the layer number, the worse the quality of the individual in the department.

[0027] (2) Calculate the gap between the layer numbers of each individual in the efficiency department and the safety department. This gap is used to describe the imbalance of the individual in the two departments. , the formula is: , Specifically, the larger the gap in layer numbers, the higher the degree of imbalance of the individual, which means that the quality difference between departments is greater and the overall balance is worse. These individuals will bias the algorithm's search and increase the difficulty of searching for the optimal solution in multiple departments.

[0028] (3) The layer number difference is greater than the set threshold Individuals with a value of less than or equal to the set threshold are defined as tilted individuals. Individuals with are defined as non-inclined individuals. Set the threshold for: , in, Indicates the maximum value of the layer number difference of all individuals in the population. The control parameter controls the proportion of tilted individuals. It is an adaptive parameter, and its value is set to: , in, Indicates the number of evaluations currently consumed. Indicates the maximum number of evaluations, you can see The value starts at 1 and gradually decreases to 0.2.

[0029] (4) After dividing the population into tilted individuals and non-tilted individuals, add a very large value to all the target values ​​of the tilted individuals so that the two types of individuals can be distinguished during subsequent re-stratification (in practice, add 10 times the maximum target value), so that the tilted individuals move to the upper right in the target space. This adjustment can effectively reduce the quality of tilted individuals in individual selection and reduce the probability of being doubled.

[0030] (5) After adjusting the target value, execute the multi-party multi-objective sorting algorithm to obtain the final layer number value of all individuals under the multi-party multi-objective problem , and calculate the individual exclusion distance according to the layer number According to the final layer number value of each individual in the combined population R and exclusion distance To sort, The smaller the better. The bigger the better. The youngest goes first. The largest one is placed first, that is, sorted from best to worst. After sorting, select individuals as the next generation population, among which is the initial population size.

[0031] Individual reproduction process: The individual reproduction process solves the problem of Generate offspring population This method uses the mutation operator of the differential evolution algorithm to perform individual mutation, and combines the multi-department balance strategy to adjust the individual mutation behavior to ensure that the compiled individual can better balance multiple goals of multiple departments.

[0032] (1) First, according to the Gaussian distribution and Cauchy distribution The formula for generating the parameter F and CR values ​​required for contemporary individual evolution is: , , in, The range of is (0,1], the range of CR is [0,1]; when When it is less than or equal to zero, regenerate until the value The value of is in the feasible region (0,1], when When the value is greater than 1, The value is set to 1; for ,when When the value of is not within the feasible region [0,1], it is set to the nearest boundary value, that is, When it is less than 0, When the value is set to 0 and CR is greater than 1, The value is set to 1.

[0033] (2) Use the improved mutation strategy to mutate individuals. The formula is: , in, and Respectively represent The parent individual Dimension value and The variant Dimension value, represents the scaling factor, is a randomly selected individual in the population The value of the dimension, is a randomly selected individual from the population and the eliminated parent individuals in the For the value of After each population update, the eliminated parent population individuals are selected. If full, replace The individual closest to the individual eliminated by the current parent generation is selected, otherwise it is directly added . is the current parent individual Select the best solution from the worst square No. Dimension value, the specific settings are as follows: First, obtain the sequence number of the worst department based on the individual's layer number vector in multiple departments : , Among them, argmax is a function that finds the parameter (set) of the function. is the layer number vector of individuals. Secondly, a set of individuals in the first layer of the multi-party multi-objective sorting is selected from the population, that is, the optimal solution to the problem; then, the individuals in the first layer are selected from the individual set. The optimal individual under the department (in department, the individual with layer number 1 is the best), forming a set , finally, from the set Randomly select an individual as the current individual The best individual on the worst square Participate in differential mutation.

[0034] (3) After differential mutation, each variant With its parent Perform crossover operation to obtain offspring individuals , the offspring individual In the Values ​​on dimensions : , Where CR is the crossover probability, represents a randomly generated decimal in the range [0,1]. In addition, if the offspring individual If the random numbers in all dimensions are less than CR, a dimension is randomly selected. , so that the offspring individuals in this dimension The value on Equal to the variant individual in this dimension The value on : .

[0035] (4) After completing the mutation crossover operation, the algorithm uses the reflection mechanism to modify value, prevent The values ​​exceed the upper and lower limits of the problem, as shown below: , in, Indicates The first offspring The value on the dimension, is the upper bound of the problem, is the lower bound of the problem, It is the remainder method.

[0036] After the termination condition is met, the individual that is a non-dominated solution in all sectors is selected from the population, which is the optimal solution found by the algorithm.

[0037] Result evaluation: The result evaluation of UAV path planning uses the multi-party inverse generalization distance to evaluate the algorithm performance. This target value is based on the inverse generalization distance of the multi-objective optimization problem, namely: , in, It is used to evaluate the quality of the algorithm in finding the optimal solution that meets both efficiency and safety requirements. represents the individual reverse generalization distance under each department, Indicates the number of departments (here the value is 2).

[0038] This embodiment is tested in a simulation environment. The test problems are based on the test problems in the document "Chen K, Luo W, Lin X, et al. Evolutionary Biparty Multiobjective UAV Path Planning: Problems and Empirical Comparisons [J]. IEEE Transactions on Emerging Topics in Computational Intelligence, 2024.8 (3): 2433~2445". There are 6 test problems in total. These test problems share a terrain map, but the goals of the efficiency department and the safety department in different problems are different. Table 1 shows the specific goals of these problems.

[0039] Table 1

[0040] Among other things, the efficiency department focuses on the following goals: is the total length of the UAV dispatch path, is the total height change of the planned path, is the amount of fuel consumed by the drone to fly a given path, It is the sum of the minimum distances from the planned path to a series of hovering points. The safety department focuses on the following goals: is the accident risk of UAV to people and vehicles during flight, is the risk to property during the flight of the UAV, is the noise pollution produced by the UAV. All of the above goals need to be minimized.

[0041] After experiments, we can get the performance table of this method and other comparison algorithms, as shown in Table 2 below.

[0042] Table 2

[0043] It can be seen from the table that in all simulation problems, the performance results of this method are the best, which is about 10% higher than the second best results.

[0044] The present invention provides a multi-party multi-target UAV path planning method based on a multi-sector balance strategy. There are many methods and ways to implement the technical solution. The above is only a preferred implementation of the present invention. It should be pointed out that for ordinary technicians in this technical field, without departing from the principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention. All components not specified in this embodiment can be implemented by existing technologies.

Claims

1. A multi-party and multi-objective UAV path planning method based on a multi-sector balance strategy, characterized in that: The following steps are involved: Step 1, model the multi-party and multi-objective UAV path planning problem as a multi-party and multi-objective continuous optimization problem; The multi-party multi-objective continuous optimization problem is expressed as: ， ， ， in, It is a modeled multi-party multi-target UAV path search problem; given the search path Down, and They represent the target set that the efficiency department is concerned about and the target set that the safety department is concerned about, and T represents transposition; Concerned about efficiency department goals; Of concern to security departments goals; Step 2: Using an evolutionary optimization algorithm based on a multi-sector balance strategy to solve the multi-party and multi-objective continuous optimization problem.

2. The method according to claim 1, characterized in that Step 2 includes: generating an initial population based on the problem information, where each individual in the population is a planned path from the starting point to the end point; , lower limit On the basis of, the initial population is constructed according to the random number generator. The initial population contains N individuals, and the generation method is: ， Where D is the dimension of the problem, It is a random number generator used to generate decimals in the range [0,1]. Represents the value of the jth dimension of the ith individual in the population; Based on the above, the target value of each individual in the population under the efficiency department and the safety department is calculated; Then, the population begins to evolve cyclically, which is divided into two steps: individual selection and individual reproduction. Individual selection and individual reproduction are continuously performed until the termination condition is met.

3. The method according to claim 2, characterized in that In step 2, the individual selection includes: Let the parent population be P and the child population be Q in each iteration, then define the merged population , using a multi-sector balance strategy for individual selection, the specific operations include: Step 2-11, for each individual in R , the individual is a planned drone path, all target values ​​of the individual in all departments are obtained according to the target value calculation method, and the fast non-dominated sorting method is performed in turn to obtain the layer number vector L of the individual in each department: ， in, is the set of goals that individuals care about in the efficiency sector The layer number value below, A collection of goals that individuals in the security sector are concerned about The layer number value below; Step 2-12, calculate the difference in the level number of each individual under the efficiency department and the safety department : ， Step 2-13: Set the layer number difference to be greater than the set threshold Individuals with a value of less than or equal to the set threshold are defined as tilted individuals. Individuals with for: ， in, Indicates the maximum value of the layer number difference of all individuals in the population. For the control parameters, set them as: ， in, Indicates the number of evaluations currently consumed. Indicates the maximum number of evaluations; Step 2-14, after dividing the population into tilted individuals and non-tilted individuals, add the value X1 to all target values ​​of the tilted individuals, X1 is greater than all target values ​​of the individuals in the population, so that the tilted individuals move to the upper right in the target space, deteriorate the fitness values ​​of the tilted individuals in the efficiency sector and the safety sector, and make the fitness values ​​of the tilted individuals in the efficiency sector and the safety sector worse than the target values ​​of the non-tilted individuals; Step 2-15: After adjusting the target value, execute the multi-party multi-objective sorting algorithm to obtain the final layer number value of all individuals under the multi-party multi-objective problem. , and calculate the individual exclusion distance according to the layer number ; According to the final layer number value of each individual in the combined population R and exclusion distance Sort, and then select individuals as the next generation population, among which is the initial population size.

4. The method according to claim 3, characterized in that In step 2, the individual reproduction includes: Steps 2-21, respectively, according to Gaussian distribution and Cauchy distribution The scaling factor F and crossover probability CR required to generate the evolution of contemporary individuals are given by: ， ， in The range of is (0,1], the range of CR is [0,1]; when When it is less than or equal to zero, regenerate until the value The value of is in the feasible region (0,1], when When the value is greater than 1, The value is set to 1; for ,when When the value of is not within the feasible region [0,1], it is set to the nearest boundary value, that is, When it is less than 0, When the value is set to 0 and CR is greater than 1, The value is set to 1; Step 2-22, use the improved mutation strategy to mutate the individual, the formula is: ， in, and Respectively represent The parent individual Dimension value and The variant Dimension value, represents the scaling factor, is a randomly selected individual in the population The value of the dimension, is a randomly selected individual from the population and the eliminated parent individuals in the The value on the dimension; after each population update is completed, the eliminated parent population individuals are selected and put together with the current population, and an individual is randomly selected as ; is the current parent individual Select the best solution from the worst square No. Dimension value, the specific settings are as follows: First, obtain the sequence number of the worst department based on the individual's layer number vector in multiple departments ; Secondly, select from the population the set of individuals that are the first level in the multi-party multi-objective sorting algorithm, that is, the optimal solution to the problem found by the algorithm in the current iteration; After that, select the individuals in the first layer from the individuals in the first layer. The best individuals under the department form a set ; in the Under the department, the individual with layer number 1 is the best; Finally, from the collection Randomly select an individual as the current individual The best individual on the worst square Participate in differential variation; Step 2-23, after differential mutation, each variant With the parent individual Perform crossover operation to obtain offspring individuals , calculate the offspring individuals In the Values ​​on dimensions ; If the offspring If the random numbers in all dimensions are less than CR, a dimension is randomly selected. , so that the offspring individuals in this dimension The value on Equal to the variant individual in this dimension The value on : ; Step 2-24, after completing the mutation crossover operation, use the reflection mechanism to modify value, prevent The value exceeds the upper and lower limits of the problem.

5. The method according to claim 4, characterized in that Step 2-22, serial number The calculation formula is: ， Among them, argmax is the function to obtain the sequence number of the maximum value. is the layer number vector of the individual.

6. The method according to claim 5, characterized in that Step 2-23, use the following formula to calculate the offspring individuals In the Values ​​on dimensions : ， in, is the crossover probability, Represents a randomly generated decimal number in the range [0,1].

7. The method according to claim 6, characterized in that In step 2-24, the formula of the reflection mechanism is: ， in, Indicates The first offspring The value on the dimension, It is the remainder method.

8. The method according to claim 7, characterized in that In step 2, after the termination condition is met, the individual with multi-party multi-objective layer number 1 is selected from the population, which is the optimal solution found by the algorithm.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores program codes, and when the program codes are executed by the processor, the processor executes the steps of the method according to any one of claims 1 to 8.

10. A storage medium, characterized in that: A computer program or instruction is stored, and when the computer program or instruction is run on a computer, the steps of the method according to any one of claims 1 to 8 are executed.

Images