A warehouse picking route optimization method and device

Abstract

The present invention relates to a warehouse picking route optimization method and device, wherein the method comprises: obtaining a list of goods to be picked, and finding the initial position of each goods; using the trajectory sampling points of all goods as decision variables to complete the encoding of individuals, and using the initial position as the first sampling point of the trajectory; in order to achieve multiple conflicting optimization goals such as less total picking time, shorter total route length, and fewer people picking at the same time, a first method is used to optimize the trajectory sampling points, and a trajectory sampling point sequence of each goods is extracted to generate the trajectory of each goods, wherein in the first method, the contribution of each variable group to the problem is quantified into two types of contribution: convergence contribution and diversity contribution, and the variable group is composed of multiple decision variables. Compared with the prior art, the present invention quantifies the contribution of each variable group to the convergence and diversity of the problem, and designs an improvement strategy for each population, which can improve the optimization degree of path planning.

Description

Technical Field

[0001] The present invention relates to the field of warehouse picking path planning, and in particular to a warehouse picking route optimization method and device. Background Art

[0002] Different from the standardized new product sales, each product in the second-hand product sales field is unique, with each product having its own price and condition. Based on this, the warehousing and logistics of large-scale second-hand products are much more difficult than new product e-commerce, especially in the second-hand luxury goods sales field. Due to the high prices of luxury goods, the requirements for the accuracy of product scheduling are higher.

[0003] Based on this, when multiple second-hand luxury goods are sold at the same time, the goods need to be allocated at the same time. At this time, it involves route planning for picking goods from the warehouse. How to successfully move all goods from the storage place to the target packaging station and achieve the total picking time, total route length and the number of people picking at the same time is the goal that technical personnel in this field have been pursuing unremittingly.

[0004] Although there are some multi-objective optimization methods in the prior art, there is a conflict between the total picking time, the total route length, and the number of people picking at the same time. Based on this, the existing general multi-objective optimization methods are often powerless when faced with such multi-objective conflicts.

[0005] Of course, there are also some methods in the prior art that can optimize conflicting multiple objectives. For example, Chinese patent CN116362330A discloses a dual-population co-evolution method and application for decision variable grouping optimization. The method defines CP and DP, and the method includes: S1, randomly initializing populations CP and DP: S2, obtaining the grouping results of convergence variables and diversity variables of existing multi-objective optimization algorithms, and grouping mixed variables; S3, based on link learning technology, dividing convergence variables into independent convergence variable groups: S4, CP mating selection process; S5, DP mating selection process: S6, in the CP evolution process, performing convergence crossover and mutation operations: S7, in the DP evolution process, performing diversity crossover and mutation operations: S8, CP and DP environment selection process: S9, the master-slave paradigm can be used for information compensation for the new population after each iteration of CP and DP.

[0006] However, the above multi-objective optimization method still has the following disadvantages:

[0007] 1. Although the above method refines the classification of mixed variables, it still only divides the variables into multiple categories and cannot quantify the properties of the variables. Therefore, when it is applied to picking multiple second-hand goods at the same time, the properties of the variables cannot be well utilized in the subsequent optimization process.

[0008] 2. Although the above method uses two independently optimized populations to balance convergence and diversity, in terms of the complementary advantages of the two populations (that is, each population uses the information / individuals of the other population to make up for its own shortcomings), since it adopts the master-slave paradigm for compensation, it only improves the quality of the main population, which affects the balance between convergence and diversity to a certain extent. Since multiple second-hand goods are picked at the same time, the starting position of each product is unique, so the result obtained is not optimal. Summary of the invention

[0009] The purpose of the present invention is to provide a warehouse picking route optimization method and device.

[0010] The purpose of the present invention can be achieved by the following technical solutions:

[0011] A warehouse picking route optimization method, comprising:

[0012] Get the list of goods to be picked and find the initial location of each item;

[0013] The trajectory sampling points of all goods are used as decision variables to complete the encoding of individuals, and the initial position is used as the first sampling point of the trajectory;

[0014] In order to achieve multiple conflicting optimization goals such as less total picking time, shorter total route length, and fewer people picking at the same time, the first method is used to optimize the trajectory sampling points, and the trajectory sampling point sequence of each product is extracted to generate the trajectory of each product. In the first method, the contribution of each variable group to the problem is quantified into two types of contribution: convergence contribution and diversity contribution, and the variable group consists of multiple decision variables.

[0015] The trajectory sampling point is a three-dimensional data group, including the product number, the time taken for the same product to go from the previous sampling point to the current sampling point, and the location number.

[0016] The first method specifically includes:

[0017] Step 1): Randomly initialize the population P cv and population P dv , where population P cv The population P is a convergence-oriented optimization population based on the ideal point distance. dv To optimize the population for convergence-oriented optimization based on reference vectors;

[0018] Step 2): Based on the improved decision variable analysis algorithm, the decision variables are divided into independent variable groups according to the interactive properties of the variables, and the contribution of each variable group to the problem is quantified as convergence contribution and diversity contribution;

[0019] Step 3): According to the convergence contribution and diversity contribution, cv and population P dv Allocate the variable group to be optimized for this iteration;

[0020] Step 4): In population P cv During the optimization process, according to the ideal point distance, the population P cv and population P dv In the process, parent individuals are selected with a certain probability to generate offspring populations and carry out environmental selection;

[0021] Step 5): In population P dv During the optimization process, according to the Pareto dominance relationship and the reference vector, from the population P dv and population P cv In the process, parent individuals are selected with a certain probability to generate offspring populations and carry out environmental selection;

[0022] Step 6): In population P cv During the improvement process, the population P is selected according to the angle between individuals cv One of the two most similar individuals in the , and select the population P according to the ideal point distance dv For individuals with poor convergence, exchange them and repeat the above process a certain number of times;

[0023] Step 7): In population P dv During the improvement process, the merged population P cv and population P dv And again, environmental selection is performed;

[0024] Step 8): Repeat steps 3) to 7) until the preset number of iterations is reached to generate an optimal solution set.

[0025] The specific steps based on the improved decision variable analysis algorithm are as follows:

[0026] Step 2-1): Based on the variable interactivity method, the decision variables are divided into multiple decision variable groups.

[0027] Step 2-2): For the variable group V, randomly select an individual from the population;

[0028] Step 2-3): perturb the corresponding variables in the variable group in the individual a certain number of times to generate a perturbed solution population P';

[0029] Step 2-4): Perform non-dominated sorting on the perturbed solution population P' and record the maximum frontier number and the number of non-dominated solutions as the convergence contribution od cv Diversity Contribution dv ;

[0030] Step 2-5): Repeat steps 2-2) to 2-4) until the preset sampling times are reached, and calculate the convergence contribution od cv Diversity Contribution dv The mean of

[0031] Step 2-6): Repeat steps 2-2) to 2-5), selecting a different variable group V each time, until the analysis of each variable group is completed;

[0032] Step 2-7): Use the maximum value of the two types of contributions and the theoretical minimum value 1 to the convergence contribution od cv Diversity Contribution dv Normalize.

[0033] The convergence contribution and diversity contribution are respectively used as the variable groups assigned to the population P in a single iteration. cv and population P dv The probability of optimization.

[0034] The population P cv and population P dv There is information sharing with certain limitations in the optimization process.

[0035] In step 6), the population P cv and population P dv The optimization process includes:

[0036] Step 6-1): According to the ideal point distance, from the population P cv and population P dv Select 2N parent individuals P from * cv With the parent individual P * dv , and with a certain probability, the parent individual P * cv The parent individuals in are replaced by the corresponding parent individuals P * dv individuals in

[0037] Step 6-2): For the parent individual P * cv Perform crossover mutation and environmental selection to obtain the offspring population P' cv ;

[0038] Step 6-3): According to the Pareto dominance relationship and the reference vector, from the population P dv With population P cv Select 2N parent individuals P from * dv With P * cv , and with a certain probability, the parent individual P * dv The parent individuals in are replaced by the corresponding parent individuals P * cv individuals in

[0039] Step 6-4): For the parent individual P * dv Perform crossover mutation and environmental selection to obtain the offspring population P' dv .

[0040] The population P dv With population P cv The offspring convergence population P' after the optimization process is completed cv and the offspring diversity population P' dv Information compensation is performed to improve the population quality, which specifically includes two aspects of information compensation paradigms: facing the offspring convergence population P' cv The paradigm based on individual perspective and ideal distance and the offspring diversity population P' dv The paradigm based on Pareto dominance relationship and reference vector.

[0041] The offspring convergent population P' cv The specific steps of the information compensation paradigm are as follows:

[0042] Step 8-1): Calculate the offspring convergence population P' cv The angle between each pair of individuals in the target space and select one of the two individuals with the smallest angle;

[0043] Step 8-2): Calculate the offspring diversity population P' dv The distance between each individual and the ideal point and select the individual with the smallest distance;

[0044] Step 8-3): Exchange these two individuals;

[0045] Step 8-4): Repeat steps 8-1) to 8-3) until the preset number of exchanges is reached to obtain the improved offspring convergence population P" cv .

[0046] A warehouse picking route optimization device includes a memory, a processor, and a program stored in the memory. When the processor executes the program, the method described above is implemented.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] 1. Different from the existing decision variable analysis-based algorithms that simply divide all decision variables into convergence variables, diversity variables or mixed variables, this application quantifies the contribution of each variable group to the convergence and diversity of the problem, which can be used for subsequent optimization resource allocation and strategy selection, and quantifies the contribution of decision variables to convergence and diversity, providing more information for subsequent optimization strategy selection.

[0049] 2. Different from the existing large-scale multi-objective evolutionary algorithms that sequentially optimize the convergence and diversity of the population in each iteration, the present invention proposes a dual-population co-evolution paradigm, which achieves independent and parallel optimization of convergence and diversity through the complementary evolution of two independent populations and inter-population co-evolution.

[0050] 3. The offspring population information compensation paradigm can utilize the complementary characteristics of the two populations to improve the overall quality of the population and guide the optimization direction.

[0051] 4. The dual-population co-evolution method based on decision variable analysis can be applied to the optimization of the optimal route for warehouse picking on live e-commerce platforms, and has a significant optimization effect on multiple conflicting objectives such as the total picking time, the total route length, and the number of people picking at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 A dual population evolution method based on decision variable analysis of the present invention;

[0053] Figure 2 It is a population distribution diagram of the dual population evolution method based on decision variable analysis in an embodiment of the present invention on the LSMOP8 test problem with 3 targets and 500 variables;

[0054] Figure 3 This is a population distribution diagram of the dual population evolution method based on decision variable analysis in an embodiment of the present invention on the WFG2 test problem with 10 targets and 500 variables. DETAILED DESCRIPTION

[0055] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0056] The present application provides a warehouse picking route optimization device, comprising:

[0057] Get the list of goods to be picked and find the initial location of each item;

[0058] The trajectory sampling points of all goods are used as decision variables to complete the encoding of individuals, and the initial position is used as the first sampling point of the trajectory;

[0059] In order to achieve multiple conflicting optimization goals such as less total picking time, shorter total route length, and fewer people picking at the same time, the first method is used to optimize the trajectory sampling points, and the trajectory sampling point sequence of each product is extracted to generate the trajectory of each product. In the first method, the contribution of each variable group to the problem is quantified into two types of contribution: convergence contribution and diversity contribution, and the variable group consists of multiple decision variables.

[0060] The present application is based on a dual population evolution method based on decision variable analysis. The method adopts an improved decision variable analysis algorithm, which groups variables and quantifies the contribution of each variable group to the problem into two types of contribution: convergence contribution and diversity contribution. The method simultaneously maintains two populations with complementary advantages and co-evolution: the population for convergence optimization (Population for Convergence-related Variable, P cv ) and Population for Diversity-related Variable, P dv ). cv and P dv Each has its own independent evolutionary paradigm: convergence-oriented optimization of population P based on ideal point distance cv and the convergence-oriented optimization population P based on reference vector dv , and dynamically allocate the optimization tasks to two populations according to the two types of contribution; the P cv and P dv Information complementarity among populations is achieved through three co-evolutionary interaction mechanisms: restricted mating choice, information sharing, and information compensation.

[0061] The above improved decision variable analysis algorithm mainly includes two parts: decision variable grouping and decision variable analysis. The specific steps are as follows:

[0062] ① Decision variable grouping: Based on the variable interactivity method, the decision variables are divided into multiple decision variable groups;

[0063] ② Decision variable analysis: Based on the random perturbation technique, the sample population is obtained, and according to the maximum frontier number and the number of non-dominated solutions of the sample population, the convergence contribution of each decision variable group is calculated. cv Diversity Contribution dv .

[0064] The above optimization task is dynamically allocated by dividing the convergence contribution of each decision variable group into cv Diversity Contribution dv As a single iteration, the decision variable group is assigned to P cv and P dv Probabilistic realization of the optimization.

[0065] The above P cv and P dv The following two collaborative interaction mechanisms are used in the evolution process to achieve the sharing and complementarity of evolutionary information:

[0066] ①Information sharing mechanism. cv and P dv During the optimization process, each population selects 2N parent individuals from the two populations based on its own selection strategy, and crosses over between populations with a certain probability, so that the newly generated offspring population has a certain probability of inheriting the advantages of the two parent populations.

[0067] ② Information compensation mechanism. To address the problem of evolutionary information loss caused by decision variable grouping, P cv and P dv After each iteration, the optimized population performs information compensation to improve the population quality. Specifically, it includes two aspects of information compensation paradigm: cv The paradigm based on individual perspective and ideal distance and the P' dv The paradigm based on Pareto dominance relationship and reference vector. Specific embodiments

[0069] Dual population evolution method based on decision variable analysis Figure 1 As shown in the figure, this method adopts an improved decision variable analysis algorithm, which groups variables and quantifies the contribution of each variable group to the problem into two types of contribution: convergence contribution and diversity contribution; this method also maintains two populations with complementary advantages and co-evolution: the population for convergence optimization (Population for Convergence-related Variable, P cv ) and Population for Diversity-related Variable, P dv ). cv and P dv Each has its own independent evolutionary paradigm: convergence-oriented optimization of population P based on ideal point distance cv and the convergence-oriented optimization population P based on reference vector dv , and dynamically allocate the optimization tasks to two populations according to the two types of contribution; the P cv and Pdv The complementation of information between populations is achieved through three co-evolutionary interaction mechanisms: restricted mating selection, information sharing, and information compensation. The specific method includes the following steps:

[0070] Step 1) Randomly initialize the population P cv and P dv ;

[0071] Step 2) Based on the improved decision variable analysis algorithm, the decision variables are divided into independent variable groups according to the interactive properties of the variables, and the contribution of each variable group to the problem is quantified as convergence contribution and diversity contribution;

[0072] Step 3) According to the convergence contribution and diversity contribution, P cv and P dv Allocate the variable group to be optimized for this iteration;

[0073] Step 4) In P cv During the optimization process, according to the ideal point distance from P cv and P dv The parent individuals are selected with a certain probability to generate the offspring population and perform environmental selection, which specifically includes the following steps:

[0074] Step 4-1) According to the ideal point distance, from P cv With P dv Select 2N parent individuals P from * cv With P * dv , and with a certain probability, P * cv The parent individuals in are replaced by the corresponding P * dv individuals in

[0075] Step 4-2) For the parent individual P * cv Perform crossover mutation and environmental selection to obtain the offspring population P' cv ;

[0076] Step 5) In P dv During the optimization process, according to the Pareto dominance relationship and the reference vector, from P dv and P cv The parent individuals are selected with a certain probability to generate the offspring population and perform environmental selection, which specifically includes the following steps:

[0077] Step 5-1) According to the Pareto dominance relationship and the reference vector, from P dv With P cv Select 2N parent individuals P from *dv With P * cv , and with a certain probability, P * dv The parent individuals in are replaced by the corresponding P * cv individuals in

[0078] Step 5-2) For the parent individual P * dv Perform crossover mutation and environmental selection to obtain the offspring population P' dv ;

[0079] Step 6) In P cv During the improvement process, P is selected according to the angle between individuals. cv One of the two most similar individuals in the , and select P according to the ideal point distance dv For individuals with poor convergence, they are exchanged and the above process is repeated a certain number of times, which specifically includes the following steps:

[0080] Step 6-1) Calculate P' cv The angle between each pair of individuals in the target space and select one of the two individuals with the smallest angle;

[0081] Step 6-2) Calculate the offspring diversity population P' dv The distance between each individual and the ideal point and select the individual with the smallest distance;

[0082] Step 6-3) Exchange these two individuals;

[0083] Step 6-4) Repeat step 8-1) - step 8-3) until the preset number of exchanges is reached to obtain the improved offspring convergence population P" cv ;

[0084] Step 7) In P dv During the improvement process, merge P cv and P dv And again, environmental selection is performed;

[0085] Step 8) Repeat steps 3) to 7) until the preset number of iterations is reached to generate an optimal solution set.

[0086] This embodiment simulates and verifies the above-mentioned dual population evolution method based on decision variable analysis on the PlatEMO platform on the Windows 11 system, and independently repeats the experiment 20 times on the DTLZ1, 3, 7, WFG2, 3, 7, 9 and LSMOP1, 5, 8 test problems with 3, 5, 10 targets and 200, 500 variables. Each experiment uses 240,0000 and 680,0000 function calculation times respectively according to the variable scale. The experimental results show that compared with several most advanced multi-objective evolutionary algorithms, this embodiment has achieved 35 optimal IGD indicators on 60 test cases with different settings and different test problems. The experimental results prove the effectiveness of this embodiment. Among them, taking the comparison of NSGA-III, MOEA / DVA and DGEA algorithms as an example, the results shown in Tables 1 and Figure 2-Figure 3 result.

[0087] Table 1 Comparison of IGD mean values ​​of four methods on 24 large-scale test cases

[0088]

[0089] In this embodiment, the specific application process is as follows: first, the total time for picking, the total length of the route, and the number of people picking at the same time are used as the problem optimization objectives to establish a large-scale multi-objective optimization model for optimizing the optimal route for warehouse picking on the live e-commerce platform; then, the number of storage locations is used as the chromosome length, and the personnel number and storage location number are used as the gene position to realize the encoding of the population individuals; finally, P is initialized according to the encoding method cv and P dv And Figure 1 The dual-population evolution process shown in the figure performs restricted crossover selection, information sharing and information compensation between groups until the termination condition of the algorithm is met, and outputs the Pareto optimal solution set with multiple conflicting objectives such as total picking time, total route length and number of people picking at the same time, that is, the optimal route for warehouse picking obtained by the algorithm.

[0090] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning, or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

[0091] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.

Claims

1. A warehouse picking route optimization method, characterized in that: include: Get the list of goods to be picked and find the initial location of each item; The trajectory sampling points of all goods are used as decision variables to complete the encoding of individuals, and the initial position is used as the first sampling point of the trajectory; In order to achieve multiple conflicting optimization goals such as less total picking time, shorter total route length, and fewer people picking at the same time, the first method is used to optimize the trajectory sampling points, and the trajectory sampling point sequence of each product is extracted to generate the trajectory of each product. In the first method, the contribution of each variable group to the problem is quantified into two types of contribution: convergence contribution and diversity contribution, and the variable group consists of multiple decision variables.

2. A warehouse picking route optimization method according to claim 1, characterized in that: The trajectory sampling point is a three-dimensional data group, including the product number, the time taken for the same product to go from the previous sampling point to the current sampling point, and the location number.

3. A warehouse picking route optimization method according to claim 1, characterized in that: The first method specifically includes: Step 1): Randomly initialize the population P cv and population P dv , where population P cv The population P is a convergence-oriented optimization population based on the ideal point distance. dv To optimize the population for convergence-oriented optimization based on reference vectors; Step 2): Based on the improved decision variable analysis algorithm, the decision variables are divided into independent variable groups according to the interactive properties of the variables, and the contribution of each variable group to the problem is quantified as convergence contribution and diversity contribution; Step 3): According to the convergence contribution and diversity contribution, cv and population P dv Allocate the variable group to be optimized for this iteration; Step 4): In population P cv During the optimization process, according to the ideal point distance, the population P cv and population P dv In the process, parent individuals are selected with a certain probability to generate offspring populations and carry out environmental selection; Step 5): In population P dv During the optimization process, according to the Pareto dominance relationship and the reference vector, from the population P dv and population P cv In the process, parent individuals are selected with a certain probability to generate offspring populations and carry out environmental selection; Step 6): In population P cv During the improvement process, the population P is selected according to the angle between individuals cv One of the two most similar individuals in the , and select the population P according to the ideal point distance dv For individuals with poor convergence, exchange them and repeat the above process a certain number of times; Step 7): In population P dv During the improvement process, the merged population P cv and population P dv And again, environmental selection is performed; Step 8): Repeat steps 3) to 7) until the preset number of iterations is reached to generate an optimal solution set.

4. A warehouse picking route optimization method according to claim 3, characterized in that: The specific steps based on the improved decision variable analysis algorithm are as follows: Step 2-1): Based on the variable interactivity method, the decision variables are divided into multiple decision variable groups. Step 2-2): For the variable group V, randomly select an individual from the population; Step 2-3): perturb the corresponding variables in the variable group in the individual a certain number of times to generate a perturbed solution population P'; Step 2-4): Perform non-dominated sorting on the perturbed solution population P' and record the maximum frontier number and the number of non-dominated solutions as the convergence contribution od cv Diversity Contribution dv ; Step 2-5): Repeat steps 2-2) to 2-4) until the preset sampling times are reached, and calculate the convergence contribution od cv Diversity Contribution dv The mean of Step 2-6): Repeat steps 2-2) to 2-5), selecting a different variable group V each time, until the analysis of each variable group is completed; Step 2-7): Use the maximum value of the two types of contributions and the theoretical minimum value 1 to the convergence contribution od cv Diversity Contribution dv Normalize.

5. A warehouse picking route optimization method according to claim 4, characterized in that: The convergence contribution and diversity contribution are respectively used as the variable groups assigned to the population P in a single iteration. cv and population P dv The probability of optimization.

6. A warehouse picking route optimization method according to claim 4, characterized in that: The population P cv and population P dv There is information sharing with certain limitations in the optimization process.

7. A warehouse picking route optimization method according to claim 6, characterized in that: In step 6), the population P cv and population P dv The optimization process includes: Step 6-1): According to the ideal point distance, from the population P cv and population P dv Select 2N parent individuals P from * cv With the parent individual P * dv , and with a certain probability, the parent individual P * cv The parent individuals in are replaced by the corresponding parent individuals P * dv individuals in Step 6-2): For the parent individual P * cv Perform crossover mutation and environmental selection to obtain the offspring population P' cv ; Step 6-3): According to the Pareto dominance relationship and the reference vector, from the population P dv With population P cv Select 2N parent individuals P from * dv With P * cv , and with a certain probability, the parent individual P * dv The parent individuals in are replaced by the corresponding parent individuals P * cv individuals in Step 6-4): For the parent individual P * dv Perform crossover mutation and environmental selection to obtain the offspring population P' dv .

8. A warehouse picking route optimization method according to claim 6, characterized in that: The population P dv With population P cv The offspring convergence population P' after the optimization process is completed cv and the offspring diversity population P' dv Information compensation is performed to improve the population quality, which specifically includes two aspects of information compensation paradigms: facing the offspring convergence population P' cv The paradigm based on individual perspective and ideal distance and the offspring diversity population P' dv The paradigm based on Pareto dominance relationship and reference vector.

9. A warehouse picking route optimization method according to claim 6, characterized in that: The offspring convergent population P' cv The specific steps of the information compensation paradigm are as follows: Step 8-1): Calculate the offspring convergence population P' cv The angle between each pair of individuals in the target space and select one of the two individuals with the smallest angle; Step 8-2): Calculate the offspring diversity population P' dv The distance between each individual and the ideal point and select the individual with the smallest distance; Step 8-3): Exchange these two individuals; Step 8-4): Repeat steps 8-1) to 8-3) until the preset number of exchanges is reached to obtain the improved offspring convergence population P" cv .

10. A warehouse picking route optimization device, comprising a memory, a processor, and a program stored in the memory, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 9 is implemented.

Images