A method for optimizing warehouse location allocation and storage path of multi-state material storage system

Abstract

The application discloses a kind of multi-state material storage system's warehouse position allocation and warehousing path coordination optimization method, including establishing warehouse position allocation multi-objective optimization model, model includes warehouse time model, shelf stability model, material correlation model, and fusion material multi-state storage constraint and batch management constraint;Improved grey wolf algorithm is used to solve warehouse position allocation multi-objective optimization model, and the information of single material is input as input parameter, and the optimal warehouse position allocation result is output.The application establishes the warehouse position allocation multi-objective optimization model considering the characteristics of multi-state material, realizes the scientificity and rationalization of material storage;Design the cooperative optimization mechanism of warehouse position allocation and path planning, realize the maximization of overall operation efficiency;It is suitable for the storage management of multi-state, small-batch material such as new product development of hydraulic pump, significantly improves the warehousing operation efficiency and warehousing management level of multi-state material storage system in new product development of hydraulic pump.

Description

Technical Field

[0001] This invention relates to the field of intelligent warehousing and logistics optimization technology, specifically to a method for collaborative optimization of storage location allocation and inbound path in a multi-state material storage system. Background Technology

[0002] In the development and assembly of new hydraulic pumps, materials are characterized by multiple states (such as qualified products, products awaiting inspection, and oil-sealed products), multiple batches, and small quantities. Traditional warehouse management systems struggle to efficiently handle the storage and handling of such complex materials. Existing research often treats location allocation and path planning as two independent problems, lacking overall collaborative optimization, leading to low warehouse operation efficiency and frequent robot path conflicts. Furthermore, existing algorithms are prone to getting trapped in local optima when solving multi-objective optimization problems and are ill-suited to the dynamic storage requirements of materials in multiple states.

[0003] Currently, the storage location allocation problem is typically solved using optimization algorithms such as genetic algorithms and particle swarm optimization. However, these algorithms suffer from slow convergence speed and a tendency to get trapped in local optima when dealing with multi-objective and multi-constraint problems. Path planning problems often employ algorithms such as A* and ant colony optimization, but in multi-robot collaborative environments, path conflicts are prominent, and effective coordination mechanisms are lacking. Summary of the Invention

[0004] The purpose of this invention is to provide a method for coordinating and optimizing storage location allocation and inbound path in a multi-state material storage system, so as to solve the technical problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] S1. Establish a multi-objective optimization model for warehouse location allocation. The model includes an inbound / outbound time model, a shelf stability model, and a material correlation model, and integrates multi-state storage constraints and batch management constraints for materials.

[0007] S2. The improved Grey Wolf algorithm is used to solve the multi-objective optimization model of warehouse location allocation. The material information of the inbound order is used as the input parameter, and the optimal warehouse location allocation result is output. The warehouse location allocation result includes the output material and its corresponding warehouse location coordinates.

[0008] S3. Based on the storage location allocation results, construct a storage system environment topology model;

[0009] S4. Based on the storage system environment topology model, the improved K-means algorithm is used to group all materials in the inbound order and generate material grouping results.

[0010] S5. Based on the material grouping results and the environmental topology model, the improved A* algorithm is used to plan the optimal handling path for each group of materials.

[0011] S6. Introduce a conflict coordination strategy based on a time window model to determine the final transportation path.

[0012] Furthermore, the specific steps of step S1 are as follows:

[0013] S11. Construct an inbound / outbound efficiency model :

[0014]

[0015]

[0016] in, Indicates the types of materials entering and leaving the warehouse. , These represent the distances traveled by the bin robot in the x, y, and z directions during the material p receiving process. This represents the Manhattan distance of material p from the entrance / exit to the storage location. This indicates the outbound frequency of material p, where a is the storage location length, b is the storage location width, and h is the storage location height. , Let p be the storage location coordinates. These refer to the widths of horizontal and vertical tunnels, respectively. This is the default height for the bin robot's robotic arm. This represents the total number of rows on the shelf.

[0017] S12. Construct a shelf stability model :

[0018]

[0019] in, Let p be the mass of material and h be the height of the storage location;

[0020] S13. Construct a material relationship model :

[0021]

[0022] in, Let be the co-occurrence frequency of material p and material q. This refers to the number of times material p and material q appear together in the BOM during the development cycle of a single new product. This refers to the total number of BOMs within a single new product development cycle. Let P be the center coordinate of material p, and S be the set of materials that have been put into storage.

[0023] S14. Construct material batch management constraints:

[0024]

[0025] in, This represents the distance 'p', representing the material in batch 'batch', to the entrance / exit. Its calculation formula is the same as the Manhattan distance in the inbound efficiency model. The calculation formula, To minimize positive numbers;

[0026] S15. Construct multi-state storage constraints for materials:

[0027]

[0028] in, Let N be the decision variable for assigning material p to storage location N. It is the region to which storage location N belongs. Material p belongs to the region; I represents the indicator function, when When, that is, when the two regions are the same, A value of 0 indicates that allocation is allowed; otherwise, 0 indicates that allocation is not allowed. This is the total number of storage locations in the system.

[0029] Furthermore, the specific steps of step S2 are as follows:

[0030] S21. The improved Grey Wolf algorithm uses a Sine-Tent hybrid chaotic mapping to initialize the population.

[0031] S22. Calculate the initial fitness values ​​using a multi-objective weighted fitness function, transforming the requirements of the multi-objective optimization model for storage allocation into quantifiable evaluation criteria. Determine the hierarchy in the population based on the fitness values: α wolf, β wolf, δ wolf, and ω wolf. Among them, α, β, and δ wolves serve as leaders to guide the population's search direction.

[0032] S23. Check if the current iteration count is less than the maximum iteration count. If it is less than the maximum iteration count, perform process iteration, dynamically adjust the search step size through the adaptive nonlinear cosine convergence factor, and update the wolf pack position and the method of adjusting the position to guide ω wolf to approach the high-quality solution of α, β, δ wolf.

[0033] S24. Introduce population diversity assessment and diversity injection strategies, using the element difference degree between individuals or Euclidean distance as the diversity assessment index. When the index is lower than the preset threshold, randomly select some components in the solution vector and replace them with feasible elements in the solution space that have not been fully explored.

[0034] S25. If the current iteration count reaches the predetermined frequency, then perform an elite region search strategy on the neighborhood of α, β, and δ wolves and record the current iteration information; otherwise, continue the iteration process after recording the current iteration information.

[0035] S26. When the number of iterations reaches the preset maximum value, or the rate of change of fitness values ​​of multiple generations of elite schemes is lower than the convergence threshold, the optimization process is terminated, and the optimal storage location allocation result corresponding to α wolf is output.

[0036] Furthermore, the method for constructing the environmental topology model in step S3 is as follows: abstracting the intersection of the storage system lanes into topology nodes and the lanes into edges; establishing the mapping relationship between nodes and storage areas; using an adjacency matrix to represent the connectivity relationship between nodes; mapping the material location onto the edges of the topology graph as material mapping points; and establishing the connectivity relationship between material mapping points and nodes.

[0037] Furthermore, the specific steps of step S4 are as follows:

[0038] S41. Divide all materials in the receiving slip into groups. The formula for dividing the materials into groups is:

[0039]

[0040] in, For material types, The number of groups after division;

[0041] S42. Based on the topology nodes in the storage system environment topology model, randomly select K initial cluster centers. , ... ;

[0042] S43. Following the K-medoids update rule based on path cost, the optimal cluster center is found through multiple iterations. During the process of finding the optimal cluster center through multiple iterations, when the relative rate of change of the total time before and after the iteration is less than a threshold... Stop iteration when the optimal grouping is reached, and output the optimal grouping as the material grouping result. The path cost formula is:

[0043]

[0044] in, This refers to the storage location corresponding to material p. This refers to the storage location corresponding to material q. Let i be the distance between storage locations i and j. For the turning cost weighting coefficient, For material p, the number of path turns in storage locations q. Let p and q be the distance cost between storage locations.

[0045] Furthermore, the specific steps of step S5 are as follows:

[0046] S51. Extract material node information within the material grouping results in sequence, and define the initial open list and close list;

[0047] S52. Starting from the entrance / exit, calculate the path cost of the material, select the node with the minimum cost as the current node, and add it to the close list;

[0048] S53. Select the connected nodes of the current node and determine whether the connected nodes are in the open list. If not, add them to the open list.

[0049] S54. Calculate the path cost of connected nodes, select the node with the lowest cost, move it to the close list, and replace it with the current node;

[0050] S55. Determine if the close list has covered all nodes. If it has covered all nodes, proceed to S56; otherwise, return to S52 until all nodes are covered.

[0051] S56. For the closed list of all covered nodes, start from the last covered target node, backtrack to the initial node through the parent node table, and generate the optimal transport path.

[0052] S57. Execute S51 to S56 until all groups have completed the generation of the optimal transport path.

[0053] Furthermore, the improved A* algorithm is specifically as follows:

[0054] Calculate the actual cost and heuristic cost estimation And based on its definition, improve the path cost function of the A* algorithm. The calculation formulas are as follows:

[0055]

[0056]

[0057]

[0058] in, and These represent nodes numbered n and n-1 respectively, t( , ) is the predecessor node To the current node The actual path cost, Preorder node To the current node Actual steering cost, min( This is the estimated minimum number of turns. This is the cost of turning around.

[0059] Furthermore, the specific steps of step S6 are as follows:

[0060] S61. Define encounter conflicts based on the time window model, specifically as follows:

[0061]

[0062] in, Indicates in the task and The set of encounters and conflicts between them This indicates the task that the bin robot R is responsible for. Occupy slave node to Time window, This indicates the task that the bin robot R is responsible for. Occupy slave node to Time window;

[0063] S62. Define intersection conflicts based on the time window model, specifically as follows:

[0064]

[0065] in, , and Let i, j, and h represent the nodes numbered i, j, and h, respectively. Indicates in the task and The set of intersecting conflicts between them and These represent the tasks performed by the bin robot R. and Middle occupancy node Time window, This indicates the task that the bin robot R is responsible for. Occupy slave node to Time window, This indicates the task that the bin robot R is responsible for. Occupy slave node to Time window;

[0066] S63. Determine the types of conflicts in the path using the time window model, and adopt a waiting strategy for them. Add a waiting time window or avoidance strategy to the path nodes and adjust the path nodes.

[0067] Beneficial effects: This invention establishes a multi-objective optimization model for storage location allocation that comprehensively considers the characteristics of materials in multiple states, thereby realizing the scientific and rational management of material storage; it designs a collaborative optimization mechanism for storage location allocation and path planning, thereby maximizing overall operational efficiency; it is applicable to the storage management of multi-state, small-batch materials such as those used in the development of new hydraulic pumps, and significantly improves the efficiency of warehousing operations and the level of warehouse management for multi-state material storage systems in the development of new hydraulic pumps. Attached Figure Description

[0068] To more clearly illustrate the technical solutions of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0069] Figure 1 This is a flowchart of the collaborative optimization method for warehouse location allocation and inbound path of the present invention;

[0070] Figure 2 A schematic diagram of the topological environment constructed in this invention;

[0071] Figure 3 This is a flowchart of steps S2 to S6 of the present invention;

[0072] Figure 4 This is a flowchart of the improved gray wolf algorithm of the present invention;

[0073] Figure 5 This is a schematic diagram of the material information on the warehouse receipt of the present invention;

[0074] Figure 6 This is a schematic diagram of the optimal storage location allocation result of the present invention;

[0075] Figure 7 This is a schematic diagram illustrating the encounter and intersection conflict scenarios of the bin robots of this invention. Detailed Implementation

[0076] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0077] like Figures 1-7 As shown, the present invention provides a method for collaborative optimization of storage location allocation and inbound path in a multi-state material storage system, the specific steps of which are as follows:

[0078] S1. Establish a multi-objective optimization model for warehouse location allocation. The model includes an inbound / outbound time model, a shelf stability model, and a material correlation model, and integrates multi-state storage constraints and batch management constraints for materials.

[0079] Based on the multiple states of material storage needs, such as oil sealing and time monitoring, and the multiple states of material attributes, the warehouse is strictly divided into several storage locations. The storage location attributes and material attributes are defined by coding. Materials must be stored in the designated storage areas and cannot be stored across areas.

[0080] One storage location can only store one type of bin. The bins are of the same size and can be divided into compartments or not, so that they can hold one type or a batch of materials.

[0081] In this embodiment, the warehousing system has only one entrance / exit. All materials enter and exit the warehouse through this entrance / exit, and are carried by bins. Two bin robots work together to transport these materials within the warehouse. When there are many materials on the inbound order, one bin robot is responsible for carrying empty bins. After the bins are sorted, the other bin robot carries the non-empty bins and delivers them to their designated storage locations.

[0082] The warehouse aisle includes five horizontal aisles between racks, one vertical aisle near the entrance / exit, and one inner vertical aisle (with the same width as the horizontal aisles). At the intersection of the vertical aisle near the entrance / exit and the middle horizontal aisle, there is a material inlet / outlet. The material is transferred to or sent out of the storage system by a roller conveyor.

[0083] In this embodiment, to illustrate the implementation of step S1, the specific steps are as follows:

[0084] S11. Construct an inbound / outbound efficiency model :

[0085]

[0086]

[0087] in, Indicates the types of materials entering and leaving the warehouse. , These represent the distances traveled by the bin robot in the x, y, and z directions during the material p receiving process. This represents the Manhattan distance of material p from the entrance / exit to the storage location. This indicates the outbound frequency of material p, where a is the storage location length, b is the storage location width, and h is the storage location height. , Let p be the storage location coordinates. These refer to the widths of horizontal and vertical tunnels, respectively. This is the default height for the bin robot's robotic arm. This represents the total number of rows on the shelf.

[0088] S12. Construct a shelf stability model :

[0089]

[0090] in, Let p be the mass of material and h be the height of the storage location;

[0091] S13. Construct a material relationship model :

[0092]

[0093] in, Let be the co-occurrence frequency of material p and material q. This refers to the number of times material p and material q appear together in the BOM during the development cycle of a single new product. This refers to the total number of BOMs within a single new product development cycle. Let S be the center coordinates of material p, and S be the set of materials already in storage. In this embodiment, based on actual conditions, materials with a co-occurrence frequency greater than a threshold are considered. Materials that are (custom-defined) are considered as already in the warehouse. .

[0094] S14. Construct material batch management constraints:

[0095] Some assembly materials have batch attributes, which facilitates the management of materials affected by different production times or processing techniques. During the development of new products, in order to prevent damage to the material properties and to distinguish the impact of different batches of materials on the assembly effect, a strict first-in-first-out (FIFO) model is implemented to control materials with batches. For materials with batches, the storage location of materials with larger batches that enter the warehouse later is kept slightly further away from the entrance and exit than that of materials with smaller batches. Combined with regular relocation of materials during non-operational hours, the materials with batches are prevented from moving further and further away from the entrance and exit, thereby achieving dynamic equilibrium.

[0096]

[0097] in, This represents the distance 'p', representing the material in batch 'batch', to the entrance / exit. Its calculation formula is the same as the Manhattan distance in the inbound efficiency model. The calculation formula, To minimize positive numbers.

[0098] S15. Construct multi-state storage constraints for materials:

[0099]

[0100] in, The decision variable for assigning material p to storage location N (1 if the material is assigned to a storage location, 0 otherwise). It is the region to which storage location N belongs. Material p belongs to the region; I represents the indicator function, when When, that is, when the two regions are the same, A value of 0 indicates that allocation is allowed; otherwise, 0 indicates that allocation is not allowed. This refers to the total number of storage locations in the system, ensuring that each type of material must be assigned to one and only one storage location that meets the area requirements, avoiding duplicate or missed assignments.

[0101] S2. The improved Grey Wolf algorithm is used to solve the multi-objective optimization model of warehouse location allocation. The material information of the inbound order is used as the input parameter, and the optimal warehouse location allocation result is output. The warehouse location allocation result includes the output material and its corresponding warehouse location coordinates.

[0102] A certain inbound order contains several materials. After data preprocessing, the following can be obtained: Figure 5 The inbound order material information shown includes name, material code, quantity, batch, and status. The quantity information corresponds to the actual weight of the material. Using the inbound order material information as input parameters, the optimal storage location allocation result is output, such as... Figure 6 As shown.

[0103] In this embodiment, to illustrate the implementation of step S2, the specific steps are as follows:

[0104] S21. The improved Grey Wolf algorithm initializes the population using a Sine-Tent hybrid chaotic mapping. Initialization parameters include population size and maximum number of iterations. By equally weighting the Sine and Tent chaotic sequences, the randomness and solution space traversal capability of the hybrid sequence are enhanced. The algorithm generates [the following parameters] through the two chaotic mappings: The chaotic sequence of numbers is then mapped to the empty storage location number N that meets the conditions, thereby ensuring the rationality and diversity of the initial population. The specific calculation is as follows:

[0105]

[0106]

[0107] in, Let be the value of the nth generation chaotic sequence, with a range of (0,1). This is the Sine chaotic sequence generated in the (n+1)th generation. For the (n+1)th generation, the Tent chaotic sequence Let k be the chaotic value in the set of chaotic sequences. for The location of the chaotic sequence mapping to the storage location number. This represents the number of available storage locations.

[0108] S22. Calculate the initial fitness values ​​using a multi-objective weighted fitness function, transforming the requirements of the multi-objective optimization model for storage allocation into quantifiable evaluation criteria. Determine the hierarchy in the population based on the fitness values: α wolf, β wolf, δ wolf, and ω wolf. Among them, α, β, and δ wolves serve as leaders to guide the population's search direction.

[0109] S23. Check if the current iteration count is less than the maximum iteration count. If it is less than the maximum iteration count, perform process iteration, dynamically adjust the search step size through the adaptive nonlinear cosine convergence factor a, and update the wolf pack position and the adjustment method to guide ω wolves closer to the high-quality solutions of α, β, and δ wolves.

[0110] For the population convergence direction, a nonlinear cosine convergence factor 'a' is dynamically adjusted using a search strategy. Initially, 'a' is larger and decreases rapidly; in the middle stage, it decreases slightly faster; and in the later stage, it decreases smaller and more steadily. This ensures a wider search range in the early stages, enabling a global search, while in the later stages, it focuses on the neighborhood of the optimal solution for a more refined search. The specific calculation is as follows:

[0111]

[0112] in, for Number of iterations This represents the maximum number of iterations.

[0113] S24. To prevent population homogenization, i.e., excessive shrinkage towards the leader's position, which leads to clustering in the neighborhood of local optima and hinders the algorithm's ability to escape local optima, thus affecting the search efficiency for the global optimum, a population diversity assessment and diversity injection strategy is introduced. The diversity assessment index is the degree of element difference between individuals or the Euclidean distance. When the index falls below a preset threshold, some components of the solution vector are randomly selected and replaced with feasible elements in the solution space that have not been fully explored. This breaks population homogenization and forces the algorithm to explore new solution spaces.

[0114] By periodically assessing the similarity among individuals in the population, when the homogeneity exceeds a threshold, a diversity injection operation is performed. This involves chaotically perturbing the reservoir coordinate components of some ω wolves, introducing new solutions to break local convergence. The assessment and diversity injection formulas are as follows:

[0115]

[0116]

[0117] in, The average degree of difference among all individual pairs. For population size, Individuals and individuals , As an indicator function, when an individual and The If not all elements in a dimension are set to 1, then the element is set to 0. For a set of randomly selected replacement positions, , , ' represents the initial value of the original individual to be optimized in the k-th dimension, the random value of the unutilized feasible solution region, and the final value of the new individual in the k-th dimension, respectively.

[0118] S25. If the current iteration count reaches the predetermined frequency, then the elite region search strategy is applied to the neighborhood of α, β, and δ wolves, and the current iteration information is recorded to enhance the local exploitation capability of the algorithm and continue the loop. Otherwise, after recording the current iteration information, the iteration process continues.

[0119] This scheme combines an elite region search strategy to generate new solutions through a fine-grained neighborhood search of the current elite solution. Whether to retain the new solution is determined by whether the fitness has improved. This combination of two strategies enhances the ability to escape local optima while ensuring the algorithm fully explores local optima, thus improving its ability to search for the global optimum. The specific formula is as follows:

[0120]

[0121] in, This is a new solution generated after searching the elite region. For the elite of the current population, Indicates elite solution The Dimensional element swapping operation, Indicates the first The dimension performs a random replacement operation.

[0122] S26. When the number of iterations reaches the preset maximum value, which is set according to the actual situation, or when the fitness value change rate of multiple generations of elite schemes is lower than the convergence threshold, the convergence threshold is also set according to the actual situation, the optimization process is terminated, and the optimal storage location allocation result corresponding to α wolf is output.

[0123] S3. Based on the storage location allocation results, construct a storage system environment topology model;

[0124] In this embodiment, the method for constructing the environmental topology model is as follows: the intersection of the storage system lanes is abstracted as a topology node, and the lanes are abstracted as edges; a mapping relationship between nodes and storage areas is established; an adjacency matrix is ​​used to represent the connectivity relationship between nodes; the material location is mapped to the edge of the topology graph as a material mapping point; and a connectivity relationship between the material mapping point and the node is established.

[0125] This embodiment follows the topology construction principle, setting all aisle intersections as nodes, primarily including all intersections of vertical and horizontal aisles, and designating the aisles themselves as edges. These topology nodes guide the bin robot from the entrance / exit to the target storage location or aisle entrance. Once the bin robot enters the correct aisle, it performs precise navigation using the nodes mapped to the storage location within the aisle. This embodiment has 10 rows of shelves, each with 13 storage locations, four vertical layers, totaling 520 storage locations. After partitioning, the rightmost three storage locations in each row are separated, creating a topology environment as follows: Figure 2 As shown, its connected adjacency matrix C is:

[0126]

[0127] S4. Based on the storage system environment topology model, the improved K-means algorithm is used to group all materials in the inbound order and generate material grouping results.

[0128] The clustering objective of the improved K-means algorithm in this embodiment is to minimize the sum of path costs within groups. The specific steps are as follows:

[0129] S41. Divide all materials in the receiving slip into groups. The formula for dividing the materials into groups is:

[0130]

[0131] in, For material types, This represents the number of groups after the division.

[0132] S42. Based on the topology nodes in the storage system environment topology model, randomly select K initial cluster centers. , ... ;

[0133] S43. Following the K-medoids update rule based on path cost, the optimal cluster center is found through multiple iterations. During the process of finding the optimal cluster center through multiple iterations, when the relative rate of change of the total time before and after the iteration is less than a threshold... The iteration stops when the threshold is reached. The threshold can be determined according to the usage; in this example, 0.5 is used. The optimal group is output as the material grouping result. The path cost formula is:

[0134]

[0135] in, This refers to the storage location corresponding to material p. This refers to the storage location corresponding to material q. Let i be the distance between storage locations i and j. For the turning cost weighting coefficient, For material p, the number of path turns in storage locations q. Let p and q be the distance cost between storage locations.

[0136] After the material storage location is allocated in this embodiment, since the bin robot can carry a maximum of 4 bins at a time, it will first randomly group them to generate two groups similar to [1,2,3,4] and [5], and calculate the path cost for iteration to generate the optimal group, such as [1,2] and [3,4,5].

[0137] S5. Based on the material grouping results and the environmental topology model, the improved A* algorithm is used to plan the optimal handling path for each group of materials.

[0138] The improved A* algorithm in this embodiment is specifically as follows:

[0139] Calculate the actual cost g'( ) and heuristic cost estimation h'( ), and improve the path cost function of the A* algorithm according to its definition. The calculation formulas are as follows:

[0140]

[0141]

[0142]

[0143] in, and These represent nodes numbered n and n-1 respectively, t( , ) is the predecessor node To the current node The actual path cost, Preorder node To the current node Actual steering cost, min( This is the estimated minimum number of turns. This is the cost of turning around.

[0144] In this embodiment, the improved A* algorithm, after determining the groups, plans the transport path for each group in the following order:

[0145] S51. Extract material node information within the material grouping results in sequence, and define the initial open list and close list;

[0146] S52. Starting from the entrance / exit, calculate the path cost of the material, select the node with the minimum cost as the current node, and add it to the close list;

[0147] S53. Select the connected nodes of the current node and determine whether the connected nodes are in the open list. If not, add them to the open list.

[0148] S54. Calculate the path cost of connected nodes, select the node with the lowest cost, move it to the close list, and replace it with the current node;

[0149] S55. Determine if the close list has covered all nodes. If it has covered all nodes, proceed to S56; otherwise, return to S52 until all nodes are covered.

[0150] S56. For the closed list of all covered nodes, start from the last covered target node, backtrack to the initial node through the parent node table, and generate the optimal transport path.

[0151] S57. Execute S51 to S56 until all groups have completed the generation of the optimal transport path.

[0152] This invention obtains the optimal node path in the topological environment by evaluating the turning cost, such as... —> —> —> —> —> —> —> —> .

[0153] S6. Introduce a conflict coordination strategy based on a time window model to determine the final handling path; resolve path conflict issues during multi-robot collaborative operations. Conflicts that may occur between bin robots during collaborative operations can be broadly categorized into two types: encounter conflicts and intersection conflicts.

[0154] The specific conflict strategy introduced in this embodiment is as follows:

[0155] S61. Encounter Conflict: Assume that the bin robots are R1 and R2. Both R1 and R2 have tasks in the same aisle, and the target storage locations are both behind the other. This situation may lead to deadlock or collision of the bin robots. The encounter conflict based on the time window model is defined as follows:

[0156]

[0157] in, Indicates in the task and The set of encounters and conflicts between them This indicates the task that the bin robot R is responsible for. Occupy slave node to Time window, This indicates the task that the bin robot R is responsible for. Occupy slave node to Time window;

[0158] S62. Intersection Conflict: When two bin robots pass through the same node at the same time, an intersection will occur. In particular, if R1 turns in the opposite direction to R2, it will lead to a subsequent encounter conflict. The intersection conflict based on the time window model is defined as follows:

[0159]

[0160] in, , and Let i, j, and h represent the nodes numbered i, j, and h, respectively. Indicates in the task and The set of intersecting conflicts between them and These represent the tasks performed by the bin robot R. and Middle occupancy node Time window, This indicates the task that the bin robot R is responsible for. Occupy slave node to Time window, This indicates the task that the bin robot R is responsible for. Occupy slave node to The time window.

[0161] S63. Determine the types of conflicts in the path using a time window model, and apply waiting strategies to them. Add waiting time windows or avoidance strategies to path nodes, and adjust the path nodes accordingly. —> —> —> —> —> —> —> —> .

[0162] The waiting strategy in this embodiment is as follows:

[0163] Please refer to Figure 7 This scenario involves encounters and intersections between bin robots. In an intersection conflict, if bin robot R1 turns, it will lead to an encounter conflict. Therefore, if the current node of bin robot R1... If the direction of the subsequent node is different (i.e., it needs to turn), only the bin robot R1 turns, and the other bin robot R2 does not need to turn, and the operation priority of the bin robot R1 is low, then the bin robot R1 will wait in place for a short time, so that the other bin robot R2 can enter the conflict node first. After the conflict node is released, the bin robot R1 will continue to move.

[0164] If all operations are turning operations or none are turning operations, determine the maximum value between the sum of the handling completion time and sorting time of one bin robot R1, and the sum of the waiting time and the handling completion time of another bin robot R2. If the latter is larger, the former needs to wait, so as to minimize the waste of waiting time throughout the process.

[0165] In the event of a collision, the time-based decision-making method in the short-term waiting strategy is adopted. Assuming that the robot that needs to wait is R1 (with low task priority), R1 should wait at the nearest free path node, that is, not at any node on the subsequent task path of R2, so as to avoid obstructing R2's movement. Alternatively, the task path can be replanned, and the optimal solution can be selected by judging the time taken by both methods, thereby avoiding path conflicts.

[0166] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0167] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for collaborative optimization of storage location allocation and inbound path in a multi-state material storage system, characterized in that, Includes the following steps: S1. Establish a multi-objective optimization model for warehouse location allocation. The model includes an inbound / outbound time model, a shelf stability model, and a material correlation model, and integrates multi-state storage constraints and batch management constraints for materials. S2. The improved Grey Wolf algorithm is used to solve the multi-objective optimization model of warehouse location allocation. The material information of the inbound order is used as the input parameter, and the optimal warehouse location allocation result is output. The warehouse location allocation result includes the output material and its corresponding warehouse location coordinates. S3. Based on the storage location allocation results, construct a storage system environment topology model; S4. Based on the storage system environment topology model, the improved K-means algorithm is used to group all materials in the inbound order and generate material grouping results. S5. Based on the material grouping results and the environmental topology model, the improved A* algorithm is used to plan the optimal handling path for each group of materials. S6. Introduce a conflict coordination strategy based on a time window model to determine the final transportation path.

2. The method for coordinated optimization of storage location allocation and inbound path according to claim 1, characterized in that, The specific steps of step S1 are as follows: S11. Construct an inbound / outbound efficiency model : in, Indicates the types of materials entering and leaving the warehouse. , These represent the distances traveled by the bin robot in the x, y, and z directions during the material p receiving process. This represents the Manhattan distance of material p from the entrance / exit to the storage location. This indicates the outbound frequency of material p, where a is the storage location length, b is the storage location width, and h is the storage location height. , Let p be the storage location coordinates. These refer to the widths of horizontal and vertical tunnels, respectively. This is the default height for the bin robot's robotic arm. This represents the total number of rows on the shelf. S12. Construct a shelf stability model : in, Let p be the mass of material and h be the height of the storage location; S13. Construct a material relationship model : in, Let be the co-occurrence frequency of material p and material q. This refers to the number of times material p and material q appear together in the BOM during the development cycle of a single new product. This refers to the total number of BOMs within a single new product development cycle. Let P be the center coordinate of material p, and S be the set of materials that have been put into storage. S14. Construct material batch management constraints: in, This represents the distance 'p', representing the material in batch 'batch', to the entrance / exit. Its calculation formula is the same as the Manhattan distance in the inbound efficiency model. The calculation formula, To minimize positive numbers; S15. Construct multi-state storage constraints for materials: in, Let N be the decision variable for assigning material p to storage location N. It is the region to which storage location N belongs. Material p belongs to the region; I represents the indicator function, when When, that is, when the two regions are the same, A value of 0 indicates that allocation is allowed; otherwise, 0 indicates that allocation is not allowed. This is the total number of storage locations in the system.

3. The method for collaborative optimization of storage location allocation and inbound path according to claim 1, characterized in that: The specific steps of step S2 are as follows: S21. The improved Grey Wolf algorithm uses a Sine-Tent hybrid chaotic mapping to initialize the population. S22. Calculate the initial fitness values ​​using a multi-objective weighted fitness function, transforming the requirements of the multi-objective optimization model for storage allocation into quantifiable evaluation criteria. Determine the hierarchy in the population based on the fitness values: α wolf, β wolf, δ wolf, and ω wolf. Among them, α, β, and δ wolves serve as leaders to guide the population's search direction. S23. Check if the current iteration count is less than the maximum iteration count. If it is less than the maximum iteration count, perform process iteration, dynamically adjust the search step size through the adaptive nonlinear cosine convergence factor, and update the wolf pack position and the method of adjusting the position to guide ω wolf to approach the high-quality solution of α, β, δ wolf. S24. Introduce population diversity assessment and diversity injection strategies, using the element difference degree between individuals or Euclidean distance as the diversity assessment index. When the index is lower than the preset threshold, randomly select some components in the solution vector and replace them with feasible elements in the solution space that have not been fully explored. S25. If the current iteration count reaches the predetermined frequency, then perform an elite region search strategy on the neighborhood of α, β, and δ wolves and record the current iteration information; otherwise, continue the iteration process after recording the current iteration information. S26. When the number of iterations reaches the preset maximum value, or the rate of change of fitness values ​​of multiple generations of elite schemes is lower than the convergence threshold, the optimization process is terminated, and the optimal storage location allocation result corresponding to α wolf is output.

4. The method for coordinated optimization of storage location allocation and inbound path according to claim 1, characterized in that: The method for constructing the environmental topology model in step S3 is as follows: abstract the intersection of the storage system lanes as topology nodes, and the lanes as edges; establish the mapping relationship between nodes and storage areas; use an adjacency matrix to represent the connectivity relationship between nodes; and map the material locations onto the edges of the topology graph as material mapping points. And establish the connection relationship between material mapping points and nodes.

5. The method for collaborative optimization of storage location allocation and inbound path according to claim 1, characterized in that: The specific steps of step S4 are as follows: S41. Divide all materials in the receiving slip into groups. The formula for dividing the materials into groups is:

6. Among them, For material types, The number of groups after division; S42. Based on the topology nodes in the storage system environment topology model, randomly select K initial cluster centers. , ... ; S43. Following the K-medoids update rule based on path cost, the optimal cluster center is found through multiple iterations. During the process of finding the optimal cluster center through multiple iterations, when the relative rate of change of the total time before and after the iteration is less than a threshold... Stop iteration when the optimal grouping is reached, and output the optimal grouping as the material grouping result. The path cost formula is: in, This refers to the storage location corresponding to material p. This refers to the storage location corresponding to material q. Let i be the distance between storage locations i and j. For the turning cost weighting coefficient, For material p, the number of path turns in storage locations q. Let p and q be the distance cost between storage locations.

7. The method for collaborative optimization of storage location allocation and inbound path according to claim 1, characterized in that: The specific steps of step S5 are as follows: S51. Extract material node information within the material grouping results in sequence, and define the initial open list and close list; S52. Starting from the entrance / exit, calculate the path cost of the material, select the node with the minimum cost as the current node, and add it to the close list; S53. Select the connected nodes of the current node and determine whether the connected nodes are in the open list. If not, add them to the open list. S54. Calculate the path cost of connected nodes, select the node with the lowest cost, move it to the close list, and replace it with the current node; S55. Determine if the close list has covered all nodes. If it has covered all nodes, proceed to S56; otherwise, return to S52 until all nodes are covered. S56. For the closed list of all covered nodes, start from the last covered target node, backtrack to the initial node through the parent node table, and generate the optimal transport path. S57. Execute S51 to S56 until all groups have completed the generation of the optimal transport path.

8. The method for collaborative optimization of storage location allocation and inbound path according to claim 1, characterized in that: The improved A* algorithm is specifically as follows: Calculate the actual cost and heuristic cost estimation And based on its definition, improve the path cost function of the A* algorithm. The calculation formulas are as follows: in, and These represent nodes numbered n and n-1 respectively, t( , ) is the predecessor node To the current node The actual path cost, Preorder node To the current node Actual steering cost, min( This is the estimated minimum number of turns. This is the cost of turning around.

9. The method for collaborative optimization of storage location allocation and inbound path according to claim 1, characterized in that: The specific steps of step S6 are as follows: S61. Define encounter conflicts based on the time window model, specifically as follows: in, Indicates in the task and The set of encounters and conflicts between them This indicates the task that the bin robot R is responsible for. Occupy slave node to Time window, This indicates the task that the bin robot R is responsible for. Occupy slave node to Time window; S62. Define intersection conflicts based on the time window model, specifically as follows: in, , and Let i, j, and h represent the nodes numbered i, j, and h, respectively. Indicates in the task and The set of intersecting conflicts between them and These represent the tasks performed by the bin robot R. and Middle occupancy node Time window, This indicates the task that the bin robot R is responsible for. Occupy slave node to Time window, This indicates the task that the bin robot R is responsible for. Occupy slave node to Time window; S63. Determine the types of conflicts in the path using the time window model, and adopt a waiting strategy for them. Add a waiting time window or avoidance strategy to the path nodes and adjust the path nodes.

Images