A warehouse storage location-based outbound path control method and device

Abstract

The application relates to a warehouse goods-out path control method and device based on a warehouse goods location, belongs to the technical field of intelligent transportation, and solves the problem that the existing goods-out path control solves the path and shelf optimization problem in a linear superposition mode, which can cause the solution to fall into a local optimum. The method comprises the following steps: calculating the similarity between goods by using the code generated by a historical goods-out batch; combining the scattered goods according to the similarity to generate a goods group; calculating and saving the goods-out path corresponding to each shelf location, and randomly placing the goods group into the shelf location based on the goods-out path; using an improved A* algorithm to plan the AGV path, using an improved heuristic function and a turning correction cost parameter to improve the total efficiency of the warehouse logistics system, calculating the fitness function value under different goods-out schemes, searching for an optimal scheme in a continuous iteration process, and determining the goods-out path of the AGV while determining the placement of the goods group in the warehouse. The dynamic updating algorithm is used to generate an optimized transportation scheme considering the factors of the goods location placement.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, and in particular to a method and apparatus for outbound path control based on warehouse storage locations. Background Technology

[0002] The advent of the intelligent era is spurring the logistics industry towards intelligent development. Warehousing operations, as a key link in logistics, directly affect the overall logistics level and operational efficiency. Automated Guided Vehicles (AGVs) are a core component of intelligent transportation systems, possessing significant research value and application prospects in both commercial and military applications, attracting considerable attention from researchers.

[0003] The effective operation of AGVs in smart warehousing relies heavily on path planning, especially multi-AGV path planning. This requires considering constraints such as time and distance to avoid conflicts between AGVs and with obstacles. Multi-AGV path planning has become a key research area in the field.

[0004] Smart warehousing optimization generally consists of two parts: rack optimization and path optimization. Rack optimization addresses the relationship between goods and racks, optimizing the placement of goods, while path optimization primarily seeks the optimal path planning for AGVs. Currently, most smart warehousing optimization studies only address these two parts independently. In practical warehousing applications, this can only be addressed through linear superposition, leading to the problem easily getting trapped in local optima. Summary of the Invention

[0005] Based on the above analysis, the present invention aims to provide a method and apparatus for outbound path control based on warehouse location, in order to solve the problem that existing outbound path control methods, which solve path and shelf optimization problems in a linear superposition manner, are prone to getting trapped in local optima.

[0006] On one hand, embodiments of the present invention provide an outbound path control method based on warehouse locations, comprising: encoding scattered goods to be outbound according to historical shipment batches, and calculating the similarity between goods using the codes generated by the historical shipment batches; combining the scattered goods according to the similarity to generate goods groups, wherein each goods group in the waiting-to-be-inbound state is regarded as a shelf unit, and the shelf unit corresponds one-to-one with the shelf location; calculating and storing the outbound path corresponding to each shelf location, and placing the goods group into the shelf location based on the outbound path; and using an improved A* algorithm for AGV path planning, using an improved heuristic function and turning correction cost parameter to improve the overall efficiency of the warehousing and logistics system, calculating the fitness function value under different outbound schemes, searching for the optimal scheme in the process of continuous iteration, and determining the inbound placement of the goods group and the outbound path of the AGV according to the optimal scheme.

[0007] The beneficial effects of the above technical solution are as follows: The A* algorithm has high path search efficiency and accuracy. Combined with reasonable cargo placement, it provides assistance for outbound route planning. The dynamic update algorithm generates an optimized transportation plan that takes into account cargo placement factors.

[0008] Further improvements to the above method include calculating the fitness function value under different outbound schemes using the following formula. α , β :

[0009] f ( x )= αf path + βf other ,

[0010] in, f ( x Let be the overall objective function for collaborative optimization. x Represents any node, f path Total time spent on outbound processing for all tasks. f other For the overhead of parts other than the AGV path planning, α and β These are the influence coefficients, and α + β =1; When the number of tasks M is greater than the total number of AGVs Q, it takes a certain amount of time for the AGV to return from the target node to the starting node. The total time spent on all tasks leaving the warehouse is calculated using the following formula:

[0011] ,

[0012] When the total number of AGVs Q is greater than or equal to the number of tasks M, one AGV is allocated to each task, and the total time spent on outbound operations for all tasks is calculated using the following formula:

[0013] ,

[0014] ,

[0015] in, µ This is the penalty coefficient for the AGV to return from the target node to the starting node. f m,q The time spent by a vehicle during a particular dispatch, of which ( i , j ) is the target node. f c The waiting time after traffic congestion occurs. These are dynamic parameters. The estimated path distance from the starting node to the target node via the current node n. f * (n) / v Proportional v It's the AGV speed. f i It is the AGV that has arrived. i Theoretical time required for the point f j It is the AGV that has arrived. j Theoretical time required for the point f I-i It is the AGV that has arrived. i The waiting time required when encountering obstacles during the process. f J-j It is the AGV that has arrived. j The waiting time required when encountering obstacles during the process.

[0016] Further improvements to the above method, using an improved heuristic function and a turning correction cost parameter to improve the overall efficiency of a warehousing and logistics system, include: using a weighted Manhattan distance as the improved heuristic function, making the distance estimation cost close to the shortest distance to reduce the number of traversed nodes; and introducing the turning correction cost parameter into the improved heuristic function to reduce the number of path turns.

[0017] Further improvements to the above method, using weighted Manhattan distance as the improved heuristic function, to make the distance estimation cost approach the shortest distance to reduce the number of traversed nodes, include: using the following weighted Manhattan distance for A * The algorithm's h * (n) Improvements:

[0018]

[0019] Among them, by selecting parameters λ 1 and λ 2. Improve A * In algorithmic path planning, path search efficiency is 0 < λ 1<1, 0< λ 2<1,( x n , y n ) is the current node, and ( x end , y end ) is the target node.

[0020] A further improvement to the above method, introducing the turning correction cost parameter into the improved heuristic function to reduce the number of path turns, includes: through the current node ( x n , y n ), parent node ( x n-1 , y n-1 ), child nodes x n+1 , y n+1 ) and the target node ( x end , y end Based on the positional relationship of the AGV, the following formula is used to determine whether the AGV has changed direction, where, when m 1 and m When 2 are equal, the AGV is in a straight-line driving state; and when m 1 and m When the 2 values ​​are not equal, the AGV changes direction:

[0021] ,

[0022] Based on the above judgment results, the heuristic function is modified as follows:

[0023]

[0024] in, φ The turning correction cost parameter ensures that the AGV prioritizes straight-line travel and 0 < φ <1.

[0025] Further improvements to the above method involve encoding scattered goods based on historical shipment batches, and calculating the similarity between goods using the codes generated from these historical shipment batches. This includes: processing warehouse outbound data to remove irrelevant data and extracting the product number and batch number; using 1 and 0 to indicate whether the product number was shipped in the current batch, where each product is a vector composed of batch numbers in each dimension; and calculating the similarity between the vectors using the following cosine similarity formula:

[0026] ;

[0027] The absolute value of the similarity between vectors minus 1 is taken as the similarity between goods. The smaller the absolute value, the higher the similarity between the goods. x Let be a vector of a type of goods. y Let be a vector of another type of goods. x i For a type of goods shipped in the i-th batch... y i For another type of goods shipped in the i-th batch, n For vectors x and y The dimension of.

[0028] Further improvements to the above method, using the improved A* algorithm for AGV path planning, include: modeling the AGV's working environment using a grid map method; inputting the AGV's starting node and target node; creating an Openlist linked list for storing nodes to be detected and a Closelist linked list for storing detected nodes; placing the coordinates of the AGV's starting node into the Openlist linked list; traversing all nodes in the Openlist linked list, finding the node with the smallest improved cost function value and setting it as the current node; moving the current node into the Closelist linked list; detecting the current node's four neighboring nodes, ignoring nodes already in the Closelist or obstacle nodes, and then determining whether the remaining neighboring nodes are in the Openlist linked list; when the remaining neighboring nodes are in the Openlist linked list, determining whether the AGV has passed the current node's calculated cost function value. g ( n Is the value smaller, where if the g ( n If the value is smaller, then the parent node of the remaining neighboring nodes is set as the current node, and the calculation is recalculated. g ( n ) value and f * ( n ) value, where,f * ( n () represents the process from the starting node through the current node. n The estimated path distance to the target node; g ( n () represents the distance from the starting node to the current node n The actual path distance; and whether the target node of the AGV has been found, wherein when the target node is found, the parent node of the AGV is sequentially searched from the target node to the starting node of the AGV, and the obtained path is the AGV driving path based on the improved A* algorithm.

[0029] Based on the further improvement of the above method, when the target node of the AGV is not found, it is further determined whether the Openlist is an empty list. When the Openlist is not an empty list, the following steps are returned: traverse all nodes in the Openlist until the Openlist is an empty list.

[0030] Based on a further improvement to the above method, when the remaining neighboring nodes are not in the Openlist linked list, the remaining neighboring nodes are added to the Openlist linked list; the current node is set as its parent node, and the values ​​of each remaining neighboring node are calculated respectively. f * ( n )value, g ( n ) value and h * ( n ) value, where, h * ( n ) is from the current node n The estimated path distance to the target node; and the next step: determining whether the target node of the AGV has been found.

[0031] On the other hand, embodiments of the present invention provide an outbound path control device based on warehouse locations, comprising: a similarity acquisition module, used to encode scattered goods to be outbound according to historical shipment batches, and calculate the similarity between goods using the codes generated by the historical shipment batches; a goods group generation module, used to combine the scattered goods according to the similarity to generate goods groups, wherein each goods group in the waiting-to-be-inbound state is regarded as a shelf unit, and the shelf unit corresponds one-to-one with the shelf position; a goods group setting module, used to calculate and save the outbound path corresponding to each shelf position, and place the goods group into the shelf position based on the outbound path; and an AGV path planning module, used to perform AGV path planning using an improved A* algorithm, improve the overall efficiency of the warehousing and logistics system using an improved heuristic function and turning correction cost parameters, calculate the fitness function value under different outbound schemes, search for the optimal scheme in the process of continuous iteration, and determine the inbound placement of goods groups and the outbound path of the AGV according to the optimal scheme.

[0032] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0033] 1. The A* algorithm has high path search efficiency and accuracy. Combined with reasonable cargo placement, it provides assistance for outbound route planning. The dynamic update algorithm generates an optimized transportation plan that takes into account cargo placement factors.

[0034] 2. By considering location planning and path planning together and combining optimal path function analysis, the shelves with high similarity are dispersed, making it easier for AGVs to find goods quickly at any station. High-frequency goods are placed in places where paths do not overlap, which can ensure that when there are many tasks, AGVs will not be blocked due to overlapping paths, thereby improving the overall efficiency of warehouse inbound and outbound operations.

[0035] 3. Improved cost estimation function f * ( n Cost estimation function compared to the previous one f ( n For ), it mainly concerns its heuristic function. h ( n The algorithm has been improved by using a weighted Manhattan distance as a heuristic function, making the distance estimation cost closer to the shortest distance and reducing the number of nodes traversed. Additionally, a turning correction cost parameter has been introduced into the heuristic function to reduce the number of path turns.

[0036] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0037] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0038] Figure 1 is a flowchart of the outbound path control method based on warehouse location according to an embodiment of the present invention;

[0039] Figure 2 A flowchart of the product similarity algorithm according to an embodiment of the present invention;

[0040] Figure 3 This is a schematic diagram of node positions according to an embodiment of the present invention;

[0041] Figure 4 This is a comparison diagram of three paths according to an embodiment of the present invention;

[0042] Figure 5 A flowchart illustrating AGV path planning based on the improved A* algorithm according to an embodiment of the present invention; and

[0043] Figure 6 This is a block diagram of an outbound path control device based on warehouse location according to an embodiment of the present invention. Detailed Implementation

[0044] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0045] refer to Figure 1A specific embodiment of the present invention discloses an outbound path control method based on warehouse location, comprising: in step S102, encoding the scattered goods to be outbound according to historical shipment batches, and calculating the similarity between goods using the codes generated by historical shipment batches; in step S104, combining the scattered goods according to the similarity to generate goods groups, wherein each goods group in the waiting-to-be-inbound state is regarded as a shelf unit, and the shelf unit corresponds one-to-one with the shelf position; in step S106, calculating and saving the outbound path corresponding to each shelf position, and randomly placing the goods group into the shelf position based on the outbound path; and in step S108, using the improved A* algorithm for AGV path planning, using the improved heuristic function and turning correction cost parameter to improve the overall efficiency of the warehousing and logistics system, calculating the fitness function value under different outbound schemes, searching for the optimal scheme in the process of continuous iteration, and determining the inbound placement of goods groups and the outbound path of AGVs according to the optimal scheme.

[0046] Compared with existing technologies, the outbound path control method based on warehouse location provided in this embodiment has higher path search efficiency and accuracy. Combined with reasonable warehouse location placement, it provides assistance for outbound path planning. A dynamic update algorithm is used to generate an optimized transportation plan that takes into account warehouse location factors.

[0047] The following text will refer to Figure 1 The steps of the outbound path control method based on warehouse location according to embodiments of the present invention are described in detail.

[0048] In step S102, the scattered goods to be shipped are coded according to historical shipment batches, and the similarity between goods is calculated using the codes generated from historical shipment batches. Specifically, coding the scattered goods according to historical shipment batches and calculating the similarity between goods using the codes generated from historical shipment batches includes: processing the warehouse goods outbound data to remove irrelevant data and extracting the product number and product batch; using 1 and 0 to indicate whether the product number is shipped in this batch, where each product is a vector composed of batch numbers in each dimension; and calculating the similarity between vectors using the following cosine similarity formula:

[0049] ;

[0050] The absolute value of the similarity between vectors minus 1 is used as the similarity between items. The smaller the absolute value, the higher the similarity between the items. x Let be a vector of a type of goods. y Let be a vector of another type of goods. x i For a type of goods shipped in the i-th batch... yi For another type of goods shipped in the i-th batch, n For vectors x and y dimensionality x , y It can be 0 or 1, where 0 indicates that the goods have not been shipped and 1 indicates that the goods have been shipped.

[0051] In step S104, scattered goods are grouped according to similarity to generate product groups. Each product group in the pending warehousing state is considered a shelf unit, and each shelf unit corresponds one-to-one with a shelf location. Specifically, grouping scattered goods according to similarity to generate product groups includes: obtaining an upper triangular matrix A of similarity between goods based on the similarity between goods, and calculating a product group based on the values ​​of the upper triangular matrix, where 2 represents the product category, 5 represents the maximum number of shipment batches, 1 represents shipment, and 0 represents no shipment.

[0052] .

[0053] In step S106, the outbound path corresponding to each shelf location is calculated and saved, and the product groups are randomly placed into shelf locations based on the outbound path. Specifically, by considering location planning and path planning together, and combining optimal path function analysis, shelves with high similarity are dispersed, making it easier for AGVs to quickly find goods at any station. Furthermore, frequently used goods are placed in locations where paths do not overlap, ensuring that AGV congestion is not caused by overlapping paths when there are many tasks, thus improving the overall warehouse inbound and outbound efficiency.

[0054] In step S108, the improved A* algorithm is used for AGV path planning. The improved heuristic function and turning correction cost parameter are used to improve the overall efficiency of the warehousing and logistics system. The fitness function value under different outbound schemes is calculated. The optimal scheme is searched in the process of continuous iteration. The outbound path of the AGV is determined at the same time as the placement of goods group in the warehouse based on the optimal scheme.

[0055] Specifically, AGV path planning using the improved A* algorithm includes: modeling the AGV's working environment using a grid map method; inputting the AGV's starting node and target node; creating an Openlist linked list to store nodes to be detected and a Closelist linked list to store detected nodes; adding the coordinates of the AGV's starting node to the Openlist linked list; traversing all nodes in the Openlist linked list, finding the node with the smallest improved cost function value and setting it as the current node; moving the current node into the Closelist linked list; detecting the current node's four neighboring nodes, ignoring nodes already in the Closelist or obstacle nodes, and then determining whether the remaining neighboring nodes are in the Openlist linked list; when the remaining neighboring nodes are in the Openlist linked list, determining whether the AGV has passed the current node and calculated the cost function value. g ( n Is the value smaller, where if g ( n If the value is smaller, set the parent node of the remaining neighboring nodes as the current node and recalculate. g ( n ) value and f * ( n ) value; when the remaining neighboring nodes are not in the Openlist linked list, add the remaining neighboring nodes to the Openlist linked list; set the current node as its parent node, and calculate the value of each remaining neighboring node. f * ( n )value, g ( n ) value and h * ( n ) value, where, f * ( n () represents the process from the starting node through the current node. n The estimated path distance to the target node; g ( n () represents the distance from the starting node to the current node n The actual path distance; h * ( n ) is from the current node nThe estimated path distance to the target node is calculated, and the process proceeds to the following steps: determining whether the target node of the AGV has been found. If the target node is found, the path from the target node to the starting node of the AGV is sequentially searched, and the obtained path is the AGV travel path based on the improved A* algorithm. If the target node is not found, the process continues to check if the Openlist is empty. If the Openlist is not empty, the process returns to the following steps: traversing all nodes in the Openlist until the Openlist is empty.

[0056] Improving the overall efficiency of a warehousing and logistics system by using an improved heuristic function and a turning cost correction parameter includes: using a weighted Manhattan distance as an improved heuristic function to make the distance estimation cost close to the shortest distance to reduce the number of traversed nodes; and introducing a turning cost correction parameter into the improved heuristic function to reduce the number of path turns.

[0057] Using weighted Manhattan distance as an improved heuristic function to make the distance estimation cost approach the shortest distance and reduce the number of traversed nodes includes: using the following weighted Manhattan distance for A * Improve the algorithm h(n): , where by selecting parameters λ 1 and λ 2. Improve A * In algorithmic path planning, path search efficiency is 0 < λ 1<1, 0< λ 2<1,( x n , y n ) is the current node, and ( x end , y end ) is the target node.

[0058] Introducing a turning correction cost parameter into the improved heuristic function to reduce the number of path turns includes: via the current node ( x n , y n ), parent node ( x n-1 , y n-1 ), child nodes x n+1 , y n+1 ) and target node ( x end , y endBased on the positional relationship, the following formula is used to determine whether the AGV has turned, where the AGV is in a straight-line driving state when m1 and m2 are equal; and when m1 and m2 are equal... m 1 and m When the 2 values ​​are not equal, the AGV will change direction:

[0059] ,

[0060] Based on the above judgment results, the heuristic function is modified as follows:

[0061]

[0062] in, φ The turning correction cost parameter ensures that the AGV prioritizes straight-line travel and 0 < φ <1.

[0063] Calculating the fitness function value under different outbound schemes involves using the following formula to calculate the fitness function value. α and β : f ( x )= αf path + βf other ,in, f ( x Let be the overall objective function for collaborative optimization. x Represents any node, f path Total time spent on outbound processing for all tasks. f other For the overhead of parts other than AGV path planning, α and β These are the influence coefficients, and α + β =1; When the number of tasks M is greater than the total number of AGVs Q, it takes a certain amount of time for the AGVs to return from the endpoint to the starting point. The total time spent on all tasks leaving the warehouse is calculated using the following formula:

[0064] ,

[0065] When the total number of AGVs Q is greater than or equal to the number of tasks M, one AGV is assigned to each task, and the total time spent on outbound operations for all tasks is calculated using the following formula:

[0066] ,

[0067]

[0068] in, µ This is the penalty coefficient for the AGV to return from the endpoint to the starting point.f m,q The time spent by a vehicle during a particular dispatch, of which ( i , j ) is the target point. f c The waiting time after traffic congestion occurs. For dynamic parameters, dynamic parameters The estimated path distance from the starting node to the target node via the current node n. f * ( n ) / v Proportional v It's the AGV speed. f i It is the AGV that has arrived. i Theoretical time required for the point f j It is the AGV that has arrived. j Theoretical time required for the point f I-i It is the AGV that has arrived. i The waiting time required when encountering obstacles during the process. f J-j It is the AGV that has arrived. j The waiting time required when encountering obstacles during the process.

[0069] refer to Figure 6 A specific embodiment of the present invention discloses an outbound path control device based on warehouse storage locations, comprising: a similarity acquisition module 602, used to encode scattered goods according to historical shipment batches and calculate the similarity between goods using the codes generated from historical shipment batches; a goods group generation module 604, used to combine scattered goods according to similarity to generate goods groups, wherein each goods group in the waiting-to-be-inbound state is regarded as a shelf unit, and the shelf unit corresponds one-to-one with the shelf position; a goods group setting module 606, used to calculate and save the outbound path corresponding to each shelf position, and randomly place the goods group into the shelf position based on the outbound path; and an AGV path planning module 608, used to perform AGV path planning using an improved A* algorithm, improve the overall efficiency of the warehousing and logistics system using an improved heuristic function and turning correction cost parameters, calculate the fitness function value under different outbound schemes, search for the optimal scheme in the process of continuous iteration, and determine the inbound placement of goods groups and the outbound path of AGVs based on the optimal scheme.

[0070] The following text will refer to Figures 2 to 5 The outbound path control method based on warehouse location according to embodiments of the present invention will be described in detail with specific examples.

[0071] Specific examples of this application can be achieved through the following technical means:

[0072] Inspired by the concept of co-evolution, this paper analyzes the interrelationships between goods, shelves, and AGV path planning, and proposes a co-evolutionary algorithm for shelf planning and path planning. This algorithm achieves collaborative optimization of shelf planning and AGV ergonomic planning. The co-optimization algorithm is based on A... * The algorithm (A*(A-Star) algorithm is the most efficient direct search method for finding the shortest path in a static road network, and it is also an effective algorithm for solving many search problems) is improved. Its invention lies in considering the location planning and path planning together, and combining the optimal path function analysis to disperse the shelves with high similarity, so that AGVs can find the goods quickly at any station. In addition, high-frequency goods are placed in places where the paths do not overlap. This can ensure that when there are many tasks, AGVs will not be blocked due to path overlap, thus improving the overall efficiency of warehouse inbound and outbound operations.

[0073] Specifically, A * The algorithm has high path search efficiency and accuracy. Combined with reasonable cargo placement, it provides assistance for outbound route planning. The dynamic update algorithm generates an optimized transportation plan that takes into account cargo placement factors.

[0074] The collaborative optimization algorithm first encodes scattered goods based on historical shipment batches. It then calculates the similarity between goods using the codes generated from the batches, and combines the scattered goods into product groups. Each product group is considered a shelf unit, awaiting warehousing, and corresponds one-to-one with a storage location. Next, before storing the product groups, the algorithm calculates the corresponding outbound path for each storage location, recording and saving the outbound path corresponding to the shelf location. Based on the above calculation results, the algorithm enters the collaborative optimization module, where the order in which product groups are placed into storage locations is free. This random placement scheme constitutes the storage location factor in the collaborative optimization process. Different placement schemes of product groups will generate different transportation routes when a shipment task arrives; these uncertain transportation routes are the path factor in the optimization process. Finally, the improved A... * The algorithm processes multiple AGV path planning algorithms, using improved heuristic functions and turning correction cost parameters to improve the overall efficiency of the warehousing and logistics system. It calculates the fitness function value under different outbound schemes, searches for the optimal scheme in the process of continuous iteration, and determines the outbound path at the same time as determining the placement of goods groups in the warehouse.

[0075] 1. Collaborative optimization mathematical model

[0076] By analyzing the coupling relationships among various components in the intelligent warehousing process, a mathematical model for collaborative optimization of storage locations is proposed. This model differs from traditional intelligent warehousing optimization algorithms in that it integrates path planning and shelf optimization into a unified whole, expressing their relationship using a mathematical formula. Specific variables and constraints are described below, where f(X) is the overall objective function for collaborative optimization. f path Total time spent on outbound processing for all tasks. f other For the overhead of the algorithm other than AGV vehicle path planning, α and β These are the influence coefficients, and α + β =1. N is the total number of AGV vehicles. n i This is the current AGV vehicle number. i =1,2…N, i ∈N. M is the total number of shipment tasks. m i This is the current shipment task number. i =l,2…M, i ∈N. G is the total number of goods. g i For the current item number, i =1,2…G, i ∈N. i , j () indicates the current coordinate position. i , j ∈N, i <=1, j <=JS represents the total number of shelves. µ (0< µ <1) is the penalty coefficient for a vehicle returning from the endpoint to the starting point. The specific variables and their constraints are described below. f ( x ) is the overall objective function for collaborative optimization (where x (represents any node) f path Total time spent on outbound shipments for all tasks ( f path (Calculate using formulas 4 and 10 together). f other For the overhead of the algorithm other than AGV vehicle path planning, α and β These are the influence coefficients, and α + β =1. N is the total number of AGV vehicles. n i This is the current AGV vehicle number.i =1,2…N, i ∈N. M is the total number of shipment tasks. m i This is the current shipment task number. i =l,2…M, i ∈N. G is the total number of goods. g i For the current item number, i =1,2…G, i ∈N. i , j () indicates the current coordinate position. i , j ∈N, i <=1, j <=JS represents the total number of shelves. µ (0< µ <1) is the penalty coefficient for vehicles returning from the end point to the starting point.

[0077] Specifically, the overall objective of the collaborative optimization in this paper is expressed as Equation 1:

[0078] f ( X )= αf path + βf other , α + β= 1Formula 1

[0079] Among them, α and β are influence coefficients. The α coefficient is the weight of "shortest path", which focuses on finding the shortest route by the algorithm. The β system is the weight of "additional overhead", which focuses on the additional time consumption such as collisions, conflicts, and turns. That is, it focuses on ensuring that the transportation paths selected by multiple AGVs do not overlap with each other as much as possible to avoid conflicts.

[0080] When M > Q, that is, when the number of tasks exceeds the total number of vehicles, it takes a certain amount of time for a vehicle to return from the destination to the starting point, as shown in Formula 2:

[0081] Formula 2

[0082] When M <= Q, that is, when the number of vehicles is greater than the number of tasks, one AGV vehicle is assigned to each task, as shown in Formula 3:

[0083] Formula 3

[0084] f m,q The time spent by a vehicle during a particular dispatch, of which ( i , j ) is the target point. f cThe time waited after the vehicle is blocked is a dynamic parameter. The dynamic parameter is proportional to the estimated path distance from the starting node through the current node n to the target node f * ( n ) / v is directly proportional to v where is the vehicle speed of the AGV, f i is the theoretical time required for the AGV to reach i the point f j is the theoretical time required for the AGV to reach j the point f I-i is the waiting time required when the AGV encounters an obstacle during the process of reaching i the point f J-j is the waiting time required when the AGV encounters an obstacle during the process of reaching j the point. See Formula 4 for details

[0085] Formula 4<00管理系统及方法,其特征在于,包括:管理系统,用于管理系统资源;资源管理方法,用于管理系统资源。

[0086] f other It mainly includes two parts. One is the calculation time of the relevance between items f gi,gj which is used for item classification, and the other is the consumption time of the shelf optimization algorithm f S where i +1 < G, see Formula 5:

[0087] Formula 5

[0088] Next, we will describe in detail the implementation idea of the collaborative optimization algorithm for shelf planning and AGV path planning proposed in this paper from three parts: the item similarity algorithm, the multi-AGV path planning algorithm, and the collaborative optimization algorithm for shelf planning and AGV path planning. The item similarity algorithm and the multi-AGV path planning algorithm will provide support for the final collaborative optimization algorithm for shelf path.

[0089] 2. Item Similarity Algorithm

[0090] This paper processes the outbound data of warehouse goods in actual operation and maintenance. Irrelevant data such as operation time and product name are removed, and data such as product number (i.e., the unique identifier of the product) and batch (the same batch of goods can be understood as goods entering and leaving the warehouse within the same time period) are extracted for data analysis. 0 and 1 are used to indicate whether a product with a given number is shipped in that batch. Each product is a vector composed of various dimensions of the batch number. The similarity between these vectors is calculated using a cosine similarity algorithm to obtain the product similarity. Specifically, the specific process of the product similarity calculation method based on cosine similarity is as follows: Figure 2 As shown.

[0091] First, calculate the maximum number of shipment batches based on warehouse data. Record the shipment information of the goods to form a vector table. For example, the vector value of goods number 2 is (1,0,1,1,0), where 1 represents shipment and 0 represents no shipment. The order represents the number of shipments, with (1,0,1,1,0) representing shipments in batches 1, 3, and 4 respectively. Other goods follow the same pattern. Accumulate the total number of shipments to calculate the shipment frequency. Calculate the cosine value between two goods based on the cosine formula. Since the cosine of a goods and itself is 1, it indicates that the two goods are most similar. Therefore, subtract 1 from the cosine value of the goods and take its absolute value. The smaller the difference from 1, the higher the similarity. Use this new value to represent the similarity between the two goods, ultimately obtaining an upper triangular matrix of goods similarity. Based on the values ​​of this matrix, a goods group can be calculated. For example, if the triangular matrices of both product groups are (1,0,1,1,0), then in Formula 6, n=5, x1=1, x2=0, x3=1, x4=1, x5=0; y1=1, y2=0, y3=1, y4=1, y5=0:

[0092] If the two product groups are exactly the same, then the two product groups are identical.

[0093] Formula 6

[0094] 3. Based on the improved A * AGV path planning algorithm

[0095] Based on traditional A * The AGV path planning algorithm has advantages such as high path search efficiency and simple implementation during the path search process. However, the path search efficiency of AGVs is greatly affected by the heuristic function. For traditional AGVs... * The heuristic function in the algorithm is improved in the following ways.

[0096] (1) Improvement of heuristic functions

[0097] When A * When the heuristic function h(n) of the algorithm is calculated using the Manhattan distance,h ( n The closer the value is to the shortest path distance from the current node to the target node, the higher the algorithm's search efficiency. However, in reality, this often leads to... h ( n This refers to situations where the distance between the current node and the target node differs significantly. In actual AGV operation, this is used... AL x , AL y ) indicates the current node ( x n , y n ) to target node ( x end , y end Actual path distance, A * When the algorithm searches for a path, h ( n | will appear in ) x n - x end |< AL x or | y n - y end |< AL y ,at this time h ( n The distance is much shorter than the actual path distance, A. * The algorithm searches for a large number of nodes, thus reducing A. * Algorithm search path efficiency.

[0098] For traditional A * Algorithms commonly suffer from the problem of traversing a large number of nodes in AGV path planning. This patent uses a weighted Manhattan distance to address this issue. * Algorithm h ( n Improvements were made, and the improved version h ( n )for:

[0099] Formula 7

[0100] Choose the appropriate λ 1 and λ A value of 2 can improve A to some extent. * The efficiency of path search in algorithm path planning, but for the improvement A using weighted Manhattan distance as the heuristic function. *The algorithm-planned path still suffers from an excessive number of turns. This paper introduces a turning correction cost parameter into the weighted Manhattan distance as a heuristic function. The specific method is as follows:

[0101] Through the current node ( x n , y n ), parent node ( x n-1 , y n-1 ), child nodes x n+1 , y n+1 ) and target node ( x end , y end Based on the positional relationship, Formula 8 is used to determine whether the AGV has changed direction.

[0102] Formula 8

[0103] If the result calculated by formula 8 m 1 and m If the two values ​​are equal, the AGV will travel in a straight line; otherwise, the AGV will turn. Figure 3 As shown.

[0104] Based on the judgment result, the heuristic function is... h * ( n )Correction:

[0105] Formula 9

[0106] in, φ The turning correction cost parameter ensures that the AGV prioritizes straight-line travel and 0 < φ <1.

[0107] Improved cost estimation function f * ( n The expression is as shown in Formula 10:

[0108] Formula 10

[0109] Improved cost estimation function f * ( n Cost estimation function compared to the previous one f ( n For ), it mainly concerns its heuristic function. h ( nThe algorithm has been improved by using a weighted Manhattan distance as a heuristic function, making the distance estimation cost closer to the shortest distance and reducing the number of nodes traversed. Additionally, a turning correction cost parameter has been introduced into the heuristic function to reduce the number of path turns.

[0110] (2) Turning correction cost parameter

[0111] Most AGV path planning algorithms typically use path length as the evaluation metric, but the increased time cost due to excessive path turns is rarely considered. For example... Figure 4 As shown, there are three paths for an AGV to reach the target node B from the starting node A: path 1, path 2, and path 3. These three paths have equal Manhattan distances and represent the shortest distances. However, paths 2 and 3 have multiple turning points, while path 1 has only one turning point. Therefore, path 1 not only has the shortest distance but also the best time. Thus, considering turning costs during AGV path planning is necessary. Specifically, when a single AGV is performing a task, the path distance is generally short. If the AGV makes too many turns, the task execution time will significantly increase, and frequent turns will also increase the AGV's energy consumption to some extent.

[0112] This article uses Formula 9... h * ( n The calculation of turning correction cost parameter is introduced. φ This increases the cost of AGV path turning, thus reducing the number of turns in the planned path, as follows: φ The selection directly affects h * ( n The magnitude of the value affects the cost estimation function. f * ( n The value of the turning correction cost parameter is important. If the turning correction cost parameter is too large, the AGV may detour; if the turning correction cost parameter is too small, it usually has no corrective effect; if a suitable turning correction cost parameter is selected, it can ensure that the AGV prioritizes straight travel and avoid detours.

[0113] (3) Based on the improvement of A * Algorithm for AGV path planning design

[0114] This article is based on improved A * The single AGV path planning process of the algorithm is as follows: Figure 5 As shown.

[0115] Where: the coordinates of the starting node of the AGV are ( xstart , y start The coordinates of the target node are ( ); x end , y end ); f * , g , h * These are respectively from formula 9 f * ( n ), g ( n ), h * ( n ).

[0116] The specific steps are as follows:

[0117] Step 1: Model the AGV working environment using the grid map method. Input the starting node and target node of the AGV, and create Openlist and Closelist linked lists to store the nodes to be detected and the nodes that have been detected, respectively.

[0118] Step 2: Set the coordinates of the AGV's starting node ( x start , y start Add it to the Openlist linked list;

[0119] Step 3: Traverse all nodes in the Openlist linked list and find f. * Find the node with the smallest (n) value and set it as the current node;

[0120] Step 4: Move the current node into the Closelist linked list;

[0121] Step 5: Detect the four neighboring nodes (east, south, west, north) of the current node. Ignore nodes that are already in the Closelist or are obstacles. Then determine whether the remaining neighboring nodes are in the Openlist. If the remaining neighboring nodes are not in the Openlist, proceed to step 6; otherwise, proceed to step 7.

[0122] Step 6: Add the remaining neighboring nodes to the Openlist linked list, and set the current node as its parent node. Calculate the values ​​for each node. f * ( n )value, g ( n ) value and h * ( n) value, then proceed to step 8;

[0123] Step 7: Determine the AGV's position after passing the current node (calculated). g ( n Is the value smaller? g ( n If the value is smaller, then the parent node of that node is set as the current node, and the calculation is recalculated. g ( n ) value and f * ( n )value;

[0124] Step 8: Determine if the target node of the AGV has been found. If the target node has been found, proceed to step 9; otherwise, proceed to step 10.

[0125] Step 9: From the target node of the AGV, sequentially find the parent node to the starting node of the AGV. The obtained path is the path based on the improved AGV. * The algorithm determines the AGV's travel path and proceeds to step 11;

[0126] Step 10: Determine if the Openlist linked list is empty. If it is empty, proceed to step 11; otherwise, proceed to step 3.

[0127] Step 11: The algorithm ends.

[0128] The purpose of this invention is to overcome the aforementioned problems in the prior art and propose an outbound control method based on warehouse location planning and path planning. This collaborative optimization algorithm is an improvement on the A* algorithm. Its inventive point is that after considering location planning and path planning together, and combining optimal path function analysis, the shelves with high similarity are dispersed, making it easy for AGVs to find goods quickly at any station. Furthermore, high-frequency goods are placed in places where paths do not overlap, which can ensure that when there are many tasks, AGV congestion will not occur due to path overlap, thereby improving the overall efficiency of warehouse inbound and outbound operations.

[0129] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0130] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for controlling outbound routes based on warehouse storage locations, characterized in that, include: The scattered goods to be shipped are coded according to the historical shipment batches, and the similarity between the goods is calculated using the codes generated by the historical shipment batches. The scattered goods are combined according to the similarity to generate a goods group, wherein each goods group in the waiting-to-be-entered state is regarded as a shelf unit, and the shelf unit corresponds one-to-one with the shelf position; Calculate and save the outbound path corresponding to each shelf location, and place the product group into the shelf location based on the outbound path; and An improved A* algorithm is used for AGV path planning. An improved heuristic function and turning correction cost parameters are employed to improve the overall efficiency of the warehousing and logistics system. Fitness function values ​​are calculated under different outbound schemes. The optimal scheme is searched through continuous iteration, and based on this optimal scheme, the placement of goods groups in the warehouse and the outbound path of the AGV are determined simultaneously. Calculating the fitness function value under different outbound schemes involves using the following formula to calculate the fitness function value. α , β : f ( x )= αf path + βf other ， in, f ( x Let be the overall objective function for collaborative optimization. x Represents any node, f path Total time spent on outbound processing for all tasks. f other For the overhead of parts other than the AGV path planning, α and β These are the influence coefficients, and α + β =1; When the number of tasks M is greater than the total number of AGVs Q, it takes a certain amount of time for the AGVs to return from the target node to the starting node. The total time spent on all tasks leaving the warehouse is calculated using the following formula: ， When the total number of AGVs Q is greater than or equal to the number of tasks M, one AGV is allocated to each task, and the total time spent on outbound operations for all tasks is calculated using the following formula: ， ， in, µ This is the penalty coefficient for the AGV to return from the target node to the starting node. f m,q The time spent by a vehicle during a particular dispatch, of which ( i , j ) is the target node. f c The waiting time after traffic congestion occurs. These are dynamic parameters. The starting node passes through the current node n Estimated path distance to the target node f * (n) / v Proportional v It's the AGV speed. f i It is the AGV that has arrived. i Theoretical time required for the point f j It is the AGV that has arrived. j Theoretical time required for the point f I-i It is the AGV that has arrived. i The waiting time required when encountering obstacles during the process. f J-j It is the AGV that has arrived. j The waiting time required when encountering obstacles during the process; Improving the overall efficiency of a warehousing and logistics system by using an improved heuristic function and a turning correction cost parameter includes: using a weighted Manhattan distance as the improved heuristic function, making the distance estimation cost close to the shortest distance to reduce the number of traversed nodes; and introducing the turning correction cost parameter into the improved heuristic function to reduce the number of path turns.

2. The outbound path control method based on warehouse location according to claim 1, characterized in that, Using weighted Manhattan distance as the improved heuristic function, so that the distance estimation cost is close to the shortest distance to reduce the number of nodes traversed, includes: Using the following weighted Manhattan distance to pair A * The algorithm's h * (n) Improvements: Among them, A is improved by selecting parameters λ1 and λ2. * In algorithmic path planning, path search efficiency is 0 < λ 1<1, 0< λ 2<1,( x n , y n ) is the current node, and ( x end , y end ) is the target node.

3. The outbound path control method based on warehouse location according to claim 2, characterized in that, Introducing the turning correction cost parameter into the improved heuristic function to reduce the number of path turns includes: Through the current node ( x n , y n ), parent node ( x n-1 , y n-1 ), child nodes x n+1 , y n+1 ) and the target node ( x end , y end Based on the positional relationship of the AGV, the following formula is used to determine whether the AGV has changed direction, where, when m 1 and m When 2 are equal, the AGV is in a straight-line driving state; and when m 1 and m When the 2 values ​​are not equal, the AGV changes direction: ， Based on the above judgment results, the heuristic function is modified. in, φ The turning correction cost parameter ensures that the AGV prioritizes straight-line travel and 0 < φ <1.

4. The outbound path control method based on warehouse location according to claim 1, characterized in that, The process of encoding scattered goods based on historical shipment batches and calculating the similarity between goods using the codes generated from the historical shipment batches includes: The warehouse outbound data is processed to remove irrelevant data and extract the product number and batch number; The item number is identified by 1s and 0s as whether it was shipped in this batch, where each item is a vector consisting of batch numbers in each dimension; and The similarity between the vectors is calculated using the following cosine similarity formula: ； The absolute value of the similarity between vectors minus 1 is taken as the similarity between goods. The smaller the absolute value, the higher the similarity between the goods. x Let be a vector of a type of goods. y Let be a vector of another type of goods. x i For a type of goods shipped in the i-th batch... y i Let n be the dimension of vectors x and y, representing another type of goods shipped in the i-th batch.

5. The outbound path control method based on warehouse location according to claim 3, characterized in that, AGV path planning using the improved A* algorithm includes: The working environment of the AGV is modeled using a grid map method. The starting node and target node of the AGV are input, and an Openlist linked list for storing nodes to be detected and a Closelist linked list for storing detected nodes are created respectively. The coordinates of the starting node of the AGV are added to the Openlist linked list; Traverse all nodes in the Openlist linked list, find the node with the smallest improved cost function value and set it as the current node; Move the current node into the Closelist linked list; Detect the four neighboring nodes of the current node, ignore nodes already in the Closelist or obstacle nodes, and then determine whether the remaining neighboring nodes are in the Openlist. When the remaining neighboring nodes are in the Openlist linked list, determine whether the AGV has passed the current node and calculated the result. g ( n Is the value smaller, where if the g ( n If the value is smaller, then the parent node of the remaining neighboring nodes is set as the current node, and the calculation is recalculated. g ( n ) value and f * ( n ) value, where, f * ( n () is from the starting node through the current node n The estimated path distance to the target node; g ( n () represents the distance from the starting node to the current node n The actual path distance; and Determine whether the target node of the AGV has been found. When the target node is found, sequentially search from the target node of the AGV to the starting node of the AGV. The obtained path is the AGV driving path based on the improved A* algorithm.

6. The outbound path control method based on warehouse location according to claim 5, characterized in that, If the target node of the AGV is not found, the system continues to check whether the Openlist linked list is empty. When the Openlist is not an empty list, return to the following steps: traverse all nodes in the Openlist until the Openlist is an empty list.

7. The outbound path control method based on warehouse location according to claim 5, characterized in that, When the remaining neighboring nodes are not in the Openlist linked list, add the remaining neighboring nodes to the Openlist linked list; Set the current node as its parent node, and calculate the values ​​of each of the remaining neighboring nodes. f * ( n )value, g ( n ) value and h * ( n ) value, where, h * ( n ) is from the current node n The estimated path distance to the target node; and Proceed to the next step: Determine whether the target node of the AGV has been found.

8. A warehouse location-based outbound route control device, characterized in that, include: The similarity acquisition module is used to encode the scattered goods to be shipped based on the historical shipment batches, and to calculate the similarity between the goods using the codes generated by the historical shipment batches. The product group generation module is used to combine the scattered goods according to the similarity to generate product groups, wherein each product group in the waiting-to-be-entered state is regarded as a shelf unit, and the shelf unit corresponds one-to-one with the shelf position; The product group setting module is used to calculate and save the outbound path corresponding to each shelf location, and place the product group into the shelf location based on the outbound path; and The AGV path planning module utilizes an improved A* algorithm for AGV path planning. It improves the overall efficiency of the warehousing and logistics system by employing an improved heuristic function and turning correction cost parameters. The module calculates the fitness function values ​​under different outbound plans, searches for the optimal solution through continuous iteration, and determines the placement of goods groups in the warehouse and the outbound path of the AGV based on the optimal solution. Calculating the fitness function value under different outbound schemes includes using the following formula to calculate the fitness function value. α , β : f ( x )= αf path + βf other ， in, f ( x Let be the overall objective function for collaborative optimization. x Represents any node, f path Total time spent on outbound processing for all tasks. f other For the overhead of parts other than the AGV path planning, α and β These are the influence coefficients, and α + β =1; When the number of tasks M is greater than the total number of AGVs Q, it takes a certain amount of time for the AGVs to return from the target node to the starting node. The total time spent on all tasks leaving the warehouse is calculated using the following formula: ， When the total number of AGVs Q is greater than or equal to the number of tasks M, one AGV is allocated to each task, and the total time spent on outbound operations for all tasks is calculated using the following formula: ， ， in, µ This is the penalty coefficient for the AGV to return from the target node to the starting node. f m,q The time spent by a vehicle during a particular dispatch, of which ( i , j ) is the target node. f c The waiting time after traffic congestion occurs. These are dynamic parameters. The starting node passes through the current node n Estimated path distance to the target node f * (n) / v Proportional v It's the AGV speed. f i It is the AGV that has arrived. i Theoretical time required for the point f j It is the AGV that has arrived. j Theoretical time required for the point f I-i It is the AGV that has arrived. i The waiting time required when encountering obstacles during the process. f J-j It is the AGV that has arrived. j The waiting time required when encountering obstacles during the process; Improving the overall efficiency of the warehousing and logistics system by using an improved heuristic function and a turning correction cost parameter includes: using a weighted Manhattan distance as the improved heuristic function to make the distance estimation cost close to the shortest distance to reduce the number of traversed nodes; and introducing the turning correction cost parameter into the improved heuristic function to reduce the number of path turns.

Images