A multi-AGV scheduling method for intelligent warehousing based on improved A-Star algorithm

By improving the A-Star algorithm, combining the raster method and turn cost function, the path planning of intelligent storage of multiple AGVs is optimized, which solves the problems of excessive turn times and imbalance in the accuracy and time in traditional algorithms, and improves the operation efficiency of AGVs.

CN115328140BActive Publication Date: 2024-11-26ZHEJIANG UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202211006074.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-08-22
Publication Date
2024-11-26
Estimated Expiration
2042-08-22

AI Technical Summary

Technical Problem

The traditional A-Star algorithm has too many turns when planning paths in intelligent storage multi-AGV environments, and the accuracy and time requirements of algorithm search are unbalanced, which affects the operation efficiency of AGV.

Method used

Using the improved A-Star algorithm, the evaluation function is constructed through raster method environmental modeling, and the turn cost function and heuristic function weight factor α is introduced, and the path planning is optimized to reduce the number of turns and balance the search accuracy and time.

Benefits of technology

It effectively reduces the number of turns of AGV, reduces energy consumption and time, improves the operating efficiency of AGV, and flexibly adjusts the α value in different scenarios to balance the search accuracy and time of the algorithm.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115328140B_ABST
    Figure CN115328140B_ABST
Patent Text Reader

Abstract

A smart warehousing multi-AGV scheduling method based on an improved A-Star algorithm comprises the following steps: S1. Based on the working environment of smart warehousing, a grid method is used to model the environment; S2. An evaluation function is constructed and the A-Star algorithm is improved; S3. An improved A-Star algorithm is designed to solve the model and obtain the optimal scheduling result. The smart warehousing multi-AGV scheduling method based on the improved A-Star algorithm disclosed by the present invention introduces corner optimization and dynamic measurement optimization, and considers the turning cost in the cost function g(n) of the A* algorithm, so as to ensure that the number of turns of the AGV is minimized without increasing the distance, thereby reducing the time of AGV task operation and reducing the energy consumption of the AGV, and improving the operation efficiency of the AGV; and the heuristic function weight factor α is introduced into the heuristic function h(n) of the A* algorithm for dynamic adjustment, so that the search accuracy and time of the algorithm are balanced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of intelligent logistics, and in particular to an intelligent warehousing multi-AGV scheduling method based on an improved A-Star algorithm. Background Art

[0002] For a long time, people have been accustomed to focusing on the production field to pay attention to costs, but with the development of modern mass production, this focus has gradually shifted to the circulation field. According to statistics, in the United States, only 5% of the time in the entire production process is used for processing and manufacturing, and 95% of the time is used for logistics processes such as transportation and storage; in Japan, logistics costs account for about 10% to 12% of the total cost of products; in my country, logistics costs account for about 70% of the difference between the purchase and sale prices of goods, and the logistics costs reflected on the books alone account for 40% of the total cost of goods. It can be seen that logistics occupies an important position in the operation and management of enterprises. As the main component of realizing the important storage function of logistics, the research of intelligent warehousing system is of great significance. Automated guided vehicles (AGVs) have become indispensable handling equipment in intelligent warehousing systems due to their advantages in reliability, flexibility, and adaptability. Therefore, the efficiency of AGV has a very important impact on the operating efficiency of intelligent warehousing systems. The efficiency of AGV depends largely on the scheduling method of AGV. At present, when the traditional A-Star algorithm is performing path planning, there may be multiple turns in the path from the starting point to the end point. In the actual operation process, when the AGV encounters a turn, it needs to stop and change direction, which will cause energy and time waste; and in different application scenarios, the accuracy and time requirements of the algorithm search are different. Therefore, it is necessary to conduct in-depth research and improvement on the A-Star algorithm applied to the multi-AGV environment of intelligent warehousing. Summary of the invention

[0003] The present invention aims to overcome the above-mentioned shortcomings of the prior art and provide an intelligent warehousing multi-AGV scheduling method based on the improved A-Star algorithm to solve the problems of excessive number of turns in the path planning process and unbalanced accuracy and time requirements of the algorithm search in the prior art.

[0004] An intelligent warehousing multi-AGV scheduling method based on an improved A-Star algorithm comprises the following steps:

[0005] S1. Based on the working environment of intelligent warehousing, the grid method is used for environmental modeling;

[0006] S2. Construct evaluation function and improve A-Star algorithm;

[0007] S3. Design an improved A-Star algorithm to solve the model and obtain the optimal scheduling result.

[0008] S1: Based on the working environment of intelligent warehousing, the grid method is used for environmental modeling, including:

[0009] According to the actual warehouse working environment, the warehouse's plane map is divided into several grids of the same size, obstacles and passable areas are marked, and a grid map is established.

[0010] S2: The construction of the evaluation function and improvement of the A-Star algorithm specifically includes:

[0011] S21: The evaluation function of the A-Star algorithm is constructed as

[0012] f(n)=g(n)+h(n) (1)

[0013] Among them, f(n) is the evaluation function, which represents the total cost from the starting node P to the intermediate node n and then to the target node P; g(n) represents the cost of traversing from the starting node P to the intermediate node n; h(n) is the heuristic function, which represents the cost from the current node n to the target node Q.

[0014] Considering the actual warehouse working environment station spacing and AGV travel direction, it is determined that AGV can only move in the up, down, left, and right directions and cannot move in the diagonal direction. Therefore, Manhattan distance is selected as the heuristic function. Let the current node be (x n ,y n ), the target node is (xs, y s ), then the heuristic function is

[0015] h(n)=|x s -x n |+|y s -y n | (2)

[0016] S22: Consider the cost function of AGV turning, so that the path optimization tends to be the path with fewer turns. Set the current node to (x n ,y n ), the parent node is (x n-1 ,y n-1 ) and its child nodes are (x n+1 ,y n+1 ), the node information can be used to determine the turn.

[0017] S221: When (x n -x n-1 )(y n+1 -y n ) and (x n+1 -x n )(y n -y n-1 ) are equal, it means that the AGV goes straight through the current node without turning.

[0018] S222: When (x n -x n-1 )(y n+1 -y n ) and (x n+1 -x n )(y n -y n-1 ) are not equal, it means that the AGV turns at the current node.

[0019] S223: Let 1 be the basic cost of AGV moving one cell, k represents the turning cost coefficient of AGV, and the value range is 0.1-0.2. Then the actual cost function g(n) is

[0020]

[0021] S23: Introduce the heuristic function weight factor α for dynamic adjustment. At this time, the evaluation function of the A-Star algorithm is f(n) = g(n) + α·h(n) (4)

[0022] Among them, when α=0, αh(n)=0, that is, f(n)=g(n), then the cost value of the solution of the A* algorithm is the same as the cost value of the Dijkstra solution, and the A* algorithm degenerates into the Dijkstra algorithm. At this time, the search accuracy of f(n) is high, and its result must be the optimal solution, but its search efficiency is low and requires a lot of search time; when 0<α≤1, α·h(n) is less than or equal to the actual cost value, as the α value increases, the accuracy and time of the solution of the A* algorithm reach the optimal; when α=1, the A* algorithm reaches the optimal; when α>1, α·h(n) is greater than the actual cost value, as the α value increases, the search speed of the A* algorithm is accelerated, and the required search time is less, but the algorithm search accuracy is poor at this time, and the search result is not necessarily the optimal solution. For the requirements of different scenarios, the α value needs to be flexibly adjusted to balance the relationship between the algorithm's search accuracy and time.

[0023] S3: The improved A-Star algorithm is used to solve the model and obtain the optimal scheduling result, which specifically includes:

[0024] S31: Set the starting point S0 and the target point V0.

[0025] S32: Initialize the Open list and Close list. The Open list represents the set of paths to be searched, and the Close list represents the set of valid paths that have been searched. The starting point S0 is placed in the Open list set first.

[0026] S33: Start searching for paths. Find the grids that can be reached around the starting point S0 in the grid map (four grids on the top, bottom, left and right), add these grids to the Open list collection, and set their parent node to S0.

[0027] S34: Delete the starting point S0 from the Open list set, and put the starting point S0 into the Close list set.

[0028] S35: Calculate the actual distance g(n) of all grid nodes in the Open list set, and calculate the estimated function α·h(n) from each node to the target point, and add the two together to obtain the estimated distance f(n) from the next node to the target point.

[0029] S36: Select the grid i with the lowest f(n) value from the Open list set, delete it from the Open list set, and put it into the Close list set.

[0030] S37: Check all adjacent and reachable grids of grid i, ignoring obstacles and grids in the Close list set.

[0031] S371: If these grids are not in the Open list set, add them to the Open list set, calculate the f(n) values ​​of these grids, and set the parent node to i;

[0032] S372: If an adjacent grid j is already in the Open list, calculate the g(n) value of the new path from S0 to grid j (i.e., the path passing through i). If the new g(n) value is lower, modify the parent node to grid i, recalculate the f(n) value, and the h(n) value does not need to be changed; if the new g(n) value is higher, it means that the new path consumption is higher, and the value does not change.

[0033] S38: Continue to find the one with the smallest f(n) value from the Open list set, delete it from the Open list set, add it to the Close list set, and then go to step 7.

[0034] S39: End judgment: When the target point V0 appears in the Open list set, it means that the path has been found; when there is no data in the Openlist set, it means that there is no suitable path.

[0035] Compared with the prior art, the present invention has the following beneficial technical effects:

[0036] The intelligent warehousing multi-AGV scheduling method based on the improved A-Star algorithm disclosed in the present invention introduces corner optimization and dynamic weight optimization, takes the turning cost into consideration in the cost function g(n) of the A* algorithm, and ensures that the number of turns of the AGV is minimized without increasing the distance, thereby reducing the running time of the AGV task and reducing the energy consumption of the AGV, thereby improving the operating efficiency of the AGV; and introduces the heuristic function weight factor α into the heuristic function h(n) of the A* algorithm for dynamic adjustment, so that the search accuracy and time of the algorithm are balanced. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical method of the present invention, the following is a brief introduction to the drawings used in the implementation method. The following drawings only show certain embodiments of the present invention and should not be regarded as limiting the scope. For ordinary technicians in this field, other drawings can be obtained based on the structure of the drawings without paying creative labor.

[0038] Figure 1 It is a node position diagram for judging turning of the present invention;

[0039] Figure 2 is a flow chart of the improved A-Star algorithm of the present invention;

[0040] Figure 3 It is a case result diagram in an embodiment of the present invention.

[0041] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0042] In order to have a clearer understanding of the purpose, technical features and effects of the invention, the implementation method of the invention will be specifically described below with reference to the accompanying drawings of the present invention.

[0043] The technical solution of the invention includes the following method:

[0044] S1. Based on the working environment of intelligent warehousing, the grid method is used for environmental modeling;

[0045] S2. Construct evaluation function and improve A-Star algorithm;

[0046] S3. Design an improved A-Star algorithm to solve the model and obtain the optimal scheduling result.

[0047] In S1, the working environment based on intelligent warehousing is modeled using the grid method, which specifically includes:

[0048] According to the actual warehouse working environment, the warehouse's plane map is divided into several grids of the same size, obstacles and passable areas are marked, and a grid map is established.

[0049] The evaluation function is constructed in S2 and the A-Star algorithm is improved, which specifically includes:

[0050] S21: The evaluation function of the A-Star algorithm is constructed as

[0051] f(n)=g(n)+h(n) (1)

[0052] Among them, f(n) is the evaluation function, which represents the total cost from the starting node P to the intermediate node n and then to the target node P; g(n) represents the cost of traversing from the starting node P to the intermediate node n; h(n) is the heuristic function, which represents the cost from the current node n to the target node Q.

[0053] Considering the actual warehouse working environment station spacing and AGV travel direction, it is determined that AGV can only move in the up, down, left, and right directions and cannot move in the diagonal direction. Therefore, Manhattan distance is selected as the heuristic function. Let the current node be (x n ,y n ), the target node is (x s ,y s ), then the heuristic function is

[0054] h(n)=|x s -x n |+|y s -y n | (2)

[0055] S22: Consider the cost function of AGV turning, so that the path optimization tends to be the path with fewer turns. Figure 1 As shown, set the current node to (x n ,y n ), the parent node is (x n-1 ,y n-1 ) and its child nodes are (x n+1 ,y n+1 ), the node information can be used to determine the turn.

[0056] S221: When (x n -x n-1 )(y n+1 -y n ) and (x n+1 -x n )(y n -y n-1 ) are equal, it means that the AGV goes straight through the current node without turning.

[0057] S222: When (x n -x n-1 )(y n+1 -y n ) and (x n+1 -x n )(y n -y n-1 ) are not equal, it means that the AGV turns at the current node.

[0058] S223: Let 1 be the basic cost of AGV moving one cell, k represents the turning cost coefficient of AGV, and the value range is 0.1-0.2. Then the actual cost function g(n) is

[0059]

[0060] S23: Introduce the heuristic function weight factor α for dynamic adjustment. At this time, the evaluation function of the A-Star algorithm is

[0061] f(n)=g(n)+α·h(n) (4)

[0062] Among them, when α=0, αh(n)=0, that is, f(n)=g(n), then the cost value of the solution of the A* algorithm is the same as the cost value of the Dijkstra solution, and the A* algorithm degenerates into the Dijkstra algorithm. At this time, the search accuracy of f(n) is high, and its result must be the optimal solution, but its search efficiency is low and requires a lot of search time; when 0<α≤1, α·h(n) is less than or equal to the actual cost value, as the α value increases, the accuracy and time of the solution of the A* algorithm reach the optimal; when α=1, the A* algorithm reaches the optimal; when α>1, α·h(n) is greater than the actual cost value, as the α value increases, the search speed of the A* algorithm is accelerated, and the required search time is less, but the algorithm search accuracy is poor at this time, and the search result is not necessarily the optimal solution. For the requirements of different scenarios, the α value needs to be flexibly adjusted to balance the relationship between the algorithm's search accuracy and time.

[0063] The improved A-Star algorithm designed in S3 is used to solve the model and obtain the optimal scheduling result, such as Figure 2 As shown, specifically including:

[0064] S31: Set the starting point S0 and the target point V0.

[0065] S32: Initialize the Open list and Close list. The Open list represents the set of paths to be searched, and the Close list represents the set of valid paths that have been searched. The starting point S0 is placed in the Open list set first.

[0066] S33: Start searching for paths. Find the grids that can be reached around the starting point S0 in the grid map (four grids on the top, bottom, left and right), add these grids to the Open list collection, and set their parent node to S0.

[0067] S34: Delete the starting point S0 from the Open list set, and put the starting point S0 into the Close list set.

[0068] S35: Calculate the actual distance g(n) of all grid nodes in the Open list set, and calculate the estimated function α·h(n) from each node to the target point, and add the two together to obtain the estimated distance f(n) from the next node to the target point.

[0069] S36: Select the grid i with the lowest f(n) value from the Open list set, delete it from the Open list set, and put it into the Close list set.

[0070] S37: Check all adjacent and reachable grids of grid i, ignoring obstacles and grids in the Close list set.

[0071] S371: If these grids are not in the Open list set, add them to the Open list set, calculate the f(n) values ​​of these grids, and set the parent node to i;

[0072] S372: If an adjacent grid j is already in the Open list, calculate the g(n) value of the new path from S0 to grid j (i.e., the path passing through i). If the new g(n) value is lower, modify the parent node to grid i, recalculate the f(n) value, and the h(n) value does not need to be changed; if the new g(n) value is higher, it means that the new path consumption is higher, and the value does not change.

[0073] S38: Continue to find the one with the smallest f(n) value from the Openlist set, delete it from the Open list set, add it to the Close list set, and then go to step 7.

[0074] S39: End judgment: When the target point V0 appears in the Open list set, it means that the path has been found; when there is no data in the Openlist set, it means that there is no suitable path.

[0075] For an example of this approach, see Figure 3 :

[0076] Create a 20×20 grid map, where one module represents a 1×1 grid. The white area is the passable area, the black area is the obstacle, the green dot is the starting point, and the yellow dot is the target point.

[0077] The shortest path is obtained by using the technical solution described in this application. Figure 3 described.

[0078] The above description is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form described herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, environments and modifications, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art shall not deviate from the spirit and scope of the present invention, and shall be within the scope of protection of the claims attached to the present invention.

Claims

1. An intelligent warehousing multi-AGV scheduling method based on improved A-Star algorithm, characterized in that: include: S1: Based on the working environment of intelligent warehousing, the grid method is used for environmental modeling, including: Step 11: According to the actual warehouse working environment, divide the warehouse plane map into several grids of the same size, mark obstacles and passable areas, and create a grid map; S2: Construct an evaluation function and improve the A-Star algorithm, including: S21: The evaluation function of the A-Star algorithm is constructed as follows: f(n)=g(n)+h(n) (1) Among them, f(n) is the evaluation function, which represents the total cost from the starting node P to the intermediate node n and then to the target node P; g(n) represents the cost of traversing from the starting node P to the intermediate node n; h(n) is the heuristic function, which represents the cost of the current node n to the target node Q; Considering the actual warehouse working environment station spacing and AGV travel direction, it is determined that AGV can only move in the up, down, left, and right directions and cannot move in the diagonal direction. Therefore, Manhattan distance is selected as the heuristic function; let the current node be (x n ,y n ), the target node is (x s ,y s ), then the heuristic function is: h(n)=|x s -x n |+|y s -y n | (2) S22: Consider the cost function of AGV turning, so that the path optimization tends to the path with fewer turns; set the current node to (x n ,y n ), the parent node is (x n-1 ,y n-1 ) and its child nodes are (x n+1 )(y n+1 ), the node information can be used to determine the turn; S221: When (x n -x n-1 )(y n+1 -y n ) and (x n+1 -x n )(y n -y n-1 ) are equal, it means that the AGV goes straight through the current node without turning; S222: When (x n -x n-1 )(y n+1 -y n ) and (x n+1 -x n )(y n -y n-1 ) are not equal, it means that the AGV turns at the current node; S223: Let 1 be the basic cost of AGV moving one cell, k represents the turning cost coefficient of AGV, and the value range is 0.1-0.

2. Then the actual cost function g(n) is S23: Introduce the heuristic function weight factor α for dynamic adjustment. At this time, the evaluation function of the A-Star algorithm is f(n)=g(n)+α·h(n) (4) Among them, when α=0, αh(n)=0, that is, f(n)=g(n), then the cost value of the solution obtained by the A* algorithm is the same as the cost value obtained by the Dijkstra algorithm, and the A* algorithm degenerates into the Dijkstra algorithm. At this time, the search accuracy of f(n) is high, and its result must be the optimal solution, but its search efficiency is low and requires a lot of search time; when 0<α≤1, α·h(α) is less than or equal to the actual cost value. As the α value increases, the accuracy and time of the solution obtained by the A* algorithm reach the optimal value; when α=1, the A* algorithm reaches the optimal value; when α>1, α·h(n) is greater than the actual cost value. As the α value increases, the search speed of the A* algorithm increases, and the required search time is reduced, but the algorithm search accuracy is poor at this time, and the search result is not necessarily the optimal solution; for the requirements of different scenarios, the α value needs to be flexibly adjusted to balance the relationship between the algorithm's search accuracy and time; S3: Design an improved A-Star algorithm to solve the model and obtain the optimal scheduling result, including: S31: Set the starting point S0 and the target point V0; S32: Initialize the Open list and Close list; the Open list represents the set of paths to be searched, and the Close list represents the set of valid paths that have been searched; and put the starting point S0 into the Open list set first; S33: Start searching for paths; find the grids that can be reached around the starting point S0 in the grid map (four grids on the top, bottom, left and right), add these grids to the Open list set, and set their parent node to S0; S34: Delete the starting point S0 from the Open list set, and put the starting point S0 into the Close list set; S35: Calculate the actual distance g(n) of all grid nodes in the Open list set, and calculate the estimated function α·h(n) from each node to the target point, and add the two together to obtain the estimated distance f(n) from the next node to the target point; S36: Select the grid i with the lowest f(n) value from the Open list set, delete it from the Open list set, and put it into the Open list set; S37: Check all adjacent and reachable grids of grid i, ignoring obstacles and grids in the Close list set; S371: If these grids are not in the Open list set, add them to the Open list set, calculate the f(n) values ​​of these grids, and set the parent node to i; S372: If an adjacent grid j is already in the Open list set, calculate the g(n) value of the new path from S0 to grid j (i.e., the path passing through i); if the new g(n) value is lower, change the parent node to grid i, recalculate the f(n) value, and the h(n) value does not need to be changed; if the new g(n) value is higher, it means that the new path has a higher cost, and the value does not change; S38: Continue to find the one with the smallest f(n) value from the Open list set, delete it from the Open list set, add it to the Close list set, and then go to step 7; S39: End judgment: When the target point V0 appears in the Open list set, it means that the path has been found; when there is no data in the Open list set, it means that there is no suitable path.