AGV path planning device and method based on order clustering

Abstract

The invention belongs to the technical field of intelligent scheduling, and particularly discloses an AGV path planning device and method based on order clustering, the device comprises an order splitting module, an intelligent control module and a material management module which are connected in sequence, and the output end of the intelligent control module is further connected with an AGV scheduling module; the order splitting module is used for clustering and segmenting original orders to obtain an order splitting result; the intelligent control module carries out AGV path planning according to an order splitting result; the material management module manages warehouse-in and warehouse-out of materials according to the AGV path planning information; and the AGV scheduling module schedules the AGV to execute the operation task and controls the start and stop of the AGV according to the AGV path planning information. The AGV path planning method based on order clustering solves the problems that an existing AGV path planning method based on order clustering is insufficient in capacity of cooperative work of multiple AGVs and real-time path adjustment, and lacks an efficient scheduling strategy when facing complex order requirements and dynamic environment changes.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent scheduling, and particularly relates to an AGV path planning device and method based on order clustering. Background Art

[0002] With the continuous development of automated logistics technology, the application of Automated Guided Vehicles (AGVs) in the warehousing and logistics industries has been gradually widely promoted. AGVs have the characteristics of automation, intelligence, and high efficiency, and can effectively reduce labor costs, improve transportation efficiency, and play an important role in material handling, warehouse management, etc. However, in practical applications, when multiple AGVs work in the same warehouse or distribution center at the same time, how to reasonably plan their paths, avoid conflicts, and improve efficiency is still a difficult problem. Traditional AGV path planning methods mostly rely on static path setting or optimization based on a specific goal, lacking dynamic response and optimization to order requirements, resulting in problems such as empty driving and path congestion in actual operations, affecting the overall efficiency and resource utilization rate of the warehouse.

[0003] In recent years, AGV path planning based on orders has gradually become a new research direction. This method dynamically plans the path by reasonably clustering orders and comprehensively considering information such as the spatial distribution and demand time of orders, thereby reducing empty paths, avoiding collisions and path crossings between AGVs, and improving transportation efficiency. However, most of the existing AGV path planning methods based on order clustering still have problems with insufficient ability for multi-AGV collaborative work and real-time path adjustment. Especially when facing complex order requirements and dynamic environmental changes, they lack efficient scheduling strategies. Therefore, how to reasonably schedule multiple AGVs through an intelligent path planning method, combined with the real-time needs of orders and the spatial layout of the warehouse, to improve the operation efficiency of the AGV system is still a technical challenge faced currently. Summary of the Invention

[0004] The purpose of the present invention is to solve the problems of insufficient ability of the existing AGV path planning methods based on order clustering for multi-AGV collaborative work and real-time path adjustment, and the lack of efficient scheduling strategies when facing complex order requirements and dynamic environmental changes, and to propose an AGV path planning device and method based on order clustering.

[0005] The technical solution of the present invention is as follows: In the first aspect, an AGV path planning device based on order clustering includes an order splitting module, an intelligent control module, and a material management module connected in sequence, and the output end of the intelligent control module is also connected to an AGV scheduling module;

[0006] The order splitting module is used to obtain the original order information, cluster and split the original order, and output the order splitting result;

[0007] The intelligent control module is used to receive the order splitting result and perform AGV path planning according to the order splitting result, and output the AGV path planning information;

[0008] The material management module is used to receive the AGV path planning information and manage the inbound and outbound of materials according to the AGV path planning information;

[0009] The AGV scheduling module is used to receive the AGV path planning information, and according to the AGV path planning information, schedule the AGV to execute the job task and control the start and stop of the AGV.

[0010] In a second aspect, an AGV path planning method based on order clustering includes the following steps:

[0011] Obtain the original order information, and cluster and split the original order according to the clustering algorithm to obtain the order splitting result;

[0012] According to the order splitting result, use the improved ant colony optimization algorithm to formulate the AGV path planning information.

[0013] Preferably, the step of clustering and splitting the original order according to the clustering algorithm to obtain the order splitting result specifically includes the following steps:

[0014] Perform One-Hot encoding on the non-numerical data of the original order to obtain the encoded order;

[0015] Use the improved particle swarm optimization algorithm to perform K-means clustering on the encoded order information to obtain the clustering result;

[0016] Split the order according to the clustering result to obtain the order splitting result.

[0017] Preferably, the step of using the improved particle swarm optimization algorithm to perform K-means clustering on the encoded order information specifically includes the following steps:

[0018] Load the data, obtain the quantity D of materials taken by the order, and obtain the maximum load UB of the AGV;

[0019] Set the number of clusters to M according to the number M of AGVs;

[0020] Set the maximum number of iterations, the population number, the upper bound ub = UB of the initial position and the lower bound lb = 0 of the initial position, and set the dimension of the initial search space to the quantity D of materials taken by the order;

[0021] Randomly initialize the population positions within the range of the upper bound and lower bound of the initial position, and set the initial velocity V = 0;

[0022] Calculate the objective function value, which is the sum of the shortest Euclidean distances from each material to each cluster center;

[0023] Randomly pair and compare particles two by two, iteratively update the objective function value until the maximum number of iterations is reached, and output the global optimal solution, that is, the clustering result; where a particle represents a possible cluster center.

[0024] Preferably, the step of randomly pairing and comparing particles two by two, iteratively updating the objective function value until the maximum number of iterations is reached, and outputting the global optimal solution specifically includes the following steps:

[0025] Randomly pair and compare particles two by two, and update the position of the inferior particle among the two particles;

[0026] Update the objective function value corresponding to the particle according to the updated position of the particle;

[0027] If the updated objective function value is better than the objective function value before update, then in the next iteration, the position update formula of particle x2 is:

[0028]

[0029] where w represents the inertia weight, IT represents the current iteration number, MaxIter represents the maximum number of iterations, V′(x2,j) represents the updated velocity of particle x2 in the next iteration, V(x2,j) represents the velocity of particle x2 before update in the next iteration, X(b,j) represents the position of the global optimal solution in the jth dimension, r2 and r3 represent random variables generated in the range [0, 1], with r2 = rand(), r3 = rand(), x1 represents the superior particle among the two particles, X(x1,j) represents the position of particle x1 before update in the next iteration, and round represents the rounding function;

[0030] If the objective function value before update is better than the updated objective function value, then in the next iteration, the position update formula of particle x2 is:

[0031] X′(x2,j) = UB / 2 - X(x2,j)

[0032] where X′(x2,j) represents the position of particle x2 after update in the next iteration, X(x2,j) represents the position of particle x2 before update in the next iteration, and j represents the jth dimension;

[0033] Repeat updating the particle positions and objective function values until the maximum number of iterations is reached, output the particle positions, and obtain the global optimal solution.

[0034] Preferably, the method of formulating AGV path planning information using an improved ant colony optimization algorithm according to the order splitting result specifically includes the following steps:

[0035] Get the order splitting results, set the maximum number of iterations, population size, pheromone evaporation factor, heuristic evaporation factor and pheromone;

[0036] The position of the ant colony is randomly initialized, and the dimension of the population is the number of materials in the order split result;

[0037] Calculate the objective function value, which is the sum of the Euclidean distances of all paths;

[0038] Obtain the initial global optimal solution based on the objective function value;

[0039] The ants are randomly paired and compared, and the positions of the ants are iteratively updated until the maximum number of iterations is reached. The final global optimal solution is output to obtain the AGV path planning information.

[0040] Preferably, the random pairing comparison of ants is performed in pairs, and the ant positions are updated iteratively until a maximum number of iterations is reached, and a final global optimal solution is output, which specifically includes the following steps:

[0041] Randomly compare pairs of ants and update the position of the worse ant y2;

[0042] According to the updated position of ant y2, update the corresponding objective function value;

[0043] If the updated objective function value is better than the objective function value before the update, calculate the probability P of the ant choosing to move from node i to node j in the next iteration. ij :

[0044]

[0045] Among them, τ ij represents the pheromone concentration on path (i, j), represents the heuristic information factor on path (i, j), τ ik represents the pheromone concentration on the path (i, k), represents the heuristic information factor on the path (i, k), and Z represents a set of integers;

[0046] Let the ants follow the probability P ij Transfer and then renew pheromones:

[0047]

[0048] Among them, τ i ' jdenotes the pheromone concentration on the updated path (i, j), ρ denotes the pheromone evaporation factor, α denotes the pheromone weight, τ ij g denotes the pheromone concentration of the global optimal solution g, denotes the pheromone concentration of y1, f(y1) denotes the objective function value of y1, and f(g) denotes the objective function value of the global optimal solution g;

[0049] If the objective function value before update is better than that after update, then in the next iteration, the position update formula of ant y2 is:

[0050]

[0051] where t denotes a temporary variable used to store randperm(T,2), randperm(T,2) denotes randomly selecting two different numbers from the dimension T of the population, t1 denotes a temporary variable used to store Y(y2,t(1)), Y(y2,t(1)) denotes the position of ant y2 in the t(1) dimension, Y(y2,t(2)) denotes the position of ant y2 in the t(2) dimension, t(1) denotes the first of the two numbers randomly selected from the dimension T of the population, and t(2) denotes the second of the two numbers randomly selected from the dimension T of the population;

[0052] Repeat updating the ant positions and pheromones until the maximum number of iterations is reached, output the path with the highest ant position and pheromone concentration to obtain the global optimal solution, that is, the final AGV path planning information.

[0053] The beneficial effects of the present invention are:

[0054] By performing clustering analysis on order demands, the present invention reasonably organizes the path planning of multiple AGVs, effectively solving the problems of empty driving and path congestion in traditional AGV path planning. The present invention can adjust the driving paths of AGVs according to the spatial distribution of orders, reduce conflicts and path intersections between AGVs, improve the overall transportation efficiency, and provide an efficient, intelligent, and flexible AGV path planning solution for the logistics and warehousing industries. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 Shown is a schematic structural diagram of an AGV path planning device provided in Embodiment 1 of the present invention.

[0056] Figure 2 Shown is a flowchart of a method for AGV path planning based on order clustering provided in Embodiment 2 of the invention. DETAILED DESCRIPTION OF THE INVENTION

[0057] Exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be understood that the embodiments shown and described in the drawings are merely exemplary, intended to illustrate the principles and spirit of the present invention, and not to limit the scope of the present invention.

[0058] Example 1:

[0059] As Figure 1 shown, an AGV path planning device based on order clustering includes an order splitting module, an intelligent control module, and a material management module connected in sequence. The output end of the intelligent control module is also connected to an AGV scheduling module;

[0060] The order splitting module is used to obtain the original order information, cluster and split the original order, and output the order splitting result;

[0061] The intelligent control module is used to receive the order splitting result and perform AGV path planning according to the order splitting result, and output the AGV path planning information;

[0062] The material management module is used to receive the AGV path planning information and manage the warehousing and outwarehousing of materials according to the AGV path planning information;

[0063] The AGV scheduling module is used to receive the AGV path planning information, and according to the AGV path planning information, schedule the AGV to execute the operation task and control the start and stop of the AGV.

[0064] Example 2:

[0065] Based on Example 1, as Figure 2 shown, an embodiment of the present invention provides an AGV path planning method based on order clustering, including the following steps:

[0066] S1. Obtain the original order information, and cluster and split the original order according to the clustering algorithm to obtain the order splitting result, which specifically includes the following steps:

[0067] Perform One-Hot encoding on the non-numerical data of the original order to obtain the encoded order;

[0068] Use the improved particle swarm optimization algorithm to perform K-means clustering on the encoded order information to obtain the clustering result, which specifically includes the following steps:

[0069] S11. Load the data, obtain the quantity D of materials taken by the order, and obtain the maximum load UB of the AGV;

[0070] S12. Set the number of clusters to M according to the number M of AGVs;

[0071] S13. Set the maximum number of iterations MaxIter = 1000, the population size N to 20, the upper bound of the initial position ub = UB, and the lower bound of the initial position lb = 0, and set the dimension of the initial search space to the number of materials D in the order;

[0072] S14. Randomly initialize the population positions within the range of the upper and lower bounds of the initial positions, and set the initial velocity V = 0;

[0073] S15. Calculate the objective function value, which is the sum of the shortest Euclidean distances from each material to each cluster center;

[0074] S16. Randomly pair and compare two particles. Here, a particle represents a possible cluster center. Iteratively update the objective function value until the maximum number of iterations is reached, and output the global optimal solution, that is, the clustering result. Specifically:

[0075] Randomly pair and compare two particles, and update the position of the worse particle among the two particles;

[0076] Update the objective function value corresponding to the particle according to the updated position of the particle;

[0077] Judge whether the updated objective function value is better than the objective function value before the update. If so, set the flag flags(x2) = 0, indicating that the position of the particle after the update is better than the position of the particle before the update. Then, in the next iteration, the position update formula of particle x2 is:

[0078]

[0079] where w represents the inertia weight, IT represents the current iteration number, MaxIter represents the maximum number of iterations, V′(x2,j) represents the updated velocity of particle x2 in the next iteration, V(x2,j) represents the velocity of particle x2 before the update in the next iteration, X(b,j) represents the position of the global optimal solution in the j-th dimension, r2 and r3 represent random variables generated in the range [0, 1], with r2 = rand(), r3 = rand(), x1 represents the better particle among the two particles, X(x1,j) represents the position of particle x1 before the update in the next iteration, and round represents the rounding function;

[0080] If the objective function value before the update is better than the objective function value after the update, set the flag flags(x2) = 1, indicating that the position of the particle before the update is better than the position of the particle after the update. Then, in the next iteration, the position update formula of particle x2 is:

[0081] X′(x2,j) = UB / 2 - X(x2,j)

[0082] Among them, X′(x2,j) represents the position of particle x2 after update in the next iteration, X(x2,j) represents the position of particle x2 before update in the next iteration, and j represents the j-th dimension;

[0083] Repeat the update of particle positions and objective function values until the maximum number of iterations is reached, output the particle positions, and obtain the global optimal solution.

[0084] Split the orders according to the clustering results to obtain the order splitting results.

[0085] S2. According to the order splitting results, use the improved ant colony optimization algorithm to formulate the AGV path planning information, which specifically includes the following steps:

[0086] S21. Obtain the order splitting results, set the maximum number of iterations MaxIter = 1000, the population size N to 20, the pheromone evaporation factor ρ = 0.05, and the heuristic evaporation factor and the pheromone

[0087] S22. Randomly initialize the positions of the ant colony, and the dimension of the population is the number of materials in the order splitting results;

[0088] S23. Calculate the objective function value, that is, the sum of the Euclidean distances of all paths;

[0089] S24. Obtain the initial global optimal solution according to the objective function value;

[0090] S25. Randomly pair and compare two ants, iteratively update the positions of the ants until the maximum number of iterations is reached, output the final global optimal solution, and obtain the AGV path planning information, which specifically includes the following steps:

[0091] Randomly pair and compare two ants, and update the position of the worse ant y2 among the two ants;

[0092] Update the corresponding objective function value according to the updated position of ant y2;

[0093] Judge whether the updated objective function value is better than the objective function value before update. If so, calculate the probability P for an ant to select from node i to node j in the next iteration ij :

[0094]

[0095] where τ ij represents the pheromone concentration on path (i, j), represents the heuristic information factor on path (i, j), τ ik represents the pheromone concentration on path (i, k), Denote the heuristic information factor on path (i, k), and Z represents the set of integers;

[0096] Let the ant transfer according to probability P ij and then update the pheromone:

[0097]

[0098] where τ i ′ j represents the pheromone concentration on path (i, j) after update, ρ represents the pheromone evaporation factor, α represents the pheromone weight, and τ ij g represents the pheromone concentration of the global optimal solution g, represents the pheromone concentration of y1, f(y1) represents the objective function value of y1, and f(g) represents the objective function value of the global optimal solution g;

[0099] If the objective function value before update is better than that after update, then in the next iteration, the position update formula of ant y2 is:

[0100]

[0101] where t represents the temporary variable for saving randperm(T, 2), randperm(T, 2) represents randomly selecting two different numbers from the dimension T of the population, t1 represents the temporary variable for saving Y(y2, t(1)), Y(y2, t(1)) represents the position of ant y2 in dimension t(1), Y(y2, t(2)) represents the position of ant y2 in dimension t(2), t(1) represents the first number randomly selected from the two numbers in the population dimension T, and t(2) represents the second number randomly selected from the two numbers in the population dimension T;

[0102] Repeat updating the ant position and pheromone until the maximum number of iterations is reached, output the path with the highest ant position and pheromone concentration to obtain the global optimal solution, that is, the final AGV path planning information.

[0103] Those of ordinary skill in the art will realize that the embodiments described herein are for helping readers understand the principles of the present invention and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific deformations and combinations without departing from the essence of the present invention based on the technical revelations disclosed in the present invention, and these deformations and combinations are still within the protection scope of the present invention.

Claims

1. An AGV path planning device based on order clustering, characterized in that, It includes an order splitting module, an intelligent control module and a material management module which are connected in sequence, and the output end of the intelligent control module is also connected to the AGV scheduling module; The order splitting module is used to obtain original order information, cluster and segment the original order, and output the order splitting result; The intelligent control module is used to receive the order splitting result, perform AGV path planning according to the order splitting result, and output AGV path planning information; The material management module is used to receive the AGV path planning information and manage the warehousing and outbound transportation of materials according to the AGV path planning information; The AGV scheduling module is used to receive the AGV path planning information, and according to the AGV path planning information, schedule the AGV to perform the task and control the start and stop of the AGV.

2. An AGV path planning method applied to the AGV path planning device based on order clustering described in claim 1, characterized in that, The following steps are involved: Obtain the original order information, and cluster and segment the original order according to the clustering algorithm to obtain the order splitting result; According to the order splitting results, the improved ant colony optimization algorithm is used to formulate the AGV path planning information.

3. The AGV path planning method of the AGV path planning device based on order clustering according to claim 2, characterized in that, The method of clustering and segmenting the original order according to the clustering algorithm to obtain the order segmentation result specifically includes the following steps: One-hot encode the non-numeric data of the original order to obtain the encoded order; The improved particle swarm optimization algorithm is used to perform K-means clustering on the encoded order information to obtain the clustering results; The orders are split according to the clustering results to obtain the order splitting results.

4. The AGV path planning method of the AGV path planning device based on order clustering according to claim 3, wherein The improved particle swarm optimization algorithm is used to perform K-means clustering on the encoded order information, specifically including the following steps: Load data, obtain the quantity D of materials taken by the order, and obtain the maximum load UB of the AGV; Set the number of clusters to M according to the number of AGVs M; Set the maximum number of iterations, the number of populations, the upper bound of the initial position ub = UB and the lower bound of the initial position lb = 0, and set the dimension of the initial search space to the number of materials taken by the order D; Randomly initialize the population position within the range of the initial position upper bound and the initial position lower bound, and set the initial velocity V = 0; Calculate the objective function value, where the objective function value is the sum of the shortest Euclidean distances from each material to each cluster center; Particles are randomly paired and compared, and the objective function value is iteratively updated until the maximum number of iterations is reached, and the global optimal solution, i.e., the clustering result, is output; wherein a particle represents a possible clustering center.

5. The AGV path planning method of the AGV path planning device based on order clustering according to claim 4, characterized in that The random particle pairing is compared, and the objective function value is iteratively updated until the maximum number of iterations is reached, and the global optimal solution is output. The following steps are involved: Randomly compare two particles and update the position of the worse particle. According to the updated position of the particle, update the objective function value corresponding to the particle; If the updated objective function value is better than the objective function value before the update, then in the next iteration, the position update formula of particle x2 is: where \( w \) represents the inertia weight, \( IT \) represents the current iteration number, \( MaxIter \) represents the maximum iteration number, \( V'(x2,j) \) represents the velocity of particle \( x2 \) after update in the next iteration, \( V(x2,j) \) represents the velocity of particle \( x2 \) before update in the next iteration, \( X(b,j) \) represents the position of the global optimal solution in the \( j \)-th dimension, \( r2 \) and \( r3 \) represent random variables generated in the range \([0, 1]\), with \( r2 = rand() \) and \( r3 = rand() \), \( x1 \) represents the better particle among the two particles, \( X(x1,j) \) represents the position of particle \( x1 \) before update in the next iteration, and \( round \) represents the rounding function; If the objective function value before update is better than that after update, then in the next iteration, the position update formula for particle \( x2 \) is: \( X'(x2,j) = UB / 2 - X(x2,j) \) where \( X'(x2,j) \) represents the position of particle \( x2 \) after update in the next iteration, \( X(x2,j) \) represents the position of particle \( x2 \) before update in the next iteration, and \( j \) represents the \( j \)-th dimension; Repeat updating the particle positions and the objective function values until the maximum iteration number is reached, output the particle positions, and obtain the global optimal solution.

6. The AGV path planning method of the AGV path planning device based on order clustering according to claim 2, wherein, Said to formulate the AGV path planning information using the improved ant colony optimization algorithm according to the order splitting result, specifically including the following steps: Obtain the order splitting result, and set the maximum iteration number, population size, pheromone evaporation factor, heuristic evaporation factor, and pheromone; Randomly initialize the positions of the ant colony, and the dimension of the population is the number of materials in the order splitting result; Calculate the objective function value, that is, the sum of the Euclidean distances of all paths; Obtain the initial global optimal solution according to the objective function value; Randomly compare pairs of ants, iteratively update the ant positions until the maximum iteration number is reached, output the final global optimal solution, and obtain the AGV path planning information.

7. The AGV path planning method of the AGV path planning device based on order clustering according to claim 6, characterized in that Said randomly comparing pairs of ants, iteratively updating the ant positions until the maximum iteration number is reached, and outputting the final global optimal solution, specifically including the following steps: Randomly compare pairs of ants, and update the position of the worse ant \( y2 \) among the two ants; Update the corresponding objective function value according to the position of ant \( y2 \) after update; If the value of the updated objective function is better than that of the objective function before the update, then calculate the probability P that the ant chooses to move from node i to node j in the next iteration ij : Among them, τ ij represents the pheromone concentration on path (i, j), represents the heuristic information factor on path (i, j), τ ik represents the pheromone concentration on path (i, k), represents the heuristic information factor on path (i, k), and Z represents the set of integers; Let the ants transfer according to the probability P ij and then update the pheromone: Among them, τ i ′ j represents the pheromone concentration on the updated path (i, j), ρ represents the pheromone evaporation factor, α represents the pheromone weight, and τ ij g represents the pheromone concentration of the global optimal solution g, represents the pheromone concentration of y1, f(y1) represents the objective function value of y1, and f(g) represents the objective function value of the global optimal solution g; If the objective function value before update is better than that after update, then in the next iteration, the position update formula for ant \( y2 \) is: where \( t \) represents a temporary variable used to store \( randperm(T,2) \), \( randperm(T,2) \) represents randomly selecting two different numbers from the dimension \( T \) of the population, \( t1 \) represents a temporary variable used to store \( Y(y2,t(1)) \), \( Y(y2,t(1)) \) represents the position of ant \( y2 \) in the \( t(1) \)-th dimension, \( Y(y2,t(2)) \) represents the position of ant \( y2 \) in the \( t(2) \)-th dimension, \( t(1) \) represents the first of the two numbers randomly selected from the dimension \( T \) of the population, and \( t(2) \) represents the second of the two numbers randomly selected from the dimension \( T \) of the population; Repeat updating the ant positions and pheromones until the maximum number of iterations is reached, output the ant positions and the path with the highest pheromone concentration to obtain the global optimal solution, that is, the final AGV path planning information.

Images