An optimized scheduling method and system for intelligent warehouse picking tasks

Abstract

This invention relates to the field of intelligent warehousing task scheduling technology, and in particular to an optimized scheduling method and system for intelligent warehousing picking tasks. The optimized scheduling method for intelligent warehousing picking tasks includes: reading information from an e-commerce platform to obtain an ID-based order set; based on the ID-based order set, obtaining an optimal order cluster set using a preset K-means clustering algorithm; matching the optimal order cluster set with picking workstations to obtain a product picking ID sequence; obtaining an optimized AGV scheduling task sequence using a multi-strategy whale algorithm based on the product picking ID sequence and preset AGV picking constraints; and scheduling the AGVs according to the optimized AGV scheduling task sequence. This invention provides a highly efficient, low-idle-load optimized scheduling method for picking tasks.

Description

Technical Field

[0001] This invention relates to the field of intelligent warehousing task scheduling technology, and in particular to an optimized scheduling method and system for intelligent warehousing picking tasks. Background Technology

[0002] To address the inefficiencies and high costs of traditional logistics, improving the sorting efficiency of goods orders has become a crucial aspect of logistics warehousing. Automated Guided Vehicles (AGVs), also known as mobile robots, are intelligent agents integrating multiple sensors that can autonomously make decisions and complete specific tasks by sensing their external environment. They are primarily used in automated transportation within logistics warehousing systems, replacing other manual picking and handling tasks, significantly improving goods handling efficiency and shortening logistics transit time.

[0003] Multi-AGV collaboration can be categorized into three forms: centralized, distributed, and hybrid. Optimization methods for order tasks include linear programming, market mechanisms, behavior-based algorithms, and swarm intelligence. In the centralized allocation scheme, a different planning strategy from offline path generation and online path planning is considered. A multi-robot path planning strategy based on a path planning layer and a collision avoidance layer is proposed: In the first layer, only the path planning of a single robot is considered, ignoring the influence of other robots, and the planned path may result in severe collisions with other robots; in the second layer, collision avoidance rules are used to solve local collisions among multiple robots. These collision avoidance rules are complex and require pre-determined human intervention, making them unchangeable based on the current characteristics of the multi-robot system.

[0004] In the research of distributed control systems, a multi-robot distributed path planning method based on neighborhood artificial potential fields is proposed. This method utilizes a distributed architecture to plan the trajectory of mobile robots and employs an improved artificial market method to achieve good trajectory planning for the multi-robot system, thereby improving its robustness. Hybrid task allocation combines the advantages of centralized and distributed systems while mitigating some of their shortcomings; however, hybrid systems are more complex and face greater design and application challenges.

[0005] The market-based algorithm borrows the fair bidding mechanism from auction theory in economics, treating each task as a commodity to be auctioned, with each robot acting as a bidder to acquire tasks. Based on the pricing mechanism during negotiations between the robot and the workpiece agent, a corresponding algorithm was designed. Experiments show that this algorithm outperforms traditional scheduling rules in improving overall system performance, but performs poorly in terms of average latency and maximum latency.

[0006] In the existing technology, there is a lack of a highly efficient and low-idle-load optimized scheduling method for picking tasks. Summary of the Invention

[0007] This invention provides an optimized scheduling method and system for intelligent warehouse picking tasks. The technical solution is as follows:

[0008] On the one hand, an optimized scheduling method for intelligent warehouse picking tasks is provided, which is implemented by electronic devices and includes:

[0009] S1. Read information from the e-commerce platform to obtain a set of ID-based orders;

[0010] S2. Based on the ID-based order set, obtain the optimal order cluster set using a preset K-means clustering algorithm;

[0011] S3. Match the optimal cluster set of orders with the picking workstations to obtain the product picking ID sequence;

[0012] S4. Based on the product picking ID sequence and the preset AGV picking constraints, obtain the AGV optimized scheduling task sequence through the multi-strategy whale algorithm;

[0013] S5. Schedule the AGVs according to the optimized AGV scheduling task sequence.

[0014] Optionally, the step of reading e-commerce platform information to obtain an ID-based order set includes:

[0015] Read information from the e-commerce platform to obtain a collection of product orders;

[0016] The product order set is assigned an ID number to obtain an ID-based order set.

[0017] Optionally, the step of obtaining the optimal cluster set of orders based on the ID-based order set using a preset K-means clustering algorithm includes:

[0018] S21. Perform cluster center selection operation based on the ID-based order set to obtain cluster centers;

[0019] S22. Calculate the distance between each order and the cluster center based on the ID-based order set and the cluster center;

[0020] S23. Based on the distance between each order and the cluster center, the ID-based order set is clustered to obtain an order cluster set;

[0021] S24. Repeat steps S21-S23 until the number of iterations reaches its maximum, then obtain the optimal cluster set of orders.

[0022] Optionally, the step of matching the optimal cluster set of orders with picking workstations to obtain a sequence of product picking IDs includes:

[0023] Each order in the optimal order cluster set is matched with a picking workstation to obtain a set of matched workstations;

[0024] For each workstation in the set of matched workstations, the matched order clusters are analyzed to obtain the product picking ID sequence.

[0025] Optionally, the step of obtaining an optimized AGV scheduling task sequence based on the goods picking ID sequence and preset AGV picking constraints using a multi-strategy whale algorithm includes:

[0026] S41. Based on the product picking ID sequence and the preset AGV picking constraints, initialize the multi-strategy whale algorithm to obtain the AGV scheduling task sequence.

[0027] S42. Calculate the idle distance generated by the scheduling tasks in the AGV scheduling task sequence to obtain the AGV scheduling idle distance;

[0028] S43. Update the AGV scheduling task sequence to obtain the AGV update scheduling task sequence;

[0029] S44. Repeat steps S41-S43. When the AGV scheduling idle distance is the minimum, obtain the AGV optimized scheduling task sequence.

[0030] Optionally, updating the AGV scheduling task sequence to obtain an updated AGV scheduling task sequence includes:

[0031] The AGV scheduling task sequence is used as the current population, and the fitness is calculated to obtain the population fitness.

[0032] Based on the population fitness, the current population is divided into population A and population B;

[0033] Based on population A, obtain the AGV scheduling task sequence in population A; the AGV scheduling task sequence in population A is optimized by a preset improved search algorithm to obtain the AGV update scheduling task sequence of population A.

[0034] Based on the population B, the AGV scheduling task sequence in population B is obtained; the AGV scheduling task sequence in population B is optimized by a preset improved multi-strategy optimization algorithm to obtain the AGV update scheduling task sequence of population B.

[0035] Based on the AGV update scheduling task sequence of population A and the AGV update scheduling task sequence of population B, an AGV update scheduling task sequence is obtained.

[0036] The preset improved search algorithm adds an improved search model to the original whale algorithm's search strategy; the improved search model uses individuals in population A as the search center to search for the positions of other individuals in the surrounding area.

[0037] The preset improved multi-strategy optimization algorithm is an improved optimization algorithm based on the optimization guidance mathematical model of the Harry Eagle algorithm and the gene replacement mathematical model of the differential evolution algorithm.

[0038] On the other hand, an optimized scheduling system for intelligent warehouse picking tasks is provided. This system is applied to an optimized scheduling method for intelligent warehouse picking tasks. The system includes electronic equipment, picking workstations, and AGVs, wherein:

[0039] The electronic device is used to read information from the e-commerce platform to obtain an ID-based order set; based on the ID-based order set, it obtains the optimal order cluster set through a preset K-means clustering algorithm; and according to the product picking ID sequence and preset AGV picking constraints, it obtains the AGV optimized scheduling task sequence through a multi-strategy whale algorithm.

[0040] The picking workstation is used to match the optimal cluster set of orders with the picking workstation to obtain the product picking ID sequence;

[0041] The AGV is used to optimize and schedule task sequences based on the AGV and execute scheduling.

[0042] Optionally, the electronic device is further configured to:

[0043] Read information from the e-commerce platform to obtain a collection of product orders;

[0044] The product order set is assigned an ID number to obtain an ID-based order set.

[0045] Optionally, the electronic device is further configured to:

[0046] S21. Perform cluster center selection operation based on the ID-based order set to obtain cluster centers;

[0047] S22. Calculate the distance between each order and the cluster center based on the ID-based order set and the cluster center;

[0048] S23. Based on the distance between each order and the cluster center, the ID-based order set is clustered to obtain an order cluster set;

[0049] S24. Repeat steps S21-S23 until the number of iterations reaches its maximum, then obtain the optimal cluster set of orders.

[0050] Optionally, the picking workstation is further used for:

[0051] Each order in the optimal order cluster set is matched with a picking workstation to obtain a set of matched workstations;

[0052] For each workstation in the set of matched workstations, the matched order clusters are analyzed to obtain the product picking ID sequence.

[0053] Optionally, the electronic device is further configured to:

[0054] S41. Based on the product picking ID sequence and the preset AGV picking constraints, initialize the multi-strategy whale algorithm to obtain the AGV scheduling task sequence.

[0055] S42. Calculate the idle distance generated by the scheduling tasks in the AGV scheduling task sequence to obtain the AGV scheduling idle distance;

[0056] S43. Update the AGV scheduling task sequence to obtain the AGV update scheduling task sequence;

[0057] S44. Repeat steps S41-S43. When the AGV scheduling idle distance is the minimum, obtain the AGV optimized scheduling task sequence.

[0058] Optionally, the electronic device is further configured to:

[0059] The AGV scheduling task sequence is used as the current population, and the fitness is calculated to obtain the population fitness.

[0060] Based on the population fitness, the current population is divided into population A and population B;

[0061] Based on population A, obtain the AGV scheduling task sequence in population A; the AGV scheduling task sequence in population A is optimized by a preset improved search algorithm to obtain the AGV update scheduling task sequence of population A.

[0062] Based on the population B, the AGV scheduling task sequence in population B is obtained; the AGV scheduling task sequence in population B is optimized by a preset improved multi-strategy optimization algorithm to obtain the AGV update scheduling task sequence of population B.

[0063] Based on the AGV update scheduling task sequence of population A and the AGV update scheduling task sequence of population B, an AGV update scheduling task sequence is obtained.

[0064] The preset improved search algorithm adds an improved search model to the original whale algorithm's search strategy; the improved search model uses individuals in population A as the search center to search for the positions of other individuals in the surrounding area.

[0065] The preset improved multi-strategy optimization algorithm is an improved optimization algorithm based on the optimization guidance mathematical model of the Harry Eagle algorithm and the gene replacement mathematical model of the differential evolution algorithm.

[0066] On the other hand, an electronic device is provided, comprising a processor and a memory, wherein the memory stores at least one instruction, which is loaded and executed by the processor to implement the above-described optimized scheduling method for intelligent warehouse picking tasks.

[0067] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, and the at least one instruction is loaded and executed by a processor to implement the above-described optimized scheduling method for intelligent warehouse picking tasks.

[0068] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0069] This invention proposes an optimized scheduling method for intelligent warehouse picking tasks. It employs the K-means clustering algorithm to effectively reduce the number of times AGVs move shelves in the intelligent warehousing system; and utilizes a multi-strategy whale algorithm to shorten the empty-load transport distance of AGVs and reduce logistics turnaround time. This invention can be applied to the construction of novel automated intelligent warehousing systems, and the proposed optimized scheduling method based on order characteristics can greatly improve order picking efficiency. This invention is a high-efficiency, low-empty-load optimized scheduling method for picking tasks. Attached Figure Description

[0070] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0071] Figure 1 This is a flowchart of an optimized scheduling method for intelligent warehouse picking tasks provided in an embodiment of the present invention;

[0072] Figure 2 This is a schematic diagram of an intelligent warehousing environment provided in an embodiment of the present invention;

[0073] Figure 3 This is a schematic diagram of an AGV shelf handling system provided in an embodiment of the present invention;

[0074] Figure 4 This is a block diagram of an optimized scheduling system for intelligent warehouse picking tasks provided in an embodiment of the present invention;

[0075] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0076] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0077] This invention provides an optimized scheduling method for intelligent warehouse picking tasks. This method can be implemented by an electronic device, which can be a terminal or a server. Figure 1 The flowchart shown is an optimized scheduling method for intelligent warehouse picking tasks. The processing flow of this method may include the following steps:

[0078] S1. Read information from the e-commerce platform to obtain a set of ID-based orders.

[0079] Optionally, information from the e-commerce platform can be read to obtain a set of ID-based orders, including:

[0080] Read information from the e-commerce platform to obtain a collection of product orders;

[0081] Assign ID numbers to the product order set to obtain an ID-based order set.

[0082] In one feasible implementation, the intelligent warehousing system of this invention revolutionarily upgrades the traditional "person-to-goods" model to a "goods-to-person" model, adapting to the miniaturized characteristics of modern logistics and distribution units. The control center of the intelligent warehousing system, composed of electronic equipment, possesses powerful computing capabilities. Its main tasks are to complete the real-time management and updating of stored goods, the management and control of warehousing hardware equipment, and the allocation and scheduling of order tasks. It is the control core of the intelligent warehousing system and an indispensable component of the system.

[0083] When order information for goods is retrieved from an e-commerce platform, the location of the goods on the corresponding shelf in the smart warehouse is determined based on the order information. The goods in the order set are then assigned IDs, resulting in an ID-based order set. The smart warehouse environment is as follows: Figure 2 As shown.

[0084] The top section houses the control center for the intelligent warehouse, comprised of electronic devices, used to schedule and control the hardware within the warehouse. The leftmost section is the picking service station, where pickers wait for AGVs to transport goods from relevant shelves to the station to complete orders. The central area is the storage rack area, with each square representing a specific type of storage rack, neatly arranged within the designated area. Circular icons represent AGVs. When idle, AGVs either wait for task assignment in the idle waiting area below or charge in the charging area on the right. When in operation, AGVs can move freely within the warehouse environment, using planned, collision-free paths to transport the storage racks represented by the squares to the picking workstations.

[0085] A grid-based method was used to model the storage shelving area, treating the environment as a two-dimensional plane divided into multiple equally sized grids. Computer-generated numbers "0" and "1" were used to determine the presence of obstacles within the grid environment, and each grid was numbered to describe its relative position. The grid map consisted of two types of grids: free space grids, representing units where the mobile robot could traverse, and obstacle grids, representing the locations of obstacles in the current environment, areas where the mobile robot typically could not move.

[0086] S2. Based on the ID-based order set, obtain the optimal order cluster set through the preset K-means clustering algorithm.

[0087] Optionally, based on the ID-based order set, the optimal order cluster set is obtained using a pre-defined K-means clustering algorithm, including:

[0088] S21. Perform cluster center selection operation based on the ID-based order set to obtain the cluster centers;

[0089] S22. Calculate the distance between each order and the cluster center based on the ID-based order set and cluster center;

[0090] S23. Based on the distance between each order and the cluster center, cluster the ID-based order set to obtain the order cluster set;

[0091] S24. Repeat steps S21-S23 until the number of iterations reaches its maximum, then obtain the optimal cluster set of orders.

[0092] In one feasible implementation, this invention employs the EIQ (E: Order Entry, I: Item, Q: Quantity) programming method to comprehensively analyze the order characteristics of e-commerce platforms based on the diversity of network users, providing a theoretical basis for optimizing order allocation and sorting. Combined with big data survey results, e-commerce platform orders are characterized by low EN values, low EQ values, large fluctuations in IK values, and large fluctuations in IQ values; that is, e-commerce platform order tasks exhibit characteristics of small batches, multiple categories, large batches, and significant fluctuations.

[0093] Due to the large order volume, assigning picking tasks to picking stations using random or first-come, first-served (FCFS) strategies can easily lead to multiple handling of the same type of goods on the shelves, increasing the workload of Automated Guided Vehicles (AGVs) and reducing picking efficiency. Therefore, to further improve AGV handling efficiency, it is necessary to first process the order tasks received by the smart warehouse, assigning goods with similar order contents to the same picking station for picking, thereby improving sorting efficiency.

[0094] In intelligent warehousing systems, since AGVs typically travel at a constant speed while the picking speed of workers at workstations is relatively fixed, the objective of order task allocation is to minimize the rack handling distance. Let k AGVs handle the handling of n racks in the warehousing system; then the total distance traveled by all AGVs is S, expressed mathematically as follows (1):

[0095]

[0096] Where, d s This represents the distance traveled by the s-th AGV from the idle area to the shelf area for the first time. Since the locations of the idle waiting area, sorting station, and shelf area are fixed, therefore... This is usually a fixed value and can be ignored. d i The distance the AGV needs to travel to move the shelf to the picking station is twice that distance because the shelf needs to be restored. i This represents the number of times the shelf is handled. Since the same shelf may be needed by different picking stations, the shelf may need to be handled multiple times. For all picking stations m in the entire intelligent warehousing system, the objective function is as follows (2):

[0097]

[0098] When all order tasks are assigned to a picking station, the items that need to be picked at that station are already determined. That is, before and after order task assignment, the shelves to be picked are already determined, i.e., the distance d between the picking stations. i It has been determined. The difference in AGV transport distance comes from the number of times the shelves are moved. Therefore, in order to minimize the total distance of AGV picking orders, it is necessary to minimize the number of times the shelves are moved when assigning order tasks. Therefore, the objective function is transformed into the following equation (3):

[0099]

[0100] When orders with identical items are assigned to the same picking station, the AGV can handle multiple orders with the same items in a single shelf movement, reducing the number of shelf movements. Therefore, to minimize shelf movement by analyzing order characteristics, K-means clustering is used to cluster orders with identical items before assigning them to the same picking station. To quantify the differences between orders, O = {x1, x2, ... x...} k}, x1, x2...x k The degree of similarity between x1, x2, ... x2 ∈ [1, 120], where x1, x2, ... x2 ∈ [1, 120]. k For the product ID, the mathematical expression of the quantification model used is shown in equation (4) below:

[0101]

[0102] Among them, MO ij For order O i and O j The intersection of item IDs, i.e., the union of IDs of the same items in two orders, is calculated by MO. ij The cardinality SI, i.e., card(MO) ij Determine the similarity between two orders. When MO ij When the cardinality SI is empty, it is 0, and the similarity between the two orders is 0. When two orders contain the same item, the cardinality SI is 1, and the order similarity is 1. The more items that contain the same item, the larger the cardinality SI of the set, the higher the order similarity, and the greater the probability of them being clustered into the same picking workstation.

[0103] There are many different goods between different orders, and the empty travel distance of AGV between different shelves also needs to be considered. That is, after the AGV completes this shelf handling task, it restores the shelf from the picking workstation to the initial position and starts to move the next shelf. When two shelves are close to each other, the AGV handling distance can be further reduced. According to the shelf layout rules of the warehousing system, the shelves with small ID differences are closer together. Therefore, when clustering, the differences between orders also need to be considered, that is, the distance between the goods of two orders in the intelligent warehousing system in the storage location. Its quantitative model (5) is as follows:

[0104]

[0105] in, Representing order O j Compared to order O i The supplement, Representing order O i Compared to order O j The complement of SO. ij This represents the union of two complements, i.e., the set of IDs where the goods in the two orders are completely different. Let set SO be an example. ij The dispersion of the ID in the warehousing system represents the difference in the distance between the required shelf locations of two orders in the warehousing system, i.e., SO ij The standard deviation of ID in the set is used as the difference DI between two orders. A smaller standard deviation indicates that the shelves containing the required goods for the two orders are located in a locally concentrated area. A larger standard deviation indicates that the shelves containing the required goods are more dispersed. Where N represents SO ij The total number of items in the set, u is the mean of item IDs, and x is the total number of items in the set. i The item ID in the set.

[0106] Randomly select m orders as initial cluster centers. Calculate the distance between all orders and the m cluster centers using order similarity (SI) and order dissimilarity (DI), and assign each order to the cluster set of orders closest to it. After each clustering iteration, recalculate the centroid of each order cluster set, which is the centroid of all samples in that order category. Calculate the number of AGV handling operations required for the current order cluster set, and repeat the above steps until the maximum number of iterations is met, resulting in m order cluster sets. Here, m corresponds to the number of picking workstations.

[0107] S3. Match the optimal cluster set of orders with the picking workstations to obtain the order picking ID sequence.

[0108] Optionally, the optimal cluster set of orders is matched with picking workstations to obtain a sequence of item picking IDs, including:

[0109] Each order cluster in the optimal order cluster set is matched with a picking workstation to obtain the set of matched workstations;

[0110] For each workstation in the matched workstation set, analyze the clustering of matched orders to obtain the product picking ID sequence.

[0111] In one feasible implementation, the sorting service station in this invention is located in a fixed position in the smart warehouse and is where staff pick up the goods from the orders.

[0112] When the AGV transports the shelves to the sorting station, the picking workers will retrieve the corresponding quantity of goods according to the order task to complete the picking task, and update the order completion status and warehouse inventory in real time by scanning the barcodes of the goods. The picking workstation analyzes the matched order clusters to obtain the sequence of item IDs waiting to be picked.

[0113] S4. Based on the product picking ID sequence and the preset AGV picking constraints, obtain the AGV optimized scheduling task sequence through the multi-strategy whale algorithm.

[0114] In one feasible implementation, during the handling process, the destination of the goods to be picked is fixed at the picking workstation that accepts the order, and all shelves in the warehousing system are always in fixed positions, and the contents of the order are also determined.

[0115] Therefore, the starting and ending points of the goods required for the picking station are fixed. Without considering other AGV congestion paths, the shortest path distance for the AGV to transport the goods from the shelf to the picking station is fixed. Therefore, a way to further shorten the AGV's travel distance is to reduce the idle distance the AGV travels in the shelf area during task execution, i.e., to change the order in which the AGVs perform transport tasks. If the shelf locations of adjacent tasks in the AGV task list are far apart, it will increase the AGV's idle distance in the shelf area, prolonging order completion time, increasing the overall travel distance, and reducing transport efficiency.

[0116] Optionally, based on the product picking ID sequence and preset AGV picking constraints, an optimized AGV scheduling task sequence is obtained using a multi-strategy whale algorithm, including:

[0117] S41. Based on the product picking ID sequence and the preset AGV picking constraints, initialize the multi-strategy whale algorithm to obtain the AGV scheduling task sequence.

[0118] S42. Calculate the idle distance generated by the scheduling tasks in the AGV scheduling task sequence to obtain the AGV scheduling idle distance;

[0119] S43. Update the AGV scheduling task sequence to obtain the updated AGV scheduling task sequence.

[0120] S44. Repeat steps S41-S43 until the AGV scheduling idle distance is minimized, then obtain the AGV optimized scheduling task sequence.

[0121] One feasible implementation method is, for example Figure 3 A schematic diagram of AGV shelf handling shows the total travel distance when two AGVs jointly perform four item picking tasks. L1 and L2 represent the travel distances of the two AGVs when handling the shelf. Figure 3 In the left image, the first AGV transports shelves 12 and 2, and the second AGV transports shelves 14 and 4. The total travel distance of the two AGVs is 60, and the unloaded distance is 10. Figure 3 In the right-hand diagram, the first AGV transports shelves 12 and 14, while the second AGV transports shelves 2 and 4. The total travel distance of the two AGVs is 54, and the empty-load distance is 4. This shows that although the total number and location of shelves required by the picking station remain constant, different AGV transport task sequences result in significant differences in the total travel distance. The reduction in total travel distance stems from the reduction in empty-load distance. Therefore, by scheduling shelf transport tasks at the picking station and rationally allocating shelf transport task sequences to each AGV, it is possible to effectively reduce AGV empty-load and shorten order completion time.

[0122] To maximize handling efficiency, a reasonable task sequence needs to be developed for each AGV to reduce its idle distance. Based on the layout of the storage rack area, AGV motion constraints and quantity, and upper and lower limits of AGV task volume, AGV task scheduling is modeled and analyzed.

[0123] In an intelligent warehousing system, this model has the following characteristics: Due to factors such as equipment procurement costs and transportation requirements, the total number of AGVs is limited. AGVs, constrained by their own power consumption and multi-AGV collaboration, can only perform a limited number of handling tasks within a given timeframe. For orders at the same picking workstation, the required storage racks only need to be moved by the AGVs once, allowing the picking operator to retrieve the required quantity of that type of goods for that workstation. Based on the warehouse system's rack layout, goods with similar IDs are located adjacent to each other in the warehouse. The goal of this model is to minimize the idle distance (LC) of all AGVs, reduce the total AGV handling distance, and improve rack handling speed.

[0124] Based on the above analysis, the required item IDs of the same picking workstation are sorted in ascending order to obtain the workstation's shelf demand sequence W. c ={x1, x2...x n}, x1∈x2<...<xn The scheduling problem for AGV rack handling tasks is then transformed into: based on the constraint of the AGV handling task volume, ensuring that the AGV sequentially intercepts the workstation rack sequence W. c When a certain amount of shelving handling tasks are performed, the AGV's empty travel distance LC is minimized. The evaluation function and constraints for the AGV task scheduling model are constructed as shown in equation (6) below:

[0125]

[0126] st l min ≤l≤l max

[0127] Let l be the amount of goods handled by the AGV, then x i，j =(x i，j，1 x i，j，2 (1≤i≤n,1≤j≤l) represents the position of the j-th cargo in the task list of the i-th AGV, d(x i，j x i，j+1 Let represent the empty distance traveled by the i-th AGV in the storage rack area when it moves to the (j+1)-th item after completing the transport of the j-th item. The constraint indicates that there is a limit to the number of racks the AGV can handle within a certain period, meaning the amount of tasks the AGV can acquire is limited.

[0128] Optionally, the AGV scheduling task sequence is updated to obtain an updated AGV scheduling task sequence, including:

[0129] The AGV scheduling task sequence is used as the current population, and the fitness is calculated to obtain the population fitness.

[0130] Based on population fitness, the current population is divided into population A and population B;

[0131] Based on population A, obtain the AGV scheduling task sequence in population A; optimize the AGV scheduling task sequence in population A using a preset improved search algorithm to obtain the AGV update scheduling task sequence for population A.

[0132] Based on population B, obtain the AGV scheduling task sequence in population B; optimize the AGV scheduling task sequence in population B using a preset improved multi-strategy optimization algorithm to obtain the AGV update scheduling task sequence in population B.

[0133] Based on the AGV update scheduling task sequence of population A and the AGV update scheduling task sequence of population B, the AGV update scheduling task sequence is obtained.

[0134] In one feasible implementation, the objective function of formula (6) in this invention is a typical NP-hard problem, and an approximate solution to this problem can be achieved by using a heuristic optimization algorithm. This invention proposes a multi-strategy whale optimization algorithm (MWOA) that balances the exploration and development capabilities of the original whale optimization algorithm and adjusts the direction of the algorithm's search to avoid getting trapped in local optima.

[0135] In the proposed multi-strategy whale optimization algorithm, to maintain good search capabilities during iterative solving, the search strategy of this invention splits the population into two fixed-number populations based on the fitness value of the objective function. The top 40% of individuals with the best fitness values ​​in the overall population are assigned to the first population, named Population A, which is suitable for an elite search strategy to perform local searches and improve solution accuracy. The remaining 60% of individuals are assigned to the second population, named Population B, which is suitable for a general search strategy to enhance solution diversity. The two populations are suited to different search strategies, ensuring that the algorithm does not lose its global or local search capabilities during iterative updates of parameters.

[0136] Once an iterative search is completed, an individual's fitness value changes. The population is then re-divided based on the individual's ranking within the overall population. In other words, individuals continuously switch between the two populations and apply different search strategies based on their fitness value within the overall population.

[0137] The preset improved search algorithm adds an improved search model to the original whale algorithm's search strategy; the improved search model uses individuals in population A as the search center and searches for the positions of other individuals in the surrounding area.

[0138] In one feasible implementation, the fitness of individual whales in population A is among the better parts of the overall population, making them suitable for a local search strategy. Therefore, in addition to using the local search strategy in the original whale algorithm, individual whales can also search their surrounding area with themselves as the center to find a better position. The mathematical expression of the improved search model is shown in equation (7) below:

[0139]

[0140] ε is any random value between [0,1]. This search method is added to the local search strategy, and it is parallel to the local search strategy in the original whale algorithm. The improved search model applicable to population A is shown in equation (7). The mathematical expression of the improved search algorithm applicable to population A is shown in equation (8):

[0141]

[0142] In this process, depending on the different random probabilities p, one of two search strategies is randomly selected to search for the optimal solution. Represents the current best individual. Representing the current individual, represents the distance between the current individual whale and the optimal individual, b is the logarithmic spiral constant, l is a random number between [-1, 1], and e is the natural constant.

[0143] Among them, the preset improved multi-strategy optimization algorithm is an improved optimization algorithm based on the optimization guidance mathematical model of the Harry Eagle algorithm and the gene replacement mathematical model of the differential evolution algorithm.

[0144] In one feasible implementation, population B employs an improved multi-strategy optimization algorithm that combines the Harry Hawk (HHO) algorithm and the Differential Evolution (DE) multi-strategy model. The soft-surrounding design in the HHO algorithm strategy references the average position of the current population and guides the whale optimization algorithm to continuously explore towards the optimal solution, accelerating the algorithm's convergence speed. The DE algorithm utilizes the idea of ​​gene substitution. First, a gene template is constructed by proportionally cross-crossing and replacing the genes of three random individuals. Then, new individuals are created by randomly exchanging some genes in the current individuals using the template.

[0145] This method can retain some information of the current individual and, to a certain extent, accept the solution information of other individuals, thus achieving the fusion of multiple information and helping to avoid the algorithm getting trapped in local optima. The mathematical expression of the model of population B is shown in equation (9) below:

[0146]

[0147] in, and It is a parameter vector. The vector decreases linearly from 2 to 0 during the iteration process. The components of the vector are random numbers with values ​​between [0, 1]. It is the average position of the population. It is any individual in the population; N is the population size; r1 and r2 are random numbers between 0 and 1; LB and UB are the upper and lower limits of the objective function's range; E = 2E0(1-t / T) represents the energy parameter; X r1 (t), X r2 (t), X r3(t) represents the position of any individual in the population; M is a sequence containing only 0s and 1s, with a length comparable to the search dimension, used to select the gene to be retained in the template gene; Fa is the crossover ratio factor used to generate the template gene. Individuals in population B use the optimal one of the three search strategies in model formula (9) to search for the optimal solution.

[0148] S5. Schedule the AGVs according to the AGV optimization scheduling task sequence.

[0149] In one feasible implementation, the AGV in this invention is mainly used for cargo handling in warehousing and logistics systems. It achieves autonomous path planning by providing given task start and end information. It is an important execution unit for improving the automatic sorting and rapid handling of warehouse goods and comprehensively improving the operational efficiency of warehousing, transportation, and distribution.

[0150] In an intelligent warehousing system, after a picking workstation receives an assigned order, it needs to distribute it to an AGV for shelving transport to complete the picking task. Ignoring the distance from the idle area to the starting point of the first transport task, the AGV's transport process is as follows: it receives a batch of order tasks from the picking workstation (requiring the transport of at least one item), travels from the idle area to the shelving location of the first transport task, transports the shelving to the picking workstation that requires the item, and after the picking is completed, returns the shelving to its original location in the shelving area. Then, it proceeds to the shelving location of the next task to begin the next transport task.

[0151] This invention compares the minimum number of handling operations obtained by two methods under different order volumes. By investigating and analyzing the order characteristics of e-commerce platforms, the system randomly generates orders, ensuring that each order contains a minimum of 1 item and a maximum of 5 items. The total number of orders increases from 400 to 2000, with each increase being 400 orders. The maximum number of orders that each picking station can pick within a certain period is 50.

[0152] All data are recorded in Table 1 (Table 1: Statistical Results of Comparative Experiments). As the number of orders increased rapidly, the number of handling operations and the total number of handling operations for both task allocation strategies began to increase. However, the total number of AGV handling operations calculated using the K-means task allocation strategy based on order characteristics was less than that of the fixed allocation strategy. This proves that clustering order tasks based on the similarity characteristics between orders and assigning orders with high similarity to the same picking workstation can effectively reduce the total number of AGV handling operations on the racks.

[0153] Furthermore, as shown in Table 1, the K-means task allocation strategy based on order characteristics reduced the number of handling operations by 13.44% to 20.79%, with an average reduction of 17.15%, thus verifying the effectiveness of the K-means task allocation strategy.

[0154] Table 1

[0155]

[0156] Assume the maximum number of tasks an AGV can perform within a certain period is 10, meaning it can perform the handling of a maximum of 10 shelves; the minimum number of tasks an AGV can perform within a certain period is 4, meaning the AGV will only start working after acquiring handling tasks for at least 4 shelves.

[0157] By comparing the travel distances of the Multi-Strategy Whale Optimization (MWOA) algorithm with different order volumes compared to the AGV's travel distance when always maintaining the maximum task volume (Max-C) and minimum task volume (Min-C), the efficiency of MWOA in solving the NP-hard problem of AGV task scheduling is verified. After scheduling the total orders at the picking workstation, the AGV task sequence is obtained, and the idle distance of the AGV when performing the handling tasks according to this sequence is calculated. The results are recorded in Table 2 (Table 2: Statistical Results of Comparative Experiments).

[0158] Table 2

[0159]

[0160]

[0161] As shown in Table 2, when each AGV is required to acquire handling tasks from the picking station according to its maximum or minimum task load, the total travel distance of the AGVs is consistently higher than that of the MWOA strategy, which allows for free task allocation to the AGVs. This demonstrates that the task scheduling strategy of the MWOA algorithm fully considers the distance relationships between warehouse racks, reducing the travel distance of individual AGVs handling racks.

[0162] Meanwhile, as the number of orders increases, the total travel distance of the AGV increases, and the reduction in total travel distance using the MWOA scheduling algorithm increases. In other words, when the number of orders is large, the MWOA scheduling algorithm has a greater impact on the performance of the warehousing system.

[0163] This invention proposes an optimized scheduling method for intelligent warehouse picking tasks. It employs the K-means clustering algorithm to effectively reduce the number of times AGVs move shelves in the intelligent warehousing system; and utilizes a multi-strategy whale algorithm to shorten the empty-load transport distance of AGVs and reduce logistics turnaround time. This invention can be applied to the construction of novel automated intelligent warehousing systems, and the proposed optimized scheduling method based on order characteristics can greatly improve order picking efficiency. This invention is a high-efficiency, low-empty-load optimized scheduling method for picking tasks.

[0164] Figure 4 This is a block diagram illustrating an optimized scheduling system for intelligent warehouse picking tasks, according to an exemplary embodiment. The system is applied to an optimized scheduling method for intelligent warehouse picking tasks, with reference to... Figure 4 The system includes electronic equipment, picking workstations, and AGVs, among which:

[0165] Electronic device 410 is used to read information from e-commerce platforms and obtain an ID-based order set; based on the ID-based order set, it obtains the optimal order cluster set through a preset K-means clustering algorithm; and according to the product picking ID sequence and preset AGV picking constraints, it obtains the AGV optimized scheduling task sequence through a multi-strategy whale algorithm.

[0166] Picking workstation 420 is used to match the optimal cluster set of orders with the picking workstation to obtain the order picking ID sequence;

[0167] AGV430 is used to optimize and schedule task sequences based on AGVs and execute scheduling.

[0168] Optionally, the electronic device 410 is further used for:

[0169] Read information from the e-commerce platform to obtain a collection of product orders;

[0170] Assign ID numbers to the product order set to obtain an ID-based order set.

[0171] Optionally, the electronic device 410 is further used for:

[0172] S21. Perform cluster center selection operation based on the ID-based order set to obtain the cluster centers;

[0173] S22. Calculate the distance between each order and the cluster center based on the ID-based order set and cluster center;

[0174] S23. Based on the distance between each order and the cluster center, cluster the ID-based order set to obtain the order cluster set;

[0175] S24. Repeat steps S21-S23 until the number of iterations reaches its maximum, then obtain the optimal cluster set of orders.

[0176] Optionally, the picking workstation 420 is further used for:

[0177] Each order cluster in the optimal order cluster set is matched with a picking workstation to obtain the set of matched workstations;

[0178] For each workstation in the matched workstation set, analyze the clustering of matched orders to obtain the product picking ID sequence.

[0179] Optionally, the electronic device 410 is further used for:

[0180] S41. Based on the product picking ID sequence and the preset AGV picking constraints, initialize the multi-strategy whale algorithm to obtain the AGV scheduling task sequence.

[0181] S42. Calculate the idle distance generated by the scheduling tasks in the AGV scheduling task sequence to obtain the AGV scheduling idle distance;

[0182] S43. Update the AGV scheduling task sequence to obtain the updated AGV scheduling task sequence.

[0183] S44. Repeat steps S41-S43 until the AGV scheduling idle distance is minimized, then obtain the AGV optimized scheduling task sequence.

[0184] Optionally, the electronic device 410 is further used for:

[0185] The AGV scheduling task sequence is used as the current population, and the fitness is calculated to obtain the population fitness.

[0186] Based on population fitness, the current population is divided into population A and population B;

[0187] Based on population A, obtain the AGV scheduling task sequence in population A; optimize the AGV scheduling task sequence in population A using a preset improved search algorithm to obtain the AGV update scheduling task sequence for population A.

[0188] Based on population B, obtain the AGV scheduling task sequence in population B; optimize the AGV scheduling task sequence in population B using a preset improved multi-strategy optimization algorithm to obtain the AGV update scheduling task sequence in population B.

[0189] Based on the AGV update scheduling task sequence of population A and the AGV update scheduling task sequence of population B, the AGV update scheduling task sequence is obtained.

[0190] The preset improved search algorithm adds an improved search model to the original whale algorithm's search strategy; the improved search model uses individuals in population A as the search center and searches for the positions of other individuals in the surrounding area.

[0191] Among them, the preset improved multi-strategy optimization algorithm is an improved optimization algorithm based on the optimization guidance mathematical model of the Harry Eagle algorithm and the gene replacement mathematical model of the differential evolution algorithm.

[0192] This invention proposes an optimized scheduling method for intelligent warehouse picking tasks. It employs the K-means clustering algorithm to effectively reduce the number of times AGVs move shelves in the intelligent warehousing system; and utilizes a multi-strategy whale algorithm to shorten the empty-load transport distance of AGVs and reduce logistics turnaround time. This invention can be applied to the construction of novel automated intelligent warehousing systems, and the proposed optimized scheduling method based on order characteristics can greatly improve order picking efficiency. This invention is a high-efficiency, low-empty-load optimized scheduling method for picking tasks.

[0193] Figure 5 This is a schematic diagram of the structure of an electronic device 500 provided in an embodiment of the present invention. The electronic device 500 may vary considerably due to different configurations or performance. It may include one or more central processing units (CPUs) 501 and one or more memories 502. The memory 502 stores at least one instruction, which is loaded and executed by the processor 501 to implement the steps of the above-mentioned optimized scheduling method for intelligent warehouse picking tasks.

[0194] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including instructions that can be executed by a processor in a terminal to complete the aforementioned optimized scheduling method for intelligent warehouse picking tasks. For example, the computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, or optical data storage device.

[0195] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0196] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An optimized scheduling method for intelligent warehouse picking tasks, characterized in that, The method includes: S1. Read information from the e-commerce platform to obtain a set of ID-based orders; S2. Based on the ID-based order set, obtain the optimal order cluster set using a preset K-means clustering algorithm; In this process, K-means clustering was used to cluster orders with the same items and then assign them to the same picking workstation. This was done to quantify the differences between orders. The degree of similarity between them, among which For the product ID, the quantification model used is: ； in, For orders and The intersection of item IDs, i.e., the union of IDs of the same items in the two orders, is calculated... cardinality ,Right now Determine the similarity between two orders; During clustering, the order differences are also calculated, specifically the distance between the goods in two orders within the smart warehousing system. A quantification model is used as follows: ； in, Representative orders Relative to orders The supplement, Representative orders Relative to orders The supplement, This represents the union of two complements, i.e., the set of IDs where the goods in the two orders are completely different; [the set is missing here]. The dispersion of the ID represents the difference in the distance between the required shelf locations of two orders in the warehousing system, that is... The standard deviation of ID in the set is used as the difference between two orders. ; for Total amount of goods in the collection This represents the average of the product IDs. The item ID in the collection; S3. Match the optimal cluster set of orders with the picking workstations to obtain the product picking ID sequence; S4. Based on the product picking ID sequence and preset AGV picking constraints, obtain an optimized AGV scheduling task sequence using a multi-strategy whale algorithm, including: S41. Based on the product picking ID sequence and the preset AGV picking constraints, initialize the multi-strategy whale algorithm to obtain the AGV scheduling task sequence. S42. Calculate the idle distance generated by the scheduling tasks in the AGV scheduling task sequence to obtain the AGV scheduling idle distance; S43. Update the AGV scheduling task sequence to obtain an updated AGV scheduling task sequence, including: The AGV scheduling task sequence is used as the current population, and the fitness is calculated to obtain the population fitness. Based on the population fitness, the current population is divided into population A and population B; Based on population A, obtain the AGV scheduling task sequence in population A; the AGV scheduling task sequence in population A is optimized by a preset improved search algorithm to obtain the AGV update scheduling task sequence of population A. Based on the population B, the AGV scheduling task sequence in population B is obtained; the AGV scheduling task sequence in population B is optimized by a preset improved multi-strategy optimization algorithm to obtain the AGV update scheduling task sequence of population B. Based on the AGV update scheduling task sequence of population A and the AGV update scheduling task sequence of population B, an AGV update scheduling task sequence is obtained. S44. Repeat steps S41-S43 until the AGV scheduling idle distance is minimized, and obtain the AGV optimized scheduling task sequence. S5. Schedule the AGVs according to the optimized AGV scheduling task sequence.

2. The optimized scheduling method for intelligent warehouse picking tasks according to claim 1, characterized in that, The step of reading e-commerce platform information to obtain a set of ID-based orders includes: Read information from the e-commerce platform to obtain a collection of product orders; The product order set is assigned an ID number to obtain an ID-based order set.

3. The optimized scheduling method for intelligent warehouse picking tasks according to claim 1, characterized in that, The step of obtaining the optimal cluster set of orders based on the ID-based order set using a preset K-means clustering algorithm includes: S21. Perform cluster center selection operation based on the ID-based order set to obtain cluster centers; S22. Calculate the distance between each order and the cluster center based on the ID-based order set and the cluster center; S23. Based on the distance between each order and the cluster center, the ID-based order set is clustered to obtain an order cluster set; S24. Repeat steps S21-S23 until the number of iterations reaches its maximum, then obtain the optimal cluster set of orders.

4. The optimized scheduling method for intelligent warehouse picking tasks according to claim 1, characterized in that, The step of matching the optimal cluster set of orders with the picking workstations to obtain the product picking ID sequence includes: Each order in the optimal order cluster set is matched with a picking workstation to obtain a set of matched workstations; For each workstation in the set of matched workstations, the matched order clusters are analyzed to obtain the product picking ID sequence.

5. The optimized scheduling method for intelligent warehouse picking tasks according to claim 1, characterized in that, The preset improved search algorithm adds an improved search model to the original whale algorithm's search strategy; the improved search model uses individuals in population A as the search center and searches for the positions of other individuals in the surrounding area.

6. The optimized scheduling method for intelligent warehouse picking tasks according to claim 1, characterized in that, The preset improved multi-strategy optimization algorithm is an improved optimization algorithm based on the optimization guidance mathematical model of the Harry Eagle algorithm and the gene replacement mathematical model of the differential evolution algorithm.

7. An optimized scheduling system for intelligent warehouse picking tasks, characterized in that, The system includes electronic equipment, picking workstations, and AGVs, wherein: The electronic device is used to read information from the e-commerce platform to obtain an ID-based order set; based on the ID-based order set, it obtains the optimal order cluster set through a preset K-means clustering algorithm; and according to the product picking ID sequence and preset AGV picking constraints, it obtains the AGV optimized scheduling task sequence through a multi-strategy whale algorithm. In this process, K-means clustering was used to cluster orders with the same items and then assign them to the same picking workstation. This was done to quantify the differences between orders. The degree of similarity between them, among which For the product ID, the quantification model used is: ； in, For orders and The intersection of item IDs, i.e., the union of IDs of the same items in the two orders, is calculated... cardinality ,Right now Determine the similarity between two orders; During clustering, the order differences are also calculated, specifically the distance between the goods in two orders within the smart warehousing system. A quantification model is used as follows: ； in, Representative orders Relative to orders The supplement, Representative orders Relative to orders The supplement, This represents the union of two complements, i.e., the set of IDs where the goods in the two orders are completely different; [the set is missing here]. The dispersion of the ID represents the difference in the distance between the required shelf locations of two orders in the warehousing system, that is... The standard deviation of ID in the set is used as the difference between two orders. ; for Total amount of goods in the collection This represents the average of the product IDs. The item ID in the collection; The electronic device is further used for: S41. Based on the product picking ID sequence and the preset AGV picking constraints, initialize the multi-strategy whale algorithm to obtain the AGV scheduling task sequence. S42. Calculate the idle distance generated by the scheduling tasks in the AGV scheduling task sequence to obtain the AGV scheduling idle distance; S43. Update the AGV scheduling task sequence to obtain an updated AGV scheduling task sequence, including: The AGV scheduling task sequence is used as the current population, and the fitness is calculated to obtain the population fitness. Based on the population fitness, the current population is divided into population A and population B; Based on population A, obtain the AGV scheduling task sequence in population A; the AGV scheduling task sequence in population A is optimized by a preset improved search algorithm to obtain the AGV update scheduling task sequence of population A. Based on the population B, the AGV scheduling task sequence in population B is obtained; the AGV scheduling task sequence in population B is optimized by a preset improved multi-strategy optimization algorithm to obtain the AGV update scheduling task sequence of population B. Based on the AGV update scheduling task sequence of population A and the AGV update scheduling task sequence of population B, an AGV update scheduling task sequence is obtained. S44. Repeat steps S41-S43 until the AGV scheduling idle distance is minimized, and obtain the AGV optimized scheduling task sequence. The picking workstation is used to match the optimal cluster set of orders with the picking workstation to obtain the product picking ID sequence; The AGV is used to optimize and schedule task sequences based on the AGV and execute scheduling.

Images