Order processing decision-making method and device based on medicine circulation industry and storage medium

Abstract

The invention relates to an order processing decision-making method and device based on the medicine circulation industry and a storage medium, and is applied to the technical field of medicine logistics automatic warehousing, and the method comprises the steps: through calculating a similarity coefficient between orders, distributing the orders with a high coupling degree to the same sorting table, and enhancing the possibility of material box reuse; aiming at an internal order of each sorting table, a single sorting table order sorting algorithm based on neighborhood search is adopted, the order sequence is optimized through neighborhood operations such as local exchange and fragment overturning by taking minimization of the number of times of material box replacement as a target, and the order sorting quality and stability are improved; on the basis of completion of order sorting in each sorting table, global synchronous and conflict-free execution logic in each stage is taken as a core, state expansion is performed in combination with sorting requirements of a plurality of sorting tables, a conflict state of the workbins is effectively eliminated through construction of a global candidate state set and execution of Cartesian product combination and conflict detection, and the sorting efficiency of the workbins is improved through global sorting of the workbins. And parallel non-coupling efficient sorting is realized.

Description

Technical Field

[0001] The present invention relates to the field of pharmaceutical logistics automated warehousing technology, and in particular to an order processing decision method, device and storage medium based on the pharmaceutical distribution industry. Background Art

[0002] Compared to other industries, pharmaceutical logistics places extremely stringent time constraints. Efficiently completing the picking of large numbers of break-even orders within a limited timeframe presents a significant challenge for the industry. The "goods-to-person" picking system (SBS / RS) based on multi-level shuttles is widely used for break-even picking in the pharmaceutical industry due to its efficient inbound and outbound operations and rapid order processing capabilities. This system uses multi-level shuttles and elevators to remove bins from high-density storage shelves and transport them via conveyor lines to picking stations, where pickers select the goods. However, in the pharmaceutical industry, as specialized commodities, the distribution of pharmaceuticals is subject to strict batch number management regulations. According to the requirements of the "Pharmaceutical Business Quality Management Standards", the entry and exit of drugs must follow the principle of first-in-first-out (Batch Number). Drugs from different batches cannot be stored in the same bin. These rules mean that for drugs with the same batch number, drugs with later batches can only be released after the previous batch number has been released. When the last bin of a batch is needed by orders from multiple picking stations, bin coupling will occur. This coupling will reduce the picking efficiency to a certain extent. In addition, the high order ratio of orders in the pharmaceutical industry will also aggravate this situation to a certain extent.

[0003] In SBS / RS, the system uses multi-layer shuttle vehicles to extract the required material boxes from high-density shelves, and transports them to the picking station via the conveyor line, and the pickers complete the order picking operation. After each material box is picked, the system decides whether to temporarily store it in the cache or directly return it to the warehouse based on whether the next order still needs the goods in the material box. The system is usually equipped with multiple picking stations, and a single picking station adopts the a vs b mode, that is, a material box picking position and b order box positions are set. However, in actual operations, a single picking station needs to be replenished with the next batch of orders after the current batch of orders is picked. The unreasonable order sequence will lead to frequent material box replacement, resulting in long waiting time and significantly reduced efficiency. For scenarios where multiple picking stations operate in parallel, when different picking stations pick orders, it may happen that two different orders from two picking stations require the same SKU In the case of a material box, order coupling occurs. Once the SKU The material box is locked by a picking station. Only after the picking station releases the material box, other picking stations can pick the material box. SKU , before the material box is released, the other picking stations can only wait. This will not only cause a waste of picking resources, but also cause the inefficiency of the entire picking system.

[0004] Existing research on order allocation mainly focuses on optimizing the processing speed of picking stations and minimizing shelf visits, but has not yet considered the inter-order drug SKU The impact of overlapping order characteristics on order allocation. Second, existing order sorting research has mostly focused on shelf / location-level sorting optimization, lacking consideration of bin replacement frequency and waiting bottlenecks. Third, joint optimization of order allocation, order sorting, and storage unit sorting has mostly focused on resource coordination in RMFS systems, while research on bin coupling in SBS / RS systems is limited. In the pharmaceutical industry, bin coupling is particularly significant due to the high-frequency demand and high-precision picking characteristics of drugs. Without a sound sorting strategy, this can lead to frequent waiting times and system congestion. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide an order processing decision method, device and storage medium based on the pharmaceutical distribution industry to solve the problem that in the existing technology, the research on order allocation mainly focuses on the optimization of the processing speed of the picking station and the minimization of the shelf access volume, but has not yet considered the drug distribution between orders. SKU The impact of overlapping characteristics on order allocation. Secondly, existing order sorting research mostly focuses on sorting optimization at the shelf / location level, lacking consideration of the replacement frequency of bins and waiting bottlenecks. Thirdly, the joint optimization of order allocation, order sorting, and storage unit sorting mostly focuses on resource coordination of RMFS systems, while there is a lack of research on bin coupling problems in SBS / RS systems.

[0006] According to a first aspect of an embodiment of the present invention, a method for order processing decision-making in the pharmaceutical distribution industry is provided, the method comprising: For all orders in a batch to be processed, a seed order is assigned to each picking station according to the similarity coefficient between any two orders; For the remaining orders that have not been assigned as seed orders, the number of goods of the same type as the seed orders assigned to each picking station is calculated as the coupling factor for each order, and each remaining order is assigned to the picking station with the largest coupling factor to complete the order allocation; For the multiple orders assigned to each picking station, a random method is used to generate multiple candidate order sequences to form an initial candidate set. Each candidate order sequence in the initial candidate set is locally optimized using neighborhood search to obtain a preliminary local optimal order sequence. Perform non-adjacent order swapping and local segment flipping operations on the preliminary local optimal order sequence to obtain a final candidate set, and select the order sequence with the least number of bin replacements from the final candidate set as the optimal order sequence for the picking station; Within each picking station, an initial state set of each picking station is obtained according to the optimal order sequence; each picking station independently performs a picking operation to generate a local candidate state set of each picking station; The local candidate state sets of each picking station are combined through Cartesian products to form a global candidate state combination; each generated global candidate state is tested for bin conflicts. If there is no bin conflict between picking stations, the global candidate state is legal; if there is a bin conflict, the global candidate state is excluded; Use comprehensive evaluation function to score the legal global candidate states, sort the scores from low to high, and retain the top candidate with the lowest score. K A global candidate state; For reservations K The global candidate states are cyclically expanded and pruned until all order picking tasks are completed according to the optimal order sequence of each picking station or the maximum number of iterations is reached, and the final global candidate state set is obtained; The state with the lowest comprehensive evaluation function score is selected from the final global candidate state set as the global optimal state, and the bin sorting sequence corresponding to the global optimal state is output.

[0007] Preferably, The method of allocating a seed order to each picking station according to the similarity coefficient between any two orders includes: Calculate the similarity coefficient between any two orders based on the number of repeated ordered goods between them, and obtain a similar data set; Selecting two orders with the largest similarity coefficients from the similarity data set as seed orders for a picking station, and deleting all similarity coefficients related to the two selected orders from the similarity data set; Among the remaining similarity coefficients, continue to select the two orders with the largest similarity coefficients as the seed orders for the next picking station, and delete all similarity coefficients related to the two selected orders; repeat this step until a seed order is assigned to each picking station.

[0008] Preferably, it also includes: Set the difference in the number of orders allocated to two adjacent picking stations to not exceed the preset threshold; When the order quantity assigned to any picking station just meets the difference between the order quantity of the adjacent picking station and the preset threshold, the orders that need to be assigned to the picking station will be sorted according to the coupling factor and assigned to the picking station with the second largest coupling factor. If the picking station with the second largest coupling factor has also met the order quantity, the order will be assigned to the picking station with the third largest coupling factor, so that the order quantity under each picking station is relatively balanced.

[0009] Preferably, The method of performing local optimization on each candidate order sequence in the initial candidate set by using neighborhood search to obtain a preliminary local optimal order sequence includes: For each candidate order sequence in the initial candidate set, enumerate all adjacent order pairs in the sequence in sequence; According to the order of adjacent order pairs, the order of adjacent order pairs is exchanged in sequence. After each exchange of the order pairs, the minimum number of bin replacement times of the new sequence after the order pair order is exchanged is calculated using the bin replacement number determination algorithm based on beam search. If the minimum number of bin replacements of the new sequence after the exchange is less than the minimum number of bin replacements of the sequence before the exchange, the new sequence is accepted; Continue to exchange adjacent orders in the updated sequence until the order exchange of all adjacent orders cannot reduce the minimum number of bin replacements, and obtain the preliminary local optimal order sequence.

[0010] Preferably, The preliminary local optimal order sequence is subjected to non-adjacent order exchange and local segment flipping operations in sequence, and the final candidate set obtained includes: Perform a sequence swap on any two non-adjacent order pairs in the preliminary local optimal order sequence, and obtain the minimum bin replacement number of the sequence after the sequence swap; if the minimum bin replacement number of the sequence after the swap is less than the minimum bin replacement number of the sequence before the swap, then accept the swap; repeat this process until the sequence swap of any two non-adjacent order pairs cannot reduce the minimum bin replacement number; For the order sequence after the non-adjacent order exchange, select any continuous subsequence, perform a reverse operation on all orders in the subsequence, and obtain the minimum number of bin replacements of the sequence after the reverse operation. If the minimum number of bin replacements of the sequence after the reverse operation is less than the minimum number of bin replacements of the sequence before the reverse operation, the exchange is accepted; repeat this process until the reverse operation of any continuous subsequence cannot reduce the minimum number of bin replacements, and the final candidate set is obtained.

[0011] Preferably, it also includes: During the process of performing neighborhood search and local optimization on the order sequence, or performing non-adjacent order swapping operations on the order sequence, or performing local segment flipping operations on the order sequence, an acceptance probability is set, and the acceptance probability decreases step by step with the number of iterations; When the minimum number of bin replacements in the updated order sequence is greater than or equal to the minimum number of bin replacements in the order sequence before the update, the update is accepted with the set acceptance probability and the local optimum is jumped out.

[0012] Preferably, The calculation of the minimum number of bin replacement times using the bin replacement number determination algorithm based on the beam search includes: Get the total number of orders assigned to any picking station and assign each order to the picking station c order positions, and obtain the order sequence for each order position; Set the initial state set of the picking station, the initial state of the picking station includes: the current order of each order position of the picking station to be picked SKU List, current bin, index of next standby order, cumulative bin replacement count initialized to 0, record bin arrival sequence; Determine whether each state in the initial state set of the picking station meets the preset completion state, the preset completion state includes: SKU All are picked and the waiting orders are exhausted; If any state meets the preset completion state, the state is saved to the completion state list; otherwise, the state is expanded and all possible picking operations are enumerated to obtain the candidate state set for the next stage; the state changes in the candidate state set compared to the state before expansion include: The order line value of the order position corresponding to the picked order line is reduced by 1. When the value is reduced to 0, it means that all current orders on the order position have been picked. A new order needs to be added from the waiting order queue. The corresponding order line value is updated to the number of rows of the new order, and the order number is increased by 1. SKU Switching is used to determine whether the material box needs to be replaced. If the material box used in the new state is different from the previous state, it means that the material box has been replaced, and the cumulative number of material box replacements is increased by 1; Each state in the candidate state set is sorted from low to high according to the score calculated by the comprehensive evaluation function. For states with the same score, they are sorted in the order of generation. The comprehensive evaluation function calculation includes: obtaining the comprehensive evaluation function score by using the cumulative number of bin replacements, the remaining picking workload, the remaining number of standby orders, and a preset weight index in the state; Before selecting the comprehensive evaluation function score K states as the state set for the next stage; If there is a state that meets the completion conditions in the new state set, it is saved to the completion state list and the expansion ends; if there is no state that meets the completion conditions in the new state set; Then continue to expand each state in the new state set and select the state with the highest score of the comprehensive evaluation function. KThe state continues to be the state set of the next stage; it is judged again whether there is a state that meets the completion condition in the new state set; this step is repeated until a state that meets the completion condition exists in the new state set after a certain expansion, or the preset number of iterations is reached, or all states in the new state set after a certain expansion cannot be expanded, and the expansion is ended; If there is a completed state, the state with the lowest comprehensive evaluation function score is selected from the completed states as the final optimal state; If there is no completed state, the state with the lowest comprehensive evaluation function score is selected from the state set obtained in the last iteration as the final optimal state; The order sequence corresponding to the final optimal state is the optimal order sequence, and the accumulated number of material box replacements corresponding to the final optimal state is the minimum number of material box replacements for the picking station.

[0013] Preferably, The reserved K The global candidate state loop performs state expansion and pruning operations including: For reservations K The global candidate states perform a picking operation according to the optimal order sequence of each picking station to obtain the expanded global candidate state set; Perform bin conflict detection on each global candidate state in the expanded global candidate state set. If there is no bin conflict between picking stations, the global candidate state is legal; if there is a bin conflict, the global candidate state is excluded. The comprehensive evaluation function is used to score the expanded legal global candidate states, and the scores are sorted from low to high, and the top candidate with the lowest score is retained. K A global candidate state; Repeat the above steps until all order picking tasks are completed or the maximum number of iterations is reached, and the final global candidate state set is obtained.

[0014] According to a second aspect of an embodiment of the present invention, an order processing decision-making device for the pharmaceutical distribution industry is provided, the device comprising: Seed order allocation module: for all orders in a batch to be processed, a seed order is allocated to each picking station according to the similarity coefficient between any two orders; All Orders Allocation Module: For the remaining orders that have not been allocated as seed orders, the module calculates the number of goods of the same type as the seed orders allocated to each picking station as the coupling factor, and allocates each remaining order to the picking station with the largest coupling factor to complete the order allocation; Preliminary order sequence acquisition module: used to randomly generate multiple candidate order sequences for the multiple orders assigned to each picking station to form an initial candidate set, and then perform local optimization on each candidate order sequence in the initial candidate set using neighborhood search to obtain a preliminary local optimal order sequence; Final order sequence acquisition module: used to perform non-adjacent order exchange and local segment flipping operations on the preliminary local optimal order sequence in sequence to obtain a final candidate set, and select the order sequence with the least number of bin replacements from the final candidate set as the optimal order sequence for the picking station; Local candidate state acquisition module: used to obtain the initial state set of each picking station according to the optimal order sequence within each picking station; each picking station independently performs a picking operation to generate a local candidate state set of each picking station; Global candidate state acquisition module: used to form a global candidate state combination by Cartesian product of the local candidate state sets of each picking station; perform bin conflict detection on each generated global candidate state. If there is no bin conflict between picking stations, the global candidate state is legal; if there is a bin conflict, the global candidate state is excluded; Global state screening module: It is used to score the legal global candidate states using a comprehensive evaluation function, sort the scores from low to high, and retain the lowest-scoring candidate states. K A global candidate state; Cycle extension module: used to retain K The global candidate states are cyclically expanded and pruned until all order picking tasks are completed according to the optimal order sequence of each picking station or the maximum number of iterations is reached, and the final global candidate state set is obtained; A global optimal state acquisition module is used to select the state with the lowest comprehensive evaluation function score in the final global candidate state set as the global optimal state, and output the bin sorting sequence corresponding to the global optimal state.

[0015] According to a third aspect of an embodiment of the present invention, a storage medium is provided, wherein the storage medium stores a computer program, and when the computer program is executed by a host controller, each step in the above method is implemented.

[0016] The technical solutions provided by the embodiments of the present invention may have the following beneficial effects: This application calculates the similarity coefficient between orders and assigns orders with a high degree of coupling to the same picking station, thereby increasing the possibility of bin reuse and laying the foundation for subsequent optimization; for orders within each picking station, a single picking station order sorting algorithm based on neighborhood search is adopted, with the goal of minimizing the number of bin replacements. The order sequence is optimized through neighborhood operations such as local exchange and fragment flipping, and a bin replacement number determination algorithm based on beam search is used to accurately evaluate the cost of the sorting scheme, thereby improving the quality and stability of order sorting; on the basis that the orders have been assigned and the order sorting within each picking station is completed, with the global synchronization and conflict-free execution logic of each stage as the core, the picking needs of multiple picking stations are combined for state expansion, and the conflicting bin states are effectively eliminated by constructing a global candidate state set, performing Cartesian product combination and conflict detection, and achieving parallel and uncoupled efficient picking through global bin sorting.

[0017] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0019] Figure 1 This is a flowchart illustrating an order processing decision-making method based on the pharmaceutical distribution industry according to an exemplary embodiment; Figure 2 is a flow chart of a multi-picking platform bin sorting decoupling algorithm according to another exemplary embodiment; Figure 3 is a schematic diagram showing order similarity comparison within a picking station according to another exemplary embodiment; Figure 4 is a schematic diagram showing order similarity comparison between picking stations according to another exemplary embodiment; Figure 5 is a schematic diagram showing a comparison of the cumulative number of times of material box replacement according to another exemplary embodiment; Figure 6 is a schematic diagram showing a comparison of the cumulative number of bin replacements between a single picking station order sorting algorithm based on neighborhood search and a random order sorting strategy under different order plans according to another exemplary embodiment; Figure 7 is a schematic diagram of a verification solution result according to another exemplary embodiment; Figure 8 is a system diagram of an order processing decision-making device based on the pharmaceutical distribution industry according to another exemplary embodiment; In the accompanying figure: 1-seed order allocation module, 2-all order allocation module, 3-preliminary order sequence acquisition module, 4-final order sequence acquisition module, 5-local candidate state acquisition module, 6-global candidate state acquisition module, 7-global state screening module, 8-loop expansion module, 9-global optimal state acquisition module. DETAILED DESCRIPTION

[0020] Exemplary embodiments will be described in detail herein, examples of which are illustrated in the accompanying drawings. In the following description, when referring to the drawings, like numbers in different figures represent like or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present invention. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present invention, as detailed in the appended claims.

[0021] Example 1 Figure 1 FIG. 1 is a flow chart showing an order processing decision method based on the pharmaceutical distribution industry according to an exemplary embodiment. Figure 1 As shown, the method includes: S1, for all orders in a batch to be processed, a seed order is assigned to each picking station according to the similarity coefficient between any two orders; S2: For the remaining orders that have not been assigned as seed orders, calculate the number of goods of the same type as the seed orders assigned to each picking station as the coupling factor for each order, and assign each remaining order to the picking station with the largest coupling factor to complete the order allocation; S3, for the multiple orders assigned to each picking station, a plurality of candidate order sequences are generated by random method to form an initial candidate set, and each candidate order sequence in the initial candidate set is locally optimized by neighborhood search to obtain a preliminary local optimal order sequence; S4, performing non-adjacent order swapping and local segment flipping operations on the preliminary local optimal order sequence to obtain a final candidate set, and selecting the order sequence with the least number of bin replacements from the final candidate set as the optimal order sequence for the picking station; S5, within each picking station, obtaining an initial state set of each picking station according to the optimal order sequence; each picking station independently performs a picking operation to generate a local candidate state set of each picking station; S6, the local candidate state sets of each picking station are combined into a global candidate state through Cartesian product; each generated global candidate state is tested for bin conflicts. If there is no bin conflict between picking stations, the global candidate state is considered legal; if there is a bin conflict, the global candidate state is excluded; S7, use the comprehensive evaluation function to score the legal global candidate states, sort the scores from low to high, and retain the top candidate with the lowest score. K A global candidate state; S8, for the reserved K The global candidate states are cyclically expanded and pruned until all order picking tasks are completed according to the optimal order sequence of each picking station or the maximum number of iterations is reached, and the final global candidate state set is obtained; S9, selecting the state with the lowest comprehensive evaluation function score from the final global candidate state set as the global optimal state, and outputting the bin sorting sequence corresponding to the global optimal state; It is understandable that in the pharmaceutical distribution scenario, a typical goods-to-person picking system usually includes multiple picking stations operating in parallel. In order to improve the overall efficiency of the system and alleviate the resource competition problem between the picking stations, it is necessary to comprehensively consider the three key links of "order allocation strategy", "order sorting within the picking station", and "material box sorting and conflict decoupling between picking stations" in the order execution process. Based on this, this embodiment constructs a multi-picking station order execution decision algorithm framework, whose core goals are: to maximize the material box reuse rate, minimize the number of material box replacements, and eliminate material box conflicts between multiple picking stations.

[0022] It mainly includes the following three stages: Order similarity allocation strategy: First, initial allocation is performed based on order structure information. The traditional dynamic allocation strategy allocates orders to each picking station in sequence based on the order receipt time. However, this method may cause highly similar orders to be dispersed, increasing the fragmentation of SKU distribution. Therefore, this embodiment proposes an order similarity allocation strategy. By calculating the SKU overlap between orders, orders with a higher coupling degree are allocated to the same picking station, thereby enhancing the possibility of material box reuse and laying the foundation for subsequent optimization.

[0023] Single picking station order sorting algorithm based on neighborhood search: For each internal order of the picking station, the order processing order is further optimized after the allocation is completed. This embodiment designs a single picking station order sorting algorithm based on neighborhood search, with the goal of minimizing the number of material box replacements. The order sequence is optimized through neighborhood operations such as local exchange and fragment flipping, and a material box replacement number determination algorithm based on beam search is used to accurately evaluate the cost of the sorting scheme, thereby improving the quality and stability of order sorting.

[0024] Decoupling algorithm for sorting bins at multiple picking stations: Based on the fact that orders have been allocated and order sorting has been completed within each picking station, this embodiment further proposes a decoupling algorithm for sorting bins at multiple picking stations. This algorithm takes the global synchronization and conflict-free execution logic of each stage as its core, and combines the picking requirements of multiple picking stations to perform state expansion. By constructing a global candidate state set, performing Cartesian product combination and conflict detection, it effectively eliminates bin conflict states, and achieves parallel and non-coupling efficient picking through global bin sorting.

[0025] in, 1. Order similarity allocation strategies specifically include: The dynamic order allocation strategy is a more traditional allocation method. Orders are allocated to each picking station in sequence according to the time they are received. Combined with the results of order batching, the specific steps of the existing dynamic allocation strategy are as follows: The orders in a batch are allocated to each picking station according to the receiving time, where the number of picking stations is n , the total order quantity is q , when the picking station number is less than or equal to When the picking station is assigned orders, when the picking station number is greater than When the picking station is assigned Each picking station picks orders in turn according to the assigned orders. If any order is completed during the picking process, the next order will be added to continue picking until all orders in the batch are picked.

[0026] However, in reality, there is a certain degree of correlation within the internal structure of orders. If highly correlated orders are assigned to a picking station, then such a raw material box can meet the needs of multiple orders. Therefore, this embodiment proposes an order similarity allocation strategy, which allocates orders based on the correlation coefficient between orders, so that orders with high correlation are assigned to the same picking station, thereby improving overall picking efficiency by increasing the re-picking rate of a single raw material box. The specific allocation steps are as follows: (1) Assume that the number of orders that a single picking station can process at the same time is , the total order quantity is , the number of picking stations is , calculate the similarity coefficient between any two orders in a batch , that is, the number of repeated ordered goods between two orders, added to the dataset In descending order, select The corresponding two orders are used as seed orders for a picking station and are deleted from the data set. All the corresponding correlation coefficients are then Continue to select the one with the largest similarity coefficient As the seed order for the next picking station, repeat this process until n Each picking station is assigned a seed order (each seed order corresponds to two similar orders); (2) After each picking station determines the seed order, it is assigned the remaining orders. The remaining orders that are not assigned as seed orders form the order pool. , for each order, calculate the number of the same goods types as the seed order assigned to the picking station as the coupling factor , for each order , assign it to The corresponding picking station repeats this step repeatedly until all orders are assigned to the picking station. The maximum number of orders assigned to each picking station is as follows: When the picking station number is less than or equal to When the picking station is assigned orders, when the picking station number is greater than When the picking station is assigned orders; in simple terms, the number of orders between two adjacent picking stations cannot exceed 1. If the order quantity of a picking station has met the quantity requirement, the next order assigned to this picking station will be assigned to the picking station with the second highest coupling factor. If the order quantity of the picking station with the second highest coupling factor also meets the requirement, the order will be postponed to the picking station with the third highest coupling factor, and so on. 2. Single picking station order sorting algorithm based on neighborhood search: In the "goods-to-person" pharmaceutical picking system, although orders have been assigned to various picking stations according to certain rules, due to the There are differences in types and quantities. In the actual picking process, if the order processing sequence is not arranged reasonably, it may lead to frequent replacement of material boxes. Each replacement of material boxes is not only accompanied by equipment switching operations, but may also cause problems such as conveyor line blockage and picking waiting, thereby significantly reducing the system operation efficiency.

[0027] Therefore, after completing the order allocation, this embodiment further focuses on the order sorting problem within a single picking station. The core goal is to reasonably sort the order execution sequence so that The number of material box replacements is minimal, which improves the continuous operation capacity of the picking platform and the system throughput efficiency. Specifically, reasonable order sorting can make the same items in multiple orders Centralized picking can maximize the reuse rate and continuous use time of the material boxes without affecting the order accuracy.

[0028] Based on the above motivations, this embodiment designs an order sorting optimization algorithm based on neighborhood search. After setting the initial order sequence, it gradually performs local perturbations and global search evaluations on it, and uses feedback from the number of bin replacements to screen the sorting solutions. Specifically, it includes: (1) Generate initial candidate solutions Randomly generate orders in the current batch Sort the candidates to form the initial candidate set: For the current batch of picking stations First, we use a random method to generate candidate order rankings, and record any candidate ranking as: (1) To ensure the diversity of the solution space, randomly generate The candidate orders are sorted to form the initial candidate set: (2) (2) Preliminary optimization of adjacent order exchanges For each candidate solution, swap adjacent orders pair by pair and calculate the number of bin replacements (evaluated by the bin replacement number determination algorithm based on beam search). If the swap reduces the number of bin replacements, accept the new sequence and continue optimizing until adjacent swaps cannot improve further: In order to reduce the number of bin replacements, each candidate sequence is locally optimized using neighborhood search. , enumerate all adjacent order pairs in the sequence , swap positions one by one in order to generate a new sequence , the sequence obtained after the exchange , using the algorithm based on beam search to determine the number of bin replacements to calculate the minimum number of bin replacements ,like: (3) Then accept the new sequence , update the current candidate ranking, let: (4) Continue to exchange adjacent orders for the updated sequence until all adjacent exchanges cannot reduce the So far, the initial local optimal sequence is obtained, which is recorded as ; (3) Neighborhood expansion optimization Based on the local optimal solution obtained from the preliminary optimization, two extension operations are used for further exploration. The first extension is non-adjacent order swapping. Two non-adjacent orders in the sequence are randomly selected and swapped. The number of bin replacements is re-evaluated and accepted if the result is improved. The second extension is local segment flipping. Selected continuous subsequences in the sequence are reversed and the number of bin replacements is re-evaluated. If a better result is obtained, it is accepted. Specifically, the following operations are performed: 1. Non-adjacent order exchange right Any one of them meets Order pair , exchange positions to generate new sequences , for each swapped sequence, calculate and decide whether to accept; 2. Local fragment flip In sequence Select any continuous subsequence from position arrive (in ), perform reverse operation on all orders in the segment to form a new sequence , enumerate all possible and Combination, for each flip result, the same calculation and decide whether to accept; (4) Acceptance criteria and exit mechanism During the iteration process, to avoid falling into local optimality, a mechanism is set up to accept slightly inferior solutions with a certain probability, and the acceptance probability is gradually reduced as the number of iterations increases. This allows the algorithm to escape local traps and also stably converge to a better solution in the later stages. Specifically, In order to avoid the expansion operation being restricted by local optimality, an acceptance probability is set (This probability decreases gradually with the number of iterations), for a certain exchange operation: like , then the probability Accept the update, jump out of the local optimum, otherwise strictly accept Operation that reduces the value.

[0029] (5) Selection and output of optimal sequence After multiple iterations and expansion operations, the sequence with the minimum number of bin changes is selected from the final retained sequence as the output, which is the optimal arrangement plan for the single picking station order, including: After completing the neighborhood expansion operation, the updated candidate sequence is recorded as ; After the neighborhood search, the initial candidate set Updated to: (5) Each After the partial improvement of order arrangement, the number of box replacements is reduced to a lower level under a given decision. The order sequence with the minimum number of bin replacements is selected as the final order sorting solution, that is: (6) The sorting scheme This is the optimal order sorting solution for this single picking station, and the corresponding number of bin replacements is the theoretical minimum.

[0030] It should be emphasized that the above-mentioned algorithm for determining the number of bin replacements based on beam search calculates the minimum number of bin replacements specifically including: In order to accurately evaluate the number of bin replacements under any order sequence, it is necessary to comprehensively consider multi-dimensional factors such as the "number of order slots" and the "arrival of standby orders". To this end, this embodiment adopts a bin replacement number determination algorithm based on cluster search. In the multi-stage picking process simulation, dynamic state transition and cluster pruning are used to efficiently solve the approximately optimal bin arrival sequence and the corresponding minimum number of replacements.

[0031] Main ideas: The picking process on the single picking platform is divided into several consecutive time stages (each stage corresponds to the picking of an order line). Initially, an order is assigned to each order slot in sequence. When all orders on a certain order slot are picked, the next order is added from the waiting order queue; In order to avoid exponential state explosion in each picking stage, the algorithm uses a comprehensive evaluation function to evaluate the state at each stage and retains the previous state. states (controlled by the width of the beam search), pruning is performed when the retained state is extended to the next stage, and the comprehensive evaluation function considers the cumulative number of bin replacements , the remaining order quantity and the remaining workload to ensure that the search direction is mainly aimed at reducing the number of material box replacements.

[0032] Status description and initial definition: A batch of An order is assigned to a picking station for picking, and the picking process is divided into multiple consecutive time stages (denoted as ), each stage corresponds to a picking time period of an order line, and a picking station is set Each order position is bound to at most one order to be picked at any time. It should be noted that one order corresponds to multiple order lines, and each order line corresponds to a SKU , each SKUCorresponding to a material box; In the initial stage, the The orders are assigned to each order position in turn, forming the initial state of the picking station, and the state vector is used to represent the stage No. The state consists of four parts, which are represented by symbols: (7) Among them, the collection Indicates the number of order lines that have not been picked in the current order at each order position. If no order is assigned to an order position, then define ; Indicates the bin number currently used by the picking station, used to determine whether bin replacement occurs during the picking process; The cumulative number of bin replacements from the initial state to the current state; Indicates the position index of the next order to be added in the waiting order queue. The waiting order sequence is recorded as: (8) Therefore, the initial stage (i.e. ) can be written as: (9) in Indicates the The total quantity of the order line; State extension mechanism: Only one order line is picked in each phase. After the picking operation is completed, the status changes in two ways: Order status update: The order position of the corresponding picked order line is The value is reduced by 1; when the value is reduced to 0, it means that all orders on the order position have been completed. At this time, a new order needs to be added from the waiting order queue. Update to the number of rows of the new order and make Increase by 1; Bin replacement judgment: Based on whether the picking operation occurs Switching, so as to judge whether the material box needs to be replaced. If the material box used in the new state is different from that in the previous state, it means that the material box has been replaced. .

[0033] The two-way update is abstractly described by the state transition formula: (10) in,

[0034] After each picking operation is completed, the system moves from one stage to the next. Each stage corresponds to one order line picking operation and status update. After each stage, a set of candidate states is generated. The entire state transition process can be summarized into three scenarios based on the different conversion situations of the picking task. Table 1 below describes these three scenarios, showing the state transition process and the impact of each operation on the state update: Table 1 State transition description

[0035] State pruning strategy for beam search: Although the picking operation in a single stage only corresponds to the processing of one order line at the current order position, due to the existence of multiple decision branches, let the state set retained in the current stage be , then each state , by enumerating all possible picking operations, we can get the candidate state set for the next stage: (11) in Indicates status Perform state transfer to generate a set of all possible successor states. Since each state may produce multiple different state branches in a picking operation, as the number of stages increases As the number of candidate states increases, it tends to grow exponentially. To suppress the computational burden brought by the exponential growth, a beam search strategy is used to perform pruning operations to speed up the efficiency of the algorithm. After the state expansion in each stage, the candidate states are sorted by a comprehensive evaluation function, and only the top ranked states are retained. The state enters the next stage, thereby limiting the maximum number of states retained in each stage; The comprehensive evaluation function is constructed as follows: (12) in Expressed as the cumulative number of material box replacements; Indicates the remaining picking workload; ( ) represents the remaining quantity of the waiting order; is a positive weight. The goal is to minimize Although the evaluation function takes multiple factors into consideration, the goal of this embodiment is to minimize the cumulative number of bin replacements. Therefore, in the parameter setting, a higher Weights to ensure that the search is primarily along the decreasing direction; In the candidate state set For each state calculate Then, sort by score from low to high. For those with the same score, sort by generation order, and finally select the top ranked ones in the sorting results. The states constitute the state set of the next stage ; To sum up, the specific process of the algorithm for determining the number of bin replacements based on beam search is as follows: Step 1: Initialization (1) Calculate the total number of orders and assign an initial order to each order position (select the first orders); (2) Set the initial state, including: the current order positions to be picked List; current container; index of next backlog order; cumulative container replacement count Initialized to 0; record the arrival sequence of the material box.

[0036] Step 2: State expansion and candidate state generation For each state in the current state set, the following processing is performed: (1) If the status has met the completion conditions (all order positions All orders are picked and the waiting orders are exhausted), and the status is saved to the completion status list; (2) Otherwise, expand the state, generate all possible successor states (each state corresponds to a picking operation), and collect the candidate state set.

[0037] Step 3: Prune and update the state set (1) Sort the candidate state set according to the comprehensive evaluation function (considering the cumulative number of replacements, the number of remaining standby orders, and the remaining picking workload); (2) Before selection states as the state set for the next stage; (3) If there is a state that meets the completion conditions in the new state set, it is saved to the completion state list and the expansion is ended early; (4) Otherwise, repeat “Step 2” and “Step 3” until the maximum number of iterations is reached or all states cannot be expanded.

[0038] Step 4: Result Selection If there is a completed state, that is, when all the order positions The lists are all empty, and the supplementary orders have been added and the picking is completed. In this state, no more picking operations can be performed because all tasks have been completed. The state with the lowest comprehensive evaluation score is selected as the final optimal state; if there is no completed state, the state with the lowest score is selected from the current state set (if the order of the material boxes is unreasonable, it may cause a certain order sequence to not be completed, for example, the material box required by the current order line has been used before).

[0039] Step 5: Return to the final state Returns the optimal state, which includes the optimal bin arrival sequence and the corresponding minimum cumulative bin replacement times B .

[0040] 3. Multi-picking platform material box sorting decoupling algorithm: In an actual multi-picking station parallel operation environment, if different picking stations need the same material box at the same time, coupling conflicts are very likely to occur, seriously affecting the picking efficiency. To effectively solve this problem, this embodiment further proposes a multi-picking station material box sorting decoupling algorithm, which considers the needs of all picking stations at the same time in each picking stage, and quickly screens out conflict-free and efficient material box sorting schemes for decoupling between multiple picking stations by constructing and expanding the global state space and using the pruning mechanism of beam search.

[0041] As attached Figure 2 As shown, the specific steps of the algorithm include: (1) Global state initialization Inside each picking station, the items to be picked are initialized according to the existing order sorting scheme. List, current material box, waiting order information and material box replacement times and other information.

[0042] (2) Parallel phased state expansion During each picking phase, each picking station independently performs a "single picking station state transfer" to generate its own local candidate state set. Each state represents the updated result after a bin picking operation, specifically including: In the case of a single picking station, each stage only needs to update the picking status of a single order line. However, when it is expanded to multiple picking stations, each stage becomes a parallel process in which all picking stations advance synchronously. That is, all picking stations must perform their own picking actions at the same time, and each picking station has its own local state at a certain stage. : (13) Among them, the collection Indicates picking platform No. The number of order lines in the current order that have not been picked yet on the order slot. If no order is assigned to an order slot, define ; Indicates the current picking station The bin number used to determine whether a bin change occurs during the picking process; The cumulative number of bin replacements from the initial state to the current state; Indicates the position index of the next order position to be filled in the waiting order queue.

[0043] For each picking station in the system , based on the current state , independently execute the state transfer process of a single picking platform logic, according to the single picking platform state transfer scenario, for the picking platform Generate a set of feasible local candidate states for the next stage: (14) Each local candidate state has the following information: (15) (3) Global state combination and conflict detection The local states generated independently by each picking station are combined into a global state through a Cartesian product. For each generated global state, a bin conflict check is performed. If there is no bin conflict between picking stations, the state is legal and proceeds to the next evaluation step; if there is a bin conflict, the state is excluded, specifically including: After each picking station generates a local candidate state, the Cartesian product operation is performed on the candidate state sets of all picking stations to construct a complete global candidate state set. Each generated global candidate state It can be expressed as: (16) Among them, the global cumulative number of bin replacements It is the sum of the number of replacements of each picking station in the current candidate state and the cumulative number of replacements before, specifically expressed as: (17) in is the number of picking stations.

[0044] In the multi-picking platform parallel operation scenario defined in this algorithm, "conflict" clearly means: if and only if, at the same stage, different picking platforms select the same material box (i.e., pick the same SKU ), a conflict is considered to have occurred, which can be expressed as follows: (18) To avoid coupling problems, conflicts need to be filtered: For each global candidate state , extract the bins determined in each picking stage respectively to form a global bin set: (19) Then, detect the collection after deduplication The number of elements and the number of picking stations whether they are equal; 1. If the number of elements is equal to ,Right now , it means that the bins selected by each picking station are unique in this stage, and the global candidate state meets the non-conflict constraint and can enter the next evaluation stage; 2. If the number of elements is less than ,Right now , it means that at least two picking stations have selected the same material box. At this time, the global state will be immediately excluded from the candidate state set and will not participate in the subsequent state evaluation and pruning process.

[0045] Ensure that all candidates entering the evaluation phase are in the correct status All strictly meet the global constraints of no conflict among multiple picking stations, and finally form a legal global candidate set .

[0046] (4) Beam search pruning optimization A comprehensive evaluation function is used to score the legal global candidate states. The evaluation function comprehensively considers the number of bin replacements, the remaining picking quantity and the number of standby orders, and only retains the top candidates with the best scores through sorting. Enter the next stage in each state to avoid too many search states, including: After completing the conflict filtering, the legal global candidate states are scored and sorted according to the comprehensive evaluation function. The comprehensive evaluation function is as follows: (20) in, Represents the global cumulative number of bin replacements; Indicates that all picking stations have not yet completed The total number of rows; Indicates the remaining quantity of waiting orders for all picking stations; is a positive weight, the goal is to minimize ; Then, the legal global candidate states are sorted by the beam search mechanism, and only the top states with the lowest scores are retained. The search process enters the next stage, thereby controlling the search scale and effectively improving the search efficiency.

[0047] (5) Iterative advancement and termination judgment The state expansion and pruning process is executed cyclically until all order picking tasks are completed or the maximum number of iterations is reached.

[0048] (6) Output the optimal solution Select the state with the lowest score among all completed states, and output the corresponding bin sorting sequence and the cumulative number of bin replacements.

[0049] In order to verify the feasibility and effectiveness of the above solution of this application, this embodiment also discloses the following specific implementation cases: Experimental analysis of order similarity allocation strategy: To verify the effectiveness of the order similarity allocation strategy, random order data with different similarity levels (0.1 to 0.9) were generated, and the proposed similarity allocation strategy was compared with the traditional dynamic random allocation strategy. The experimental setting was 80 orders and 4 picking stations; After running the above strategies, record the average similarity of orders within the picking platform under different similarity parameters (measure the similarity between orders on the same picking platform). The results are shown in the attached figure. Figure 3 And attached Figure 4 shown.

[0050] Attachment Figure 3 The results show that when the overall similarity of orders is gradually improved, the different allocation strategies have an impact on the similarity of orders within the picking station. When the overall similarity is low (such as similarity 0.1-0.2), the difference between the two strategies in the similarity of orders within the picking station is not significant. As the similarity gradually increases, the order similarity allocation strategy gradually shows a higher mean value of similarity within the station, indicating that this strategy is more inclined to Orders with large intersections are placed on the same picking table, thus reducing the number of orders in the pharmaceutical order picking scenario. The role of dispersion.

[0051] Attachment Figure 4 The trend of the change of order similarity between “picking stations” is shown. As the overall similarity increases, the increase in the similarity of orders between different picking stations by the traditional dynamic allocation strategy is relatively large, indicating that under this strategy, orders with the same or similar Orders with similar orders may be distributed more dispersedly to multiple picking stations; while the order similarity allocation strategy significantly reduces the dispersion, making the overlap between different picking stations smaller, which is more in line with the requirement for order independence under pharmaceutical batch constraints.

[0052] In summary, the order similarity allocation strategy can achieve the allocation effect of "similar orders on the same platform and different orders across platforms" at a high degree of similarity, which is more practical for reducing interference between different batches of orders in the medical scenario.

[0053] Experimental analysis of single-picking platform order sorting algorithm based on neighborhood search: Experimental design and data description: In order to verify the effectiveness and stability of the single-picking platform order sorting algorithm based on neighborhood search proposed in this application, this embodiment designs a numerical experiment. The specific experimental settings are as follows: Order data generation: Generate 20 orders for one picking station. Can overlap between different orders, that is, within each order No duplication, but allowed between different orders Repeat to simulate the same order in different environments demand.

[0054] Picking station configuration: Set the number of order slots that can be processed in parallel on the picking station to 4. At the start of picking, the first 4 orders are assigned to the 4 order slots for processing. After all pickings for a certain order slot are completed, new orders are added from the waiting list to the vacant order slots.

[0055] Algorithm parameters: Determine the algorithm for the number of bin replacements based on the beam search, and set different Parameter, to evaluate the balance between search accuracy and computational efficiency, with values ​​ranging from {1, 10, 15}.

[0056] A random order sorting strategy is set as the comparison baseline. To ensure the stability of the random strategy, each sorted random strategy is repeated 30 times and the average value is taken. are respectively taken as (1000, 1, 1), where Tend to focus on reducing the number of box changes, and For unfinished orders and remaining items to be picked The number is also given a certain weight, which is much smaller than .

[0057] Experimental results and discussion Impact analysis on order sorting optimization Run the algorithm and record the different The minimum cumulative number of bin replacements in each picking stage during the parameter search process The changes in each strategy (different The cumulative number of bin replacements obtained by the beam search Comparison, as shown in the attached Figure 5 As shown: Under a single order sequencing scheme, different The parameters are selected differently in each picking stage, which leads to the final cumulative number of bin replacements. There are obvious differences, Since larger strategies can evaluate more states at each stage, they can select solutions with better state quality in the early stages, making the final The value decreased significantly, and Smaller strategies may only obtain poor states in stage selection due to limited search range, resulting in a high B value at convergence. The settings directly affect the diversity and quality of state selection at each stage, which in turn determines the degree of optimization and stability of the final result.

[0058] Comprehensive comparative analysis of single-picking platform order sorting algorithm based on neighborhood search and random strategy: Attachment Figure 6 The cumulative number of bin replacements between the neighborhood search-based single picking platform order sorting algorithm proposed in this paper and the random order sorting strategy under different order plans is compared.

[0059] From the attached Figure 6 It is obvious that no matter which Regardless of the parameters, the single-picking station order sorting algorithm based on neighborhood search always performs well in reducing the cumulative number of bin replacements, and its advantages far exceed those of the random order sorting strategy.

[0060] When using a larger When , beam search can find the optimal path in a wider state space, thereby achieving a lower cumulative number of bin replacements in most order plans ( value), when the smaller Or when relying entirely on random strategies, the algorithm is more likely to fall into local optimality due to the limited local search range, resulting in The value fluctuates greatly in different order arrangements, and the overall optimization effect is poor.

[0061] Therefore, the single picking platform order sorting algorithm based on neighborhood search can not only stably achieve low value, and further improved the results of the final ranking scheme by increasing the number of search states.

[0062] Stability analysis of the single-picking platform order sorting algorithm based on neighborhood search for random initial sequences: To further verify the stability of the algorithm, multiple independent experiments were conducted using the same data configuration (20 orders, with 4 order digits per platform). In each experiment, a different initial order sequence was randomly generated and the proposed single-picking platform order sorting algorithm based on neighborhood search was applied to solve the problem. As attached Figure 7 The optimal value obtained for each run is shown. Value line with mean and (Standard Deviation) error band, it can be found that most of the running The values ​​fluctuate around a relatively small range, indicating that the algorithm can achieve good convergence effect for different random initial solutions under the same parameter settings; a relatively slightly higher value appears in a very small number of runs. This usually means that the search failed to successfully transition to a better solution during the early stages of the local optimum or the acceptance of inferior solutions, but overall such cases are rare. It can be concluded that the algorithm of this application is relatively less dependent on the initial solution and exhibits a certain degree of stability and robustness.

[0063] Experimental analysis of the decoupling algorithm for sorting bins on multiple picking platforms: Experimental design and data description In order to evaluate the actual performance of the multi-picking station material box sorting decoupling algorithm proposed in this application, 9 sets of multi-picking station parallel experimental scenarios were constructed. Each scenario contains 4 independent picking stations, and each picking station is equipped with 4 order positions. The generation method and scale of order data are the same as above to ensure the consistency and comparability of experimental conditions.

[0064] The three specific comparative analysis strategies are as follows: (1) Random strategy: Each picking station is independently selected by randomly selecting the bins for picking, and the bin picking order of each picking station is obtained; (2) Single picking station cluster search strategy: Execute the bin replacement number determination algorithm based on cluster search for each picking station (without considering the conflicts between different picking stations) to obtain the bin picking order for each picking station; (3) Decoupling strategy for sorting bins at multiple picking stations: In each parallel stage, the local state of each picking station is expanded, and a global candidate is formed through Cartesian product combination. Then, the combination of "duplicate SKUs in the same stage" is filtered out to ensure that the final picking sequence has no conflict, and set The value is 10 to filter the status and count the last number of bin replacements and coupling degree.

[0065] To quantify the comprehensive performance of each algorithm, the following key indicators are recorded: (1) Coupling times: indicates the number of conflicts caused by the same SKU used by each picking station in the same stage; (2) Number of bin replacements: The total number of bin replacements at the four picking stations reflects the picking efficiency; (3) Algorithm running time: the time (in seconds) it takes for the algorithm to start and generate the final sorting solution.

[0066] Experimental results and discussion: The three strategies were applied to 9 groups of test data respectively. According to the experimental results, the random strategy cannot avoid the same stage in the multi-picking platform parallel scenario. Conflict, resulting in a relatively high number of couplings, and due to random selection It is more difficult to control the frequency of bin replacement, which results in a high number of bin replacements. Although its operating time is the shortest, the overall picking efficiency and conflict level are not good.

[0067] Single picking station cluster search optimizes the picking order within a single picking station, which can significantly reduce the number of bin replacements. However, it does not coordinate between picking stations, so conflicts in the parallel phase cannot be avoided. Compared with the random strategy, the degree of coupling is reduced, but a certain frequency of repetition may still occur. .

[0068] In the decoupling strategy for sorting bins at multiple picking stations, since the “same bins at the same stage” are filtered at each stage, SKU " is a candidate combination. Ideally, the number of couplings can be reduced to 0 (that is, no conflict at all). Experimental results also show that the "number of couplings" of this strategy is significantly lower than that of the first two strategies, and there is no conflict. At the same time, it also has obvious advantages in reducing the number of bin replacements. However, due to the need for larger-scale state combinations (Cartesian product expansion) and conflict detection and evaluation, its algorithm running time is relatively the longest among the three.

[0069] In summary, the multi-picking station bin sorting decoupling algorithm can achieve the lowest coupling times and better bin replacement times in most test scenarios, showing obvious picking quality advantages; if the actual application environment has higher requirements on the amount of conflicts and bin scheduling quality, the algorithm will be more competitive.

[0070] Example 2 Figure 8 2 is a system diagram illustrating an order processing decision-making device for the pharmaceutical distribution industry according to another exemplary embodiment, the device comprising: Seed order allocation module 1: for all orders in a batch to be processed, a seed order is allocated to each picking station according to the similarity coefficient between any two orders; All Orders Allocation Module 2: For the remaining orders that have not been allocated as seed orders, for each order, calculate the number of goods of the same type as the seed orders allocated to each picking station as a coupling factor, and allocate each remaining order to the picking station with the largest coupling factor to complete the order allocation; Preliminary order sequence acquisition module 3: for generating multiple candidate order sequences for the multiple orders assigned to each picking station using a random method to form an initial candidate set, and performing local optimization on each candidate order sequence in the initial candidate set using neighborhood search to obtain a preliminary local optimal order sequence; Final order sequence acquisition module 4: used to perform non-adjacent order swapping and local segment flipping operations on the preliminary local optimal order sequence to obtain a final candidate set, and select the order sequence with the least number of bin replacements from the final candidate set as the optimal order sequence for the picking station; Local candidate state acquisition module 5: used to obtain the initial state set of each picking station according to the optimal order sequence within each picking station; each picking station independently performs a picking operation to generate a local candidate state set of each picking station; Global candidate state acquisition module 6: used to form a global candidate state combination by Cartesian product of the local candidate state sets of each picking station; perform bin conflict detection on each generated global candidate state. If there is no bin conflict between picking stations, the global candidate state is legal; if there is a bin conflict, the global candidate state is excluded; Global state screening module 7: It is used to score the legal global candidate states using a comprehensive evaluation function, sort the scores from low to high, and retain the lowest-scoring candidate states. K A global candidate state; Cyclic extension module 8: used to retain K The global candidate states are cyclically expanded and pruned until all order picking tasks are completed according to the optimal order sequence of each picking station or the maximum number of iterations is reached, and the final global candidate state set is obtained; The global optimal state acquisition module 9 is used to select the state with the lowest comprehensive evaluation function score in the final global candidate state set as the global optimal state, and output the bin sorting sequence corresponding to the global optimal state.

[0071] Example 3: This embodiment provides a storage medium storing a computer program. When the computer program is executed by a host controller, each step of the above method is implemented. It is understandable that the storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0072] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0073] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are used for descriptive purposes only and should not be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" is at least two.

[0074] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0075] It should be understood that various components of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof may be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0076] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0077] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing module, or each unit may exist physically separately, or two or more units may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or in the form of software functional modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may also be stored in a computer-readable storage medium.

[0078] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0079] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0080] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. An order processing decision-making method based on the pharmaceutical distribution industry is characterized by: The method comprises: For all orders in a batch to be processed, a seed order is assigned to each picking station according to the similarity coefficient between any two orders; For the remaining orders that have not been assigned as seed orders, the number of goods of the same type as the seed orders assigned to each picking station is calculated as the coupling factor for each order, and each remaining order is assigned to the picking station with the largest coupling factor to complete the order allocation; For the multiple orders assigned to each picking station, a random method is used to generate multiple candidate order sequences to form an initial candidate set. Each candidate order sequence in the initial candidate set is locally optimized using neighborhood search to obtain a preliminary local optimal order sequence. Perform non-adjacent order swapping and local segment flipping operations on the preliminary local optimal order sequence to obtain a final candidate set, and select the order sequence with the least number of bin replacements from the final candidate set as the optimal order sequence for the picking station; Within each picking station, an initial state set of each picking station is obtained according to the optimal order sequence; each picking station independently performs a picking operation to generate a local candidate state set of each picking station; The local candidate state sets of each picking station are combined through Cartesian products to form a global candidate state combination; each generated global candidate state is tested for bin conflicts. If there is no bin conflict between picking stations, the global candidate state is legal; if there is a bin conflict, the global candidate state is excluded; Use comprehensive evaluation function to score the legal global candidate states, sort the scores from low to high, and retain the top candidate with the lowest score. K A global candidate state; For reservations K The global candidate states are cyclically expanded and pruned until all order picking tasks are completed according to the optimal order sequence of each picking station or the maximum number of iterations is reached, and the final global candidate state set is obtained; The state with the lowest comprehensive evaluation function score is selected from the final global candidate state set as the global optimal state, and the bin sorting sequence corresponding to the global optimal state is output.

2. The method according to claim 1, characterized in that The method of allocating a seed order to each picking station according to the similarity coefficient between any two orders includes: Calculate the similarity coefficient between any two orders based on the number of repeated ordered goods between them, and obtain a similar data set; Selecting two orders with the largest similarity coefficients from the similarity data set as seed orders for a picking station, and deleting all similarity coefficients related to the two selected orders from the similarity data set; Among the remaining similarity coefficients, continue to select the two orders with the largest similarity coefficients as the seed orders for the next picking station, and delete all similarity coefficients related to the two selected orders; repeat this step until a seed order is assigned to each picking station.

3. The method according to claim 2, characterized in that Also includes: Set the difference in the number of orders allocated to two adjacent picking stations to not exceed the preset threshold; When the order quantity assigned to any picking station just meets the difference between the order quantity of the adjacent picking station and the preset threshold, the orders that need to be assigned to the picking station will be sorted according to the coupling factor and assigned to the picking station with the second largest coupling factor. If the picking station with the second largest coupling factor has also met the order quantity, the order will be assigned to the picking station with the third largest coupling factor, so that the order quantity under each picking station is relatively balanced.

4. The method according to claim 3, characterized in that The method of performing local optimization on each candidate order sequence in the initial candidate set by using neighborhood search to obtain a preliminary local optimal order sequence includes: For each candidate order sequence in the initial candidate set, enumerate all adjacent order pairs in the sequence in sequence; According to the order of adjacent order pairs, the order of adjacent order pairs is exchanged in sequence. After each exchange of the order pairs, the minimum number of bin replacement times of the new sequence after the order pair order is exchanged is calculated using the bin replacement number determination algorithm based on beam search. If the minimum number of bin replacements of the new sequence after the exchange is less than the minimum number of bin replacements of the sequence before the exchange, the new sequence is accepted; Continue to exchange adjacent orders in the updated sequence until the order exchange of all adjacent orders cannot reduce the minimum number of bin replacements, and obtain the preliminary local optimal order sequence.

5. The method according to claim 4, characterized in that The preliminary local optimal order sequence is subjected to non-adjacent order exchange and local segment flipping operations in sequence, and the final candidate set obtained includes: Perform a sequence swap on any two non-adjacent order pairs in the preliminary local optimal order sequence, and obtain the minimum bin replacement number of the sequence after the sequence swap; if the minimum bin replacement number of the sequence after the swap is less than the minimum bin replacement number of the sequence before the swap, then accept the swap; repeat this process until the sequence swap of any two non-adjacent order pairs cannot reduce the minimum bin replacement number; For the order sequence after the non-adjacent order exchange, select any continuous subsequence, perform a reverse operation on all orders in the subsequence, and obtain the minimum number of bin replacements of the sequence after the reverse operation. If the minimum number of bin replacements of the sequence after the reverse operation is less than the minimum number of bin replacements of the sequence before the reverse operation, the exchange is accepted; repeat this process until the reverse operation of any continuous subsequence cannot reduce the minimum number of bin replacements, and the final candidate set is obtained.

6. The method according to claim 5, characterized in that Also includes: During the process of performing neighborhood search and local optimization on the order sequence, or performing non-adjacent order swapping operations on the order sequence, or performing local segment flipping operations on the order sequence, an acceptance probability is set, and the acceptance probability decreases step by step with the number of iterations; When the minimum number of bin replacements in the updated order sequence is greater than or equal to the minimum number of bin replacements in the order sequence before the update, the update is accepted with the set acceptance probability and the local optimum is jumped out.

7. The method according to claim 6, characterized in that The calculation of the minimum number of bin replacement times using the bin replacement number determination algorithm based on the beam search includes: Get the total number of orders assigned to any picking station and assign each order to the picking station c order positions, and obtain the order sequence for each order position; Set the initial state set of the picking station, the initial state of the picking station includes: the current order of each order position of the picking station to be picked SKU List, current bin, index of next standby order, cumulative bin replacement count initialized to 0, record bin arrival sequence; Determine whether each state in the initial state set of the picking station meets the preset completion state, the preset completion state includes: SKU All are picked and the waiting orders are exhausted; If any state meets the preset completion state, the state is saved to the completion state list; otherwise, the state is expanded and all possible picking operations are enumerated to obtain the candidate state set for the next stage; the state changes in the candidate state set compared to the state before expansion include: The order line value of the order position corresponding to the picked order line is reduced by 1. When the value is reduced to 0, it means that all current orders on the order position have been picked. A new order needs to be added from the waiting order queue. The corresponding order line value is updated to the number of rows of the new order, and the order number is increased by 1. SKU Switching is used to determine whether the material box needs to be replaced. If the material box used in the new state is different from the previous state, it means that the material box has been replaced, and the cumulative number of material box replacements is increased by 1; Each state in the candidate state set is sorted from low to high according to the score calculated by the comprehensive evaluation function. For states with the same score, they are sorted in the order of generation. The comprehensive evaluation function calculation includes: obtaining the comprehensive evaluation function score by using the cumulative number of bin replacements, the remaining picking workload, the remaining number of standby orders, and a preset weight index in the state; Before selecting the comprehensive evaluation function score K states as the state set for the next stage; If there is a state that meets the completion conditions in the new state set, it is saved to the completion state list and the expansion ends; if there is no state that meets the completion conditions in the new state set; Then continue to expand each state in the new state set and select the state with the highest score of the comprehensive evaluation function. K The state continues to be the state set of the next stage; it is judged again whether there is a state that meets the completion condition in the new state set; this step is repeated until a state that meets the completion condition exists in the new state set after a certain expansion, or the preset number of iterations is reached, or all states in the new state set after a certain expansion cannot be expanded, and the expansion is ended; If there is a completed state, the state with the lowest comprehensive evaluation function score is selected from the completed states as the final optimal state; If there is no completed state, the state with the lowest comprehensive evaluation function score is selected from the state set obtained in the last iteration as the final optimal state; The order sequence corresponding to the final optimal state is the optimal order sequence, and the accumulated number of material box replacements corresponding to the final optimal state is the minimum number of material box replacements for the picking station.

8. The method according to claim 7, characterized in that The reserved K The global candidate state loop performs state expansion and pruning operations including: For reservations K The global candidate states perform a picking operation according to the optimal order sequence of each picking station to obtain the expanded global candidate state set; Perform bin conflict detection on each global candidate state in the expanded global candidate state set. If there is no bin conflict between picking stations, the global candidate state is legal; if there is a bin conflict, the global candidate state is excluded. The comprehensive evaluation function is used to score the expanded legal global candidate states, and the scores are sorted from low to high, and the top candidate with the lowest score is retained. K A global candidate state; Repeat the above steps until all order picking tasks are completed or the maximum number of iterations is reached, and the final global candidate state set is obtained.

9. An order processing decision-making device based on the pharmaceutical distribution industry, characterized in that: The device comprises: Seed order allocation module: for all orders in a batch to be processed, a seed order is allocated to each picking station according to the similarity coefficient between any two orders; All Orders Allocation Module: For the remaining orders that have not been allocated as seed orders, the module calculates the number of goods of the same type as the seed orders allocated to each picking station as the coupling factor, and allocates each remaining order to the picking station with the largest coupling factor to complete the order allocation; Preliminary order sequence acquisition module: used to randomly generate multiple candidate order sequences for the multiple orders assigned to each picking station to form an initial candidate set, and then perform local optimization on each candidate order sequence in the initial candidate set using neighborhood search to obtain a preliminary local optimal order sequence; Final order sequence acquisition module: used to perform non-adjacent order exchange and local segment flipping operations on the preliminary local optimal order sequence in sequence to obtain a final candidate set, and select the order sequence with the least number of bin replacements from the final candidate set as the optimal order sequence for the picking station; Local candidate state acquisition module: used to obtain the initial state set of each picking station according to the optimal order sequence within each picking station; each picking station independently performs a picking operation to generate a local candidate state set of each picking station; Global candidate state acquisition module: used to form a global candidate state combination by Cartesian product of the local candidate state sets of each picking station; perform bin conflict detection on each generated global candidate state. If there is no bin conflict between picking stations, the global candidate state is legal; if there is a bin conflict, the global candidate state is excluded; Global state screening module: It is used to score the legal global candidate states using a comprehensive evaluation function, sort the scores from low to high, and retain the lowest-scoring candidate states. K A global candidate state; Cycle extension module: used to retain K The global candidate states are cyclically expanded and pruned until all order picking tasks are completed according to the optimal order sequence of each picking station or the maximum number of iterations is reached, and the final global candidate state set is obtained; A global optimal state acquisition module is used to select the state with the lowest comprehensive evaluation function score in the final global candidate state set as the global optimal state, and output the bin sorting sequence corresponding to the global optimal state.

10. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the main controller, each step of the order processing decision method based on the pharmaceutical distribution industry as described in any one of claims 1 to 8 is implemented.

Images