Multi-stage customized combination strategy-based warehouse picking optimization method, device and medium

Abstract

The present application relates to a kind of multi-stage self-defined combination strategy warehouse picking optimization method, device and medium, wherein, method includes collecting order data to be picked and preprocessing generation structured data set;Support user self-defined configuration wave division dimension and weight, complete order clustering and generate picking wave with unique identification;Divide the whole process of picking into four core stages, provide each stage strategy selection and combination interface, generate compatible multi-stage strategy combination scheme;Correlation scheme and wave task are executed in stages, real-time acquisition of field data and dynamically correct abnormal strategy;After completing picking, the core index is counted, report is generated and review output optimization suggestion, the present application solves the problem that traditional picking strategy is poor in adaptability, low efficiency, can be flexibly adapted to multiple warehouse scene, improve picking efficiency and accuracy, reduce resource consumption, and the practical value is remarkable.

Description

Technical Field

[0001] This invention relates to the technical field of warehouse picking systems, and in particular to a method, apparatus, equipment and medium for automatic configuration and calculation of multiple miscellaneous fees quotations. Background Technology

[0002] In the warehousing and logistics industry, picking is the core link connecting warehousing and order fulfillment. Its efficiency and accuracy directly affect the overall operating cost of the logistics chain, order delivery timeliness, and customer satisfaction. With the rapid development of e-commerce, warehousing orders are characterized by "multiple varieties, small batches, and high frequency." At the same time, the storage requirements and order timeliness requirements of different product categories vary significantly, and warehouse layouts are also diversified due to different site sizes and business types. This places higher demands on the adaptability and flexibility of picking optimization technologies. Existing warehousing picking optimization technologies mostly adopt fixed wave division strategies and single-stage optimization schemes: In terms of wave division, orders are often aggregated based on a single dimension, which cannot flexibly adjust the division dimension and weight according to the actual business scenario. This easily leads to uneven wave task load, duplicate picking paths, and reduced picking efficiency. In terms of picking process optimization, optimization strategies are mostly designed for single links, lacking a holistic consideration of the entire process of wave planning, path optimization, zone collaboration, and review and sorting. The strategies at each stage are loosely connected, making it difficult to form a collaborative optimization effect.

[0003] In summary, existing warehouse picking optimization technologies suffer from poor adaptability, insufficient flexibility, weak end-to-end collaboration, and a lack of dynamic adjustment and post-mortem optimization capabilities, making it difficult to meet the efficient picking needs of diverse warehousing scenarios. Therefore, developing a warehouse picking optimization technology that allows for customizable combination strategies, adapts to multi-stage collaborative optimization, and possesses dynamic adjustment capabilities has become an urgent technical problem to be solved in the current warehousing and logistics field. Summary of the Invention

[0004] The main objective of this invention is to provide a warehouse picking optimization method, apparatus, and medium with a multi-stage customizable combination strategy. By customizing wave division and combining multi-stage strategies after order preprocessing, the invention dynamically executes optimization and reviews the results to output suggestions, thereby improving picking efficiency and accuracy.

[0005] To achieve the above objectives, this invention provides a warehouse picking optimization method based on a multi-stage custom combination strategy, comprising the following steps:

[0006] Obtain the set of warehouse orders awaiting picking, extract the core attribute information of each order, clean and deduplicate the core attribute information, standardize the format and fill in missing fields to generate a structured order dataset; Configure a custom wavelet partitioning strategy, allowing users to select at least one partitioning dimension and set corresponding weights based on actual warehouse business scenarios. Based on the configured custom wavelet partitioning strategy, perform order clustering on the structured order dataset to generate multiple picking waves. Each wavelet is associated with a unique wavelet identifier and a wavelet task list. The entire picking process is divided into four core stages, and each stage provides a strategy selection and combination interface, allowing users to customize and select appropriate optimization strategies for each stage, forming a multi-stage custom strategy combination scheme. Based on the wave task list and the multi-stage custom strategy combination scheme, the picking optimization operation of each stage is executed in sequence, and the optimization strategy of the current stage is adapted and corrected in real time based on the preset adjustment strategy. After completing the picking tasks for each wave, the indicators for each wave are statistically analyzed, and a wave picking result report is generated. Based on the result report and the strategy execution data for each stage, an optimization review is conducted to form a review analysis result, and optimization suggestions for the strategy combination scheme are output.

[0007] Furthermore, the steps of obtaining the set of warehouse orders awaiting picking, extracting the core attribute information of each order, cleaning and deduplicating the core attribute information, standardizing the format, and filling in missing fields to generate a structured order dataset include: By using a pre-defined standardized API interface, the warehouse management system and the order management system are called simultaneously to collect the original data of the orders to be picked. After collection, an index is generated to form a set of orders to be picked. Based on a preset core attribute extraction library, data is extracted from the original order data in a targeted manner to generate an initial attribute dataset; The attribute dataset is uniformly converted according to a preset standard format to obtain preprocessed attribute data. Order priority is quantified into 1-5 levels, warehouse location distribution is converted into a three-level code, timeliness is converted into a standard time format, and address is split into a hierarchical format. Missing fields are identified through field integrity detection. The preprocessed attribute data is encapsulated according to a preset relational data table structure, and a unique field ID and data type description are assigned. After data consistency verification, a structured order dataset that can be directly called by subsequent wave division strategies is generated.

[0008] Furthermore, the configured custom wavelet segmentation strategy allows users to select at least one segmentation dimension and set corresponding weights based on actual warehouse business scenarios. Based on this configured custom wavelet segmentation strategy, the structured order dataset is clustered to generate multiple picking waves, with each wave associated with a unique wavelet identifier and a wavelet task list. The steps include: It provides a visual wave division strategy configuration interface, allowing users to select at least one of the following as the division dimension: order priority, product correlation, warehouse location clustering range, timeliness requirements, and picking resource load, and set the weight ratio for the selected dimension. The system automatically generates standardized wavelet partitioning settings based on the dimensions and weights selected by the user. The partitioning settings include dimension matching thresholds, weight calculation formulas, and clustering termination conditions. The preset clustering algorithm is invoked to aggregate orders in the structured order dataset based on the wave division settings, forming a preliminary wave set; The initial set of picking waves is validated for rationality. Abnormal waves with task volume exceeding a preset threshold are removed and re-clustered. After the validation is passed, multiple valid picking waves are generated. A unique wave identifier is assigned to each picking wave, and a wave task list containing order list, product information, and warehouse location distribution is generated simultaneously.

[0009] Furthermore, the entire picking process is divided into four core stages, providing a strategy selection and combination interface for each stage. This allows users to customize and select appropriate optimization strategies for each stage, forming a multi-stage customized strategy combination scheme. The steps include: The entire picking process is clearly divided into four core stages: wave planning, route optimization, zone collaboration, and review and sorting. The core task objectives and boundary scope of the four core stages are defined simultaneously. Configure a visual strategy selection and combination interface for the four core stages. The interface has a built-in candidate optimization strategy library for each stage, and the candidate strategies cover core scenarios such as task splitting, path algorithm, cross-region scheduling, and review and verification. Users can select appropriate optimization strategies from the candidate strategy library for each stage based on actual warehouse business scenarios, and can fine-tune the parameters of the selected strategies. The system integrates the selected strategies at each stage to generate standardized multi-stage custom strategy combination schemes.

[0010] Furthermore, the step of sequentially executing picking optimization operations at each stage based on the wave task list and the multi-stage custom strategy combination scheme, and real-time adapting and correcting the optimization strategy for the current stage based on a preset adjustment strategy, includes: Load the verified multi-stage custom strategy combination scheme, associate and bind the multi-stage custom strategy combination scheme with the task list of each wave according to the wave identifier, and clarify the specific execution strategy of each wave in different stages. Following the sequence of wave planning, route optimization, zone collaboration, and review sorting, the picking optimization operations at each stage are executed sequentially, and the strategy execution data at each stage is recorded simultaneously. Real-time data collection of warehouse picking operations, including picker location, equipment operating status, real-time inventory, and order change information; The collected data is compared with preset thresholds to monitor whether any thresholds are exceeded or abnormal events occur.

[0011] Furthermore, the step of comparing the collected data with a preset threshold to monitor whether any threshold exceeding or abnormal events occur includes: If an anomaly is detected, a preset adjustment strategy is triggered. The optimization strategy for the current stage is adapted and corrected in real time based on the actual data on site. After correction, the strategy undergoes a second logical compatibility check. After the verification is successful, the current picking optimization operation will continue to be executed using the revised strategy, and the strategy execution record will be updated synchronously.

[0012] Furthermore, the steps of completing the picking tasks for each wave and stage, statistically analyzing the indicators for each wave, generating a wave picking result report, and conducting optimization reviews based on the result report and the strategy execution data for each stage, and outputting optimization suggestions for the strategy combination scheme, include: After completing the picking tasks for each wave of the entire stage, the core indicators for each wave are statistically analyzed. These core indicators include picking efficiency, picking accuracy, and picking resource utilization. By integrating the metrics of each wave with the wave task list, a standardized wave picking result report is generated, which includes detailed data for each wave and summary data for the entire batch. Retrieve multi-stage strategy execution data for each wave, combine it with the wave picking result report, analyze the correlation between the strategy combination scheme and picking indicators, and identify the advantages and disadvantages in the strategy execution process; Based on the results of the post-mortem analysis, optimization suggestions are output, including adjustments to the weights of strategy dimensions, replacement of candidate strategies, and directions for parameter optimization.

[0013] The present invention also provides a warehouse picking optimization device with a multi-stage customizable combination strategy, comprising: Order information processing module: Used to synchronously call the warehouse management system and the order management system through a preset standardized API interface to collect raw data of orders to be picked, extract core attributes in a targeted manner and complete data preprocessing to generate a structured order dataset that can be called by wave division strategy; Wave partitioning strategy configuration module: It provides a visual strategy configuration interface, allowing users to select partitioning dimensions and set weights. Based on the user configuration, it generates wave partitioning rules, calls clustering algorithms to aggregate structured order datasets, and generates picking waves and wave task lists with unique identifiers after reasonableness verification. Multi-stage strategy combination configuration module: It is used to divide the entire picking process into four core stages, provide strategy selection and combination interfaces for each stage, support users to customize and select appropriate strategies, and generate an executable multi-stage custom strategy combination scheme after logical compatibility verification. Picking optimization execution module: It is used to load strategy combination schemes and associate them with wave task list, execute picking optimization operations in stages, collect on-site data in real time and monitor anomalies, trigger preset adjustment rules to adapt and correct strategies in real time, and synchronously record strategy execution data. The Results Statistics and Review Module is used to collect key indicators such as picking efficiency, accuracy, and resource utilization after each picking task is completed, generate a picking result report for each wave, conduct optimization reviews based on strategy execution data, and output optimization suggestions for strategy combination schemes.

[0014] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the warehouse picking optimization method with stage-defined combination strategy described above.

[0015] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the warehouse picking optimization method of the stage-defined combination strategy described in any of the above claims. Attached Figure Description

[0016] Figure 1 This is a flowchart of a warehouse picking optimization method with multi-stage custom combination strategies in one embodiment of the present invention; Figure 2 This is a structural block diagram of a warehouse picking optimization device with a multi-stage custom combination strategy in one embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0018] Reference Figure 1 This is a flowchart illustrating a warehouse picking optimization method based on a multi-stage custom combination strategy proposed in this invention, including the following steps: S1. Obtain the set of warehouse orders to be picked, extract the core attribute information of each order, clean and deduplicate the core attribute information, standardize the format and fill in missing fields to generate a structured order dataset; S2, Configure a custom wave division strategy, allowing users to select at least one division dimension and set corresponding weights based on actual warehouse business scenarios. Based on the configured custom wave division strategy, perform order clustering on the structured order dataset to generate multiple picking waves. Each wave is associated with a unique wave identifier and a wave task list. S3 divides the entire picking process into four core stages, providing a strategy selection and combination interface for each stage, allowing users to customize and select appropriate optimization strategies for each stage, forming a multi-stage custom strategy combination scheme. S4. Based on the wave task list and the multi-stage custom strategy combination scheme, the picking optimization operation of each stage is executed in sequence, and the optimization strategy of the current stage is adapted and corrected in real time based on the preset adjustment strategy. S5. After completing the picking tasks for each wave and stage, the indicators for each wave are statistically analyzed, and a wave picking result report is generated. Based on the result report and the strategy execution data for each stage, an optimization review is conducted to form a review analysis result, and optimization suggestions for the strategy combination scheme are output.

[0019] As described in step S1 above, the Warehouse Management System (WMS) and Order Management System (OMS) are first synchronously invoked through a pre-defined standardized API interface to collect raw data of orders to be picked, encapsulated in JSON format. An order set is then formed by generating an index based on "collection timestamp + system identifier". Subsequently, based on a pre-defined extraction rule base, core attribute information is extracted from the raw data, including key fields such as order priority, product SKU, product quantity, product storage location distribution, order timeliness requirements, and delivery address, generating an initial attribute dataset. Next, data preprocessing is performed: duplicate data is removed based on the unique order identifier, and invalid orders such as those with abnormal product quantities or expired timeliness are filtered out; format conversion is completed according to a unified standard, such as quantifying order priority into 1-5 levels and converting storage location distribution into a three-level code of "storage area-channel-storage location"; missing information is identified through field integrity detection, and requests for supplementation are automatically initiated for those that can be supplemented across systems, while those that cannot be supplemented are marked as abnormal orders and pushed to the operations end for review. Finally, the processed attribute data is packaged according to the preset data table structure, and after consistency verification, a structured order dataset is generated to ensure that it can be directly used for subsequent wave division strategies.

[0020] As described in step S2 above, a visual strategy configuration interface is provided, allowing users to select at least one dimension based on actual warehousing business scenarios, such as order priority, product relevance, storage location clustering range, timeliness requirements, and picking resource load, and set weight percentages for the selected dimensions. Based on the user configuration, the system automatically generates standardized wavelet partitioning rules containing dimension matching thresholds, weight calculation formulas, and clustering termination conditions. Then, a preset clustering algorithm is called to aggregate orders in the structured order dataset according to these rules, forming a preliminary wavelet set. Finally, the preliminary wavelet set undergoes a rationality check, eliminating abnormal waves with task volumes exceeding preset thresholds and re-clustering them. After successful verification, multiple valid picking waves are generated, each assigned a unique identifier, and a wavelet task list containing order lists, product information, and storage location distribution is generated simultaneously, providing a task foundation for subsequent multi-stage picking optimization.

[0021] As described in step S3 above, the entire picking process is clearly divided into four core stages: wave planning, path optimization, zone collaboration, and review and sorting. The core task objectives and boundaries of each stage are defined simultaneously to avoid task overlap or omissions between stages. Subsequently, a visual strategy selection and combination interface is configured for each core stage. This interface includes a built-in candidate optimization strategy library for the corresponding stage, covering core scenarios such as wave task splitting, dynamic path planning, cross-zone picking scheduling, and multi-dimensional review and verification. Users can select suitable strategies from the candidate strategy library for each stage based on actual business scenarios such as warehouse order volume, product type, and warehouse layout. Users can also fine-tune the key parameters of the selected strategies. Finally, the system integrates the selected strategies for each stage, generating a standardized multi-stage custom strategy combination scheme. The scheme undergoes logical compatibility verification to ensure that there are no conflicts between strategies at each stage. After successful verification, an executable strategy scheme is output.

[0022] As described in step S4 above, a validated multi-stage custom strategy combination scheme is loaded. The scheme is then associated and bound to the task list of each wave according to the wave identifier, clarifying the specific execution strategy for each wave in the stages of wave planning, path optimization, zone collaboration, and review sorting. Subsequently, picking optimization operations are executed sequentially according to the predetermined stage order, synchronously recording core data such as the execution progress and resource usage of each stage strategy in real time. During the process, on-site data such as the location of picking personnel, equipment operating status, real-time inventory of goods, and temporary order changes are dynamically collected through the warehouse's on-site sensing devices and system interface, and compared with preset thresholds to achieve anomaly monitoring. If an anomaly is detected, preset adjustment rules are immediately triggered, and the strategy for the current stage is adapted and corrected based on the actual on-site data. After correction, a second logical compatibility check is performed. After the check passes, the corrected strategy continues to execute the current stage operation, and the strategy execution record is updated to ensure that the picking process dynamically adapts to changes in the on-site environment.

[0023] As described in step S5 above, after the completion of each wave's picking task across all stages, the system automatically compiles core metrics for each wave, including picking efficiency metrics (such as picking volume per unit time), picking accuracy metrics (such as picking error rate), and picking resource utilization metrics (such as personnel / equipment load rate). Subsequently, the system integrates the statistical metrics of each wave with the corresponding wave task information to generate a standardized wave picking result report. The report covers detailed data for each wave and summary data for the entire batch, clearly presenting the picking performance of each wave. Next, the system retrieves multi-stage strategy execution data for each wave and analyzes the correlation between strategy combination schemes and picking metrics in conjunction with the result report, accurately identifying the advantages and disadvantages of strategy execution. Finally, based on the review analysis results, targeted optimization suggestions are output, including adjustments to the weights of strategy division dimensions, replacement of candidate strategies, and optimization directions for key parameters, forming a reference that can directly guide subsequent strategy configuration.

[0024] In one embodiment, step S1, which involves obtaining a set of warehouse orders awaiting picking, extracting the core attribute information of each order, cleaning and deduplicating the core attribute information, standardizing its format, and filling in missing fields to generate a structured order dataset, includes: S11, Collection of orders to be picked; S12, Targeted Extraction of Core Attributes; S13, Core attribute data preprocessing; S14, Generation of structured order dataset.

[0025] In practical implementation, a real-time data synchronization link is established with the Warehouse Management System (WMS) and Order Management System (OMS) through a pre-defined standardized API interface. The interface uses JSON format to encapsulate and transmit the raw data of the orders to be picked. During transmission, a data verification mechanism is enabled to ensure data integrity. After collection, a unique index is generated according to "collection timestamp + system identifier" to form a set of orders to be picked. Taking a daily operation scenario of an e-commerce warehouse as an example, this step can synchronously collect customer order information from the OMS system and product storage association information from the WMS system, covering more than 200 orders of daily necessities and cosmetics to be picked generated between 9:00 and 10:00 on the same day. Based on a pre-defined core attribute extraction rule base, targeted extraction operations are performed. The rule base has built-in field name mapping relationships, data type definitions and extraction thresholds for each core field, which can accurately extract key fields such as order priority, product SKU, product quantity, product storage location distribution, order time requirement and delivery address from the raw order data. The order priority is derived by a weighted algorithm of customer level (VIP customers have higher priority) and order timeliness requirements (same-day delivery > next-day delivery > normal timeliness). The distribution of goods storage locations is directly related to the real-time storage location code information in the WMS system, and finally an initial attribute dataset containing more than 200 records is generated. To address redundancy, anomalies, and non-standardization issues in the initial attribute dataset, standardized preprocessing was implemented. A hash index table was established based on the unique order identifier (e.g., order number OD20240520XXXX) to quickly locate and remove three duplicate order data entries. Data validity verification rules were used to filter out two invalid orders with negative quantities and one order with an expired delivery time (originally due before 24:00 on May 19th). The data was then converted to a unified warehousing data standard, quantifying order priorities into 1-5 levels (Level 1 for VIP customers' same-day delivery orders, and Level 5 for regular customers' regular time-sensitive orders). Warehouse location distribution information was uniformly converted to a three-level coding format of "warehouse area-channel-location" (e.g., A01-03-25 represents location 25 in channel 03 of warehouse area A01). Delivery time information was converted to the standard format "YYYY-MM-DDHH:MM:SS". The receiving address was split and integrated according to the "province-city-district-detailed address" hierarchy.By using a field integrity detection algorithm to identify missing core fields, for three orders with missing product location distributions, an automatic request to supplement the data is sent to the WMS system, and the location information is matched and supplemented based on the product SKU. For two orders with missing key fields requiring timeliness, they are marked as "abnormal orders pending review" and pushed to the operations management platform with missing field prompts. After verification and supplementation by operations personnel, they are reintroduced into the processing flow. The supplemented attribute data is encapsulated according to a preset relational data table structure, and each field is assigned a unique field ID, data type, and field description. After encapsulation, data consistency verification is performed to ensure that the product quantity matches the warehouse location's available quantity, and the order timeliness matches the current time, etc. Finally, a structured order dataset is generated. This dataset can be directly used for subsequent wave segmentation strategies. In the above e-commerce warehousing case, the 196 valid order data processed by step S1 provide accurate data support for subsequent wave segmentation based on the dimensions of "warehouse location cluster range + order priority".

[0026] In one embodiment, configuring a custom wavelet partitioning strategy allows users to select at least one partitioning dimension and set corresponding weights based on actual warehouse business scenarios. Based on the configured custom wavelet partitioning strategy, order clustering is performed on the structured order dataset to generate multiple picking waves. Step S2, where each wave is associated with a unique wavelet identifier and a wavelet task list, includes: S21, Custom wavelet division strategy configuration; S22, Wavelet division rules are generated; S23, Order clustering execution; S24, wave generation and identification and task list association.

[0027] In its implementation, the system provides a web-based visual configuration interface for wavelet segmentation strategies. The interface includes a dimension selection area, a weight setting slider, and a strategy preview module. Users can independently select segmentation dimensions and set corresponding weight percentages based on order characteristics, resource allocation, and timeliness requirements of their actual warehouse operations. Selectable dimensions cover core aspects such as order priority, product relevance, warehouse location clustering range, timeliness requirements, and picking resource load. Weight settings range from 0-100%, and the sum of the weight percentages of all selected dimensions is automatically verified to be 100%. For example, taking 196 structured order data entries from an e-commerce warehouse processed in step S1, this warehouse has a high proportion of daily necessities and beauty product orders, and the product concentration is high in warehouse areas A01 and A02. Pickers are assigned tasks based on warehouse area. Based on this scenario, users select "location clustering range" and "order priority" as the core partitioning dimensions through the configuration interface, setting their weights to 60% and 40% respectively. After receiving the user's configuration instructions, the system automatically converts them into standardized wave partitioning rules. The rules specify that the matching threshold for the location clustering range is "same storage area + adjacent passages," and order priority is assigned different weight coefficients according to levels 1-5. The clustering termination conditions are set as "50-80 orders per wave" and "total weight of goods per wave not exceeding 200kg." After the rules are generated, the system calls a preset improved K-means clustering algorithm to perform structural... The storage location code and order priority quantification value in the order data are used as core feature vectors. Clustering operation is performed on 196 order data. During the clustering process, the algorithm first classifies the orders into storage area group A01, storage area group A02, and cross-storage area group based on the storage location clustering range threshold. Then, it further subdivides the orders in each group based on the order priority weight coefficient to ensure that high-priority orders (level 1-2) are aggregated first and the workload is balanced. The clustering operation generates 4 preliminary wave sets, including 2 waves for storage area A01, 1 wave for storage area A02, and 1 wave for cross-storage area. The number of orders in a single wave is 68, 72, and 56, respectively. The system performs a rationality check on the initial wave set. The check dimensions include the number of orders per wave, total weight of goods, proportion of high-priority orders, and matching degree of picking resources. The check revealed that the initial cross-warehouse wave contained three Level 1 priority orders requiring urgent processing. Furthermore, this wave involved two non-adjacent warehouses, A01 and A03. Executing this would increase the picking path length. Therefore, the system triggered a re-clustering mechanism, splitting the three Level 1 priority orders into two waves in warehouse A01. Simultaneously, two low-priority redundant orders (Level 5) from the cross-warehouse wave were removed and redistributed. After the second check passes, four valid picking waves are generated, each assigned a unique identifier. A wave task list containing order details, product SKUs and quantities, warehouse location distribution, and priority ranking is generated simultaneously. This list is automatically pushed to the handheld terminals of picking personnel in the corresponding warehouses, providing a precise task carrier for subsequent phased picking optimization.

[0028] In one embodiment, step S3, which divides the entire picking process into four core stages and provides a strategy selection and combination interface for each stage, supporting users to customize and select appropriate optimization strategies for each stage to form a multi-stage customized strategy combination scheme, includes: S31, Core Phase Division and Definition; S32 provides a strategy selection and combination interface; S33, customizable selection of optimization strategies for each stage; S34, Generation and verification of multi-stage strategy combination schemes.

[0029] In its implementation, the system first clearly defines the entire picking process into four core stages: wave planning, path optimization, zone collaboration, and verification and sorting. Simultaneously, structured documents define the task boundaries, core objectives, and input / output data specifications for each stage to avoid task overlap or gaps in connection between stages. Specifically, the wave planning stage focuses on the detailed breakdown and execution order of wave tasks; the path optimization stage focuses on minimizing and maximizing the efficiency of picking paths within a single wave; the zone collaboration stage emphasizes task collaboration and resource allocation across multiple warehouses and picking personnel; and the verification and sorting stage focuses on product verification and order matching after picking is completed. Each stage forms a closed-loop, interconnected process. Based on the task objectives of each stage, the system provides a visual web-based strategy selection and combination interface. The interface is logically laid out as "stage navigation - strategy list - parameter configuration - solution preview," with each core stage corresponding to an independent strategy selection panel. The panel contains a library of candidate optimization strategies that have been validated in practice, and all candidate strategies support fine-tuning of key parameters.

[0030] Taking the business scenario corresponding to the four picking waves of the e-commerce warehouse generated by S2 as an example, the warehouse A01 and A02 areas are dedicated to daily necessities and beauty products. Pickers are fixedly assigned according to the warehouse area, and there are 32 high-priority orders of level 1-2 that need to be completed first on the same day. Based on this, the user completes the strategy configuration for each stage through the strategy selection interface: in the wave planning stage, select the "high priority first execution strategy", set the wave corresponding to level 1-2 orders (WAVE20240520001, WAVE20240520002) as the priority execution sequence, and at the same time fine-tune the parameters to split the single wave. The task granularity is set to "10-15 orders / sub-tasks"; in the path optimization stage, considering the dense shelving in warehouse areas A01 and A02, a "genetic algorithm path planning strategy" is selected, with parameters set to "path overlap rate ≤ 5%" and "minimum number of turns"; in the zone collaboration stage, a "warehouse area-specific picking personnel scheduling strategy" is selected, binding the A01 warehouse area wave to pickers 1-3 and the A02 warehouse area wave to pickers 4-5; in the review and sorting stage, considering the characteristics of daily necessities and cosmetics being mostly small items, a "multi-dimensional barcode verification strategy" is selected, adding a "batch number + SKU dual code verification" rule. After strategy selection, the system automatically integrates the selected strategies for each stage according to the "wave-stage" dimension, generating a standardized multi-stage custom strategy combination scheme. The scheme includes the execution sequence, data interaction specifications, and dependencies of each stage's strategies. Subsequently, the system initiates a logical compatibility verification mechanism, simulating execution scenarios to verify the rationality of the connection between strategies at each stage. The verification revealed a potential conflict between the "genetic algorithm path planning strategy" in the path optimization stage and the "warehouse-specific scheduling strategy" in the partitioned collaboration stage—the original path planning did not associate with the fixed work range of picking personnel, potentially leading to cross-warehouse-area path planning. The system immediately triggered a strategy adaptation reminder, automatically adding the constraint "limiting the current warehouse-area work range" to the path planning parameters. After a second verification, the conflict was eliminated. Finally, a directly executable strategy combination scheme is generated, synchronously linked to the task lists of the four picking waves, providing precise strategic basis for the phased execution of stage S4.

[0031] In one embodiment, step S4, which involves sequentially executing picking optimization operations at each stage based on the wave task list and the multi-stage custom strategy combination scheme, and real-time adapting and correcting the optimization strategy for the current stage based on a preset adjustment strategy, includes: S41, Scheme loading is associated with wave tasks; S42, phased picking optimization operation execution; S43, On-site data acquisition and anomaly monitoring; S44, optimization strategy is adapted and corrected in real time; S45, revised policy execution and record update.

[0032] In practice, the system first loads strategy combinations that have passed logical compatibility checks. It then establishes a mapping between the strategy and the task lists for each wave using unique wave identifiers (e.g., WAVE20240520001-WAVE20240520004). This generates a four-dimensional execution matrix of "wave-stage-strategy-parameter," clearly defining the specific execution standards, data interaction requirements, and timing relationships for each wave in wave planning, path optimization, zone collaboration, and review sorting stages, ensuring precise matching between strategies and tasks. The phased picking optimization operation proceeds in an orderly manner according to a preset timeline, with each stage's execution process deeply bound to the strategy. During the wave planning phase, based on the "high-priority priority execution strategy," the two high-priority waves, WAVE20240520001 and WAVE20240520002, are prioritized for scheduling. Simultaneously, tasks are broken down into smaller parts at a granularity of "10-15 orders / sub-tasks," and pushed to the corresponding pickers' handheld terminals. During the path optimization phase, based on the "genetic algorithm path planning strategy" and the parameter constraint of "limiting the current warehouse area's operational range," the shortest picking path is generated for wave WAVE20240520001 in warehouse area A01. For example, the path information (A01-03 → A01-05 → A01-07) includes shelf numbers and product location prompts, reducing the time pickers spend searching for goods. During the zoned collaboration phase, a "warehouse-specific picker scheduling strategy" synchronizes the work progress of pickers 1-3 in warehouse A01 and pickers 4-5 in warehouse A02 in real time, avoiding resource conflicts across warehouse areas. In the verification and sorting phase, a "multi-dimensional barcode verification strategy" is implemented, using barcode scanning devices to verify the SKU and batch number of goods, ensuring picking accuracy. Data from each stage is transmitted back to the system in real time, forming a dynamic execution ledger including work duration, resource usage, and task completion rate. On-site data collection and anomaly monitoring are integrated throughout the entire execution process. The system dynamically collects data such as the real-time location of pickers, AGV vehicle operating status, real-time product inventory, and temporary order changes through UWB positioning devices and equipment status sensors deployed in the warehouse, combined with pickers' handheld terminals and the WMS real-time inventory interface. Preset anomaly thresholds and judgment rules, such as "picker deviating from the planned path by more than 5 meters for 1 minute", "AGV trolley malfunction alarm", "insufficient inventory causing picking interruption", and "order expediting / cancellation / change", are all defined as abnormal events. Taking the operation scenario of warehouse A01 in this e-commerce warehouse as an example, during the execution process, the system detected that the handheld terminal of picker No. 1 issued a malfunction alarm (the device could not receive new task instructions), and there were still 12 sub-tasks in the current wave WAVE20240520001 that had not been completed, triggering the abnormal response mechanism.Upon triggering the anomaly, the system immediately retrieves the preset adjustment rules and performs strategy adaptation correction based on actual on-site data. Combining this with the resource scheduling logic of the zoned collaboration phase, it determines that picker #2's current task completion rate has reached 80% (with 2 sub-tasks remaining), indicating sufficient capacity to handle the remaining tasks. Therefore, the 12 unfinished sub-tasks of picker #1 are split into 8 tasks assigned to picker #2, and the remaining 4 tasks are assigned to picker #3. Simultaneously, the path optimization strategy parameters are adjusted to replan the optimal path for pickers #2 and #3, including the newly added tasks, ensuring that the path overlap rate remains ≤5%. After the correction, a second logic compatibility check is initiated, focusing on verifying the matching of the new task allocation with the picker's workload and the warehouse operation range. Once no conflicts are confirmed, the corrected execution instructions are generated and pushed to the handheld terminals of pickers #2 and #3 and the system management backend. The corrected strategy is executed seamlessly, and pickers #2 and #3 continue to complete their tasks along the new paths. The system updates the execution log in real time, recording the task adjustment time, the reason for the adjustment, the strategy parameters before and after the correction, and the differences in execution progress. Although wave WAVE20240520001 experienced a brief pause due to equipment failure, it still completed all picking tasks within the preset timeframe through dynamic correction, and the picking efficiency was not significantly affected. This fully verified the feasibility and practicality of the S4 step dynamic adaptation mechanism, ensuring that the picking process can flexibly respond to unexpected situations on site and guarantee the overall operational stability.

[0033] In one embodiment, step S5, after completing the picking tasks for each wave of the entire process, statistically analyzing the indicators for each wave, generating a wave picking result report, and based on the result report and the strategy execution data for each stage, conducting an optimization review to form a review analysis result and outputting optimization suggestions for the strategy combination scheme, includes: S51, Key Picking Indicators Statistics; S52, Wave picking result report generated; S53, optimize post-mortem analysis; S54, Strategy combination optimization suggestions are output.

[0034] In its implementation, the detailed and in-depth implementation process of the S5 steps is guided by the core principle of "results accumulation - review and iteration - strategy optimization." It connects the execution results of each wave of picking tasks in the S4 phase, providing precise basis for subsequent picking strategy optimization through data-driven review, forming a closed-loop management of "execution - review - optimization." After the completion of each wave of picking tasks, the system automatically triggers the indicator statistics process. The statistical data sources cover the dynamic execution ledger of the S4 phase, the goods inbound and outbound records of the WMS system, the work logs of the pickers' handheld terminals, and the verification records of the review and sorting process, ensuring the comprehensiveness and accuracy of the statistical data. The core metrics focus on three dimensions: picking efficiency, accuracy, and resource utilization. Each dimension includes multiple subdivided quantitative indicators. Picking efficiency indicators cover the picking volume per unit time (unit: orders / hour), average completion time per wave, and average picking time per order. Picking accuracy indicators include picking error rate (number of erroneous orders / total number of orders), review pass rate (number of reviewed and approved orders / total number of orders), and SKU matching accuracy. Picking resource utilization indicators include picking personnel load rate (actual working time / effective working time), AGV equipment utilization rate (equipment running time / equipment available time), and warehouse resource occupancy efficiency. Taking the four e-commerce warehouse picking waves (WAVE20240520001-WAVE20240520004) completed in the S4 phase as an example, the system statistics show that: wave WAVE20240520001 had a picking volume of 32 orders / hour, a picking error rate of 0%, and an average load rate of 85% for pickers 1-3; wave WAVE20240520003 had a picking volume of 25 orders / hour, a picking error rate of 0.8%, and a load rate of 62% for picker 4. The data of each wave are entered into the statistical database in real time and associated with the corresponding wave identifier.After the indicators are statistically analyzed, the system automatically integrates the data to generate a standardized wave picking result report. The report adopts a two-tiered structure of "single wave details + full batch summary." The single wave details section is categorized by wave identifier, clearly presenting the specific values ​​of each core indicator for that wave, the indicator achievement status (compared to preset target values), basic task information (order quantity, product type, operation time), and abnormal event records (such as equipment failure adjustments in the S4 phase). The full batch summary section calculates the average value, analyzes extreme values, and compares the differences for the indicators across the four waves, presenting the efficiency change trend of each wave through line graphs. The pie chart displays the distribution of error order types, intuitively presenting the overall performance of the entire batch picking operation. The report supports PDF export and in-system visualization, and is automatically pushed to the warehouse operations management platform, allowing operations personnel to quickly grasp the effectiveness of the picking operation. In the in-depth review and analysis phase, the system retrieves the multi-stage custom strategy combination scheme of the S3 stage, the wave division strategy configuration data of the S2 stage, and the strategy execution details data of the S4 stage, and establishes a correlation analysis model of "strategy configuration - execution process - indicator results". By comparing the strategy configuration and indicator performance of different waves, the advantages and disadvantages of strategy execution can be identified. Based on the above e-commerce warehousing case, the analysis revealed that: waves WAVE20240520001 and WAVE20240520002, which adopted a "genetic algorithm path planning strategy" with a path overlap rate of ≤5%, had a picking volume per unit time that was 28% higher than wave WAVE20240520003, which adopted the same strategy but had a path overlap rate of 8%. Wave WAVE20240520003 had a picking error rate of 0.8%. Tracing back to the execution process, it was found that the review and sorting stage only used "SKU single code verification" and did not enable the "batch number + SKU dual code verification" rule. Furthermore, the load rate of picker No. 4 in warehouse area A02 for this wave was relatively low, indicating an uneven task allocation problem. Based on the post-mortem analysis, the system outputs targeted multi-stage strategy combination optimization suggestions. The suggestions precisely match the strategy configuration links at each stage and are directly implementable. For the path optimization stage, it is recommended to uniformly adjust the path overlap rate parameter of the "genetic algorithm path planning strategy" for subsequent waves in warehouse A02 to ≤5% to improve picking efficiency. For the review and sorting stage, it is recommended to force the use of the "batch number + SKU dual code verification" rule for daily necessities and cosmetics small item orders in warehouse A02 to reduce the picking error rate. For the zone collaboration stage, it is recommended to optimize the task allocation logic, splitting some orders in similar tasks after wave WAVE20240520003 to picker No. 4 to increase the staff load rate to a reasonable range of 75%-85%. For the wave division stage, it is recommended to fine-tune the weight of the "warehouse location clustering range" dimension to 65% to further improve the aggregation degree of orders in the same warehouse area and reduce picking path redundancy.After optimization suggestions are generated, the system associates and stores them with the corresponding wave's strategy configuration data and indicator analysis report to form a strategy optimization knowledge base. The suggested content is simultaneously pushed to the strategy configuration interface, allowing operations personnel to directly call and modify the strategy combination scheme for subsequent waves.

[0035] Reference Figure 2 Here is a structural block diagram of a warehouse picking optimization device with a multi-stage custom combination strategy according to an embodiment of the present invention, comprising: Order information processing module: Used to synchronously call the warehouse management system and the order management system through a preset standardized API interface to collect raw data of orders to be picked, extract core attributes in a targeted manner and complete data preprocessing to generate a structured order dataset that can be called by wave division strategy; Wave partitioning strategy configuration module: It provides a visual strategy configuration interface, supports users to select partitioning dimensions and set weights, generates wave partitioning settings based on user configuration, calls clustering algorithms to aggregate structured order datasets, and generates picking waves and wave task lists with unique identifiers after reasonableness verification. Multi-stage strategy combination configuration module: It is used to divide the entire picking process into four core stages, provide strategy selection and combination interfaces for each stage, support users to customize and select appropriate strategies, and generate an executable multi-stage custom strategy combination scheme after logical compatibility verification. Picking optimization execution module: It is used to load strategy combination schemes and associate them with wave task list, execute picking optimization operations in stages, collect on-site data in real time and monitor anomalies, trigger preset adjustment rules to adapt and correct strategies in real time, and synchronously record strategy execution data. The Results Statistics and Review Module is used to collect key indicators such as picking efficiency, accuracy, and resource utilization after each picking task is completed, generate a picking result report for each wave, conduct optimization reviews based on strategy execution data, and output optimization suggestions for strategy combination schemes.

[0036] In summary, In this embodiment, the specific implementation of each unit in the above device embodiment is described in the above method embodiment, and will not be repeated here.

[0037] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0038] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the present invention and embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0039] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0040] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A warehouse picking optimization method based on a multi-stage custom combination strategy, characterized in that, Includes the following steps: Obtain the set of warehouse orders awaiting picking, extract the core attribute information of each order, clean and deduplicate the core attribute information, standardize the format and fill in missing fields to generate a structured order dataset; Configure a custom wavelet partitioning strategy, allowing users to select at least one partitioning dimension and set corresponding weights based on actual warehouse business scenarios. Based on the configured custom wavelet partitioning strategy, perform order clustering on the structured order dataset to generate multiple picking waves. Each wavelet is associated with a unique wavelet identifier and a wavelet task list. The entire picking process is divided into four core stages, and each stage provides a strategy selection and combination interface, allowing users to customize and select appropriate optimization strategies for each stage, forming a multi-stage custom strategy combination scheme. Based on the wave task list and the multi-stage custom strategy combination scheme, the picking optimization operation of each stage is executed in sequence, and the optimization strategy of the current stage is adapted and corrected in real time based on the preset adjustment strategy. After completing the picking tasks for each wave, the indicators for each wave are statistically analyzed, and a wave picking result report is generated. Based on the result report and the strategy execution data for each stage, an optimization review is conducted to form a review analysis result, and optimization suggestions for the strategy combination scheme are output.

2. The warehouse picking optimization method based on a multi-stage custom combination strategy according to claim 1, characterized in that, The steps of obtaining the set of warehouse orders awaiting picking, extracting the core attribute information of each order, cleaning and deduplicating the core attribute information, standardizing the format, and filling in missing fields to generate a structured order dataset include: By using a pre-defined standardized API interface, the warehouse management system and the order management system are called simultaneously to collect the original data of the orders to be picked. After collection, an index is generated to form a set of orders to be picked. Based on a preset core attribute extraction library, data is extracted from the original order data in a targeted manner to generate an initial attribute dataset; The attribute dataset is uniformly converted according to a preset standard format to obtain preprocessed attribute data. Order priority is quantified into 1-5 levels, warehouse location distribution is converted into a three-level code, timeliness is converted into a standard time format, and address is split into a hierarchical format. Missing fields are identified through field integrity detection. The preprocessed attribute data is encapsulated according to a preset relational data table structure, and a unique field ID and data type description are assigned. After data consistency verification, a structured order dataset that can be directly called by subsequent wave division strategies is generated.

3. The warehouse picking optimization method based on a multi-stage custom combination strategy according to claim 1, characterized in that, The configured custom wavelet segmentation strategy allows users to select at least one segmentation dimension and set corresponding weights based on actual warehouse business scenarios. Based on this strategy, the structured order dataset is clustered to generate multiple picking waves, with each wave associated with a unique wavelet identifier and a wavelet task list. The steps include: It provides a visual wave division strategy configuration interface, allowing users to select at least one of the following as the division dimension: order priority, product correlation, warehouse location clustering range, timeliness requirements, and picking resource load, and set the weight ratio for the selected dimension. The system automatically generates standardized wavelet partitioning settings based on the dimensions and weights selected by the user. The partitioning settings include dimension matching thresholds, weight calculation formulas, and clustering termination conditions. The preset clustering algorithm is invoked to aggregate orders in the structured order dataset based on the wave division settings, forming a preliminary wave set; The initial set of picking waves is validated for rationality. Abnormal waves with task volume exceeding a preset threshold are removed and re-clustered. After the validation is passed, multiple valid picking waves are generated. A unique wave identifier is assigned to each picking wave, and a wave task list containing order list, product information, and warehouse location distribution is generated simultaneously.

4. The warehouse picking optimization method based on a multi-stage custom combination strategy according to claim 1, characterized in that, The picking process is divided into four core stages, each with a strategy selection and combination interface. This allows users to customize and select appropriate optimization strategies for each stage, forming a multi-stage custom strategy combination scheme. The steps include: The entire picking process is clearly divided into four core stages: wave planning, route optimization, zone collaboration, and review and sorting. The core task objectives and boundary scope of the four core stages are defined simultaneously. Configure a visual strategy selection and combination interface for the four core stages. The interface has a built-in candidate optimization strategy library for each stage, and the candidate strategies cover core scenarios such as task splitting, path algorithm, cross-region scheduling, and review and verification. Users can select appropriate optimization strategies from the candidate strategy library for each stage based on actual warehouse business scenarios, and can fine-tune the parameters of the selected strategies. The system integrates the selected strategies at each stage to generate standardized multi-stage custom strategy combination schemes.

5. The warehouse picking optimization method based on a multi-stage custom combination strategy according to claim 1, characterized in that, The step of sequentially executing picking optimization operations at each stage based on the wave task list and the multi-stage custom strategy combination scheme, and real-time adapting and correcting the optimization strategy for the current stage based on a preset adjustment strategy, includes: Load the verified multi-stage custom strategy combination scheme, associate and bind the multi-stage custom strategy combination scheme with the task list of each wave according to the wave identifier, and clarify the specific execution strategy of each wave in different stages. Following the sequence of wave planning, route optimization, zone collaboration, and review sorting, the picking optimization operations at each stage are executed sequentially, and the strategy execution data at each stage is recorded simultaneously. Real-time data collection of warehouse picking operations, including picker location, equipment operating status, real-time inventory, and order change information; The collected data is compared with preset thresholds to monitor whether any thresholds are exceeded or abnormal events occur.

6. The warehouse picking optimization method based on a multi-stage custom combination strategy according to claim 1, characterized in that, The step of comparing the collected data with a preset threshold to monitor whether any threshold exceeding or abnormal events occur includes: If an anomaly is detected, a preset adjustment strategy is triggered. The optimization strategy for the current stage is adapted and corrected in real time based on the actual data on site. After correction, the strategy undergoes a second logical compatibility check. After the verification is successful, the current picking optimization operation will continue to be executed using the revised strategy, and the strategy execution record will be updated synchronously.

7. The warehouse picking optimization method based on a multi-stage custom combination strategy according to claim 1, characterized in that, After completing the picking tasks for each wave, the following steps are taken: first, statistical analysis of the indicators for each wave is performed, a wave picking result report is generated, and based on the result report and the strategy execution data for each stage, optimization reviews are conducted, and optimization suggestions for the strategy combination scheme are output. After completing the picking tasks for each wave of the entire stage, the core indicators for each wave are statistically analyzed. These core indicators include picking efficiency, picking accuracy, and picking resource utilization. By integrating the metrics of each wave with the wave task list, a standardized wave picking result report is generated, which includes detailed data for each wave and summary data for the entire batch. Retrieve multi-stage strategy execution data for each wave, combine it with the wave picking result report, analyze the correlation between the strategy combination scheme and picking indicators, and identify the advantages and disadvantages in the strategy execution process; Based on the results of the post-mortem analysis, optimization suggestions are output, including adjustments to the weights of strategy dimensions, replacement of candidate strategies, and directions for parameter optimization.

8. A warehouse picking optimization device with a multi-stage customizable combination strategy, characterized in that, include: Order information processing module: Used to synchronously call the warehouse management system and the order management system through a preset standardized API interface to collect raw data of orders to be picked, extract core attributes in a targeted manner and complete data preprocessing to generate a structured order dataset that can be called by wave division strategy; Wave partitioning strategy configuration module: It provides a visual strategy configuration interface, supports users to select partitioning dimensions and set weights, generates wave partitioning settings based on user configuration, calls clustering algorithms to aggregate structured order datasets, and generates picking waves and wave task lists with unique identifiers after reasonableness verification. Multi-stage strategy combination configuration module: It is used to divide the entire picking process into four core stages, provide strategy selection and combination interfaces for each stage, support users to customize and select appropriate strategies, and generate an executable multi-stage custom strategy combination scheme after logical compatibility verification. Picking optimization execution module: It is used to load strategy combination schemes and associate them with wave task list, execute picking optimization operations in stages, collect on-site data in real time and monitor anomalies, trigger preset adjustment rules to adapt and correct strategies in real time, and synchronously record strategy execution data. The Results Statistics and Review Module is used to collect key indicators such as picking efficiency, accuracy, and resource utilization after each picking task is completed, generate a picking result report for each wave, conduct optimization reviews based on strategy execution data, and output optimization suggestions for strategy combination schemes.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the warehouse picking optimization method of any one of claims 1 to 7, which includes a stage-customized combination strategy.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the warehouse picking optimization method of any one of claims 1 to 7, which includes a stage-customized combination strategy.

Images