A warehouse position management system of a display and storage integrated intelligent shoe wall

Abstract

The application provides a warehouse location management system for displaying an integrated intelligent shoe wall, and relates to the technical field of data processing. The system comprises: an information acquisition module, which is used for collecting the occupancy state of each location of the shoe wall, the sales frequency data of each type and size of shoes, and the path consumption when the mobile target takes shoes from the shoe wall based on a preset acquisition cycle, and generating a shoe wall state data set and a path consumption data set; a path analysis module, which is used for quantifying the path consumption of each location of the shoe wall, calculating the path consumption value, comparing it with a preset path consumption threshold group, and generating location accessibility classification data; a classification and allocation module, which is used for classifying shoes into different categories, and allocating locations to shoes of different categories in combination with the location accessibility classification data; an execution and feedback module, which is used for controlling the mobile target to complete the actual location adjustment operation of the shoes. The application improves the accuracy of real-time automatic allocation of location management.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a storage management system for an integrated intelligent shoe wall that combines display and warehousing. Background Technology

[0002] Existing integrated smart shoe walls, combining display and warehousing, typically employ a modular design, integrating shoe display and storage functions into a single structure. Users view the displayed shoes through transparent windows or electronic screens at the front end, while the back end uses intelligent hardware to store and manage the shoes. The system, in conjunction with electronic tags, automatic identification devices, and an inventory management module, enables dynamic management of the shoes. In existing technologies, after shoes are received, different models and sizes are allocated to different storage locations within the shoe wall according to preset system rules, supporting subsequent automatic replenishment and outbound operations.

[0003] In real-world applications, such as a large chain footwear retail store, frequent adjustments to the allocation of storage locations within the shoe wall are often necessary to improve product display diversity and inventory turnover. However, existing integrated shoe wall storage location allocation methods primarily rely on the system to simply partition or sequentially place shoes based on model or size, lacking the ability to dynamically adjust storage locations for high-frequency and low-frequency shoe items. This can result in high-selling shoes being assigned to difficult-to-access high-end or marginal storage locations, impacting replenishment and retrieval efficiency, and causing a decrease in the utilization rate of some storage locations. This single allocation strategy is insufficient to meet the demands of refined management and efficient operation in complex product structures and dynamic sales environments. Summary of the Invention

[0004] The purpose of this invention is to provide a storage management system for an integrated intelligent shoe wall that combines display and warehousing, aiming to solve the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] A storage management system for an integrated smart shoe wall that combines display and warehousing, the system comprising:

[0007] The information collection module is used to collect the occupancy status of each storage location in the shoe wall, the sales frequency data of each model and size of shoes, and the path consumption when a moving target takes shoes from the shoe wall, based on a preset collection cycle, and generate a shoe wall status dataset and a path consumption dataset.

[0008] The path analysis module is used to quantify the path consumption to each storage location on the shoe wall based on the path consumption dataset, calculate the path consumption value, and compare it with the preset path consumption threshold group to generate storage location accessibility classification data.

[0009] The hierarchical allocation module is used to classify shoes into different categories based on the sales frequency of shoes in the shoe wall status dataset, and to allocate storage locations to different categories of shoes in combination with the storage location accessibility hierarchical data, thereby generating a storage location allocation scheme.

[0010] The execution and feedback module is used to control the moving target to complete the actual storage location adjustment operation of the footwear according to the storage location allocation plan, and after the adjustment is completed, the adjustment result is sent back to the information collection module, supporting the system's self-learning and allocation strategy optimization.

[0011] Preferably, the path analysis module includes:

[0012] The path consumption modeling unit is used to calculate the path consumption value of each feasible path based on the path consumption dataset, taking into account the movement behavior of the moving target between various storage locations on the shoe wall, and combining the horizontal distance, vertical distance and number of turns, and to generate path consumption value distribution data.

[0013] The threshold comparison and self-learning unit is used to compare the path consumption value distribution data with the preset path consumption threshold group to generate storage location accessibility classification data.

[0014] Preferably, the hierarchical allocation module includes:

[0015] The multidimensional dynamic grading unit is used to update the popularity score of footwear in the future preset period in real time based on the shoe wall status dataset, combined with replenishment frequency and historical unsold risk factors. It dynamically divides footwear categories into high popularity, potential popularity and low popularity categories, and generates predicted footwear category labels.

[0016] The multi-objective self-optimizing matching unit is used to construct a multi-objective dynamic programming model based on the predicted footwear category labels and warehouse location accessibility classification data. It jointly models the priority access convenience, overall inventory balance, and local flow distribution of the shoe wall, continuously optimizes the allocation weights, and generates a warehouse location allocation scheme.

[0017] Preferably, based on the path consumption dataset, the path consumption value of each feasible path is calculated by comprehensively considering the horizontal distance, vertical distance, and number of turns, taking into account the movement behavior of the moving target between various storage locations on the shoe wall. This generates path consumption value distribution data for each storage location on the shoe wall, including:

[0018] Based on the path consumption dataset, the real-time path coordinate sequence generated during the actual movement of the moving target between various storage locations inside the shoe wall is extracted to form the original path dataset.

[0019] Based on the original path dataset, the horizontal distance, vertical distance, and turning point information of each path segment are extracted to form a multi-dimensional path feature set.

[0020] Based on the multidimensional path feature set, adjustable weights are assigned to each path parameter, and the comprehensive path consumption value of each path is quantified by weighted fusion to generate path consumption value distribution data.

[0021] Preferably, based on the shoe wall status dataset, combined with replenishment frequency and historical slow-moving risk factors, the popularity score of footwear products within a preset future period is updated in real time. Footwear categories are dynamically divided into high-popularity, potential-popularity, and low-popularity categories, generating predicted footwear category labels, including:

[0022] Based on the shoe wall status dataset, the current sales frequency of shoes of various models and sizes is analyzed, and the replenishment frequency and slow-moving risk factors are analyzed based on historical sales data to form a multi-dimensional popularity feature parameter set.

[0023] A popularity prediction model based on time series analysis is used to process a multi-dimensional popularity feature parameter set, predict the popularity score and trend of footwear in the future preset period, and generate popularity score prediction results.

[0024] Based on the popularity score prediction results, footwear is divided into high popularity, potential popularity and low popularity categories, and corresponding predicted footwear category labels are generated.

[0025] Preferably, based on predicted footwear category labels and warehouse location accessibility grading data, a multi-objective dynamic programming model is constructed. This model jointly models priority retrieval convenience, overall inventory balance, and local flow distribution within the shoe wall, continuously optimizing the allocation weights to generate a warehouse location allocation scheme, including:

[0026] Based on the predicted footwear category labels, warehouse location accessibility classification data, current warehouse location distribution, historical replenishment frequency, and traffic statistics of each area of ​​the shoe wall, a set of allocation optimization parameters is formed;

[0027] Based on the allocation optimization parameter set, a multi-objective optimization function combining priority access convenience, inventory balance and flow distribution is established, and dynamic balance among the objectives is achieved by setting weights;

[0028] An iterative optimization algorithm is used to dynamically adjust the weights of the multi-objective optimization function, and the optimal storage location allocation scheme is generated based on continuous simulation and feedback.

[0029] Preferably, the replenishment frequency and slow-moving risk factors are analyzed based on historical sales data, including:

[0030] Based on the shoe wall status dataset, historical sales and replenishment records of shoes of various models and sizes are extracted to form a sales frequency sequence and a replenishment frequency sequence with fixed periodic statistics.

[0031] For each shoe item, the actual number of replenishments and sales volume within a set period are statistically analyzed. The replenishment activity parameter and sales trend parameter of the shoe item within the period are obtained by weighted moving average calculation and compared with the preset threshold. If the replenishment activity parameter is less than the replenishment activity threshold and the sales trend parameter is less than the sales trend threshold, it is judged as a suspected slow-moving risk target.

[0032] For footwear products identified as potentially slow-moving risk targets, their historical shelf duration, inventory turnover cycle, and promotion frequency parameters are further retrieved and compared with the set shelf duration threshold, inventory turnover cycle threshold, and promotion frequency threshold respectively. If each parameter meets the slow-moving risk condition, it is counted as one point. Finally, the scores of all parameters that meet the conditions are accumulated as the slow-moving risk factor for the footwear product. The slow-moving risk factor ranges from zero to an integer of the number of parameters.

[0033] For footwear products that are not identified as having a risk of slow sales, the system can automatically set their slow sales risk factor to zero or a standard value.

[0034] Preferably, the method for constructing the heat prediction model includes:

[0035] Using the sales frequency, historical replenishment frequency, and slow-moving risk factors of footwear as inputs, a set of popularity feature vectors is constructed;

[0036] The heat feature vector set is input into the time series prediction model, which is trained using a long short-term memory neural network to model the heat change trend of footwear in a continuous period.

[0037] During the model training and prediction phases, cross-validation is used to evaluate the prediction performance of the time series prediction model, and error backpropagation and parameter optimization are implemented based on the prediction error to iteratively improve the model accuracy.

[0038] The final output is a prediction of the footwear's popularity rating over a predetermined period.

[0039] Preferably, the method for constructing the multi-objective optimization function includes:

[0040] Based on the allocation optimization parameter set, parameters for priority access convenience, overall inventory balance, and local flow distribution of the shoe wall are extracted to form a multi-objective optimization input vector;

[0041] Based on the multi-objective optimization input vector, initial weights are set for each objective and an objective evaluation function is established. The normalization method is used to unify the dimensions of different objectives, and a multi-objective optimization function is constructed through weighted synthesis.

[0042] Preferably, an iterative optimization algorithm is used to dynamically adjust the weights of the multi-objective optimization function, including:

[0043] Based on the constructed multi-objective optimization function and its initial weight parameters, combined with the current footwear category distribution and warehouse location allocation status, the first round of warehouse location allocation scheme is generated, and the evaluation function values ​​of each objective are calculated.

[0044] Using a genetic algorithm, the target weight parameters are perturbed to generate multiple sets of weight combinations and corresponding allocation candidate schemes in batches, and the achievement degree of each target evaluation function is evaluated.

[0045] For each allocation candidate scheme, the optimal weight combination is selected based on the achievement of the objective evaluation function, historical operation feedback and system optimization criteria. The selection result is then used as input to feed the genetic algorithm for a new round of iteration.

[0046] During multiple rounds of iterative optimization, the weight combination and candidate allocation schemes are continuously adjusted until the multi-objective optimization function converges or meets the set optimization termination conditions, and finally outputs the dynamically optimal target weight configuration and shoe wall storage location allocation scheme.

[0047] The above-described solution of the present invention has at least the following beneficial effects:

[0048] First, by periodically collecting data on the real-time occupancy status of each storage location in the shoe wall, the sales frequency of shoes, and the consumption of retrieval routes through the information collection module, the inventory structure of the shoe wall and actual user retrieval behavior can be dynamically monitored. Compared to existing technologies that simply divide storage locations based on model and size, this invention can generate shoe wall status datasets and route consumption datasets, providing solid data support for subsequent intelligent analysis and optimization decisions.

[0049] Secondly, the system employs a path analysis module to quantify retrieval paths from multiple dimensions. Combined with preset path consumption thresholds, it automatically categorizes the accessibility of each storage location, achieving precise characterization of replenishment and retrieval convenience. This effectively avoids the phenomenon of high-frequency footwear items being assigned to difficult-to-access storage locations, significantly improving response speed and customer experience during peak replenishment and sales periods.

[0050] Furthermore, the tiered allocation module dynamically categorizes footwear based on sales frequency and combines this with warehouse accessibility data to flexibly and intelligently prioritize high-volume footwear for easily accessible locations, while allocating low-frequency footwear to peripheral or relatively less important locations. This mechanism significantly improves the utilization rate of shoe wall space and reduces operational inefficiencies caused by allocating high-selling footwear to inaccessible locations.

[0051] Furthermore, the execution and feedback module can automatically or guide the adjustment of shoe storage locations based on the optimized allocation scheme, and provide real-time feedback on each adjustment result. Through continuous closed-loop learning, the system achieves self-optimization of the allocation strategy, automatically correcting shoe storage location allocation based on dynamic changes in sales and replenishment, ensuring the shoe wall always maintains optimal operational status. Overall, this invention effectively overcomes the static, rigid, and unadaptive problems of existing integrated shoe wall systems in terms of storage location allocation, achieving refined, efficient, and intelligent integrated management of shoe display and warehousing, effectively improving the inventory turnover rate and service responsiveness of retail terminals. Attached Figure Description

[0052] Figure 1 This is a flowchart of the storage management system for an integrated smart shoe wall that combines display and warehousing, provided by an embodiment of the present invention. Detailed Implementation

[0053] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0054] like Figure 1 As shown, an embodiment of the present invention proposes a storage management system for an integrated smart shoe wall that combines display and warehousing. The system includes:

[0055] The information collection module is used to collect the occupancy status of each storage location in the shoe wall, the sales frequency data of each model and size of shoes, and the path consumption when a moving target takes shoes from the shoe wall, based on a preset collection cycle, and generate a shoe wall status dataset and a path consumption dataset.

[0056] The path analysis module is used to quantify the path consumption to each storage location on the shoe wall based on the path consumption dataset, calculate the path consumption value, and compare it with the preset path consumption threshold group to generate storage location accessibility classification data.

[0057] The hierarchical allocation module is used to classify shoes into different categories based on the sales frequency of shoes in the shoe wall status dataset, and to allocate storage locations to different categories of shoes in combination with the storage location accessibility hierarchical data, thereby generating a storage location allocation scheme.

[0058] The execution and feedback module is used to control the moving target to complete the actual storage location adjustment operation of the footwear according to the storage location allocation plan, and after the adjustment is completed, the adjustment result is sent back to the information collection module, supporting the system's self-learning and allocation strategy optimization.

[0059] In this embodiment of the invention, real-time and accurate management of the status of all storage locations and footwear within the shoe wall is achieved. The system periodically acquires the occupancy status of each storage location within the shoe wall through an information collection module, enabling timely identification of which locations are idle, occupied, or awaiting adjustment, thus preventing empty or misplaced storage locations. For each model and size of footwear, the system automatically collects sales frequency data, dynamically reflecting changes in market demand and allowing for timely adjustments to the display and storage layout of the footwear. By collecting path consumption data of moving targets (such as staff or intelligent handling equipment) retrieving footwear from the shoe wall, not only can the convenience of replenishment and retrieval be objectively reflected, but a data foundation can also be provided for subsequent storage location optimization. The collected data, after being processed, forms a shoe wall status dataset and a path consumption dataset, laying a solid data foundation for subsequent intelligent analysis and optimization of the system.

[0060] The system's path analysis module can accurately quantify the movement paths during each shoe retrieval or replenishment process. It comprehensively analyzes factors such as horizontal and vertical distances and the number of turns in actual operation to generate a consumption value for each feasible path. This value is then compared with a preset path consumption threshold set in the system, thereby automatically classifying the accessibility of each shoe storage location. Through this mechanism, the system can automatically identify which storage locations are easy to access and which are more difficult, providing a scientific basis for shoe allocation and scheduling.

[0061] The system's hierarchical allocation module, based on the analysis results of the shoe wall status dataset, categorizes shoes according to sales frequency, such as high-frequency, potential, and low-frequency categories. Combined with warehouse location accessibility data, shoes with high demand or those requiring priority display are allocated to easily accessible warehouse locations, while shoes with lower sales or less frequent replenishment are assigned to less desirable or less readily accessible locations. This maximizes the optimization of replenishment and sales processes, improves front-end display, and reduces resource waste caused by moving shoes within the warehouse. Each allocation and warehouse location adjustment operation is automated by the execution and feedback module, with the adjustment results fed back to the information collection module in real time, achieving a closed-loop data flow. During continuous operation, the system can learn and adjust its allocation strategy based on feedback data, gradually achieving intelligent self-optimization of the shoe wall warehouse locations, continuously improving management efficiency and responsiveness. For example, by continuously collecting and analyzing various indicators within the operational cycle, the system can continuously fine-tune and dynamically optimize the shoe storage layout, greatly improving the overall operational efficiency and user experience of the smart shoe wall in practical application scenarios.

[0062] The information collection module specifically includes:

[0063] As a core component of the shoe wall management system, the information acquisition module typically integrates various sensing and data collection methods. To monitor the occupancy status of each storage location within the shoe wall, the system can use RFID tags, infrared sensors, pressure sensors, or visual recognition to detect in real time whether shoes are present in each location, as well as the model and size of those shoes. The system automatically iterates through all storage locations at set time intervals, generating occupancy data that includes the location number, current status, and corresponding shoe information, thus forming a shoe wall status dataset.

[0064] For footwear sales frequency data, the system periodically aggregates the actual sales volume of each model and size of footwear within a specific period through channels such as bank-sale systems, electronic shelf labels, and cloud sales records. Sales frequency statistics are not limited to single days or weeks; they can also be accumulated and trend analyzed over different periods based on business needs, enabling the system to dynamically reflect changes in market demand. All sales data is grouped and statistically analyzed in the background using footwear codes as indexes and synchronized to the shoe wall status dataset, providing an accurate basis for subsequent popularity prediction and tiered allocation.

[0065] For collecting path consumption data, the system combines existing positioning and trajectory tracking technologies, utilizing the location coordinates of moving targets (such as intelligent handling robots or manual picking equipment) along their actual movement paths within the shoe wall to record the complete motion trajectory of each shoe retrieval operation in real time. The system analyzes the horizontal and vertical distance changes along this trajectory and automatically counts the number of turning points on the path. Various parameters can be collected through integrated inertial measurement units, visual navigation, and LiDAR. Finally, based on factors such as distance and turning for each operation, the system calculates the actual path consumption and generates a path consumption dataset after aggregation. This data not only provides input for accessibility grading but also reflects the actual energy consumption and operational efficiency in real time, facilitating subsequent path optimization and warehouse location scheduling.

[0066] The execution and feedback module specifically includes:

[0067] The execution and feedback module is primarily responsible for the actual implementation of system output decisions and the data feedback of the operational closed loop. Based on the storage location allocation plan output by the hierarchical allocation module, the system automatically parses each shoe storage location adjustment instruction, clarifying the existing and target storage locations for the target shoes. For automated handling equipment, the system sends control signals to guide it along a preset path to the designated storage location, completing the grabbing, handling, and repositioning of the shoes. During operation, the execution status of the moving target is monitored in real time, including the start and end times of movement, path completion rate, and operational accuracy. If manual operation is involved, the system can issue the storage location adjustment plan to on-site operators through electronic guidance, light prompts, or screen navigation, and collect execution confirmation information.

[0068] After each warehouse location adjustment operation is completed, the system will automatically collect and verify the status of the target warehouse location and related footwear again to ensure consistency between the actual operation and the allocation plan. All adjustment results, including the transfer records of footwear from the source warehouse location to the target warehouse location, the actual consumption of the movement path, the operation time, and the success status, will be transmitted back to the information collection module in the form of structured data and synchronized to update the shoe wall status dataset and the path consumption dataset.

[0069] Through the closed-loop processing described above, the system can track the effects of each operation and accumulate data. In subsequent self-learning and allocation strategy optimization, the system analyzes the deviation between historical adjustment records and current operational results, automatically adjusting allocation scheme parameters or path determination logic. For example, if a certain type of footwear is frequently moved to remote storage locations during peak periods, making retrieval inconvenient, the system will optimize future allocation logic accordingly, prioritizing high-demand footwear in easily accessible main channels or low-consumption storage locations. Through continuous data feedback and automatic optimization, the system can dynamically adapt to changes in operational needs, achieving intelligent, refined, and efficient footwear management.

[0070] In a preferred embodiment of the present invention, the path analysis module includes:

[0071] The path consumption modeling unit is used to calculate the path consumption value of each feasible path based on the path consumption dataset, taking into account the movement behavior of the moving target between various storage locations on the shoe wall, and combining the horizontal distance, vertical distance and number of turns, and to generate path consumption value distribution data.

[0072] The threshold comparison and self-learning unit is used to compare the path consumption value distribution data with the preset path consumption threshold group to generate storage location accessibility classification data.

[0073] In this embodiment of the invention, the path analysis module consists of a path consumption modeling unit and a threshold comparison and self-learning unit, enabling precise modeling and intelligent analysis of shoe wall retrieval and replenishment paths. The path consumption modeling unit, by accessing the path consumption dataset, performs in-depth analysis of the trajectory data generated during the movement of a moving target between different storage locations within the shoe wall. This process includes real-time extraction of the horizontal distance, vertical distance, and number of turns for each actual path, forming a multi-dimensional path feature set. The system can assign adjustable weights to each parameter based on different shoe wall structures and operational needs. Based on this, a weighted fusion method is used to accurately quantify the comprehensive consumption value of each path, and all paths are normalized and distributed, ultimately generating path consumption value distribution data covering all storage locations within the shoe wall.

[0074] After obtaining the path consumption value distribution data, the threshold comparison and self-learning unit automatically compares it with a pre-set set of path consumption thresholds. The system determines the path consumption level of each storage location and outputs preliminary storage location accessibility classification data. In subsequent operation, the system can adaptively and dynamically adjust the path consumption threshold set based on path anomalies or traffic peaks that occur during actual operation. For example, during holidays or promotional peaks, the path consumption in certain areas may increase significantly due to dense pedestrian and cargo traffic. The system will automatically detect these anomalies and adjust the consumption thresholds of relevant storage locations to improve the flexibility and accuracy of the classification thresholds, ensuring that the accessibility classification results always conform to the actual operation of the shoe wall. This not only improves the accuracy of path classification but also greatly enhances the system's adaptability to dynamic operating environments, ensuring that the shoe wall management solution can respond to operational fluctuations in real time.

[0075] In a preferred embodiment of the present invention, the hierarchical allocation module includes:

[0076] The multidimensional dynamic grading unit is used to update the popularity score of footwear in the future preset period in real time based on the shoe wall status dataset, combined with replenishment frequency and historical unsold risk factors. It dynamically divides footwear categories into high popularity, potential popularity and low popularity categories, and generates predicted footwear category labels.

[0077] The multi-objective self-optimizing matching unit is used to construct a multi-objective dynamic programming model based on the predicted footwear category labels and warehouse location accessibility classification data. It jointly models the priority access convenience, overall inventory balance, and local flow distribution of the shoe wall, continuously optimizes the allocation weights, and generates a warehouse location allocation scheme.

[0078] In this embodiment of the invention, the hierarchical allocation module achieves intelligent dynamic allocation of footwear and storage locations through a multi-dimensional dynamic hierarchical unit and a multi-objective self-optimizing matching unit. The multi-dimensional dynamic hierarchical unit can continuously and dynamically update the popularity score of footwear within a preset future period based on the shoe wall status dataset, combined with the historical replenishment frequency and slow-moving risk factors for each type of footwear. For example, the system statistically analyzes the sales frequency of each model and size of footwear in real time, combines historical sales data to generate replenishment frequency and slow-moving risk factors, and thus forms a multi-dimensional popularity feature parameter set. By introducing a popularity prediction model based on time-series analysis, these parameters are comprehensively analyzed and trend predicted, accurately predicting the popularity changes of each footwear in the future period. The system then dynamically classifies footwear into high-popularity, potential-popularity, and low-popularity categories based on the popularity score prediction results, and outputs corresponding category labels, providing an accurate data foundation for subsequent allocation decisions.

[0079] The multi-objective self-optimizing matching unit constructs a multi-objective dynamic programming model based on footwear category tags and warehouse location accessibility grading data. This model includes objectives such as priority retrieval convenience, overall inventory balance, and local traffic distribution within the shoe wall. The system combines the actual weights of each objective, establishes a multi-objective optimization function by allocating an optimization parameter set, and employs iterative optimization algorithms (such as genetic algorithms and particle swarm optimization) to continuously adjust the objective weights and allocation schemes, ensuring optimal matching results between footwear and warehouse locations. For example, the system can prioritize allocating high-demand footwear to warehouse locations with convenient retrieval and high foot traffic, while allocating low-demand footwear to warehouse locations with higher storage space utilization but less frequent retrieval. Each allocation result can be dynamically adjusted based on historical feedback and real-time operational performance. Through continuous iteration, the system improves overall allocation efficiency and responsiveness, achieving intelligent self-optimization of shoe wall warehouse location management. This grading and matching mechanism effectively improves the overall efficiency of replenishment, retrieval, and display, reduces operating costs, and maximizes the fulfillment of the actual needs for intelligent and refined management in the footwear retail scenario.

[0080] In a preferred embodiment of the present invention, based on the path consumption dataset, the path consumption value of each feasible path is calculated by comprehensively considering the horizontal distance, vertical distance, and number of turns, based on the movement behavior of the moving target between various storage locations on the shoe wall, thereby generating path consumption value distribution data for each storage location on the shoe wall, including:

[0081] Based on the path consumption dataset, the real-time path coordinate sequence generated during the actual movement of the moving target between various storage locations inside the shoe wall is extracted to form the original path dataset.

[0082] Based on the original path dataset, the horizontal distance, vertical distance, and turning point information of each path segment are extracted to form a multi-dimensional path feature set.

[0083] Based on the multidimensional path feature set, adjustable weights are assigned to each path parameter, and the comprehensive path consumption value of each path is quantified by weighted fusion to generate path consumption value distribution data.

[0084] In this embodiment of the invention, the system, through a path consumption modeling unit, can accurately model and perform multi-dimensional analysis of all feasible paths within the shoe wall. During shoe retrieval or replenishment, the system collects the walking trajectory of the moving target in real time, automatically recording and reconstructing the actual coordinate changes of each movement path. This raw trajectory data lays the foundation for subsequent path feature extraction and consumption analysis. During the analysis process, the system breaks down each path into multiple detailed segments, automatically identifying the horizontal distance, vertical distance, and turning points of each segment, thereby forming a complete multi-dimensional path feature set.

[0085] For different shoe wall structures and actual operating environments, the system allows for independent and adjustable weights to be assigned to various parameters such as horizontal distance, vertical distance, and number of turns. For example, some compact shoe walls can be given a higher weight for the number of turns, thereby prioritizing the reduction of errors and energy consumption caused by frequent operations. After all parameters are weighted, the system uses a weighted fusion algorithm to perform comprehensive consumption quantification on each path, ensuring that the generated path consumption value can fully reflect the complexity and cost of actual operations. Finally, the comprehensive consumption value of all paths is further normalized to form path consumption value distribution data covering all storage locations in the shoe wall. In this way, the system can not only accurately assess the actual retrieval consumption of each storage location, but also provide a solid quantitative basis for subsequent storage location classification and optimized allocation. This solution greatly improves the scientific utilization of shoe wall space and the intelligent level of replenishment path management, effectively reduces ineffective operations by manpower and machinery, and improves overall operational efficiency.

[0086] Specifically, based on the multi-dimensional path feature set, adjustable weights are assigned to each path parameter, and a weighted fusion method is used to quantify the comprehensive path consumption value of each path, generating path consumption value distribution data, which includes:

[0087] In actual operation, the system first acquires all historical path trajectories of the moving target between various storage locations within the shoe wall using sensors, positioning units, or background trajectory analysis tools, and converts each movement operation into a detailed sequence of path coordinates. The system further performs segmented analysis of the path trajectories, extracting the horizontal and vertical distances of each segment, as well as the number of turns at path bends, thereby generating a complete multi-dimensional path feature set for each path. For different application scenarios or shoe wall structures, the system allows users to flexibly set the weight coefficients of various features according to operational priorities. For example, when operational efficiency is particularly sensitive to turning, the system can increase the weight of the turning frequency parameter; if space is relatively compact, the weight of horizontal or vertical distance can be increased.

[0088] During the weighted fusion processing stage, the system automatically iterates through all feature parameters of each path. Each parameter (such as horizontal distance, vertical distance, and number of turns) is multiplied by its corresponding weight coefficient, and then all weighted results are summed. The entire process is completed automatically by the system without human intervention.

[0089] Adjustable weights refer to the weighting factors set for each type of path parameter during path consumption value modeling based on a multi-dimensional path feature set. These factors can be dynamically adjusted based on system operating status, historical utilization efficiency feedback, or differences in the shoe wall structure, thereby improving the accuracy and adaptability of path consumption value modeling. Specifically, this includes:

[0090] After acquiring the raw path data, the system decomposes each path into several segments and extracts basic features related to the path, including but not limited to the horizontal travel distance, the vertical ascent and descent distance, and the number of turns in each path. Correspondingly, path parameters are defined as horizontal distance, vertical distance, and the number of turns. For these three parameters, the system introduces three weighting factors to adjust the contribution of each path feature to the overall path cost calculation. Unlike static weight allocation, the weights in this embodiment are adjustable; that is, their values ​​can be configured not only during system initialization but also dynamically updated based on performance feedback during system operation. For example, if the system finds that the time cost of vertical pickup operations is significantly higher than that of horizontal operations during actual operation, the system will automatically increase the weighting of its vertical pickup factor and correspondingly decrease the weights of the other two factors.

[0091] In a preferred embodiment of the present invention, based on the shoe wall status dataset and combined with replenishment frequency and historical slow-moving risk factors, the popularity score of footwear products within a preset future period is updated in real time. Footwear categories are dynamically divided into high-popularity, potential-popularity, and low-popularity categories, generating predicted footwear category labels, including:

[0092] Based on the shoe wall status dataset, the current sales frequency of shoes of various models and sizes is analyzed, and the replenishment frequency and slow-moving risk factors are analyzed based on historical sales data to form a multi-dimensional popularity feature parameter set.

[0093] A popularity prediction model based on time series analysis is used to process a multi-dimensional popularity feature parameter set, predict the popularity score and trend of footwear in the future preset period, and generate popularity score prediction results.

[0094] Based on the popularity score prediction results, footwear is divided into high popularity, potential popularity and low popularity categories, and corresponding predicted footwear category labels are generated.

[0095] In this embodiment of the invention, the multi-dimensional dynamic grading unit, through in-depth mining and analysis of the shoe wall state dataset, can achieve intelligent prediction and dynamic grading of changes in shoe popularity. The system first collects historical sales and replenishment data for all shoe products, statistically analyzes the periodic sales and replenishment of each model and size, and establishes a complete replenishment frequency sequence and sales frequency sequence. By statistically analyzing the actual number of replenishments and sales fluctuations, the system uses a weighted moving average method to calculate the replenishment activity and sales trend of each shoe product, thereby identifying which shoes are high-frequency replenished and which are low-frequency or slow-moving.

[0096] For footwear products with consistently low replenishment frequency and sluggish sales, the system combines characteristics of past slow-moving footwear, including shelf duration, inventory turnover cycle, and promotional frequency, to conduct a risk scoring model to assess the risk of slow sales, assigning a quantitative slow-moving risk factor to each product. These factors, along with the replenishment frequency parameter, form a set of popularity feature parameters, which are fed into the popularity prediction model. The system employs time-series analysis to model and train multi-dimensional feature parameters, accurately predicting the popularity score and trends of footwear products within a preset future period. Based on the prediction results, the system dynamically categorizes footwear into high-popularity, potential-popularity, and low-popularity categories, ensuring that each footwear classification reflects market changes and inventory risks in real time. This solution effectively prevents imbalances in footwear inventory structure, enables early response to potential best-selling and slow-moving products, and improves the initiative and risk control capabilities of footwear management.

[0097] In a preferred embodiment of the present invention, a multi-objective dynamic programming model is constructed based on predicted footwear category labels and warehouse location accessibility grading data. This model jointly models priority retrieval convenience, overall inventory balance, and local flow distribution within the shoe wall, continuously optimizing the allocation weights to generate a warehouse location allocation scheme, including:

[0098] Based on the predicted footwear category labels, warehouse location accessibility classification data, current warehouse location distribution, historical replenishment frequency, and traffic statistics of each area of ​​the shoe wall, a set of allocation optimization parameters is formed;

[0099] Based on the allocation optimization parameter set, a multi-objective optimization function combining priority access convenience, inventory balance and flow distribution is established, and dynamic balance among the objectives is achieved by setting weights;

[0100] An iterative optimization algorithm is used to dynamically adjust the weights of the multi-objective optimization function, and the optimal storage location allocation scheme is generated based on continuous simulation and feedback.

[0101] In this embodiment of the invention, the multi-objective self-optimizing matching unit realizes multi-objective optimization matching of shoe category labels and warehouse location accessibility classification data. The system first comprehensively forms an allocation optimization parameter set based on predicted shoe category labels, warehouse location accessibility classification data, current warehouse location distribution, historical replenishment frequency, and traffic statistics for each area of ​​the shoe wall. For each optimization objective, such as priority access convenience, overall inventory balance, and local traffic distribution of the shoe wall, the system establishes an independent evaluation function and achieves dynamic balance between objectives through normalization and weighting mechanisms. The multi-objective optimization function can flexibly adjust the weights of each objective to meet diverse management needs during different operational stages or promotional activities.

[0102] To further improve allocation efficiency, the system introduces iterative optimization algorithms, such as genetic algorithms and particle swarm optimization. Through repeated simulations and target weight perturbations, multiple candidate weight combinations and allocation schemes are generated and evaluated in batches. In each round of optimization, the system selects the optimal weight combination based on the achievement degree of the target evaluation function and historical feedback information, and feeds the selection results back to the optimization algorithm for continuous weight adjustment and allocation optimization. As the number of optimization rounds increases, the system can achieve convergence of the target evaluation function or reach the optimization termination condition, ultimately outputting the globally optimal shoe storage location allocation result under multi-objective dimensions. This mechanism not only improves the scientific and rational nature of the allocation but also ensures the flexibility and adaptability of shoe wall operations, significantly improving dynamic inventory management and replenishment smoothness in actual retail scenarios.

[0103] Specifically, based on predicted footwear category labels, warehouse location accessibility grading data, current warehouse location distribution, historical replenishment frequency, and traffic statistics for each area of ​​the shoe wall, a set of allocation optimization parameters is formed, including:

[0104] Before optimizing the allocation of shoe storage locations, the system first obtains the predicted category tags for all shoe items in the current sales cycle through a multi-dimensional dynamic grading unit, such as high popularity, potential popularity, and low popularity. Simultaneously, the path analysis module outputs the accessibility grading results for each shoe location, assigning an actual operational difficulty level to each location. The system also automatically compiles the existing shoe distribution information for each location, including the model, quantity, and popularity level of the shoes currently stored in each location. This data reflects the current spatial layout and inventory utilization status.

[0105] The system further extracts the replenishment frequency of each shoe type within a historical period, reflecting market demand and replenishment pressure, and providing decision-making references for prioritizing convenient storage locations for frequently replenished shoes. Simultaneously, the system uses embedded sensors, sales tracking, or RFID technologies to statistically analyze the access frequency or traffic distribution of different areas of the shoe wall within a set period, generating local traffic statistics for the shoe wall. This information accurately reflects customer activity and the distribution of popular storage locations in different areas, providing significant guidance for optimizing the display and retrieval locations of high-demand shoes.

[0106] All the above parameters are automatically integrated to form an allocation optimization parameter set. This parameter set not only reflects footwear demand, replenishment characteristics, and popularity distribution, but also integrates space utilization efficiency, replenishment response speed, and user behavior characteristics. Based on this parameter set, the system will use it as a key input for subsequent multi-objective dynamic programming, weight allocation, and intelligent allocation optimization, enabling warehouse location management to go beyond static zoning and achieve dynamic, intelligent, and highly adaptable allocation decisions to the actual sales and operational environment.

[0107] In a preferred embodiment of the present invention, the analysis of replenishment frequency and slow-moving risk factors based on historical sales data includes:

[0108] Based on the shoe wall status dataset, historical sales and replenishment records of shoes of various models and sizes are extracted to form a sales frequency sequence and a replenishment frequency sequence with fixed periodic statistics.

[0109] For each shoe item, the actual number of replenishments and sales volume within a set period are statistically analyzed. The replenishment activity parameter and sales trend parameter of the shoe item within the period are obtained by weighted moving average calculation and compared with the preset threshold. If the replenishment activity parameter is less than the replenishment activity threshold and the sales trend parameter is less than the sales trend threshold, it is judged as a suspected slow-moving risk target.

[0110] For footwear products identified as potentially slow-moving risk targets, their historical shelf duration, inventory turnover cycle, and promotion frequency parameters are further retrieved and compared with the set shelf duration threshold, inventory turnover cycle threshold, and promotion frequency threshold respectively. If each parameter meets the slow-moving risk condition, it is counted as one point. Finally, the scores of all parameters that meet the conditions are accumulated as the slow-moving risk factor for the footwear product. The slow-moving risk factor ranges from zero to an integer of the number of parameters.

[0111] For footwear products that are not identified as having a risk of slow sales, the system can automatically set their slow sales risk factor to zero or a standard value.

[0112] In this embodiment of the invention, the system achieves quantitative identification and hierarchical management of the risk of unsold footwear through collaborative analysis of historical sales data and replenishment data. Specifically, the system first uses a shoe wall status dataset as a foundation to automatically extract and record the historical sales quantity and replenishment operation frequency for each model and size of footwear. All raw data is periodically statistically analyzed according to actual business management needs, such as on a weekly, monthly, or custom business cycle basis, forming a stable and reliable sales frequency sequence and replenishment frequency sequence. The system can dynamically adapt to the rhythm of store operations, automatically collect and archive all historical records without manual intervention, ensuring the timeliness and completeness of the data.

[0113] After obtaining the periodic sales and replenishment frequencies, the system further processes the above sequence for each shoe product using a weighted moving average algorithm. The weighted moving average highlights recent sales and replenishment dynamics while also considering historical trends, thus obtaining replenishment activity parameters and sales trend parameters. This process achieves a sensitive response to market changes by adjusting the weight parameters. For example, when a shoe product has seen very few replenishments and a continuous decline in sales in recent periods, the system can significantly reduce its activity and trend parameters, facilitating subsequent risk identification. All parameter calculations and comparisons are completed automatically by the system without additional configuration, significantly improving the efficiency of risk assessment in large-scale shoe inventory environments.

[0114] For each shoe item, the system compares its replenishment activity parameter with the replenishment activity threshold, and its sales trend parameter with the sales trend threshold. Only when both parameters are below their respective thresholds will the system automatically classify the shoe item as a potential slow-moving risk target. Subsequently, for these suspected slow-moving shoes, the system automatically retrieves multiple management parameters such as historical shelf life, inventory turnover cycle, and promotion frequency. Each parameter is compared with the system's pre-set thresholds. For example, if a shoe item's shelf life exceeds the shelf life threshold, its inventory turnover cycle exceeds the turnover cycle threshold, and its promotion frequency is below the promotion frequency threshold, the system will award one point for each parameter that meets the slow-moving criteria. Finally, all scores are added together to form a clear slow-moving risk factor. The higher the value of this factor, the higher the risk of the shoe item being judged as having slow-moving risk by multiple management indicators. The system uses this comprehensive score and the slow-moving risk factor as a feature input for subsequent popularity prediction and dynamic classification, achieving accurate identification and intervention for slow-moving products.

[0115] For footwear items not identified as potentially slow-moving risk targets, the system automatically sets their slow-moving risk factor to zero or a standard value. This strategy ensures that the feature input dimensions of all footwear items remain consistent, allowing them to participate in subsequent popularity prediction and dynamic hierarchical modeling regardless of whether a slow-moving risk assessment result exists. For example, if a popular footwear item is never identified as a slow-moving target by the system due to continuous restocking and high-frequency sales, its slow-moving risk factor will always be zero, and it will not negatively impact the subsequent popularity prediction model. Through the aforementioned end-to-end data collection, parameter discrimination, and factor generation mechanism, the system greatly improves the automation and scientific rigor of footwear slow-moving risk identification, effectively reduces inventory backlog, improves the health of the product structure, and provides a solid technical foundation for the refined operation and risk-controllable management of the footwear market.

[0116] In a preferred embodiment of the present invention, the method for constructing the heat prediction model includes:

[0117] Using the sales frequency, historical replenishment frequency, and slow-moving risk factors of footwear as inputs, a set of popularity feature vectors is constructed;

[0118] The heat feature vector set is input into the time series prediction model, which is trained using a long short-term memory neural network to model the heat change trend of footwear in a continuous period.

[0119] During the model training and prediction phases, cross-validation is used to evaluate the prediction performance of the time series prediction model, and error backpropagation and parameter optimization are implemented based on the prediction error to iteratively improve the model accuracy.

[0120] The final output is a prediction of the footwear's popularity rating over a predetermined period.

[0121] In this embodiment of the invention, the method for constructing the popularity prediction model significantly improves the accuracy of judging footwear popularity trends and predicting categories. The system automatically constructs a popularity feature vector set using multi-dimensional features such as sales frequency, historical replenishment frequency, and slow-moving risk factors as input. By training these vectors and employing time-series prediction algorithms such as Long Short-Term Memory (LSTM) neural networks, it can capture the nonlinear fluctuations and complex trends in footwear popularity over continuous periods. A cross-validation mechanism is introduced during the model training and prediction stages to effectively avoid overfitting and improve the model's generalization ability. Backpropagation of prediction errors and real-time parameter optimization ensure the model's adaptability and robustness in dynamic market environments.

[0122] In actual operation, the system automatically retrains and updates model parameters at set intervals to ensure that the popularity score prediction always keeps pace with actual footwear sales and market changes. The popularity score output by the model not only provides a scientific basis for the dynamic grading and category label generation of footwear but also provides a data foundation for subsequent inventory allocation, replenishment strategies, and display optimization. For example, if a shoe suddenly becomes a hot seller and its sales frequency surges, the model can promptly predict it as a high-popularity category. Based on this, the system will proactively allocate the shoe to the optimal display and retrieval location, thereby improving response speed and sales opportunities. This intelligent prediction and dynamic grading capability significantly enhances the market adaptability and overall operational effectiveness of the shoe wall.

[0123] The time-series prediction model is trained using a long short-term memory neural network to model the popularity trend of footwear over a continuous period, specifically including:

[0124] The system first uses the popularity characteristics of each shoe product over multiple time periods as model inputs. These popularity characteristics include numerical data such as restocking frequency, sales activity, slow-moving risk factors, and historical popularity scores. These parameters are arranged in chronological order to form a popularity time series input matrix, which reflects the evolution of the shoe product's popularity over multiple consecutive periods.

[0125] The system constructs a neural network model based on Long Short-Term Memory (LSTM) architecture. This model consists of an input layer, multiple LSTM unit layers, a fully connected layer, and an output layer. The input layer receives the time series of popularity features; the LSTM unit layers have memory capabilities to capture short-term fluctuations and long-term trends in the time series; the fully connected layer maps the output of the LSTM units to specific popularity score prediction results; and the output layer outputs the popularity prediction value for the future target period.

[0126] During the model training phase, the system uses the popularity feature parameters from known historical periods as the input to the training set, and the corresponding actual popularity scores as the training objective. The system employs supervised learning, adjusting weights by minimizing the error between the predicted and actual popularity scores. In the early stages of training, the model initializes its parameters with small random values ​​and iteratively trains based on historical data until the model's prediction error converges.

[0127] By introducing an LSTM structure, the model can effectively model the time-relatedness and trends in footwear popularity ratings, overcoming the shortcomings of traditional neural networks in handling time-dependent problems. For example, for a seasonal footwear product, its popularity may increase cyclically in certain months. The LSTM model can identify such cyclical changes and reflect them in the prediction, thereby improving prediction accuracy.

[0128] The method involves evaluating the prediction performance of the time series forecasting model using cross-validation, and implementing error backpropagation and parameter optimization based on the prediction error to iteratively improve model accuracy. Specifically, this includes:

[0129] During model training, to avoid overfitting or underfitting due to training samples, the system uses cross-validation to evaluate model performance. Specifically, the system divides the complete historical popularity time series data into several consecutive time intervals, and uses a k-fold sliding window to alternately select a portion of the data as the validation set and the rest as the training set, repeating the training and validation operations.

[0130] During each cross-validation process, the system calculates the error between the predicted popularity score and the actual score during the training phase. Common error evaluation metrics include root mean square error (RMSE) and mean absolute error (MAE). The model then optimizes parameters such as connection weights and bias coefficients in the network structure through backpropagation based on this error value. During optimization, the system uses the gradient descent algorithm for error backpropagation, gradually correcting the model parameters by calculating the partial derivatives of the loss function with respect to each parameter until the error converges or the set number of training epochs is reached.

[0131] In addition, to further improve generalization ability, the system recalculates the overall prediction accuracy trend after each training round. If the model error does not decrease significantly in multiple training rounds, the system automatically adjusts the model hyperparameters (such as learning rate, number of layers, number of memory units, etc.) and restarts the next training round.

[0132] Finally, the system selects the set of model parameters with the smallest error as the final structure of the popularity prediction model, and outputs the popularity score prediction results of footwear in the future preset period based on this, providing a reliable basis for subsequent footwear grading and warehouse allocation.

[0133] In a preferred embodiment of the present invention, the method for constructing the multi-objective optimization function includes:

[0134] Based on the allocation optimization parameter set, parameters for priority access convenience, overall inventory balance, and local flow distribution of the shoe wall are extracted to form a multi-objective optimization input vector;

[0135] Based on the multi-objective optimization input vector, initial weights are set for each objective and an objective evaluation function is established. The normalization method is used to unify the dimensions of different objectives, and a multi-objective optimization function is constructed through weighted synthesis.

[0136] In this embodiment of the invention, the method for constructing a multi-objective optimization function achieves a high degree of scientific rigor and customization in footwear allocation decisions. Based on the allocation optimization parameter set, the system comprehensively extracts parameters related to priority access convenience (such as path consumption), overall inventory balance (such as inventory differences between storage locations), and local flow distribution parameters for shoe storage areas (such as regional access frequency), generating a multi-objective optimization input vector. For different optimization objectives, the system sets initial weights and establishes corresponding objective evaluation functions. All objectives are standardized in scale through normalization and weighting mechanisms, avoiding misjudgments caused by weight imbalances among multiple objectives.

[0137] In actual operation, the system can also dynamically update the weights of each objective based on historical allocation results and real-time feedback, maintaining an adaptive balance among the allocation objectives. For example, during peak replenishment periods, the convenience weight can be temporarily increased, while during off-seasons, the focus is on inventory balance. After each optimization iteration, the system outputs the optimized multi-objective evaluation function configuration in real time, providing an efficient basis for subsequent allocation decisions and weight adjustments. This not only ensures the scientific and fair nature of the allocation scheme but also allows for flexible adaptation to changing market and operating environments, significantly improving the intelligence and refinement of shoe wall operations.

[0138] The following is an example of a multi-objective evaluation function:

[0139] ;

[0140] in, , , ;

[0141] Where F: the final value of the multi-objective optimization function;

[0142] , , Weighting coefficients reflect the degree of importance attached to each sub-objective, and their sum is equal to 1.

[0143] Each sub-item represents the objective evaluation function for the normalized priority access convenience, overall inventory balance, and local flow distribution of the shoe wall, respectively;

[0144] : Average path consumption value for all popular footwear items;

[0145] The total number of highly popular shoe items included in the calculation;

[0146] : The path consumption value after the i-th highly popular shoe item is assigned;

[0147] , The minimum and maximum path consumption values ​​in the system, used for normalization;

[0148] B: Average standard deviation of inventory distribution;

[0149] : Number of shoe categories (e.g., high popularity, potential popularity, low popularity);

[0150] Standard deviation of inventory quantity for the k-th type of footwear across different zones;

[0151] , Minimum / maximum values ​​of inventory balance evaluation indicators;

[0152] D: Average deviation between footwear inventory flow distribution and expected demand;

[0153] Quantity of all footwear;

[0154] : The actual traffic volume of the assigned storage area for footwear item j;

[0155] The ideal regional popularity value corresponding to the popularity level of shoe item j;

[0156] , : The minimum / maximum value of this deviation evaluation index.

[0157] In a preferred embodiment of the present invention, an iterative optimization algorithm is used to dynamically adjust the weights of the multi-objective optimization function, including:

[0158] Based on the constructed multi-objective optimization function and its initial weight parameters, combined with the current footwear category distribution and warehouse location allocation status, the first round of warehouse location allocation scheme is generated, and the evaluation function values ​​of each objective are calculated.

[0159] By using swarm intelligence optimization methods such as genetic algorithms or particle swarm optimization, the target weight parameters are perturbed to generate multiple sets of weight combinations and corresponding allocation candidate schemes in batches, and the degree of achievement of the target evaluation function is evaluated for each of them.

[0160] For each allocation candidate scheme, the optimal weight combination is selected based on the achievement of the objective evaluation function, historical operation feedback and system optimization criteria. The selection result is then used as input to feed the genetic algorithm for a new round of iteration.

[0161] During multiple rounds of iterative optimization, the weight combination and candidate allocation schemes are continuously adjusted until the multi-objective optimization function converges or meets the set optimization termination conditions, and finally outputs the dynamically optimal target weight configuration and shoe wall storage location allocation scheme.

[0162] In this embodiment of the invention, an iterative optimization algorithm is used to dynamically adjust the weights of the multi-objective optimization function, providing an adaptive, efficient, and globally optimal solution for shoe warehouse allocation. When making allocation decisions, the system first combines the constructed multi-objective optimization function and its initial weight parameters with the current shoe category distribution and warehouse allocation status to automatically generate an initial allocation scheme and evaluate each objective evaluation function. Subsequently, swarm intelligence optimization techniques such as genetic algorithms or particle swarm optimization algorithms are used to perturb and combine the objective weight parameters, generating multiple candidate weights and allocation schemes in batches. The system automatically evaluates the achievement degree of each scheme on each objective evaluation function.

[0163] Based on historical feedback and target achievement performance, the system automatically selects the best-performing weight combination and uses it as feedback input for the next round of optimization. The optimization process involves multiple iterative rounds, with the system continuously adjusting and filtering candidate weights and allocation schemes until the multi-objective optimization function converges or reaches the preset optimization termination criterion. The final output is a dynamically optimal weight configuration and allocation scheme, ensuring that the shoe wall storage location allocation is always in a highly efficient, adaptive, and globally optimal state. This solution significantly improves the scientific nature of shoe replenishment, display, and inventory management, achieving automated intelligent allocation and continuous self-learning optimization, providing a solid technical foundation for cost reduction and efficiency improvement in large-scale footwear retail scenarios.

[0164] Specifically, based on the constructed multi-objective optimization function and its initial weight parameters, combined with the current footwear category distribution and warehouse location allocation status, a first-round warehouse location allocation plan is generated, and the evaluation function values ​​for each objective are calculated, including:

[0165] During the initialization phase, the system first invokes a pre-constructed multi-objective optimization function. This function is built upon multiple influencing factors, including but not limited to: predicted category labels for footwear, accessibility grading for each storage location, the current shoe wall inventory structure, and traffic information for each region. This optimization function sets a set of initial weight parameters to balance the relative importance of each objective evaluation function. These initial weights can be configured based on system experience or determined based on historical best strategies.

[0166] The system first receives the output from the popularity prediction model, obtaining the popularity tag for each shoe item in the current period. The popularity tags are divided into high-popularity, potential-popularity, and low-popularity categories, along with their corresponding popularity scores. The system uses the popularity score as one of the core criteria for prioritizing weight allocation.

[0167] Subsequently, the system loads the allocation status information of all storage locations in the current shoe wall, mainly including: the accessibility level of each storage location, its current occupancy status (idle or used), historical replenishment frequency, the traffic density level of its area, and whether it is in a system-locked state (unallocated). This status information constitutes the constraints in the optimization process.

[0168] The system constructs a target allocation path for each footwear item, with the goal of prioritizing the selection of storage location combinations that meet the following conditions from all available storage locations:

[0169] The higher the accessibility level of the storage location, the better;

[0170] The traffic density in the area is at a low to medium level, so high-traffic congestion areas should be avoided.

[0171] The storage space is currently idle, or may be subject to exchange or adjustment.

[0172] Avoid over-concentrating the same type of footwear in adjacent areas to improve the efficiency of pickup and distribution.

[0173] The dynamic programming algorithm uses shoe popularity ratings as decision variables and a storage location attribute matrix as the state space to map shoe sets to storage location sets. The algorithm solves this problem by minimizing the total allocation cost, which considers popularity weights, accessibility penalty factors, regional traffic factors, and historical exchange frequency penalties. The system constructs the state transition matrix using a step-by-step table-filling method, selecting the path with the minimum cost at each allocation decision stage as the candidate solution.

[0174] For example, for a pair of shoes with a popularity rating of 0.9, the system will prioritize allocating them to an available storage location with an accessibility level of "A" and a traffic level of "1". If multiple storage locations meet the conditions, the system will further consider the regional distribution balance and historical replenishment frequency, and prioritize the storage location with a higher replenishment frequency and a more dispersed surrounding distribution.

[0175] After completing the initial mapping of all footwear items to storage locations, the system performs a complete objective evaluation function calculation for the overall allocation scheme. For example, "priority retrieval convenience" can be measured by the average path retrieval time length, "inventory balance" can be measured by the balanced distribution level of footwear items of different popularity categories, and "footwear storage location flow distribution" can be quantified by the current flow differences in different areas of the shoe wall. All evaluation function values ​​are finally normalized and output in a unified format for subsequent optimization iterations.

[0176] Specifically, a genetic algorithm is used to perturb the target weight parameters, generating multiple sets of weight combinations and corresponding candidate allocation schemes in batches. The achievement degree of each target evaluation function is then evaluated, including:

[0177] After completing the initial allocation scheme evaluation, the system activates a genetic algorithm as the global search optimization strategy. The genetic algorithm first uses the initial weight combination as the "parent chromosome," and then generates a set of "offspring chromosomes" through perturbation operations. Each set of chromosomes corresponds to a new set of target weight combinations. The perturbation process includes operations such as random mutation and crossover based on a Gaussian distribution to ensure a thorough exploration of the weight parameter space.

[0178] For each newly generated target weight combination, the system will re-invoke the optimization function to generate corresponding candidate storage location allocation schemes. Each candidate scheme must complete the calculation of the target evaluation function value and compare its result with the preset target standard.

[0179] For example, if the system sets target thresholds such as "the average access path does not exceed X meters" and "the proportion of highly popular footwear in the area with the highest accessibility level is not less than Y%", the actual achievement of the candidate solutions can be quantified by comparing the evaluation function value with the target thresholds. The achievement of all candidate solutions is recorded uniformly for use in the next round of screening and genetic algorithm feedback.

[0180] Specifically, for each allocation candidate scheme, the optimal weight combination is selected based on the achievement degree of the objective evaluation function, historical operation feedback, and system optimization criteria. The selection result is then used as input to feed the genetic algorithm for a new round of iteration, which includes:

[0181] After generating and evaluating all candidate solutions, the system will comprehensively score and rank the candidate solutions based on the following three categories of indicators:

[0182] Objective evaluation function achievement rate: This refers to the actual completion of each objective in the multi-objective optimization function of the candidate solution. For example, if a solution simultaneously achieves the highest attainability rate for popular footwear and the highest overall inventory balance, then the solution will receive a higher score.

[0183] Historical operational feedback: This refers to the feedback information accumulated by the system over a number of past periods when running this type of weight configuration scheme, such as replenishment delay rate and picking error rate. This information is used to guide the genetic algorithm to avoid weight combinations that have historically performed poorly.

[0184] System optimization criteria include the current load status of the shoe wall, personnel flow, and business peak warning signals, which serve as contextual reference boundaries for solution selection.

[0185] Based on the above three categories of indicators, the system will ultimately select the best-performing weight combination. This weight combination should not only score the highest among the candidates in this round, but also demonstrate its improvement in stability and adaptability by comparing it with the best historical solutions.

[0186] Ultimately, this weight combination will serve as the parent chromosome input for the next-generation genetic algorithm, driving the next round of iterative optimization until the weights converge or the preset number of optimization rounds is reached.

[0187] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A bin management system for a display and warehouse integrated intelligent shoe wall, characterized in that, The system comprises: An information collection module, configured to collect, based on a preset collection period, an occupancy state of each storage location of the shoe wall, a sales frequency data of each type and size of shoes, and a path consumption of the mobile target when taking shoes from the shoe wall, to generate a shoe wall state data set and a path consumption data set; A path analysis module, configured to quantify, according to the path consumption data set, the path consumption of each storage location of the shoe wall, calculate a path consumption value, and compare the path consumption value with a preset path consumption threshold group, to generate a storage location accessibility classification data; A classification and distribution module, configured to divide the shoes into different categories according to the shoe sales frequency in the shoe wall state data set, and combine the storage location accessibility classification data to distribute the shoes of different categories to different storage locations, to generate a storage location distribution scheme; An execution and feedback module, configured to control the mobile target to complete an actual storage location adjustment operation of the shoes according to the storage location distribution scheme, and return the adjustment result to the information collection module after the adjustment is completed, to support self-learning and distribution strategy optimization of the system; The classification and distribution module comprises a multi-target self-optimization matching unit, configured to construct a multi-target dynamic programming model according to the predicted shoe category label and the storage location accessibility classification data, jointly model the priority taking convenience, overall inventory balance, and local flow distribution of the shoe wall, continuously optimize the distribution weight, and generate a storage location distribution scheme, specifically comprising: According to the predicted shoe category label, the storage location accessibility classification data, and the current storage location distribution, historical replenishment frequency, and flow statistical information of each region of the shoe wall, an allocation optimization parameter set is formed; According to the allocation optimization parameter set, a multi-target optimization function of the combination of the priority taking convenience, inventory balance, and flow distribution is established, and a dynamic balance between the targets is realized through weight setting; An iterative optimization algorithm is adopted to dynamically adjust the weight of the multi-target optimization function, and on the basis of continuous simulation and feedback, an optimal storage location distribution scheme is generated, including: According to the constructed multi-target optimization function and initial weight parameters, the current shoe category distribution and storage location distribution state are combined to generate a first-round storage location distribution scheme, and each target evaluation function value is calculated; A genetic algorithm is used to disturb the target weight parameters, a plurality of weight combinations and corresponding distribution candidate schemes are generated in batches, and the respective target evaluation function achievement degrees are evaluated; According to the target evaluation function achievement degrees, historical running feedback, and system optimization standards, the best-performing weight combination is selected from each distribution candidate scheme, and the selection result is fed back to the genetic algorithm for a new round of iteration; In the process of multiple rounds of iterative optimization, the weight combination and the distribution candidate scheme are continuously adjusted until the multi-target optimization function converges or the set optimization termination condition is met, and finally the dynamically optimal target weight configuration and the shoe wall storage location distribution scheme are output.

2. The bin management system of the integrated display and storage intelligent shoe wall according to claim 1, wherein, The path analysis module comprises: A path consumption modeling unit, configured to calculate, according to the path consumption data set, a path consumption value of each feasible path of the mobile target between each storage location of the shoe wall, and generate path consumption value distribution data by comprehensively considering the horizontal distance, vertical distance, and number of turns. The threshold comparison and self-learning unit is configured to compare the path consumption value distribution data with a preset path consumption threshold group, and generate the storage location accessibility classification data.

3. The bin management system of a display and warehouse integrated intelligent shoe wall according to claim 1, characterized in that, The hierarchical allocation module further includes: The multi-dimensional dynamic classification unit is configured to update the heat score of the shoes in a future preset period in real time according to the shoe wall state data set, combine the restocking frequency and the historical slow-selling risk factor, dynamically classify the shoes into a high-heat, potential-heat and low-heat category, and generate a predicted shoe category label.

4. The bin management system of a display and warehouse integrated intelligent shoe wall according to claim 2, characterized in that, According to the path consumption data set, the movement behavior of the mobile target between the storage locations of the shoe wall is calculated by comprehensively considering the horizontal distance, vertical distance and number of turns, the path consumption value of each feasible path is calculated, and the path consumption value distribution data of each storage location of the shoe wall is generated, including: According to the path consumption data set, the real-time path coordinate sequence generated in the actual movement process of the mobile target between the storage locations inside the shoe wall is extracted, and an original path data set is formed; According to the original path data set, the horizontal distance, vertical distance and turning point information of each path segment are extracted respectively, and a multi-dimensional path feature set is formed; According to the multi-dimensional path feature set, adjustable weights are assigned to each path parameter, and the comprehensive path consumption value of each path is quantitatively processed by using a weighted fusion method, and the path consumption value distribution data is generated.

5. The bin management system of a display and warehouse integrated intelligent shoe wall according to claim 1, wherein, According to the shoe wall state data set, the heat score of the shoes in a future preset period is updated in real time by combining the restocking frequency and the historical slow-selling risk factor, the shoes are dynamically classified into a high-heat, potential-heat and low-heat category, and a predicted shoe category label is generated, including: According to the shoe wall state data set, the current sales frequency of each model and size of shoes is analyzed, and based on the historical sales data, the restocking frequency and the slow-selling risk factor are analyzed to form a multi-dimensional heat feature parameter set; A heat prediction model based on time series analysis is used to process the multi-dimensional heat feature parameter set, predict the heat score and trend of the shoes in a future preset period, and generate a heat score prediction result; According to the heat score prediction result, the shoes are classified into a high-heat, potential-heat and low-heat category, and a corresponding predicted shoe category label is generated.

6. The bin management system of a display and warehouse integrated intelligent shoe wall according to claim 5, characterized in that, Based on the historical sales data, the restocking frequency and the slow-selling risk factor are analyzed, including: According to the shoe wall state data set, the historical sales records and restocking records of each model and size of shoes are extracted to form a sales frequency sequence and a restocking frequency sequence which are counted in a fixed period; For each shoe, the actual restocking times and sales amount in a set period are counted, the restocking activity parameter and the sales trend parameter of the shoe in the period are obtained by weighted moving average calculation, and compared with the preset threshold value, if the restocking activity parameter is less than the restocking activity threshold value and the sales trend parameter is less than the sales trend threshold value, it is determined as a suspected slow-selling risk target; For the shoes determined as suspected slow-selling risk targets, the historical shelf time, inventory turnover period and promotion frequency parameters of the shoes are further called and compared with the set shelf time threshold, inventory turnover period threshold and promotion frequency threshold. If each parameter meets the slow-selling risk condition, it is counted as one point. The final score of all parameters meeting the conditions is added up as the slow-selling risk factor of the shoes, and the value of the slow-selling risk factor is zero to an integer of the number of parameter items. For shoes not determined as slow-selling risk targets, the system can automatically set the slow-selling risk factor of the shoes to zero or a standard value.

7. The bin management system of a display and warehouse integrated intelligent shoe wall according to claim 5, wherein, The method for constructing the heat prediction model comprises: Taking the sales frequency, historical restocking frequency and slow-selling risk factor of the shoes as inputs, a heat feature vector set is constructed; The heat feature vector set is input into a time series prediction model, the time series prediction model is trained using a long short-term memory neural network, and the heat change trend of the shoes in a continuous period is modeled; In the model training and prediction stage, the cross-validation method is used to evaluate the prediction effect of the time series prediction model, and the error back propagation and parameter optimization are implemented according to the prediction effect error to iteratively improve the accuracy of the model; Finally, the heat score prediction result of the shoes in a future preset period is output.

8. The bin management system of a display and warehouse integrated intelligent shoe wall according to claim 1, wherein, The method for constructing the multi-objective optimization function comprises: According to the allocation optimization parameter set, the priority convenience parameter, the overall inventory balance parameter and the shoe wall local flow distribution parameter are extracted respectively to form a multi-objective optimization input vector; According to the multi-objective optimization input vector, the initial weight is set and the target evaluation function is established for each target, the different targets are unified in dimension by using the normalization method, and the multi-objective optimization function is constructed by weighted synthesis.

Images