Inventory distribution decision-making method and system based on dynamic multi-objective optimization

Abstract

The invention provides an inventory distribution decision-making method and system based on dynamic multi-objective optimization, and the method comprises the steps: obtaining real-time order data, goods weight data, historical demand data and environment state data of a target commodity; based on the goods weight data, generating a distribution adjustment strategy of the target goods on the goods shelf in combination with a preset goods shelf load-bearing constraint condition; the distribution adjustment strategy, the real-time order data, the historical demand data and the environment state data are combined, and a multi-target decision model with minimization of the replenishment cycle, minimization of the storage cost and optimization of risk balance as optimization targets is constructed; and outputting an optimal inventory distribution decision scheme by adopting an interactive iteration mode based on the multi-objective decision model. The inventory turnover efficiency is improved, and the operation risk is reduced.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent warehouse management, and in particular to an inventory distribution decision-making method and system based on dynamic multi-objective optimization. Background Art

[0002] In e-commerce warehouse management, inventory distribution decisions must simultaneously balance multiple objectives, including replenishment efficiency, storage costs, and shelf stability. This is especially true during peak sales periods, when demand for goods fluctuates significantly and the warehouse environment changes rapidly. Traditional static optimization methods struggle to adapt to dynamic scenarios. Therefore, an intelligent decision-making solution is urgently needed that can respond in real time to changes in orders, environmental conditions, and shelf load constraints.

[0003] An existing inventory scheduling method based on single-objective optimization predicts replenishment demand based on historical sales figures. This method, combined with shelf weight limits, generates a fixed-cycle replenishment plan and employs linear programming to find the distribution solution that minimizes storage costs. The system adjusts product placement based on pre-set rules to ensure that the weight limit per shelf is met.

[0004] This solution only uses warehousing costs as the optimization target, and treats replenishment cycle and risk control as fixed constraints. As a result, when orders suddenly increase or the environment is abnormal, the system-generated solution often sacrifices replenishment timeliness to meet cost requirements. There is also a lack of dynamic assessment of the long-term stability of the shelves, resulting in a high risk of replenishment delays and shelf safety hazards in actual operations. Summary of the Invention

[0005] The present application provides an inventory distribution decision-making method and system based on dynamic multi-objective optimization to solve the problems of low inventory turnover efficiency and high operational risks in the prior art.

[0006] In a first aspect, the present application provides an inventory distribution decision-making method based on dynamic multi-objective optimization, comprising:

[0007] Obtain real-time order data, product weight data, historical demand data, and environmental status data for target products;

[0008] Based on the product weight data and the preset shelf load-bearing constraints, a distribution adjustment strategy for the target product on the shelf is generated;

[0009] Combining the distribution adjustment strategy, the real-time order data, the historical demand data, and the environmental status data to construct a multi-objective decision model with the optimization goals of minimizing replenishment cycle time, minimizing storage costs, and optimizing risk balance;

[0010] Based on the multi-objective decision model, an optimal inventory distribution decision plan is outputted in an interactive iterative manner.

[0011] Optionally, the distribution adjustment strategy, the real-time order data, the historical demand data, and the environmental status data are combined to construct a multi-objective decision model with the optimization goals of minimizing replenishment cycle time, minimizing storage costs, and optimizing risk balance, including:

[0012] generating a migration instruction according to the distribution adjustment strategy, and converting the migration instruction into a storage space occupancy parameter;

[0013] Calculate and generate a replenishment urgency parameter based on the real-time order data and the historical demand data;

[0014] Based on the environmental status data and the shelf stability change trend, a storage risk coefficient is generated;

[0015] A multi-objective decision model is established with the replenishment urgency parameter, the storage space occupancy parameter and the storage risk coefficient as variables.

[0016] Optionally, the multi-objective decision model is established with the replenishment urgency parameter, the storage space occupancy parameter, and the storage risk coefficient as variables, including:

[0017] A first function representing minimization of replenishment cycle is constructed based on the replenishment urgency parameter, a second function representing minimization of storage cost is constructed based on the storage space occupancy parameter, and a third function representing optimization of risk balance is constructed based on the storage risk coefficient;

[0018] Constructing a fluctuation range constraint condition for the replenishment cycle based on the replenishment urgency parameter and the benchmark replenishment cycle;

[0019] Based on the storage space occupancy parameters, generating boundary conditions for storage costs;

[0020] According to the dynamic change rate of the warehousing risk coefficient, a stability threshold of risk balance is set;

[0021] Integrating the fluctuation range constraint, the boundary condition, and the stability threshold into an optimization framework including the first function, the second function, and the third function;

[0022] A two-way conflict resolution mechanism for coordinating conflicting objectives is established in the optimization framework to form a multi-objective decision model.

[0023] Optionally, a two-way conflict resolution mechanism for coordinating objective conflicts is established in the optimization framework to form a multi-objective decision model, including:

[0024] Establishing a feedback correlation relationship between the replenishment cycle and the storage cost;

[0025] Establishing a constraint intervention relationship of the risk balance on the feedback association relationship;

[0026] Establishing a dynamic coupling relationship between the storage cost and the risk balance target;

[0027] Building a bidirectional conflict resolution mechanism based on the feedback association relationship, the constraint intervention relationship, and the dynamic coupling relationship;

[0028] In the optimization framework, the conflicts among replenishment cycle, storage cost and risk balance are coordinated in real time through the two-way conflict resolution mechanism to generate a multi-objective decision model.

[0029] Optionally, generating a distribution adjustment strategy for target products on shelves based on the product weight data and in combination with preset shelf load constraints includes:

[0030] Identify the current load distribution status of the product weight data in the multi-layer space of the shelf;

[0031] Comparing the current load distribution state with the preset shelf load constraints layer by layer to identify overload areas exceeding the preset load threshold;

[0032] Generating a plan for migrating goods from the over-limited areas to non-over-limited areas based on the location distribution and degree of over-limit of the over-limited areas;

[0033] In combination with the environmental status data, the migration path and migration order in the product migration plan are adjusted, and the adjusted product migration plan is used as a distribution adjustment strategy.

[0034] Optionally, generating a plan for migrating goods from the over-limited area to the non-over-limited area according to the location distribution and the degree of over-limit of the over-limited area includes:

[0035] Determining a spatial distribution topology of out-of-gauge goods based on the location distribution of the out-of-gauge areas;

[0036] Calculating the migration amount according to the degree of exceeding the limit in the exceeding limit area;

[0037] Select a target area whose bearing margin is greater than the migration amount from all areas that are not exceeded;

[0038] Based on the spatial distribution topological relationship and the target area, a product migration plan is generated.

[0039] Optionally, outputting an optimal inventory distribution decision solution based on the multi-objective decision model in an interactive iterative manner includes:

[0040] Based on the multi-objective decision model, an initial decision solution set is generated through an initialization algorithm as input for the first interactive iteration;

[0041] In each interactive iteration, the real-time change of the environmental state data and the distribution adjustment strategy are used as dynamic constraints;

[0042] Screening out a subset of non-inferior solutions that satisfy the dynamic constraint conditions from the decision solution set;

[0043] Determining that a change in the solution space of the non-inferior solution subset is less than a preset convergence threshold;

[0044] If not, the non-inferior solution subset is expanded to obtain a non-inferior solution set, and the non-inferior solution set is used as the decision solution set to repeat the iterative process until the solution space change of the non-inferior solution subset is less than the preset convergence threshold or the maximum number of iterations is reached, and the optimal inventory distribution decision plan is generated.

[0045] In a second aspect, the present application provides an inventory distribution decision system based on dynamic multi-objective optimization, comprising:

[0046] An acquisition module is used to obtain real-time order data, product weight data, historical demand data, and environmental status data of target products when the real-time demand fluctuation of e-commerce warehouses exceeds a preset fluctuation;

[0047] A generation module, configured to generate a distribution adjustment strategy for target commodities on the shelves based on the commodity weight data and in combination with preset shelf load-bearing constraints;

[0048] A construction module is used to combine the distribution adjustment strategy, the real-time order data, the historical demand data, and the environmental status data to construct a multi-objective decision model with the optimization goals of minimizing replenishment cycle time, minimizing warehousing costs, and optimizing risk balance;

[0049] The output module is used to output the optimal inventory distribution decision plan in an interactive iterative manner based on the multi-objective decision model.

[0050] In a third aspect, the present application provides a computing device comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute an inventory distribution decision method based on dynamic multi-objective optimization as described in any one of the first aspects.

[0051] In a fourth aspect, the present application provides a computer storage medium having computer program instructions stored thereon, wherein the computer program instructions, when executed by a processor, implement an inventory distribution decision method based on dynamic multi-objective optimization as described in any one of the first aspects.

[0052] In the present application, a method for inventory distribution decision-making based on dynamic multi-objective optimization is provided, which includes: obtaining real-time order data, product weight data, historical demand data and environmental status data of target products; generating a distribution adjustment strategy for the target products on the shelves based on the product weight data and in combination with preset shelf load-bearing constraints; combining the distribution adjustment strategy, the real-time order data, the historical demand data and the environmental status data to construct a multi-objective decision-making model with the optimization objectives of minimizing replenishment cycle, minimizing warehousing costs and optimizing risk balance; and outputting the optimal inventory distribution decision plan based on the multi-objective decision-making model in an interactive iterative manner.

[0053] The technical solution provided by this application has the following beneficial effects:

[0054] This application provides comprehensive and dynamic data support for decision-making by collecting order, weight, historical demand and environmental data in real time, ensuring that the system can respond to various changes in a timely manner. Dynamically optimize the location distribution of goods based on shelf load constraints, effectively avoiding the risk of shelf overruns while improving space utilization. Integrate replenishment cycle, cost and risk goals to achieve coordinated optimization of the three, improve inventory turnover efficiency and reasonably control operating costs. Through dynamic constraint adjustment and non-inferior solution screening, the optimal solution that adapts to real-time changes is generated, greatly improving the accuracy of decision-making.

[0055] Furthermore, this application also converts the distribution adjustment strategy into space occupancy parameters, combines order data to generate replenishment urgency parameters, and associates environmental data to generate risk coefficients. A multi-objective decision-making model is established with these three as variables to achieve a closed-loop transformation from data to model.

[0056] Furthermore, through parameterized mapping and multi-objective coupling, while ensuring shelf safety, it dynamically balances replenishment efficiency and storage costs, improves order processing capabilities, and reduces equipment maintenance requirements.

[0057] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0059] Figure 1 A flowchart of an inventory distribution decision method based on dynamic multi-objective optimization provided in an embodiment of the present application;

[0060] Figure 2 A schematic diagram of the structure of an inventory distribution decision system based on dynamic multi-objective optimization provided in an embodiment of the present application;

[0061] Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0062] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0063] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0064] Existing single-objective optimization solutions for intelligent e-commerce warehouse scheduling have limitations: they focus on warehousing cost as the sole optimization objective, treating replenishment cycles and risk control as fixed constraints. Consequently, when faced with sudden increases in orders or environmental anomalies, the generated solutions are often forced to extend replenishment cycles to meet cost requirements. Furthermore, they lack a dynamic assessment mechanism for shelf stability, resulting in a compromise between operational efficiency and safety. This shortcoming stems from the fact that static optimization models fragment the collaborative relationships among multiple objectives, making them difficult to adapt to the dynamic complexity of modern warehousing.

[0065] In response to the above problems, this application proposes an inventory distribution decision method based on dynamic multi-objective optimization. By integrating order data, shelf load-bearing status and environmental monitoring information in real time, a multi-objective decision model is constructed to balance replenishment cycle, storage cost and risk. This method first quantifies the product distribution strategy into space occupancy parameters, generates a dynamic urgency index based on order fluctuation characteristics, and establishes a risk prediction mechanism by associating environmental data; then, an interactive iterative algorithm is used to automatically balance the conflicting relationships among the three objectives while meeting the hard constraints of shelf load-bearing, and output the optimal inventory distribution decision plan that adapts to real-time working conditions. This solution breaks through the limitations of traditional single-objective optimization. Through multi-source data fusion and dynamic weight adjustment mechanism, it not only ensures the timely response capability of peak orders, but also realizes the coordinated control of storage costs and safety risks, fundamentally solving the problems of efficiency loss and safety hazards caused by target fragmentation in the existing technology.

[0066] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0067] Figure 1 A flowchart of an inventory distribution decision method based on dynamic multi-objective optimization provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the method includes:

[0068] Step 101: Obtain real-time order data, product weight data, historical demand data, and environmental status data of the target product.

[0069] In step 101, real-time order data refers to the current pending order information for goods, including product type, quantity, and delivery time requirements, reflecting immediate market demand. Product weight data represents the actual weight of goods and their distribution on the shelves, used to assess shelf load capacity. Historical demand data represents the sales patterns and fluctuations of goods over past periods, used to predict replenishment needs. Environmental data represents physical parameters within the warehouse, such as temperature and humidity, and shelf vibration amplitude, used to assess warehouse safety risks.

[0070] In this embodiment of the present application, the system collects product order information, shelf weight sensor data, and environmental monitoring data in real time through IoT devices, while simultaneously retrieving historical sales records from a database. Order data is cleaned to extract effective demand characteristics, product weight data is stored by shelf layer and partition, historical data is analyzed for periodic patterns through time series analysis, and environmental data is uploaded to the central processing unit in real time via a distributed sensor network. These four types of data are standardized to form structured input, providing a multi-dimensional basis for subsequent decision-making.

[0071] For example, during a promotional period, an e-commerce company's North China warehouse center detected a sudden increase in real-time orders for air conditioners. Shelf pressure sensors indicated that the total weight of refrigerators stored on the top shelf exceeded the standard, while temperature and humidity sensors detected localized high temperatures. The system simultaneously retrieved June sales data for the product over the past three years, analyzed its historical fluctuations, and generated a comprehensive dataset based on current environmental conditions. Real-time order volume was directly acquired through the order management system, shelf weight data was collected by load cells installed on each shelf, historical data was exported from the company's ERP system, and environmental data was uploaded every five minutes by IoT terminals deployed in the warehouse.

[0072] Step 102: Based on the product weight data and in combination with preset shelf load constraints, a distribution adjustment strategy for the target product on the shelf is generated.

[0073] In step 102, the shelf load constraints represent the maximum load allowed on each shelf layer and the distribution density limit, which is used to ensure storage safety. The distribution adjustment strategy includes the operation plan of the product migration path, sequence and quantity to optimize the shelf space utilization.

[0074] In an embodiment of the present application, the real-time weight distribution of each shelf layer is first analyzed and compared layer by layer with the preset weight threshold, marking the over-limit areas and their over-limit weights. The number of goods to be relocated is calculated based on the degree of over-limit, and combined with the shelf space topology, the target location is selected as the area within the limit with sufficient weight margin. When generating migration instructions, priority is given to areas with severe over-limit, and the path planning is adjusted based on environmental data to avoid areas with high temperatures or abnormal vibrations. The final output is a distribution adjustment strategy containing specific migration steps.

[0075] For example, the system detected that the current load capacity of the top shelf exceeded the threshold, with refrigerators stored in a certain area exceeding the weight limit by 80 kg. Based on the three-dimensional model of the shelf, two vacant areas on the middle level were selected as migration targets, with remaining load capacities of 50 kg and 30 kg, respectively. Instructions were generated to migrate the items in two steps: first, move the 50 kg refrigerator to the left area on the middle level, and then move the remaining 30 kg to the right area on the middle level. Because environmental monitoring indicated that temperatures were high in some areas of the route, the system automatically adjusted the transport route to bypass the central aisle.

[0076] Step 103: Combining the distribution adjustment strategy, the real-time order data, the historical demand data, and the environmental status data, a multi-objective decision model is constructed with minimizing replenishment cycle time, minimizing storage costs, and optimizing risk balance as optimization goals.

[0077] In step 103, the replenishment cycle target is generated by generating a replenishment urgency parameter based on the fluctuation characteristics of real-time order data and historical demand data. The target value is calculated using the formula: base replenishment cycle / urgency parameter. Factors influencing this include order surges and historical sales cyclicality. The relationship is that the greater the urgency parameter, the shorter the target replenishment cycle. The warehousing cost target is generated by calculating the space occupancy parameter converted from item weight data and the load status of transportation nodes. The target value is quantified using the formula: space density × transportation cost coefficient + load penalty term. Factors influencing this include shelf space utilization and distribution center processing capacity. The relationship is that higher space density or greater load results in higher cost targets. The risk balance target is generated by generating a warehousing risk coefficient based on environmental status data and changes in the physical state of the shelves. The stability target range is set based on a safety threshold. Factors influencing this include the temperature and humidity change rate and shelf vibration amplitude. The target range is automatically tightened when the risk coefficient exceeds the threshold. The multi-objective decision-making model uses mathematical optimization methods to simultaneously coordinate three objectives: replenishment cycle time, storage costs, and risk balance. The replenishment cycle objective requires the fastest response time, the storage cost objective seeks to minimize operating expenses, and the risk balance objective ensures shelf safety and environmental stability. The model uses the spatial parameters transformed by the distribution adjustment strategy, the urgency parameters generated by order data, and the risk coefficient derived from environmental data as input variables to establish a multi-objective function system. Through dynamic weight allocation and constraint coupling mechanisms, it outputs a Pareto optimal solution set that balances these three conflicting objectives.

[0078] In this embodiment, the migration instructions in the distribution adjustment strategy are converted into space occupancy parameters to reflect changes in shelf density; the replenishment urgency parameter is calculated based on the deviation between real-time orders and historical data; and the warehouse risk coefficient is derived based on environmental data trends. Using these three as input variables, an optimization model with three objective functions is established: Where P is the replenishment urgency parameter, S is the spatial density, L is the transport load factor, ΔV is the change in risk coefficient, and w and k are weight coefficients. These three objectives are linked through a dynamic constraint network to form a collaborative optimization framework.

[0079] For example, the current replenishment urgency parameter for air conditioners is calculated to be 2.0, and the shelf space density parameter is 90kg / m 2, the transport load factor is 75%, and the risk coefficient change is 0.15. When constructing the model, weight coefficients w1 = 0.6, w2 = 0.4, and the sensitivity coefficient k = 2 were set. This generates the following objective function: replenishment cycle target f1 = 0.5, storage cost target f2 = 0.6 × 90 + 0.4 × 75 = 84, and risk balance target f3 = 0.77. The model uses iterative calculations to find a balanced solution for these three factors.

[0080] Step 104: Based on the multi-objective decision model, an optimal inventory distribution decision solution is outputted in an interactive iterative manner.

[0081] In step 104, the optimal inventory distribution decision plan is the final execution strategy selected from the Pareto solution set generated by the multi-objective model. This plan, while meeting the hard constraints of shelf load capacity and environmental safety, comprehensively considers the balance between replenishment efficiency, cost control, and risk management. Specifically, it includes a plan for adjusting product storage locations, scheduling replenishment schedules, and environmental control requirements. It is a set of operational instructions for achieving dynamic warehouse optimization. Through interactive iteration, the plan continuously approaches the optimal equilibrium state, ensuring decision-making rationality despite order fluctuations and environmental changes.

[0082] In this embodiment, a first solution set containing random feasible solutions is initialized. Constraints are updated in each iteration based on the latest environmental data and order changes, and non-inferior solutions that meet the current constraints are selected. The spatial distribution density of the solution set is calculated to retain solutions with high diversity. When the change in the solution space over multiple consecutive iterations is less than a threshold, the solution with the largest crowding distance is selected as the final solution.

[0083] For example, after five iterations, the model's set of non-inferior solutions converged from an initial 50 to 15. The current optimal solution is: replenishment cycle 1.2 days, storage cost 78 units, and risk factor 1.1. The system automatically generates execution instructions, including replenishment batch scheduling, shelf layout adjustment diagrams, and environmental control requirements.

[0084] This method achieves intelligent decision-making for inventory distribution through multi-source data fusion and dynamic multi-objective optimization. While ensuring shelf safety, it improves replenishment response speed and storage space utilization, while effectively controlling operational risks. The system can adapt to order fluctuations and environmental changes, providing efficient and reliable decision-making support for modern e-commerce warehousing.

[0085] To address the technical challenges of multi-objective collaborative optimization in e-commerce warehousing, in some embodiments, step 103: combining the distribution adjustment strategy, the real-time order data, the historical demand data, and the environmental status data to construct a multi-objective decision-making model with the optimization objectives of minimizing replenishment cycle time, minimizing warehousing costs, and optimizing risk balance, includes:

[0086] Step 201: Generate a migration instruction according to the distribution adjustment strategy, and convert the migration instruction into a storage space occupancy parameter.

[0087] In step 201, migration instructions are standardized operational commands generated based on an analysis of overrun areas on the shelves. They include the specific location of the items to be moved, the target storage area, the number of items to be moved, and a priority ranking. By quantifying the relationship between the degree of overrun and the spatial topology, these instructions transform the distribution adjustment strategy into an executable set of item displacement operations, serving as the fundamental action unit for dynamic shelf optimization. The storage space occupancy parameter, a quantified indicator of shelf space utilization derived from the product migration instructions, reflects the actual load status of each shelf area. This parameter, encompassing two dimensions, namely, load-bearing capacity per unit area and distribution uniformity, is used to assess storage space utilization efficiency.

[0088] In this embodiment, the system parses the migration instructions in the distribution adjustment strategy, extracts the weight and location information of the items to be moved, and calculates the spatial density changes in each area before and after the migration based on the shelf grid coordinates. By mapping the migration path to the three-dimensional shelf model, the load-bearing values ​​of each layer and area are dynamically updated, ultimately generating space occupancy parameters including a density distribution matrix, providing a quantitative basis for cost optimization.

[0089] Step 202: Calculate and generate a replenishment urgency parameter based on the real-time order data and the historical demand data.

[0090] In step 202, the replenishment urgency parameter is a dynamic indicator generated by combining real-time order fluctuations with historical patterns, representing the urgency of the current replenishment demand. By integrating real-time sales trends with cyclical characteristics, this parameter overcomes the lag inherent in static forecasts.

[0091] In the embodiment of the present application, a sliding window analysis is first performed on the real-time order data to calculate the deviation of the current sales volume from the historical mean value for the same period; at the same time, the seasonal and cyclical characteristics of the historical demand data are extracted as a benchmark. The real-time fluctuation amplitude is weighted and fused with the historical regularity coefficient to generate a replenishment urgency parameter with time series sensitivity, which is used to dynamically adjust the replenishment cycle target. The specific process is as follows: extract the cyclical characteristics of the historical demand data, calculate the benchmark daily average sales volume μ and the standard deviation σ; obtain the real-time order data, calculate the current sales volume x; calculate the fluctuation coefficient α=(x-μ) / σ; the replenishment urgency parameter P=β×(1+α), where β is the historical adjustment coefficient. Example: The historical average daily sales volume of air conditioners in June is μ=120 units, σ=30 units; the real-time sales volume is x=200 units; then α=(200-120) / 30=2.67; take β=1.2, and get P=1.2×3.67=4.4.

[0092] Step 203: Generate a storage risk coefficient based on the environmental status data and the shelf stability change trend.

[0093] In step 203, the shelf stability change trend refers to the dynamic characteristics of the shelf structure's safety status evolving over time, obtained by continuously monitoring the correlation between environmental parameters (temperature, humidity, vibration) and physical conditions (deformation, load distribution). Specifically, it comes from: the fluctuation patterns of data collected in real time by environmental sensors obtained through time series analysis, and the cumulative trend of micro-deformations recorded by shelf strain gauges. The two are weighted and fused to form a stability assessment curve. The warehouse risk coefficient is a composite indicator that quantifies the degree of correlation between environmental conditions and shelf stability. This coefficient predicts potential operational risks by monitoring dynamic parameters such as temperature and humidity gradients and vibration amplitude, combined with shelf physical deformation data.

[0094] In this embodiment, real-time temperature and humidity data collected by environmental sensors are filtered and their rate of change over time is calculated. Simultaneously, shelf strain gauges are used to monitor micro-deformations after load adjustment. The environmental change rate and physical deformation are normalized, and a comprehensive risk factor is generated using a pre-set weighting formula, serving as the model's safety constraint boundary.

[0095] Step 204: Establish a multi-objective decision model using the replenishment urgency parameter, the storage space occupancy parameter, and the storage risk coefficient as variables.

[0096] In this embodiment, the space occupancy parameter is mapped as an input variable of the warehousing cost function, the replenishment urgency parameter is converted into an adjustment coefficient for the cycle target, and the risk factor is used as the boundary value of the constraint condition. A nonlinear programming method is used to construct an association matrix for the three objectives, and an interactive weight adjustment mechanism is used to balance the conflicting objectives, ultimately forming a dynamically responsive decision-making model.

[0097] Here's a specific example:

[0098] During the peak season for air conditioner sales, a large e-commerce warehouse system detected a 150% increase in real-time orders for product area A. The total weight of Category B items stored on the top shelves reached 580 kg, exceeding the preset weight-bearing threshold of 500 kg. Analysis of the three-dimensional shelf model determined that the excess weight was concentrated between coordinates X10-Y15 and X15-Y20. The excess weight, calculated as 80 kg using the formula (overweight = measured weight - weight-bearing threshold), was 80 kg. Based on a distribution adjustment strategy, the system generated relocation instructions: two units of Category B, weighing a total of 96 kg, were relocated in two stages: the first 48 kg to the left section of the middle shelf, coordinates X5-Y8 to X8-Y11; the remaining 48 kg to the right section of the bottom shelf, coordinates X3-Y5 to X6-Y8. The relocation route was automatically adjusted to bypass the central aisle due to the temperature in the X12-Y14 area reaching 38°C. At the same time, the system retrieves the sales data of product A over the past three years and calculates that the average daily sales volume is 120 units, with a standard deviation of 30 units. The corresponding replenishment urgency parameter for the current sales volume of 200 units is calculated as 4.4 according to the formula P = β × (1 + (x-μ) / σ), where β is a historical adjustment coefficient of 1.2. Environmental monitoring shows that the current temperature of 38°C has a deviation coefficient of 0.086 from the safety threshold of 35°C, and the risk contribution value of the shelf vibration amplitude exceeding the standard of 0.12mm is 0.04, resulting in a comprehensive storage risk coefficient of 1.126. With a space density of 90kg / m 2 , and the transportation load rate of 75% is used as input to establish a multi-objective function: replenishment cycle target f1 = 1 / P = 0.227, storage cost target f2 = w1S + w2L = 0.6×90 + 0.4×75 = 84, risk balance target f3 = 1 / (1+kΔV) = 0.816, where k is the risk sensitivity coefficient 2, and ΔV is the risk coefficient change of 0.15.

[0099] In the embodiment of the present application, the method realizes intelligent decision-making of inventory distribution strategy through parametric modeling and dynamic optimization, improves replenishment response capability and storage resource utilization while ensuring shelf safety, and provides reliable technical support for inventory management in complex environments.

[0100] To solve the problem of multi-objective collaborative optimization in e-commerce warehousing, in some embodiments, step 204: establishing a multi-objective decision model using the replenishment urgency parameter, the storage space occupancy parameter, and the storage risk coefficient as variables, includes:

[0101] Step 301: constructing a first function representing minimizing replenishment cycle based on the replenishment urgency parameter, constructing a second function representing minimizing storage cost based on the storage space occupancy parameter, and constructing a third function representing optimizing risk balance based on the storage risk coefficient.

[0102] In step 301, the first function is a mathematical expression that reflects the efficiency of replenishment. The smaller its value, the faster the replenishment response. The second function is a mathematical expression that quantifies the cost of warehouse operation. The smaller its value, the better the cost control. The third function is a mathematical expression that evaluates the security risk. The larger its value, the higher the system stability.

[0103] In this embodiment, the system first converts the replenishment urgency parameter into an adjustment coefficient for a periodic function, allowing the function value to dynamically change with demand fluctuations. It then decomposes the space occupancy parameter into two components: storage density and transport load, constructing a linearly weighted cost function. Finally, the risk coefficient is reciprocally processed to convert it into a stability function for maximum optimization. These three functions are normalized to maintain dimensional consistency.

[0104] Step 302: Constructing a fluctuation range constraint condition for the replenishment cycle based on the replenishment urgency parameter and the benchmark replenishment cycle.

[0105] In step 302, the baseline replenishment cycle refers to the standard replenishment interval for a product under normal sales conditions. Its value is derived from a statistical analysis of the product's historical sales data, specifically by calculating the ratio of the average inventory consumption rate to the safety stock level over multiple past sales cycles. The system then adjusts this value based on factors such as product characteristics and supplier delivery cycles, and uses it as a baseline reference value for the optimization model. The fluctuation range constraint, defined by the baseline cycle and the urgency parameter, defines the upper and lower limits of the allowed variation in the replenishment cycle. These limits are determined by the baseline cycle and the urgency parameter, and serve to constrain the range of plausible optimization solutions.

[0106] In this embodiment, a baseline replenishment cycle is set based on product characteristics. This is multiplied and divided by a replenishment urgency parameter to create a dynamic constraint interval. When orders surge, the urgency parameter increases, shifting the constraint interval upwards, allowing for a shorter replenishment cycle; otherwise, the constraint interval shifts downwards, ensuring the replenishment frequency is not excessive.

[0107] Step 303: Generate boundary conditions for storage costs based on the storage space occupancy parameters.

[0108] In step 303, the boundary condition refers to the floating range allowed by the warehousing cost target, and its upper and lower limits are dynamically adjusted according to the space occupancy parameters and the current status of transportation resources to balance storage efficiency and logistics capabilities.

[0109] In this embodiment, the cost boundary is calculated by analyzing the spatial density distribution of each shelf area and combining it with the real-time load rate of the transport node. When the density in a certain area is too high or the transport load is too large, the corresponding boundary conditions are automatically relaxed to avoid over-constraints that lead to infeasible solutions.

[0110] Step 304: According to the dynamic change rate of the warehousing risk coefficient, a stability threshold of risk balance is set.

[0111] In step 304, the dynamic rate of change of the warehouse risk factor refers to the speed at which the risk factor changes over time. This rate is calculated by calculating the relative change between the current risk factor and the previous value. Specifically, a sliding time window algorithm is used to perform linear regression analysis on the risk factors of the most recent monitoring cycles. The slope is used as the rate of change indicator to determine the urgency of the risk development trend. The stability threshold is the minimum safety standard that must be met to achieve risk balance. Its value is dynamically updated based on the changing trend of the risk factor to ensure the safety of the warehouse environment.

[0112] In this embodiment, the recent slope of the risk factor is monitored. When the rate of change exceeds a preset warning value, the stability threshold is proportionally tightened. When the change is stable, the threshold is appropriately relaxed to increase optimization space. This dynamic adjustment mechanism ensures both security and flexibility.

[0113] Step 305: Integrate the fluctuation range constraint, the boundary condition, and the stability threshold into an optimization framework including the first function, the second function, and the third function.

[0114] In step 305, the optimization framework refers to a mathematical solution space that integrates three objective functions and three types of constraints. Its structural design directly affects the efficiency and effect of multi-objective optimization.

[0115] In the embodiment of this application, a hierarchical nested approach is used to construct the framework: the outermost layer is a parallel optimization space for three objective functions, the middle layer embeds a dynamic constraint network, and the inner layer sets variable interaction channels. This structure enables the coordinated processing of objectives and constraints.

[0116] Step 306: Establishing a two-way conflict resolution mechanism for coordinating objective conflicts in the optimization framework to form a multi-objective decision model.

[0117] In step 306, the bidirectional conflict resolution mechanism refers to a regulatory system that coordinates the competitive relationship between objectives, and dynamically balances the optimization requirements of different objectives through two paths: forward conduction and reverse feedback.

[0118] In this embodiment, when replenishment cycles conflict with storage costs, the mechanism first attempts to adjust the cost boundary conditions; if reconciliation remains unsuccessful, a weight redistribution process is initiated. A risk warning trigger mechanism is also implemented to prioritize safety when stability is threatened. This two-way adjustment ensures that the system can still output reasonable solutions in complex environments.

[0119] Here's a specific example:

[0120] During the peak season for air conditioner sales at a large e-commerce warehouse, the system constructed a multi-objective decision-making model based on the real-time data of the second embodiment. First, based on the replenishment urgency parameter of 4.4 and the benchmark replenishment cycle of 1.5 days, the replenishment cycle constraint range was calculated to be 0.34-6.6 days using the formulas of fluctuation lower limit = benchmark cycle / urgency parameter, fluctuation upper limit = benchmark cycle × urgency parameter. The benchmark cycle of 1.5 days is derived from the historical average turnover days of the product. For the storage space occupancy parameter of 90kg / m 2 With a transport load factor of 75%, the cost boundary condition of 56.25-75 yuan is generated using the formulas: Cost Lower Limit = Base Cost 50 × Space Density Factor 90 / 80 = 56.25, and Cost Upper Limit = Base Cost 50 × Load Factor 75 / 50 = 75. Based on the warehousing risk factor of 1.126 and its change of 0.15 over the last three hours, the risk threshold is set using the formula: Stability Threshold = Base Value 1.0 × Risk Factor 1.126 + Change Correction Value 0.15 × Sensitivity Factor 2 = 1.426. Sensitivity Factor 2 is determined through historical data analysis. The three objective functions, replenishment cycle f1 = 1 / 4.4 ≈ 0.227, warehousing cost f2 = 0.6 × 90 + 0.4 × 75 = 84, and risk balance f3 = 1 / (1 + 2 × 0.15) ≈ 0.769, where the weight coefficients 0.6 and 0.4 come from expert evaluation and the sensitivity coefficient 2 reflects risk tolerance, are integrated into the optimization framework. When the conflict resolution mechanism detects a conflict between the replenishment cycle target and the risk target, it prioritizes ensuring that the risk threshold of 1.426 is not exceeded.

[0121] In the embodiment of the present application, the method achieves a dynamic balance among replenishment efficiency, cost control and risk management by establishing a scientific multi-objective decision-making model, enabling the warehousing system to intelligently adapt to changes in market demand and environmental fluctuations, thereby improving the quality and reliability of operational decisions.

[0122] To further improve the coordination capability of multi-objective optimization, in some embodiments, step 306: establishing a bidirectional conflict resolution mechanism for coordinating objective conflicts in the optimization framework to form a multi-objective decision model includes:

[0123] Step 401: Establishing a feedback relationship between the replenishment cycle and the storage cost.

[0124] In step 401, the feedback relationship is a dynamic adjustment link between replenishment cycle time and storage costs. When the replenishment cycle shortens, the cost constraint is automatically relaxed. When the storage cost exceeds the limit, the replenishment frequency is adjusted accordingly, forming a two-way adjustment channel. This relationship is quantified through the target sensitivity matrix to determine the degree of mutual influence.

[0125] In this embodiment, the system first establishes the elasticity coefficient of replenishment cycle changes to warehousing costs, as well as the feedback coefficient of cost fluctuations on replenishment cycle, building a dynamic response model between the two. When it detects that optimizing the replenishment cycle target has led to a surge in costs, it automatically triggers adjustments to cost boundary conditions. Conversely, when cost constraints tighten, the replenishment cycle target weight is adjusted accordingly, forming a closed-loop regulation.

[0126] Step 402: Establishing a constraint intervention relationship between the risk balance and the feedback association relationship.

[0127] In step 402, the constraint intervention relationship refers to the mandatory regulatory effect of the risk balance objective on the preceding feedback link. When the risk coefficient exceeds the warning value, the safety objective is prioritized, temporarily freezing the optimization space for some cycles and costs. This relationship is implemented through the risk warning trigger mechanism.

[0128] In this embodiment, multiple risk factor thresholds are set. When the primary threshold is reached, an early warning signal is issued. At the intermediate threshold, the adjustment range of the replenishment cycle is restricted. At the advanced threshold, the system is forced to switch to safety-first mode. This hierarchical intervention ensures safety while minimizing interference with normal optimization.

[0129] Step 403: Establish a dynamic coupling relationship between the storage cost and the risk balance target.

[0130] In step 403, the dynamic coupling relationship refers to an indirect link between storage costs and risk balance objectives. This relationship reflects the economic balance between environmental control costs and safety maintenance expenses through the trade-off between the two. This relationship is quantified through a cost-risk conversion coefficient matrix.

[0131] In this embodiment, the corresponding relationship between storage cost input and risk coefficient changes in historical data is analyzed to establish the risk improvement coefficient brought about by unit cost input. During the optimization process, when the safety level needs to be improved, the corresponding minimum cost increment is automatically calculated and embedded into the model as a constraint.

[0132] Step 404: construct a bidirectional conflict resolution mechanism based on the feedback association relationship, the constraint intervention relationship, and the dynamic coupling relationship.

[0133] In this application, the feedback relationship is used as the basic regulation layer, the constraint intervention relationship is used as the safety protection layer, and the dynamic coupling relationship is used as the economic balance layer. Through layered integration, a complete mitigation mechanism is constructed. The system evaluates the status of each target in real time and automatically selects the optimal coordination strategy.

[0134] Step 405: In the optimization framework, the conflicts among replenishment cycle, storage cost and risk balance are coordinated in real time through the two-way conflict resolution mechanism to generate a multi-objective decision model.

[0135] In this embodiment, the model continuously monitors the achievement of the three objectives during operation. When conflicts are detected, the corresponding resolution rule base is invoked to generate an adjustment plan. Through iterative optimization, each objective is gradually approached to a state of equilibrium, ultimately outputting the optimal decision that balances efficiency, cost, and safety.

[0136] Here's a specific example:

[0137] During the peak season for air conditioner sales at a large e-commerce warehouse, the system implemented a two-way conflict resolution mechanism based on the optimization framework of Example 3. First, a feedback relationship was established between replenishment cycle and storage costs. When it was detected that the current replenishment cycle of 0.5 days resulted in costs reaching 74 yuan, approaching the upper limit of 75 yuan, the new cost upper limit was calculated using the formula: original upper limit = adjustment factor 1.1, resulting in a new cost upper limit of 82.5 yuan. The adjustment factor 1.1 was derived from regression analysis of similar historical scenarios. Simultaneously, risk monitoring indicated that the coefficient had risen to 1.38, approaching the threshold of 1.426. This triggered a constraint intervention relationship. The maximum allowable cycle was calculated using the formula: original cycle = risk buffer factor 1.2, resulting in a new cost of 0.6 days. The risk buffer factor 1.2 was determined by safety regulations. At this point, the dynamic coupling relationship determined that a cost increase was required to reduce risk. Using the formula: cost increment = risk difference × conversion factor, the required cost increase was 5 yuan. The risk difference was 1.426 - 1.38 = 0.046, and the conversion factor was the historical average of 108.7. After integrating these three relationships, the conflict resolution mechanism outputs a balanced solution: adjusting the replenishment cycle to 0.55 days, controlling costs to 79 yuan, and stabilizing the risk factor at 1.4. It also generates execution instructions, including moderately reducing replenishment frequency, increasing shelf ventilation, and relocating some products to alternate storage areas. The cycle adjustment value is a compromise between the conflicting parties, while the cost control value of 79 yuan is derived from the new upper limit of 82.5 yuan minus a 3.5 yuan risk adjustment margin. The final risk factor of 1.4 is confirmed to meet safety requirements through real-time monitoring. The entire decision-making process undergoes multiple rounds of iterative optimization to ensure that all three objectives are optimally positioned within acceptable limits.

[0138] In the embodiment of the present application, the method establishes an intelligent conflict resolution mechanism to enable the multi-objective optimization system to have adaptive coordination capabilities, maintain the rationality and reliability of decision-making in a complex and changing operating environment, and improve the overall efficiency of warehouse management.

[0139] To further improve the intelligent level of shelf load-bearing safety management, in some embodiments, step 102: generating a target product distribution adjustment strategy on the shelf based on the product weight data and in combination with preset shelf load-bearing constraints, includes:

[0140] Step 501: Identify the current load-bearing distribution status of the product weight data in the multi-layer space of the shelf.

[0141] In step 501, the current load distribution status refers to the actual load conditions in each area of ​​each shelf layer. A three-dimensional load heat map is constructed using real-time data collected by weight sensors to reflect the weight distribution characteristics of the goods in space. This status includes the location coordinates of each storage unit, the load value, and the relationship between adjacent areas.

[0142] In this embodiment, the system uses a network of embedded sensors on the shelf to acquire real-time weight data for each layer and section. Combined with a three-dimensional digital model of the shelf, it generates a position-coded load-bearing distribution matrix. A spatial clustering algorithm identifies high-density clusters, analyzes weight gradient trends across these areas, and generates a comprehensive load-bearing status assessment report.

[0143] Step 502: Compare the current load distribution state with the preset shelf load constraint conditions layer by layer to identify the over-limit area that exceeds the preset load threshold.

[0144] In step 502, the preset load threshold is a quantitative representation of the shelf load constraint, directly derived from the maximum allowable load per shelf layer specified in the constraint. For example, if the constraint specifies "≤ 500kg per layer," the load threshold is 500kg, which is used to specifically compare and determine overload conditions. The overload zone refers to the shelf space where the actual load exceeds the preset safety threshold. The identification results include core parameters such as the overload location coordinates, overload weight, and overload ratio. This zone is determined through a cross-analysis of spatial location and load data.

[0145] In this embodiment of the application, the system compares the real-time load distribution matrix with the preset shelf load standard matrix element by element, marking all over-limit units. Adjacent over-limit units are merged into continuous over-limit regions using a region growing algorithm. The total over-limit amount and average over-limit ratio of each region are calculated to generate a list of over-limit regions.

[0146] Step 503: Generate a plan for migrating goods from the over-limited area to the non-over-limited area based on the location distribution and the degree of over-limit of the over-limited area.

[0147] In step 503, the location distribution and degree of overweight areas are determined by comparing the current load distribution of each shelf layer with the preset load threshold layer by layer. The specific shelf layers and areas exceeding the threshold (location distribution) are then located, and the degree of overweight is quantified based on the percentage difference between the overweight and the threshold. Specifically, the location distribution is determined by mapping shelf pressure sensor data to three-dimensional coordinates, and the degree of overweight is calculated as (measured weight - threshold) / threshold × 100%. Non-overweight areas are areas on the shelf where the measured load is less than the preset load threshold. Physical proximity to overweight areas is not a requirement. The system uses shelf coordinates to locate all areas that meet the "measured load < threshold" requirement as candidate target areas. Regardless of whether they are adjacent to overweight areas, areas with sufficient load margin and optimal routing are prioritized. The product relocation plan is a specific operational plan to resolve overweight conditions, including a list of products to be relocated, target storage locations, relocation path planning, and execution priority ranking. This plan is generated using a spatial optimization algorithm to ensure that overweight issues are resolved with minimal handling effort.

[0148] In this embodiment, the system first calculates the theoretical migration capacity for each area based on a list of over-limit areas. It then searches the shelf model for candidate target areas that meet the load-bearing requirements. Using a path cost evaluation function, it selects the optimal migration combination and generates a preliminary migration plan that includes the source location, target location, number of items, and a suggested path. The plans are then executed in descending order of over-limit severity.

[0149] Step 504: Based on the environmental status data, the migration path and migration sequence in the product migration plan are adjusted, and the adjusted product migration plan is used as a distribution adjustment strategy.

[0150] In step 504, the migration path generates the shortest transport route based on the spatial topological relationship between the over-limit and low-load areas. The migration order is sorted from highest to lowest over-limit. Specifically, the path planning uses shelf navigation grid data to calculate the optimal path. The sequential strategy prioritizes items with an over-limit of >20%, followed by items with an over-limit of 10-20%. The adjusted item migration plan is the final execution strategy that incorporates environmental constraints. It adds environmental adaptability adjustments to the basic migration plan, including optimization measures such as path obstacle avoidance and operation time adjustment. This plan ensures the safety and feasibility of the migration operation.

[0151] In this embodiment, the system overlays environmental monitoring data onto a three-dimensional model of the rack to identify areas of high temperature, high humidity, or abnormal vibration. The system then optimizes the paths in the initial migration plan to avoid obstacles, adjusts operation times to avoid peak environmental conditions, and splits heavy load migrations into multiple batches when necessary. Ultimately, a distribution adjustment strategy document is generated, tagged with environmental adaptation indicators.

[0152] Here's a specific example:

[0153] During peak air-conditioning sales at a large e-commerce warehouse, the system detected that the top shelf load in product area A reached 580 kg, exceeding the preset threshold of 500 kg. The excess weight of Class B products, concentrated in areas X10-Y15 to X15-Y20, was calculated to be 80 kg, based on the formula: excess weight = measured weight - load threshold. Analyzing the three-dimensional shelf models, the remaining loads of 50 kg in the left section of the middle level and 30 kg in the right section of the bottom level were selected as relocation targets. A preliminary relocation plan was generated: two Class B products weighing 96 kg each were relocated in two stages, based on the degree of overload. The first stage involved relocating the heavier 48 kg to the left section of the middle level, coordinates X5-Y8 to X8-Y11, and the remaining 48 kg to the right section of the bottom level, coordinates X3-Y5 to X6-Y8. Environmental monitoring indicated that the temperature in aisle X12-Y14 had reached 38°C, exceeding the safety threshold. The system automatically adjusted the route to bypass the central aisle and scheduled the relocation operation for the cooler nighttime hours. Replenishment urgency parameter 2.0 calculated based on real-time order data, shelf space density 90kg / m 2 With a transport load factor of 75%, a multi-objective function was constructed: replenishment cycle objective f1 = 1 / 2.0 = 0.5, storage cost objective f2 = 0.6 × 90 + 0.4 × 75 = 84, where the weights 0.6 and 0.4 are derived from expert assessment, and risk balance objective f3 = 1 / (1 + 2 × 0.15) ≈ 0.77, with k = 2 being the risk sensitivity coefficient. After multiple rounds of iterative optimization, the final implementation plan was: Under the premise of ensuring a risk factor not exceeding 1.1, a distribution strategy with a replenishment cycle of 1.2 days and a storage cost of 78 units was adopted. The adjusted product migration plan was simultaneously implemented, completing the migration over two nights and initiating auxiliary cooling measures.

[0154] In the embodiment of the present application, the method realizes precise regulation of shelf load through intelligent load-bearing state identification and migration plan generation, optimizes space resource utilization while ensuring storage safety, and improves the operational reliability and management efficiency of large-scale storage systems.

[0155] To further improve the accuracy and feasibility of the product migration plan, in some embodiments, step 503: generating a product migration plan from the over-limited area to the non-over-limited area based on the location distribution and degree of over-limit of the over-limited area, includes:

[0156] Step 601: Based on the location distribution of the out-of-gauge area, determine the spatial distribution topology relationship of the out-of-gauge goods.

[0157] In step 601, the spatial distribution topology refers to the locational characteristics of out-of-gauge goods within the three-dimensional space of the shelf, including the geometry of the out-of-gauge area, the connection between adjacent areas, and the spatial distance from other functional areas. This relationship is represented by a graph structure of the digital shelf model, with nodes representing storage units and edges representing accessible transport paths.

[0158] In this embodiment, the system analyzes the coordinate data of oversized areas and constructs a weighted undirected graph with storage units as vertices and transport channels as edges. Using a graph traversal algorithm, the system identifies clusters of oversized goods and analyzes the spatial adjacency between oversized units, forming a topological network describing the distribution density of goods.

[0159] Step 602: Calculate the migration amount according to the degree of exceeding the limit in the exceeding area.

[0160] In step 602, the transfer amount refers to the total weight of goods that need to be transferred to eliminate the over-limit condition. This calculation takes into account both the absolute value of the over-limit and the safety margin to ensure a reasonable buffer space after the transfer. This value is the core basis for formulating the transfer plan. Transfer amount = over-limit weight × safety margin factor, where over-limit weight = measured load-bearing capacity minus load-bearing capacity threshold.

[0161] In this embodiment, the theoretical migration amount is calculated by multiplying the overweight data in the overweight area list by a preset safety adjustment factor. For continuous overweight areas, a regional merging calculation method is used, while discrete overweight points are calculated separately, ultimately summarizing and generating a migration amount requirement table for each area.

[0162] Step 603: Filter target areas whose bearing margin is greater than the migration amount from all areas that are not exceeded.

[0163] In step 603, load margin = load threshold - measured load. Target area screening involves selecting the most suitable storage location for the relocated goods from all candidate areas that meet the load requirements. Screening criteria include load margin compatibility, potential for optimizing space utilization, and convenient access to over-limit areas.

[0164] In this embodiment, the system traverses all areas within the shelf model that are not over-limited, calculating the degree of compatibility between the current inventory in each area and the migration requirements. Through a multi-level filtering mechanism, areas with insufficient inventory are first eliminated, and the transport path costs of the remaining candidate areas are then evaluated. Finally, a list of target areas is generated, sorted by their suitability.

[0165] Step 604: Generate a product migration plan based on the spatial distribution topological relationship and the target area.

[0166] In this embodiment, a heuristic search algorithm is used to generate an initial migration path based on the topological network and the target area list. The path is then verified for feasibility using shelf operation specifications, and the loading sequence is optimized based on product attributes. Ultimately, a detailed migration plan document is generated, including time scheduling, equipment scheduling, and staffing.

[0167] Here's a specific example:

[0168] During peak air conditioner sales at a large e-commerce warehouse, the system, based on the over-limit analysis results from Example 5, identified a Class B product in the top-level X10-Y15 to X15-Y20 zones that exceeded the limit by 80kg. First, through a graph structure analysis of the three-dimensional shelf model, the system determined that the five refrigerators formed a 2×3 matrix with the core over-limit nodes concentrated in the X12-Y16 to X14-Y18 zones. To calculate the migration amount based on the degree of over-limit, the system used the formula: migration amount = over-limit weight × safety factor. For an over-limit weight of 80kg, the safety factor was set to a standard value of 1.2, resulting in a theoretical migration amount of 96kg. The system traversed the shelf database to select areas within the limit and calculated the scores for each candidate zone using the formula: zone fit = margin matching × path coefficient. The middle left zone, with a margin of 100kg, scored 0.85, while the bottom right zone, with a margin of 80kg, scored 0.72. The middle left zone was selected as the primary target. Generate a migration plan based on topological relationships: sort the over-limit areas by distance, and prioritize migrating the three refrigerators in the center with a total weight of 72 kg to the middle left area X5-Y8 to X8-Y11. The remaining 24 kg will be supplemented by migrating a refrigerator at the edge.

[0169] In the embodiment of the present application, the method generates a migration plan through systematic spatial analysis and intelligent matching mechanism, which not only effectively solves the problem of shelf overruns, but also fully considers various constraints in actual operations, thereby improving the safety and execution efficiency of warehouse scheduling.

[0170] To further improve the optimization efficiency and quality of inventory decision solutions, in some embodiments, step 104: outputting an optimal inventory distribution decision solution in an interactive iterative manner based on the multi-objective decision model includes:

[0171] Step 701: Based on the multi-objective decision model, an initial decision solution set is generated through an initialization algorithm as input for the first interactive iteration.

[0172] In step 701, the initial decision solution set refers to the group of feasible solutions from the first iteration of the multi-objective optimization process. It contains several groups of candidate solutions that satisfy the basic constraints, each corresponding to a specific combination of replenishment cycle, storage cost, and risk balance objectives. This set is generated using a spatial sampling algorithm to ensure a diverse and extensive distribution of solutions.

[0173] In an embodiment of the present application, the system generates an initial solution set using the Latin hypercube sampling method based on the variable range of the multi-objective decision model. Each solution contains parameter values ​​of three dimensions, and the solutions that meet the hard constraints such as shelf load-bearing capacity and environmental safety are retained through feasibility testing to form a representative initial decision solution set. Specific process: Based on the multi-objective decision model, an initial decision solution set is generated through interactive iterative calculation, specifically including: randomly generating feasible solutions that meet the shelf load-bearing constraints and environmental state constraints as the initial population, and screening non-inferior solutions through non-dominated sorting and crowding distance calculation to form a decision solution set. For example, during the e-commerce promotion period, a warehouse initialized 100 groups of feasible solutions for home appliances, each group of solutions containing different shelf distribution schemes (such as a combination of centralized storage areas for refrigerators and decentralized storage of washing machines). After evaluating the replenishment cycle, storage cost and risk coefficient of each scheme through the model, 30 groups of non-dominated Pareto optimal solutions are retained to form a decision solution set.

[0174] Step 702: In each interactive iteration, the real-time change of the environmental status data and the distribution adjustment strategy are used as dynamic constraint conditions.

[0175] In step 702, dynamic constraints are optimization limits that are updated in real time as the environment changes and strategies adjust. They transform external changes into adjustments to model parameter boundaries, ensuring that the optimization direction aligns with actual requirements. These constraints are applied to the objective function space via a constraint propagation mechanism.

[0176] In an embodiment of the present application, the system continuously monitors environmental sensor data and execution feedback of the distribution adjustment strategy. When changes in key indicators such as temperature, humidity, or shelf vibration are detected, the risk constraint boundaries are automatically updated. According to the implementation progress of the migration plan, the value range of the space occupancy parameter is dynamically adjusted to form a set of constraint conditions that adapt to the latest status.

[0177] Step 703: Filter out a subset of non-inferior solutions that meet the dynamic constraint conditions from the decision solution set.

[0178] In step 703, the non-inferior solution subset is a set of candidate solutions that are not completely dominated by other solutions in the current iteration round and that simultaneously meet the dynamic constraints and Pareto optimality requirements. This subset is extracted through a multi-dimensional screening mechanism and represents the candidate group for the current optimal solution.

[0179] In this embodiment of the application, the system matches each candidate solution in the decision solution set with the dynamic constraints, eliminating solutions that violate the constraints. The remaining solutions are then non-dominated and selected to form a non-inferior solution subset with top-ranked and evenly distributed solutions, ensuring the quality and diversity of the solutions.

[0180] Step 704: Determine whether the solution space variation of the non-inferior solution subset is less than a preset convergence threshold.

[0181] In step 704, the solution space variation refers to the degree of difference in the distribution of non-inferior solution subsets in the objective function space between adjacent iterations, reflecting the convergence state of the optimization process. This indicator is calculated using the solution set distance metric function and is used to determine the termination condition.

[0182] In this embodiment, the average distance between the current non-inferior solution subset and the previous solution set in each target dimension is calculated to comprehensively evaluate the degree of evolution of the solution space. When the change is continuously below the threshold, the optimization process is considered to have reached a stable state.

[0183] Step 705: If not, the non-inferior solution subset is expanded to obtain a non-inferior solution set, and the non-inferior solution set is used as the decision solution set to repeat the iterative process until the change in the solution space of the non-inferior solution subset is less than the preset convergence threshold or the maximum number of iterations is reached, thereby generating the optimal inventory distribution decision plan.

[0184] In step 705, the expansion process refers to the intelligent expansion of the non-inferior solution subset to maintain the optimization momentum. It generates new solutions through mutation and recombination strategies to avoid premature convergence of the algorithm. This process balances the breadth and depth of the search.

[0185] In this embodiment, cluster analysis is performed on a subset of non-inferior solutions. New solutions are generated in sparse regions through Gaussian mutation. High-quality solutions in dense regions are cross-combined to produce improved offspring. After merging these new solutions with the atomic set, feasibility verification and non-dominated screening are performed to form a new set of even higher-quality decision solutions for further iteration.

[0186] Here's a specific example:

[0187] During the peak season for air conditioner sales at a large e-commerce warehouse, the system initiated an interactive iterative optimization based on the multi-objective decision-making model of Example 1. First, Latin hypercube sampling was used to generate 50 initial solutions. Each solution contained three parameters: replenishment cycle, storage cost, and risk factor. The replenishment cycle range was determined by the formula: lower limit = baseline cycle 1.5 days / replenishment urgency parameter 2.0 = 0.75 days, and upper limit = baseline cycle 1.5 days × replenishment urgency parameter 2.0 = 3.0 days. During the first iteration, environmental monitoring indicated that the local temperature had risen to 38°C. After adjusting the risk threshold using the formula (original threshold 1.1 × temperature impact coefficient 1.15 ≈ 1.27), the constraints were reset. After screening, 12 non-inferior solutions were retained, and the calculated solution space variance was 0.28. The system then performed Gaussian mutation and simulated binary crossover on the non-inferior solutions, expanding the solution space to 30 solutions and continuing the iteration. During the second iteration, the shelf migration strategy was updated to free up new storage space. The storage cost cap was relaxed to 82.5 yuan, 1.1 times the original value of 75 yuan. Eighteen non-inferior solutions were identified, and the variance was reduced to 0.18. During the third iteration, the risk coefficient change rate monitoring value dropped back to 0.1. The judgment criteria were adjusted using the formula: convergence threshold = base value 0.15 × rate of change attenuation coefficient 0.9 = 0.135. Finally, the variance of 15 solutions converged with a variance of 0.12. The solution with the largest crowding distance was selected as the output solution: a replenishment cycle of 1.1 days, calculated using the formula: cycle = base cycle 1.5 days / optimization coefficient 1.36. The storage cost of 79 yuan was within the relaxed constraints, and the risk coefficient of 1.08 met the new threshold. Based on this, the system generated a final implementation plan, including a specific schedule for three daily replenishments, an adjusted shelf load distribution map, and temperature control measures to ensure efficient and safe inventory management during peak promotional periods.

[0188] In the embodiment of the present application, the method uses an intelligent interactive iterative mechanism to enable inventory decision-making solutions to dynamically adapt to environmental changes and business needs, thereby improving optimization efficiency while ensuring the quality of the solution, and providing reliable multi-objective decision support for complex warehousing scenarios.

[0189] Figure 2 A structural diagram of an inventory distribution decision system based on dynamic multi-objective optimization provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the system includes:

[0190] The acquisition module 21 is used to obtain the real-time order data, product weight data, historical demand data and environmental status data of the target product when the real-time demand fluctuation of the e-commerce warehouse is greater than the preset fluctuation.

[0191] The generating module 22 is configured to generate a distribution adjustment strategy for the target commodity on the shelf based on the commodity weight data and in combination with a preset shelf load-bearing constraint condition.

[0192] The construction module 23 is used to combine the distribution adjustment strategy, the real-time order data, the historical demand data and the environmental status data to construct a multi-objective decision model with the optimization goals of minimizing the replenishment cycle, minimizing the storage cost and optimizing the risk balance.

[0193] The output module 24 is used to output the optimal inventory distribution decision plan in an interactive iterative manner based on the multi-objective decision model.

[0194] Figure 2 The inventory distribution decision system based on dynamic multi-objective optimization can be executed Figure 1 The implementation principles and technical effects of the dynamic multi-objective optimization-based inventory distribution decision-making method described in the illustrated embodiment are not further elaborated. The specific manner in which each module and unit performs operations in the dynamic multi-objective optimization-based inventory distribution decision-making system described in the aforementioned embodiment has been described in detail in the related embodiments and will not be further elaborated here.

[0195] In one possible design, Figure 2 The inventory distribution decision system based on dynamic multi-objective optimization of the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0196] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0197] The processing component 32 performs the above Figure 1 The embodiment provides an inventory distribution decision-making method based on dynamic multi-objective optimization.

[0198] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0199] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random-access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0200] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0201] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0202] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0203] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0204] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The illustrated embodiment is an inventory distribution decision method based on dynamic multi-objective optimization.

[0205] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0206] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0207] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

[0208] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for inventory distribution decision-making based on dynamic multi-objective optimization, characterized in that: include: Obtain real-time order data, product weight data, historical demand data, and environmental status data for target products; Based on the product weight data and the preset shelf load-bearing constraints, a distribution adjustment strategy for the target product on the shelf is generated; Combining the distribution adjustment strategy, the real-time order data, the historical demand data, and the environmental status data to construct a multi-objective decision model with the optimization goals of minimizing replenishment cycle time, minimizing storage costs, and optimizing risk balance; Based on the multi-objective decision model, an optimal inventory distribution decision plan is outputted in an interactive iterative manner.

2. The method according to claim 1, characterized in that The distribution adjustment strategy, the real-time order data, the historical demand data, and the environmental status data are combined to construct a multi-objective decision model with the optimization goals of minimizing replenishment cycle, minimizing storage costs, and optimizing risk balance, including: generating a migration instruction according to the distribution adjustment strategy, and converting the migration instruction into a storage space occupancy parameter; Calculate and generate a replenishment urgency parameter based on the real-time order data and the historical demand data; Based on the environmental status data and the shelf stability change trend, a storage risk coefficient is generated; A multi-objective decision model is established with the replenishment urgency parameter, the storage space occupancy parameter and the storage risk coefficient as variables.

3. The method according to claim 2, characterized in that The multi-objective decision model is established with the replenishment urgency parameter, the storage space occupancy parameter, and the storage risk coefficient as variables, including: A first function representing minimization of replenishment cycle is constructed based on the replenishment urgency parameter, a second function representing minimization of storage cost is constructed based on the storage space occupancy parameter, and a third function representing optimization of risk balance is constructed based on the storage risk coefficient; Constructing a fluctuation range constraint condition for the replenishment cycle based on the replenishment urgency parameter and the benchmark replenishment cycle; Based on the storage space occupancy parameters, generating boundary conditions for storage costs; According to the dynamic change rate of the warehousing risk coefficient, a stability threshold of risk balance is set; Integrating the fluctuation range constraint, the boundary condition, and the stability threshold into an optimization framework including the first function, the second function, and the third function; A two-way conflict resolution mechanism for coordinating conflicting objectives is established in the optimization framework to form a multi-objective decision model.

4. The method according to claim 3, characterized in that The two-way conflict resolution mechanism for coordinating conflicting objectives is established in the optimization framework to form a multi-objective decision model, including: Establishing a feedback correlation relationship between the replenishment cycle and the storage cost; Establishing a constraint intervention relationship of the risk balance on the feedback association relationship; Establishing a dynamic coupling relationship between the storage cost and the risk balance target; Building a bidirectional conflict resolution mechanism based on the feedback association relationship, the constraint intervention relationship, and the dynamic coupling relationship; In the optimization framework, the conflicts among replenishment cycle, storage cost and risk balance are coordinated in real time through the two-way conflict resolution mechanism to generate a multi-objective decision model.

5. The method according to claim 1, wherein The method of generating a distribution adjustment strategy for target products on the shelf based on the product weight data and in combination with preset shelf load-bearing constraints includes: Identify the current load distribution status of the product weight data in the multi-layer space of the shelf; Comparing the current load distribution state with the preset shelf load constraints layer by layer to identify overload areas exceeding the preset load threshold; Generating a plan for migrating goods from the over-limited areas to non-over-limited areas based on the location distribution and degree of over-limit of the over-limited areas; In combination with the environmental status data, the migration path and migration order in the product migration plan are adjusted, and the adjusted product migration plan is used as a distribution adjustment strategy.

6. The method according to claim 5, characterized in that Generating a plan for migrating goods from the over-limit area to the non-over-limit area according to the location distribution and the degree of over-limit of the over-limit area includes: Determining a spatial distribution topology of out-of-gauge goods based on the location distribution of the out-of-gauge areas; Calculating the migration amount according to the degree of exceeding the limit in the exceeding limit area; Select a target area whose bearing margin is greater than the migration amount from all areas that are not exceeded; Based on the spatial distribution topological relationship and the target area, a product migration plan is generated.

7. The method according to claim 1, characterized in that The method of outputting an optimal inventory distribution decision plan based on the multi-objective decision model in an interactive iterative manner includes: Based on the multi-objective decision model, an initial decision solution set is generated through an initialization algorithm as input for the first interactive iteration; In each interactive iteration, the real-time change of the environmental state data and the distribution adjustment strategy are used as dynamic constraints; Screening out a subset of non-inferior solutions that satisfy the dynamic constraint conditions from the decision solution set; Determining that a change in the solution space of the non-inferior solution subset is less than a preset convergence threshold; If not, the non-inferior solution subset is expanded to obtain a non-inferior solution set, and the non-inferior solution set is used as the decision solution set to repeat the iterative process until the solution space change of the non-inferior solution subset is less than the preset convergence threshold or the maximum number of iterations is reached, and the optimal inventory distribution decision plan is generated.

8. An inventory distribution decision system based on dynamic multi-objective optimization, characterized by: include: An acquisition module is used to obtain real-time order data, product weight data, historical demand data, and environmental status data of target products when the real-time demand fluctuation of e-commerce warehouses exceeds a preset fluctuation; A generation module, configured to generate a distribution adjustment strategy for target commodities on the shelves based on the commodity weight data and in combination with preset shelf load-bearing constraints; A construction module is used to combine the distribution adjustment strategy, the real-time order data, the historical demand data, and the environmental status data to construct a multi-objective decision model with the optimization goals of minimizing replenishment cycle time, minimizing warehousing costs, and optimizing risk balance; The output module is used to output the optimal inventory distribution decision plan in an interactive iterative manner based on the multi-objective decision model.

9. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an inventory distribution decision method based on dynamic multi-objective optimization as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, the inventory distribution decision method based on dynamic multi-objective optimization according to any one of claims 1 to 7 is implemented.

Images