An intelligent warehouse dynamic goods location allocation method based on goods location heat analysis

Abstract

This invention discloses a method for dynamic warehouse location allocation based on location heat analysis, belonging to the field of intelligent warehouse management technology. The method includes: Step 1: collecting location coordinates, product attributes, historical and current order data; Step 2: calculating the real-time demand heat of products and shelf location heat based on current orders, and coupling them to obtain location coupling heat; Step 3: constructing a dynamic equilibrium model with the goal of maximizing total location coupling heat and optimizing the estimated picking path, and solving to output a location allocation scheme; Step 4: executing allocation and picking tasks, collecting actual operation data, and feeding back to correct the heat calculation model to achieve closed-loop optimization. This invention solves the coupling conflict between static location allocation and dynamic order picking, realizes the linkage decision-making between location optimization and path planning, and can adapt to order fluctuations and changes in the operating environment, significantly improving warehouse operation efficiency.

Description

Technical Field

[0001] This invention relates to the field of intelligent warehouse management and logistics optimization technology, and in particular to an intelligent warehouse dynamic storage location allocation method based on storage location heat analysis. Background Technology

[0002] With the rapid development of e-commerce and flexible manufacturing systems, modern warehousing centers face orders characterized by fragmentation, real-time processing, and high concurrency. Traditional warehouse location allocation methods mainly fall into two categories: the first is allocation strategies based on static historical data, such as placing high-frequency goods near the shipping outlet or in locations with short picking paths based on the frequency of goods entering and leaving the warehouse (turnover rate); the second is allocation strategies based on fixed rules, such as first-in, first-out (FIFO) and proximity-based storage. These methods are effective in environments with stable order patterns and strong predictability.

[0003] However, in today's increasingly complex warehouse operations, the aforementioned existing technologies have revealed many inherent flaws, making it difficult to meet the demands for high-efficiency and low-cost operations. Specific problems include: First, static allocation strategies based on long-term historical data lack real-time responsiveness. Warehouse order activity changes dynamically over time; for example, demand for certain items can surge rapidly in a short period due to promotional activities, seasonal factors, or trending events. Static strategies cannot detect these real-time changes, leading to currently popular items potentially being stored in inefficient locations far from the picking area, severely slowing down overall picking efficiency. Second, existing location allocation and order picking path planning are typically two independent, sequential optimization processes. Location allocation often only considers the attributes of the goods themselves (such as size and weight) and general turnover rates, ignoring the direct impact of specific order combinations on the picker's walking path. A location that is "highly popular" in static analysis may, when faced with a specific batch of orders, result in a more circuitous and longer picking path due to its dispersed location from other selected locations. This coupling conflict between location allocation and order picking at the actual execution level has not been fully considered and resolved by existing technologies. Finally, most methods do not consider the wave-like nature and dynamic congestion of warehousing operations. In reality, orders are released to pickers in batches (waves), with multiple orders within a wave. The picking operations of the current wave will change the inventory of goods on the shelves and may cause temporary congestion of personnel or equipment in certain areas. These real-time states will affect the optimality of location allocation for the next wave. Existing technologies lack a closed-loop mechanism for continuous perception and dynamic adjustment during operation execution.

[0004] Therefore, there is an urgent need for an intelligent dynamic storage location allocation method that can deeply integrate real-time order demand, physical characteristics of storage locations, and dynamic operational status, and can adaptively and collaboratively optimize during continuous warehouse operations, so as to fundamentally improve the agility and efficiency of warehouse operations. Summary of the Invention

[0005] To achieve the above objectives, this invention provides an intelligent warehouse dynamic storage location allocation method based on storage location heat analysis, comprising the following steps:

[0006] Step 1: Collect warehouse data, which includes warehouse location coordinates, product physical attributes, historical order data, and current order pool data containing unexecuted outbound orders and goods to be received.

[0007] Step 2: Perform real-time location heat analysis based on the current order pool data. The real-time location heat analysis includes calculating the real-time demand heat of goods, the shelf location heat, and the location coupling heat that combines the real-time demand heat of goods and the shelf location heat.

[0008] Step 3: Make a location allocation decision based on the dynamic equilibrium model. The dynamic equilibrium model takes the total location coupling heat between the goods to be allocated and the available locations and the prediction results of future picking paths as the optimization objectives, solves and outputs the location allocation scheme.

[0009] Step 4: Execute the location allocation plan and picking wave tasks, collect actual picking data, dynamically update the location coupling heat calculation method based on the actual picking data, and return to Step 2 to process the new current order pool data.

[0010] Preferably, step 1 specifically includes:

[0011] Step 1.1: Collect the spatial coordinate data of all storage locations and assign a three-dimensional coordinate to each storage location. The origin of the three-dimensional coordinate is the warehouse shipping point. The first axis of the three-dimensional coordinate represents the straight-line horizontal distance from the center point of the storage location to the shipping point. The second axis of the three-dimensional coordinate represents the aisle number where the storage location is located. The third axis of the three-dimensional coordinate represents the floor number where the storage location is located.

[0012] Step 1.2: Collect physical attribute data for all goods in stock. The physical attribute data includes the length, width, height and weight of the goods. The length, width and height data are obtained by measuring the external dimensions of the goods' packaging boxes, and the weight data is obtained by weighing.

[0013] Step 1.3: Collect all executed order data within the set historical time period from the warehouse management system database. The executed order data includes the order number, the item number of each item in the order, the outbound quantity of each item, and the order creation timestamp.

[0014] Step 1.4: Collect current order pool data in real time. The current order pool data includes two parts: the first part is all inbound goods records in the warehouse management system with the status of "unassigned location". Each inbound goods record includes the goods number and the quantity to be put on the shelf; the second part is all outbound order records in the warehouse management system with the status of "picking not started". Each outbound order record includes the order number and the set of goods numbers and outbound quantities required for the order.

[0015] Preferably, in step 2, the process of calculating the real-time demand for goods includes:

[0016] Step 2.1.1: For each unique product number in the current order pool data, iterate through all outbound order records that have not started picking, accumulate the total number of outbound orders in which the product number appears in all outbound orders, and obtain the real-time outbound frequency of the product.

[0017] Step 2.1.2: For the same product number in the current order pool data, iterate through all inbound product records with "unassigned storage locations", accumulate the total number of products waiting to be put on the shelves corresponding to that product number, and obtain the real-time total number of products waiting to be put on the shelves.

[0018] Step 2.1.3: Calculate the real-time demand heat value of goods by weighting and summing the real-time outbound frequency and the real-time total amount to be put on the shelves. The outbound weight coefficient and the shelf weight coefficient used for weighting and summing are preset by the warehouse manager according to the warehouse's operation strategy of prioritizing outbound or shelf-based operations.

[0019] Preferably, in step 2, the process of calculating the heat of the shelf location includes:

[0020] Step 2.2.1: For each storage location, read the spatial coordinate data of that location and extract the first axis coordinate value as the basic value of the horizontal distance;

[0021] Step 2.2.2: Based on the lane number where the cargo location is located, i.e. the second axis coordinate value, query the preset lane conversion coefficient table to convert the lane number into a distance equivalent value. The distance equivalent value corresponding to different lane numbers reflects the detour path cost required to enter different lanes from the delivery point.

[0022] Step 2.2.3: Based on the floor height number of the storage location, i.e. the third axis coordinate value, look up the preset floor height conversion coefficient table to convert the floor height number into a height equivalent value. The height equivalent value reflects the difference in time and operational difficulty of storage locations of different floor heights when accessed manually or mechanically.

[0023] Step 2.2.4: Input the horizontal distance base value, the distance equivalent value, and the height equivalent value into a standardization function. After normalizing the three input values, the standardization function performs linear weighting according to preset distance weight, aisle weight, and floor height weight, and outputs a shelf location heat value that represents the overall ease of access to the storage location. The shelf location heat value is inversely proportional to the ease of access.

[0024] Preferably, in step 2, the process of calculating the coupling heat of the cargo location specifically involves:

[0025] Step 2.3: For each item to be allocated or stored in the current order pool data, and for each storage location in the warehouse, divide the real-time demand heat value of the item calculated in Step 2.1.3 by the shelf location heat value of the storage location calculated in Step 2.2.4. The resulting quotient is defined as the storage location coupling heat value between the item and the storage location. The higher the real-time demand heat value of the item and the lower the shelf location heat value, the higher the storage location coupling heat value, indicating that the overall benefits of placing the item in the storage location are higher.

[0026] Preferably, step 3 specifically includes:

[0027] Step 3.1: Construct a dynamic storage location allocation set. Extract all the product numbers to be allocated and their corresponding quantities to be put on the shelves from the "unallocated storage locations" in the current order pool data to form a set of products to be allocated. From the real-time inventory status of the warehousing system, filter out all the idle storage locations with zero current inventory and the non-idle storage locations with inventory below the preset transfer trigger threshold to form a set of available storage locations. The transfer trigger threshold is set according to the average outbound batch of the products.

[0028] Step 3.2: Establish a dynamic equilibrium allocation model. The objective function of the dynamic equilibrium allocation model is to maximize the total expected benefit. The formula for calculating the total expected benefit is: Total expected benefit = α × Total location coupling heat + β × Path optimization factor; where α and β are balance coefficients, and the sum of α and β is 1; the total location coupling heat is the sum of the location coupling heat values ​​of all corresponding locations after all the goods to be allocated are placed in available locations according to the allocation scheme; the calculation process of the path optimization factor includes: simulating and generating the next picking wave based on the urgency and number of outbound orders that are currently "not yet picked", determining the set of locations that each order needs to access in the simulated picking wave based on the allocation scheme, and using the nearest neighbor algorithm to calculate the estimated total length of the picking path to access the set of locations. The path optimization factor is the reciprocal of the estimated total length of the picking path.

[0029] Preferably, in step 3, the process of solving the dynamic equilibrium allocation model using a genetic algorithm includes:

[0030] Step 3.3.1: Encoding. Encode the location allocation scheme into a chromosome. The length of the chromosome is equal to the total number of items in the set of items to be allocated. Each gene position of the chromosome corresponds to a specific item to be allocated. The value of the gene is the location number in the set of available locations to which the item is allocated.

[0031] Step 3.3.2: Initialize the population by randomly generating a set number of chromosomes to form the initial population. Each chromosome represents a cargo space allocation scheme.

[0032] Step 3.3.3: Fitness assessment. Based on the objective function of the dynamic equilibrium allocation model defined in Step 3.2, calculate the total expected benefit of the allocation scheme represented by each chromosome in the population, and use the total expected benefit as the fitness value of the chromosome.

[0033] Step 3.3.4: Selection operation, using roulette wheel selection method, selects the parent chromosomes for reproduction according to the fitness value of the chromosomes in proportion;

[0034] Step 3.3.5: Crossover operation. Pair the selected parent chromosomes together and perform a single-point crossover operation with a set crossover probability. Exchange the gene segments after the crossover point of the paired chromosomes to generate new offspring chromosomes.

[0035] Step 3.3.6: Mutation operation. For all offspring chromosomes generated after crossover, the location number of a certain gene locus on the chromosome is randomly changed with a set mutation probability. The new value is randomly selected from the available location set.

[0036] Step 3.3.7: Iteration. The offspring chromosomes generated through selection, crossover, and mutation are used as a new population. Steps 3.3.3 to 3.3.6 are repeated until the preset maximum number of iterations is reached or the improvement of the optimal fitness value over multiple consecutive generations is less than the stability threshold.

[0037] Step 3.3.8: Decode the output. After the iteration terminates, decode the chromosome with the highest fitness in the population into a mapping relationship between the specific product number and the storage location number, and output it as the final storage location allocation scheme.

[0038] Preferably, step 4, the process of executing the location allocation plan and picking wave tasks, includes:

[0039] Step 4.1: The warehouse control system receives the location allocation plan output in step 3.3.8, generates detailed put-away task instructions and transfer task instructions, and sends them to automated warehousing equipment or displays them to warehouse operators; at the same time, the order management system divides a specific picking wave from the current "not yet picked" outbound orders according to the preset wave division rules. The wave division rules include the latest completion time window of the order, the overlap of goods in the order, and the upper limit of the total volume or total weight of a single wave.

[0040] Step 4.2: Warehouse operators or automated picking equipment execute the put-away task instructions, transfer task instructions and picking wave tasks; during the execution of picking wave tasks, the coordinates are recorded at fixed time intervals or key nodes through the positioning terminal worn by the personnel or the positioning system built into the equipment, forming actual picking path trajectory data, and automatically recording the system timestamp when each item is picked.

[0041] Preferably, step 4, the process of dynamically updating the calculation method of location coupling heat based on actual picking data, includes:

[0042] Step 4.3.1: Path efficiency evaluation. After the picking wave task is completed, the actual total walking distance and total operation time of the wave task are calculated based on the collected actual picking path trajectory data to obtain the actual average walking speed.

[0043] Step 4.3.2: Prediction correction. The actual average walking speed is compared with the preset standard walking speed used in step 3.2 when calculating the path optimization factor. The path prediction correction coefficient is calculated. The correction coefficient is equal to the actual average walking speed divided by the preset standard walking speed.

[0044] Step 4.3.3: Model parameter update. The calculated path prediction correction coefficient is applied to the subsequent calculation of the shelf location heat value. The specific method is as follows: In the normalization function described in step 2.2.4, the distance weight, aisle weight and floor height weight used to calculate the shelf location heat value are multiplied by the path prediction correction coefficient and then normalized. The updated weights are used for the shelf location heat calculation in the next round of step 2.

[0045] Preferably, the method further includes a cold start process for initial model training, which is executed when the method is first applied or when there is a significant change in the warehouse layout, specifically including:

[0046] Step A: Load complete historical order data and historical storage location data for a specific historical period;

[0047] Step B: Using the process from Steps 2 to 4, simulate the dynamic location allocation and picking for each day of the historical period in a simulated playback manner. However, in Step 4.3.3, update the model parameters using the path prediction correction coefficients obtained from the simulation.

[0048] Step C: After completing the simulation playback of the entire historical cycle, the final stabilized model parameters, including the outbound weight coefficient and the shelving weight coefficient in step 2.1.3, the distance weight, aisle weight and floor height weight in step 2.2.4, and the balance coefficients α and β in step 3.2, are saved as the initial default parameters of the system for subsequent real online dynamic storage location allocation.

[0049] The beneficial effects of this invention are:

[0050] 1. Existing technologies treat location allocation and picking route planning as two independent or sequential processes, potentially leading to detours in actual picking. This invention creatively introduces the core indicator of "location coupling heat," dynamically linking the real-time demand heat of goods with the physical location heat of locations (e.g., calculating the ratio of the real-time demand heat value of the goods to the shelf location heat value). This embeds a comprehensive evaluation of the popularity of goods and the advantages and disadvantages of their locations into the allocation decision. More importantly, when establishing the dynamic equilibrium allocation model, it not only maximizes the total location coupling heat but also simultaneously introduces the path prediction for a specific picking wave in the future (e.g., using the nearest neighbor algorithm to calculate the estimated total picking path length), and integrates the path optimization objective into the same objective function for solution. This ensures that the final location allocation scheme can prioritize the allocation of high-demand goods to easily accessible locations while effectively shortening the resulting actual picking walking distance, fundamentally achieving synergistic optimization of "storage" and "picking."

[0051] 2. Traditional static strategies or fixed-cycle adjustment strategies cannot respond in real time to the instantaneous changes in order hotspots. The technical solution of this invention constructs a complete closed loop of "real-time analysis - balanced decision-making - task execution - feedback update" (e.g., collecting actual picking path data and updating the heat calculation model parameters after execution in step 4), giving the system dynamic rebalancing capabilities. Specifically, after each round of allocation decision and guiding the completion of actual picking operations, the system automatically corrects the key parameters of the location heat calculation model by comparing the estimated path efficiency with the actual path efficiency (e.g., calculating the ratio of the actual average walking speed to the preset standard walking speed to obtain the path prediction correction coefficient). This correction coefficient is used to adjust the distance weight, aisle weight, and floor height weight. This feedback mechanism enables the system to perceive and learn actual conditions such as temporary aisle congestion and fluctuations in operational efficiency, and reflects these in the next round of allocation decisions. This allows warehouse management strategies to continuously align with real-world operational scenarios and possess strong adaptability to dynamic factors such as promotional activities and seasonal demand changes.

[0052] 3. This invention ultimately translates the optimization objective into quantifiable improvements in operational efficiency. From calculating the real-time demand intensity of goods based on inbound / outbound needs, to quantitatively assessing the storage and retrieval costs of locations considering multiple factors such as distance, aisle, and floor height, and then using heuristic algorithms (such as genetic algorithms) to solve a global optimization model that balances demand intensity and path, the output of this complete technical chain is a theoretically superior location allocation scheme. The implementation effect of this scheme is reflected in the following: high-demand goods are more likely to be allocated to low-cost storage and retrieval locations, directly reducing the single picking journey; at the same time, since the allocation scheme has pre-considered the compactness optimization of subsequent picking paths, it effectively reduces the picking walking distance and operation time of a single order and even the entire wave of orders. Compared with traditional static methods, the method of this invention can achieve technical effects such as a 13.1% reduction in picking walking distance and a 14.7% increase in operational efficiency, verifying the practicality and significant progress of its technical solution. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in this invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0054] Figure 1 This is a flowchart of the steps of the method of the present invention;

[0055] Figure 2 This is a flowchart of step 1 of the method of the present invention;

[0056] Figure 3This is a flowchart of step 2 of the method of the present invention for calculating the real-time demand heat of goods. Detailed Implementation

[0057] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should also be noted that, to make the embodiments more comprehensive, the following embodiments are the best and preferred embodiments, and those skilled in the art can use other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0058] Please see Figures 1-3 This invention provides an intelligent warehousing dynamic location allocation method based on location heat analysis. The warehouse management system first collects and integrates the spatial coordinate data of all locations. This coordinate data is used as the origin at the warehouse shipping point to construct a three-dimensional coordinate system including horizontal distance, aisle number, and floor height number. Simultaneously, the system collects the physical attribute data of all goods in storage, including length, width, height, and weight data obtained through measuring equipment. Furthermore, the system extracts all executed order data within a set historical time period from the database, including order details and timestamps.

[0059] Finally, the system captures the current dynamic order pool data in real time. This data consists of inbound goods records with a status of "unassigned location" and outbound order records with a status of "picking not started." This step provides a unified, real-time, and multi-dimensional data foundation for subsequent analysis, ensuring the accuracy and timeliness of decision-making.

[0060] In one possible implementation, the spatial coordinate data of the storage location is achieved by assigning a unique three-dimensional coordinate system to each storage unit, with the origin fixed at the main outbound port of the warehouse. The first axis of the coordinate system records the straight-line horizontal distance from the center point of the storage location to the origin, in meters. The second axis represents the aisle number, with different aisles represented by an integer sequence. The third axis represents the shelf height number, with an integer sequence representing the shelf level starting from the ground. The physical attribute data of the goods is automatically acquired and entered into the system when the goods are first received into the warehouse, through linkage with weighing and dimensional measurement equipment.

[0061] Historical order data is structured data exported from the order history table of the warehouse management system, filtered according to time range. Each record contains at least the order number, the item number, the quantity, and the creation time. Current order pool data is obtained by real-time querying of a specific status table in the database, ensuring synchronized awareness of inbound and outbound demand. These specific operations are all existing conventional technologies in the field of warehouse management, providing the necessary data infrastructure for realizing this invention.

[0062] In one possible implementation, firstly, all outbound order records in the current order pool that are "not yet picked" are iterated through. For each unique item number, the total outbound quantity appearing in all such orders is summed. This sum represents the real-time outbound frequency, directly reflecting the intensity of the urgent outbound demand for that item. Secondly, for the same item number, all inbound item records that are "not assigned a location" are iterated through, and the total quantity awaiting shelving is summed to obtain the real-time total quantity awaiting shelving, reflecting the upcoming inventory space demand for that item. Finally, the real-time outbound frequency and the real-time total quantity awaiting shelving are weighted and summed to calculate the final real-time demand heat value for the item.

[0063] The outbound and shelving weighting coefficients used in the weighting process are pre-configured by the system administrator based on the warehouse's operational strategy, prioritizing either outbound efficiency or shelving smoothness. For example, during peak e-commerce promotional periods, the outbound weight can be set to 0.9 and the shelving weight to 0.1, strongly guiding outbound efficiency. This calculation process dynamically captures the real-time demand for goods in both inbound and outbound operations, providing accurate demand-side input for subsequent optimal storage location.

[0064] In one possible implementation, firstly, the three-dimensional coordinates of the storage location are read. The value of the first axis of the coordinate system is directly used as the base value for the horizontal distance. Secondly, based on the lane number on the second axis of the coordinate system, a preset lane conversion coefficient table is consulted. This table converts the abstract lane number into a specific distance equivalent value. The distance equivalent value reflects the tortuousness of the path required to reach different lanes from the outlet. For example, the coefficient for lanes located at the edge of the warehouse might be 1.0, while the coefficient for lanes located in the central area, which need to traverse other lanes, might be 1.5.

[0065] Next, based on the floor height number on the third axis of the coordinate system, a pre-defined floor height conversion coefficient table is consulted to convert the floor height number into a height equivalent value. The height equivalent value reflects the impact of different floor heights on the time and difficulty of manual or mechanical operations; generally, the higher the floor height, the larger the coefficient. Finally, the basic horizontal distance value, the distance equivalent value, and the height equivalent value are input into a standardization function. This function first normalizes the three input values ​​to the interval between 0 and 1, and then performs a linear combination according to pre-defined distance weights, aisle weights, and floor height weights to calculate a shelf location heat value between 0 and 1. The larger the shelf location heat value, the higher the overall cost of storage and retrieval operations from that location, and the worse the location. This process, through multi-dimensional modeling, transforms the physical location attributes of the storage location into a calculable, unified heat index.

[0066] In one possible implementation, location coupling heat is the core bridge connecting product demand and location attributes. For any product and any location, the location coupling heat value is calculated by dividing the product's real-time demand heat value by the location's shelf location heat value. The physical meaning of this division operation is that it measures the intensity of product demand that can be supported per unit of location cost. If a product has high demand heat while a location has low location heat (i.e., a good location), then the quotient will be large, indicating that placing this high-demand product in this excellent location can generate high overall operational efficiency.

[0067] Conversely, placing high-demand goods in poorly located storage locations will reduce their market value. This calculation ensures that in subsequent allocation decisions, the system will naturally tend to match high-demand goods with highly accessible storage locations, thereby achieving initial optimization of inventory layout. This feature is simple yet effective, creatively coupling two different dimensions of popularity.

[0068] In one possible implementation, step 3 first constructs a dynamic storage location allocation set: all goods to be shelved and their quantities are extracted from the real-time order pool to form a set of goods to be allocated; simultaneously, completely idle storage locations and adjustable storage locations with inventory levels below a certain set threshold (e.g., less than 5 units) are selected from the real-time inventory of the warehousing system to form a set of available storage locations. Subsequently, a dynamic equilibrium allocation model is established. The objective function of this model aims to maximize a value called "total expected benefit".

[0069] The overall expected benefit consists of a weighted sum of two parts: the first part is the total location coupling heat, which is the sum of the coupling heat values ​​between all items to be allocated and their corresponding locations after placement according to the plan; the second part is the path optimization factor. The calculation of the path optimization factor involves simulating the next picking wave, determining the set of locations to be visited in the simulated wave based on the current allocation plan, and estimating the total path length required to complete this wave of picking using the nearest neighbor algorithm. The path optimization factor is the reciprocal of this estimated total length. The final objective function is the weighted sum of the first and second parts, with alpha and beta used to adjust the importance of the two parts. This model innovatively integrates static location attribute optimization and dynamic picking path optimization within the same decision-making framework.

[0070] In one possible implementation, step 3, which uses a genetic algorithm to solve the dynamic equilibrium allocation model, begins with encoding. A storage location allocation scheme is encoded as a chromosome, with a chromosome length equal to the number of goods to be allocated. Each gene bit represents a type of goods, and the gene value represents the storage location number assigned to it. Next, an initial population of 100 chromosomes is randomly generated. Then, iterative optimization is performed: in each generation, the fitness of each chromosome is calculated, representing the total expected benefit of the scheme it represents. Based on the fitness, a roulette wheel selection method is used to select dominant individuals as parents.

[0071] A single-point crossover operation is performed on parent individuals with a crossover probability of 0.85, exchanging a portion of their gene sequences. The locus number of a specific gene locus in the offspring individuals is randomly changed with a mutation probability of 0.02. This selection, crossover, and mutation process generates a new generation of individuals. This iterative process is repeated, with the termination condition set at reaching the maximum number of iterations (500 generations) or an improvement in optimal fitness of less than one-thousandth for 50 consecutive generations. After iteration terminates, the chromosome with the highest fitness in the current generation is decoded, yielding the optimal locus allocation scheme. Genetic algorithms, as mature heuristic global optimization algorithms, can effectively handle such combinatorial optimization problems, ensuring that an approximate optimal solution is found within a reasonable timeframe.

[0072] In one possible implementation, after receiving the optimal location allocation plan, the warehouse control system breaks it down into specific shelving task instructions and transfer task instructions, and issues them to the automated guided vehicle system or the handheld terminal of the warehouse management personnel. At the same time, the order management system, based on preset wave division rules, such as order creation time, delivery timeliness, and product overlap, delineates an actual picking wave from the order pool to be picked, and issues the wave task to the pickers.

[0073] During the picking process, the positioning module built into the picker's handheld terminal or wearable device (such as based on Bluetooth beacon or UWB technology) continuously records the movement trajectory, forming actual picking path trajectory data, and automatically associates it with the system timestamp of each picking action. This step translates the virtual allocation scheme into physical operations and simultaneously completes the collection of key data on the operational effectiveness, providing real-world feedback for system learning. Both the wave division rules and positioning data collection are existing and common technologies in warehousing operations.

[0074] In one possible implementation, the process of dynamically updating the location coupling heat calculation method in step 4 is key to the system's adaptive learning capability. First, after a picking wave is completed, the system analyzes the collected actual picking path trajectory data, calculates the actual total walking distance and total operation time for that wave, and thus obtains the actual average walking speed. Second, a prediction correction is performed: the calculated actual average walking speed is compared with the preset standard walking speed used in the path estimation in step 3; the ratio of the two is defined as the path prediction correction coefficient. If the actual speed is lower than the preset speed, the correction coefficient is less than 1, indicating that the actual operating environment is more congested or inefficient than the model predicted.

[0075] Finally, the model parameters are updated: this path prediction correction coefficient is applied to the calculation of shelf location popularity values. Specifically, in the standardized function for calculating shelf location popularity values, the distance weight, aisle weight, and floor height weight are each multiplied by this correction coefficient, and then the product is normalized again so that the sum of the three weights is still 1. After this update, the model will fine-tune its assessment of cost factors for each location in the next round of calculations, making it closer to the latest actual operational efficiency, thereby achieving self-iteration and optimization of the model.

[0076] In one possible implementation, a cold start process for system initialization is also provided. This process is executed after the method is first deployed in the warehouse or after a significant change in the warehouse layout. The cold start process loads complete historical order data and location status data for a historical period (e.g., the past 30 days). Subsequently, the system virtually executes the dynamic location allocation and picking operations for each day of these 30 days in a simulated playback manner. During the simulation, the system uses historical data to simulate "actual" operations and also performs path efficiency evaluation and model parameter update steps.

[0077] Through repeated simulations and learning using this historical data, the key parameters in the model, including the weights of outbound and shelving in the real-time demand calculation of goods, the weights of various items in the shelf location demand calculation, and the balance coefficients in the dynamic equilibrium model, will gradually converge from random initial values ​​to a set of relatively stable values. These stable values ​​are saved as the system's initial default parameters. Afterward, the system enters online operation mode and continues to fine-tune the parameters online through the feedback mechanism described in step 4. The cold start process utilizes historical data to provide a good initial starting point for the system, accelerating the online learning process and ensuring that the system possesses superior performance from the very beginning.

[0078] Example

[0079] This embodiment is set in the North China regional distribution center of a large B2C e-commerce company. The warehouse mainly handles more than 10,000 SKUs (stock keeping units) of goods such as daily necessities and small home appliances, and processes more than 50,000 orders per day. The orders have obvious volatility (such as a surge in order volume during "618" and "Double 11" promotions) and fragmentation characteristics (each order contains a small number of items, but the total number of orders is huge). The warehouse has a high-bay racking and picking aisle layout, and adopts a "person-to-goods" picking mode, with pickers using handheld terminals (PDAs) for operation.

[0080] Step 1: Data preparation and preprocessing for the warehousing system.

[0081] In a warehouse management system (WMS), prepare and collect the required basic data as follows:

[0082] 1.1 Basic Data Collection for Pallet Locations: The warehouse has 5,000 standard pallet locations. A unique three-dimensional coordinate system is defined for each location. Origin of coordinate system Set it as the main shipping outlet (i.e., the picking start point). The axis represents the straight-line horizontal distance from the center point of the storage location to the outlet. For example, the coordinates of storage location A are... This indicates that the cargo location is 15.2 meters from the outlet, located in the 3rd aisle, on the 2nd floor. The axis is the tunnel number, with a value ranging from 1 to 10. The axis represents the floor height number, ranging from 1 to 5 (floor 1 being the ground floor, and floor 5 being the highest floor). The coordinate information of all storage locations is stored in the storage location master data table of the WMS.

[0083] 1.2 Basic Product Data Collection: For each SKU, its physical attribute data is collected using a warehouse-wide integrated dimensional measurement and weighing machine upon its first entry into the warehouse. For example, for tissues with product number "SKU20240001", the data includes: length data... Width data Altitude data Weight data This data is stored in the product master data table.

[0084] 1.3 Historical Order Data Collection: Collect all completed outbound order data from the WMS order fulfillment history table within the past 90 days. Each order record includes: order number (e.g., "ORD202405200001"), a list of item numbers and their quantities included in the order (e.g., "SKU20240001:2", "SKU20240002:1"), and order creation timestamp (e.g., "2024-05-2014:30:25"). This data is used for subsequent potential model training (cold start) or analysis.

[0085] 1.4 Real-time collection of current order pool data: The system maintains two dynamic data pools in real time.

[0086] Inbound Pool: Query the details of inbound orders with a status of "Pending Shelving" in the WMS to obtain information on all goods awaiting allocation to physical storage locations. For example, there are currently 100 boxes of "SKU20240001" waiting to be shelved.

[0087] Outbound Pool: Query the details of customer orders in WMS with a status of "Pending Picking". For example, there are currently 1500 pending orders, of which 500 orders contain "SKU20240001".

[0088] Step 2: Real-time location popularity analysis based on the real-time order pool.

[0089] This step involves calculating and coupling the dynamic heat values ​​of the goods and storage locations at the current moment.

[0090] 2.1 Calculate the real-time demand for goods: Calculate the demand for each of the goods in the current outbound and inbound pools.

[0091] Regarding product "SKU20240001":

[0092] Real-time outbound frequency Out of 1500 orders awaiting picking, count the number of orders containing "SKU20240001", let's assume 500; further, summarize the total demand for this SKU across these orders, let's assume 1200 units. .

[0093] Total number of items pending release in real time In the inbound pool, the total quantity of all "SKU20240001" items awaiting shelving is 100 boxes (assuming 20 items per box, totaling 2000 items). .

[0094] Real-time demand for goods :calculate In this embodiment, the warehouse prioritizes rapid shipment and sets an outbound weighting coefficient. Listing weight coefficient .but .

[0095] 2.2 Calculate shelf location popularity: For each shelf location, calculate its comprehensive storage and retrieval cost.

[0096] With storage location A For example:

[0097] Horizontal distance base value Take directly Coordinate values .

[0098] Lane Conversion: A preset lane conversion coefficient table is used. Assuming lanes 1 and 10 are on the outermost edge, with the most convenient path, the coefficient is 1.0; lanes 2 and 9 have a coefficient of 1.1, and so on, with the middle lanes 5 and 6 having a coefficient of 1.5. If cargo location A is in lane 3, the distance equivalent value is retrieved. =1.3.

[0099] Floor Height Conversion: A preset floor height conversion coefficient table is used. Floor height 1 (ground level) is the easiest to operate, with a coefficient of 1.0; floor height 2 has a coefficient of 1.2; floor height 3 has a coefficient of 1.5; floor height 4 has a coefficient of 2.0; floor height 5 (requires climbing or using lifting equipment) has a coefficient of 3.0. Storage location A is on the 2nd floor; the height equivalent value is retrieved. =1.2.

[0100] Shelf location popularity value The calculation is performed using a standardized function. First, the three values ​​are normalized. Assuming the maximum horizontal distance in the warehouse is 50m, the normalization is then... The tunnel and floor height coefficients can be considered as equivalent values ​​and used directly. Set distance weights. weight of alleyways Floor height weight The shelf location heat value is:

[0101] ;

[0102] The higher the value, the higher the overall storage and retrieval cost of the warehouse, and the "worse" the location.

[0103] 2.3 Calculate the location coupling heat: Correlate the popularity of the goods with the popularity of the location. For goods "SKU20240001" and location A, the location coupling heat value is calculated. for:

[0104] ;

[0105] The higher this value, the "better" the currently high-demand "SKU20240001" will be placed. The higher the expected overall benefits from a cargo location A (with lower value), the better.

[0106] Step 3: Location allocation decision based on dynamic equilibrium model.

[0107] 3.1 Constructing a Dynamic Storage Location Allocation Set: Assume the current inbound pool has 50 different SKUs that need to be shelved, totaling 2000 boxes, forming a set of goods to be allocated. The system scans the status of all storage locations and finds 300 completely empty locations, and another 50 locations with only a few items remaining (stock less than 5 items, triggering a transfer). The remaining goods in these locations will be moved out and merged into other locations, thus freeing up space. These 350 locations constitute the available storage location set.

[0108] 3.2 Establishing a dynamic equilibrium allocation model: The goal of the model is to find the optimal allocation scheme, maximizing the following total expected benefits. :

[0109] ;

[0110] Part 1: Overall Storage Location Coupling Heat. This refers to the heat generated by placing all goods to be allocated according to the plan, and the relationship between each item and its assigned storage location. The sum of values. This is its weighting coefficient, which is set to 0.7 in this example.

[0111] Part Two: Path Optimization Factors. First, the system simulates and draws the next picking wave from 1500 pending orders based on order urgency (e.g., orders that must be shipped within 2 hours) and batching rules (no more than 50 orders per wave), assuming it contains 45 orders. Then, based on the current allocation plan, it determines all the storage locations (including newly allocated and existing stock locations) that these 45 orders need to access. Next, it uses a nearest neighbor algorithm to plan the shortest path to all these storage locations and calculates the estimated total length of the picking path. For example, calculated The path optimization factor is its reciprocal. . This is its weighting coefficient, set to 0.3 in this example, and... The sum is 1. The significance of this model is that it not only aims to place popular items in good locations (Part 1), but also considers whether this allocation decision is friendly to the specific picking task path that is about to occur (Part 2), thus realizing the linkage optimization of allocation and picking.

[0112] 3.3 Solving using a genetic algorithm:

[0113] Encoding: A chromosome is 50 units long (corresponding to 50 SKUs to be assigned). Gene position 1 represents the first SKU, and its gene value "A15" indicates that the SKU is assigned to the storage location numbered A15.

[0114] Initialization: 100 chromosomes are randomly generated (i.e., 100 random allocation schemes) to form the initial population.

[0115] Fitness assessment: Calculate the fitness of each chromosome. The value is used as fitness.

[0116] Selection: Using roulette wheel selection, chromosomes with higher fitness have a greater probability of being selected as parents.

[0117] Crossover: Set the crossover probability Pair the selected parents together, randomly select crossover points, exchange some gene segments, and generate new schemes.

[0118] Mutation: Set the mutation probability Randomly change the location number of a certain gene locus.

[0119] Iteration: Repeat the selection, crossover, mutation, and evaluation process. The maximum number of iterations is set to 500 generations, or the process stops when the optimal fitness improves by less than 0.1% over 50 consecutive generations.

[0120] Decoding output: After the iteration is completed, the chromosome with the highest fitness is decoded to generate a detailed "Location Allocation Scheme Table", which contains: SKU20240001->Location B05, SKU20240002->Location A01, ...

[0121] Step 4: Implementation of the allocation plan and dynamic rebalancing of heat.

[0122] 4.1 Execution of Allocation and Generation of Wave Tasks: The WCS (Warehouse Control System) receives the plan and generates task instructions:

[0123] Instructions are issued to the AGV (Automated Guided Vehicle) to transport 100 boxes of "SKU20240001" to storage location B05 and put them on the shelf.

[0124] Send a transfer instruction to the handheld terminal to guide employees to move the remaining 3 "SKU20240003" items in location C10 to location D12.

[0125] At the same time, the order management system officially generates and issues the same actual picking wave task as the simulated wave in step 3.2 to the pickers.

[0126] 4.2 Picking and Data Collection: Pickers follow the system-recommended path guided by the PDA. The PDA's built-in positioning module records coordinates once per second, forming the actual picking path trajectory. The system automatically records the timestamp of each item being scanned and picked.

[0127] 4.3 Dynamically update the popularity of storage locations:

[0128] Path efficiency assessment: After each wave is completed, the system analyzes the trajectory data to determine the actual total walking distance. Total work time The actual average walking speed .

[0129] Prediction Correction: In the path prediction step 3.2, the algorithm uses a preset standard walking speed. for (Based on historical average efficiency settings). The path prediction correction coefficient is then... . This indicates that the actual walking efficiency was lower than expected, possibly due to factors such as alleyway congestion and difficulty in retrieving and placing goods.

[0130] Model parameter update: Feed this correction coefficient back into the heat model. Update the weights in step 2.2.4: New distance weights. Similarly , .Will , , Renormalize to obtain the updated weights: Although the numerical changes are small, this process is a crucial learning mechanism. This means that after the system senses a decrease in actual walking efficiency, it will slightly increase the location cost in the next round of heat calculation. This value allows the allocation model to be more inclined to select areas with higher actual traffic efficiency in the future.

[0131] 4.4 Loop: After completing the above update, the system immediately returns to step 2, collects the latest order pool data (at this time, the inbound pool and outbound pool have been updated), and uses the optimized parameters to make a new round of dynamic location allocation decisions, thereby achieving continuous adaptive optimization.

[0132] Cold start process

[0133] After the system is first deployed in the warehouse or after a large-scale adjustment of the shelving layout, a cold start training is required to initialize the model parameters.

[0134] Step A: Load complete historical data (orders, storage location status) for the past 30 days.

[0135] Step B: Perform simulation playback on a daily basis. The system virtually executes dynamic allocation and picking for each day, and records the difference between the "actual" path efficiency (estimated based on historical operation time) and the estimated efficiency for each day. The model parameters are then iteratively updated according to the method in Step 4.3.

[0136] Step C: After 30 days of simulation training, the model parameters... The system has stabilized. These parameters are saved as the system's initial operating parameters. Afterward, the system enters online operation mode and continues online fine-tuning through step 4.3.

[0137] To demonstrate the effectiveness of the method of this invention, a comparative test was conducted over three consecutive working days in the warehouse described in this embodiment. During the test period, the average daily order volume was approximately 48,000 orders.

[0138] Comparative approach: The existing static classification storage strategy of the warehouse is adopted. This strategy divides SKUs into three categories—A (high frequency), B (medium frequency), and C (low frequency)—based on the historical outbound frequency over the past year, and stores them in areas from closest to furthest from the shipping outlet. Location allocation follows only the principle of "finding the nearest available space" and is unrelated to real-time orders. Picking routes use the traditional shortest path algorithm but are not linked to location allocation.

[0139] The method of this invention is a dynamic storage location allocation and optimization method using the complete process described above.

[0140] The comparison metric is the average operational data for a single picking wave (containing an average of 45 orders), and the results are the 3-day average, as shown in Table 1:

[0141] Table 1 Comparison between the method of the present invention and the comparative method

[0142] Comparison indicators Comparative Example (Static Categorization Storage) This invention (dynamic cargo location allocation) Increase Average picking walking distance 1450m / wave 1260m / wave Decreased by 13.1% Average order picking time 85s / single 73s / single Decreased by 14.1% Average number of lane congestion times 3.5 times / wave 2.0 times / wave Reduced by 42.9% Average picking distance for popular SKUs 28m 18m Reduced by 35.7% Warehouse operation efficiency 381 per person per class 437 per person per class An increase of 14.7%

[0143] Table notes:

[0144] "Number of times the alleyway is congested" refers to the number of times a picker pauses in the alleyway due to waiting or avoiding obstacles, which is calculated from PDA movement status data.

[0145] "Average picking distance for popular SKUs" refers to the average distance from the storage location to the shipping outlet for 10 randomly selected high-frequency SKUs of the day during the test period.

[0146] "Warehouse operational efficiency" refers to the total number of orders picked by a picker in one shift (8 hours).

[0147] Results Analysis: As shown in the table, compared with the traditional static method, the method of this invention shows significant improvements in all key indicators. Its core advantage lies in the fact that, through real-time heat coupling and path pre-optimization, it not only shortens the overall picking path and the picking path for popular items, but also reduces operational conflicts (congestion) in aisles through balanced allocation, thereby improving warehouse operational efficiency by 14.7%. This verifies the effectiveness and ingenuity of this invention in solving the technical problem of "coupling conflict between static location allocation and dynamic order picking".

[0148] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0149] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for intelligent warehouse dynamic storage location allocation based on storage location heat analysis, characterized in that, Includes the following steps: Step 1: Collect warehouse data, which includes warehouse location coordinates, physical attribute data of goods, historical order data, and current order pool data containing unexecuted outbound orders and goods to be received; Step 2: Perform real-time location heat analysis based on the current order pool data. The real-time location heat analysis includes calculating the real-time demand heat of goods, the shelf location heat, and the location coupling heat that combines the real-time demand heat of goods and the shelf location heat. Step 3: Make a storage location allocation decision based on the dynamic equilibrium model. The dynamic equilibrium model takes the total storage location coupling heat between the goods to be allocated and the available storage locations and the prediction results of the future picking path as the optimization objectives, solves and outputs the storage location allocation scheme. Step 3 specifically includes: Step 3.1: Construct a dynamic storage location allocation set. Extract all the product numbers to be allocated and their corresponding quantities to be put on the shelves from the "unallocated storage locations" inbound product records of the current order pool data to form a set of products to be allocated. From the real-time inventory status of the warehousing system, filter out all the idle storage locations with zero current inventory and the non-idle storage locations with inventory below the preset transfer trigger threshold to form a set of available storage locations. The transfer trigger threshold is set according to the average outbound batch of the products. Step 3.2: Establish a dynamic equilibrium allocation model. The objective function of the dynamic equilibrium allocation model is to maximize the total expected benefit. The formula for calculating the total expected benefit is: Total expected benefit = α × Total location coupling heat + β × Path optimization factor; where α and β are balance coefficients, and the sum of α and β is 1; the total location coupling heat is the sum of the location coupling heat values ​​of all corresponding locations after all the goods to be allocated are placed in available locations according to the allocation scheme; the calculation process of the path optimization factor includes: based on the urgency and number of the current "not yet picked" outbound orders, simulating the next picking wave to be executed; determining the set of locations that each order needs to access in the simulated picking wave based on the allocation scheme; using the nearest neighbor algorithm to calculate the estimated total length of the picking path to access the set of locations; the path optimization factor is the reciprocal of the estimated total length of the picking path; Step 4: Execute the location allocation plan and picking wave task, collect actual picking data, dynamically update the location coupling heat calculation method based on the actual picking data, and return to Step 2 to process the new current order pool data. Step 2, the process of calculating the real-time demand for goods, includes: Step 2.1.1: For each unique product number in the current order pool data, iterate through all outbound order records that have not started picking, accumulate the total number of outbound orders in which the product number appears in all outbound orders, and obtain the real-time outbound frequency of the product. Step 2.1.2: For the same product number in the current order pool data, iterate through all inbound product records with "unassigned storage locations", accumulate the total number of products waiting to be put on the shelves corresponding to that product number, and obtain the real-time total number of products waiting to be put on the shelves. Step 2.1.3: Calculate the real-time demand heat value of goods by weighting and summing the real-time outbound frequency and the real-time total amount to be put on the shelf. The outbound weight coefficient and the shelf weight coefficient used for weighting and summing are preset by the warehouse manager according to the warehouse's operation strategy of prioritizing outbound or shelf-based operations. Step 2, the process of calculating the shelf location heat, includes: Step 2.2.1: For each storage location, read the spatial coordinate data of that location and extract the first axis coordinate value as the basic value of the horizontal distance; Step 2.2.2: Based on the lane number where the cargo location is located, i.e. the second axis coordinate value, query the preset lane conversion coefficient table to convert the lane number into a distance equivalent value. The distance equivalent value corresponding to different lane numbers reflects the detour path cost required to enter different lanes from the delivery point. Step 2.2.3: Based on the floor height number of the storage location, i.e. the third axis coordinate value, look up the preset floor height conversion coefficient table to convert the floor height number into a height equivalent value. The height equivalent value reflects the difference in time and operational difficulty of storage locations of different floor heights when accessed manually or mechanically. Step 2.2.4: Input the horizontal distance base value, the distance equivalent value, and the height equivalent value into a standardization function. After normalizing the three input values, the standardization function performs linear weighting according to preset distance weight, aisle weight, and floor height weight, and outputs a shelf location heat value that represents the overall ease of access to the storage location. The shelf location heat value is inversely proportional to the ease of access. In step 2, the process of calculating the location coupling heat is as follows: Step 2.3: For each item to be allocated or stored in the current order pool data, and for each storage location in the warehouse, divide the real-time demand heat value of the item calculated in Step 2.1.3 by the shelf location heat value of the storage location calculated in Step 2.2.

4. The resulting quotient is defined as the storage location coupling heat value between the item and the storage location. The higher the real-time demand heat value of the item and the lower the shelf location heat value, the higher the storage location coupling heat value, indicating that the overall benefits of placing the item in the storage location are higher.

2. The intelligent warehousing dynamic storage location allocation method based on storage location heat analysis according to claim 1, characterized in that, Step 1 specifically includes: Step 1.1: Collect the spatial coordinate data of all storage locations and assign a three-dimensional coordinate to each storage location. The origin of the three-dimensional coordinate is the warehouse shipping point. The first axis of the three-dimensional coordinate represents the straight-line horizontal distance from the center point of the storage location to the shipping point. The second axis of the three-dimensional coordinate represents the aisle number where the storage location is located. The third axis of the three-dimensional coordinate represents the floor number where the storage location is located. Step 1.2: Collect physical attribute data for all goods in stock. The physical attribute data includes the length, width, height and weight of the goods. The length, width and height data are obtained by measuring the external dimensions of the goods' packaging boxes, and the weight data is obtained by weighing. Step 1.3: Collect all executed order data within the set historical time period from the warehouse management system database. The executed order data includes the order number, the item number of each item in the order, the outbound quantity of each item, and the order creation timestamp. Step 1.4: Collect current order pool data in real time. The current order pool data includes two parts: the first part is all inbound goods records in the warehouse management system with the status of "unassigned location". Each inbound goods record includes the product number and the quantity to be put on the shelf; the second part is all outbound order records in the warehouse management system with the status of "picking not started". Each outbound order record includes the order number and the set of product numbers and outbound quantities required for the order.

3. The intelligent warehousing dynamic storage location allocation method based on storage location heat analysis according to claim 1, characterized in that, Step 3, the process of solving the dynamic equilibrium allocation model using a genetic algorithm, includes: Step 3.3.1: Encoding. Encode the location allocation scheme into a chromosome. The length of the chromosome is equal to the total number of items in the set of items to be allocated. Each gene position of the chromosome corresponds to a specific item to be allocated. The value of the gene is the location number in the set of available locations to which the item is allocated. Step 3.3.2: Initialize the population by randomly generating a set number of chromosomes to form the initial population. Each chromosome represents a cargo space allocation scheme. Step 3.3.3: Fitness assessment. Based on the objective function of the dynamic equilibrium allocation model defined in Step 3.2, calculate the total expected benefit of the allocation scheme represented by each chromosome in the population, and use the total expected benefit as the fitness value of the chromosome. Step 3.3.4: Selection operation, using roulette wheel selection method, selects the parent chromosomes for reproduction according to the fitness value of the chromosomes in proportion; Step 3.3.5: Crossover operation. Pair the selected parent chromosomes together and perform a single-point crossover operation with a set crossover probability. Exchange the gene segments after the crossover point of the paired chromosomes to generate new offspring chromosomes. Step 3.3.6: Mutation operation. For all offspring chromosomes generated after crossover, the location number of a certain gene locus on the chromosome is randomly changed with a set mutation probability. The new value is randomly selected from the available location set. Step 3.3.7: Iteration. The offspring chromosomes generated through selection, crossover, and mutation are used as a new population. Steps 3.3.3 to 3.3.6 are repeated until the preset maximum number of iterations is reached or the improvement of the optimal fitness value over multiple consecutive generations is less than the stability threshold. Step 3.3.8: Decode the output. After the iteration terminates, decode the chromosome with the highest fitness in the population into a mapping relationship between the specific product number and the storage location number, and output it as the final storage location allocation scheme.

4. The intelligent warehousing dynamic storage location allocation method based on storage location heat analysis according to claim 3, characterized in that, Step 4, the process of executing the location allocation plan and picking wave tasks, includes: Step 4.1: The warehouse control system receives the location allocation plan output in step 3.3.8, generates detailed put-away task instructions and transfer task instructions, and sends them to automated warehousing equipment or displays them to warehouse operators; at the same time, the order management system divides a specific picking wave from the current "not yet picked" outbound orders according to the preset wave division rules. The wave division rules include the latest completion time window of the order, the overlap of goods in the order, and the upper limit of the total volume or total weight of a single wave. Step 4.2: Warehouse operators or automated picking equipment execute the put-away task instructions, transfer task instructions and picking wave tasks; during the execution of picking wave tasks, the coordinates are recorded at fixed time intervals or key nodes through the positioning terminal worn by the personnel or the positioning system built into the equipment, forming actual picking path trajectory data, and automatically recording the system timestamp when each item is picked.

5. The intelligent warehousing dynamic storage location allocation method based on storage location heat analysis according to claim 4, characterized in that, Step 4, the process of dynamically updating the calculation method of location coupling heat based on actual picking data, includes: Step 4.3.1: Path efficiency evaluation. After the picking wave task is completed, the actual total walking distance and total operation time of the wave task are calculated based on the collected actual picking path trajectory data to obtain the actual average walking speed. Step 4.3.2: Prediction correction. The actual average walking speed is compared with the preset standard walking speed used in step 3.2 when calculating the path optimization factor. The path prediction correction coefficient is calculated. The correction coefficient is equal to the actual average walking speed divided by the preset standard walking speed. Step 4.3.3: Model parameter update. The calculated path prediction correction coefficient is applied to the subsequent calculation of the shelf location heat value. The specific method is as follows: In the normalization function described in step 2.2.4, the distance weight, aisle weight and floor height weight used to calculate the shelf location heat value are multiplied by the path prediction correction coefficient and then normalized. The updated weights are used for the shelf location heat calculation in the next round of step 2.

6. The intelligent warehousing dynamic storage location allocation method based on storage location heat analysis according to claim 5, characterized in that, The method also includes a cold start process for initial model training, which is executed when the method is first applied or when there is a significant change in the warehouse layout, specifically including: Step A: Load complete historical order data and historical storage location data for a specific historical period; Step B: Using the process from Steps 2 to 4, simulate the dynamic location allocation and picking for each day of the historical period in a simulated playback manner. However, in Step 4.3.3, update the model parameters using the path prediction correction coefficients obtained from the simulation. Step C: After completing the simulation playback of the entire historical cycle, the final stabilized model parameters, including the outbound weight coefficient and the shelving weight coefficient in step 2.1.3, the distance weight, aisle weight and floor height weight in step 2.2.4, and the balance coefficients α and β in step 3.2, are saved as the initial default parameters of the system for subsequent real online dynamic storage location allocation.

Images