Method for automatically selecting and arranging drive-in type goods shelves of library body based on combinatorial optimization algorithm

Abstract

The invention discloses a drive-in type goods shelf automatic selection and arrangement method of a library body based on a combinatorial optimization algorithm. The method comprises the following specific steps: S1, collecting state data of goods shelves and goods in the library body in real time through an Internet of Things technology; s2, preprocessing the acquired data; s3, establishing a warehouse layout initial model and setting a target function; s4, generating a layout scheme by using a combinatorial optimization algorithm and performing dynamic adjustment; s5, combining an improved support vector machine and a gradient boosting tree algorithm evaluation scheme, and selecting an optimal scheme; s6, automatically adjusting the shelf layout according to the optimal scheme; and S7, updating the layout in real time according to the dynamic data, and ensuring continuous optimization and adaptability. According to the method, by combining a combinatorial optimization algorithm, technical improvement and fusion, automatic selection and arrangement of drive-in type goods shelves in the warehouse and optimization of the layout design of the goods shelves, the use efficiency of the warehouse body space is remarkably improved, the goods storage and taking cost is reduced, and the method can be widely applied to automatic warehouses and logistics systems.

Description

Technical Field

[0001] The present invention relates to the technical field of warehousing management and logistics automation, and particularly to a method for automatically selecting and arranging drive-in racks of a library body based on a combinatorial optimization algorithm. Background Art

[0002] With the rapid development of the e-commerce and logistics industries, modern warehousing management systems play an increasingly important role in the supply chain. With the increase in warehousing demand, traditional warehouse management methods have gradually been unable to meet the requirements of modern logistics systems for high efficiency, low cost, high flexibility, etc. Especially in automated warehouses and intelligent logistics systems, how to efficiently utilize warehouse space, improve the efficiency of goods storage and retrieval, and reduce operating costs has become the focus of industry attention.

[0003] Currently, many warehousing management systems still rely on manual experience or simple rule algorithms for shelf layout. The layout of the shelves is usually arranged according to factors such as the type of goods and the storage and retrieval frequency. However, this traditional method based on experience has many deficiencies. First, low space utilization is a prominent problem. Since the layout of the shelves has not been globally optimized, many areas in the warehouse are not fully utilized, resulting in vacancy and waste. Second, the storage and retrieval efficiency is not high. The arrangement order of the shelves usually does not consider the optimization of the item storage and retrieval path, and the staff needs to spend a long time walking when storing and retrieving items, resulting in a low overall operation efficiency. In addition, the traditional method has poor adaptability to the dynamic changes of warehousing requirements and cannot be adjusted in real time according to changes in the warehouse environment, goods storage and retrieval frequency, and goods type.

[0004] In order to improve warehousing management efficiency and reduce costs, in recent years, warehousing systems have gradually begun to adopt optimization algorithms for the automated design of shelf layout. The shelf layout method based on algorithms can consider the space utilization and storage and retrieval efficiency of the warehouse from a global perspective and provide a more scientific solution than traditional manual design. However, there are still some problems in the actual application of existing optimization algorithms. Traditional optimization methods often rely too much on the optimization of a single objective, such as only focusing on space utilization or storage and retrieval efficiency, while ignoring other important factors such as warehousing costs, operation time, labor costs, etc., resulting in limited comprehensive optimization effects.

[0005] In addition, existing logistics transportation algorithms have a high computational complexity when dealing with large-scale warehouses, especially for multi-objective optimization, and the time cost of the calculation process is large, which makes it difficult for the optimization process to respond to the demand changes of the warehouse in real time. In actual applications, how to closely combine these optimization algorithms with the actual needs of warehousing management systems and effectively reduce the computational complexity is still a technical problem to be solved urgently.

[0006] Therefore, aiming at the deficiencies in multi-objective optimization, computational efficiency, and practical adaptability in the prior art, there is an urgent need for an automatic selection and layout method for drive-in racks of a library body based on a combinatorial optimization algorithm to achieve efficient space utilization, improve storage and retrieval efficiency, and reduce warehousing costs, while ensuring efficient computing and real-time adaptability in a large-scale warehouse environment. Summary of the Invention

[0007] An object of the present invention is to propose an automatic selection and layout method for drive-in racks of a library body based on a combinatorial optimization algorithm. By optimizing the rack layout through the combinatorial optimization algorithm, the present invention not only improves the utilization rate of warehouse space and the material storage and retrieval efficiency, but also effectively reduces the warehousing cost. This method can automatically adjust the rack layout plan according to different warehousing requirements, adapt to the dynamic changes of large-scale automated warehouses, and is widely applied to various logistics warehousing and intelligent warehouse systems.

[0008] An automatic selection and layout method for drive-in racks of a library body based on a combinatorial optimization algorithm according to an embodiment of the present invention includes the following steps:

[0009] S1. Real-time collect the status data of the racks and goods in the library body through Internet of Things technology;

[0010] S2. Preprocess the collected real-time data, including noise removal and anomaly detection;

[0011] S3. Establish an initial model of the warehouse layout, and set the objective functions for rack selection and layout, including the utilization rate of warehouse space, material storage and retrieval efficiency, and warehousing cost;

[0012] S4. Combine the particle swarm optimization algorithm and the simulated annealing algorithm to perform multi-objective optimization and solution on the initial model, generate multiple drive-in rack layout plans, and perform dynamic adjustment according to the constraint conditions;

[0013] S5. Improve and fuse the kernel function of the support vector machine and the integrated learning characteristics of the gradient boosting tree, intelligently evaluate the storage and retrieval efficiency and warehousing cost of each layout plan, perform weighted ranking, and select the layout plan that best meets the actual requirements;

[0014] S6. Automatically adjust the layout of the drive-in racks according to the optimal plan, including the selection of racks, the placement positions, and the storage areas for various materials;

[0015] S7. In actual application, update the layout of the drive-in racks in real time according to the dynamically changing data in the warehouse to ensure the continuous optimization and adaptability of the layout plan.

[0016] Optionally, the S1 specifically includes:

[0017] S11. Install RFID tags, temperature and humidity sensors, weight sensors, and laser ranging sensors in the library body to monitor the location information, temperature and humidity, and load status of goods in real time. Each piece of goods and the shelf communicate with the data acquisition platform through Zigbee protocol and Wi-Fi protocol;

[0018] S12. Operate according to the sensor working process: The RFID tag automatically records the goods information and sends it to the server every time the goods enter or leave. The temperature and humidity sensor continuously monitors the environmental changes. The weight sensor measures the shelf load in real time. The laser ranging sensor is used to accurately measure the distance between the goods and the shelf;

[0019] S13. Use the MQTT protocol for data transmission, and utilize the lightweight message queue transmission mechanism to ensure the real-time and stability of data;

[0020] S14. The sensor data is transmitted to the edge computing unit in real time, that is, the network edge close to the data source, and the burden on the main server is reduced through local data processing for data aggregation and analysis;

[0021] S15. Store and transmit all the collected raw data in the following format:

[0022] D = [RFID, Distance, Weight, Humidity, Temperature, Time, Load];

[0023] Among them, D represents the set of collected raw data, RFID represents the unique identifier of the goods, Distance is the distance between the goods and the shelf measured by the laser ranging sensor, Weight represents the weight of the goods, Humidity and Temperature respectively represent the temperature and humidity data, Time is the timestamp of data acquisition, and Load is the load situation on the shelf.

[0024] Optionally, the S2 specifically includes:

[0025] S21. Use the k-nearest neighbor interpolation method to fill in the missing values in the sensor data. The formula is as follows:

[0026]

[0027] Among them, X new is the estimated value of the missing data, and X i is the k data values closest to the missing data;

[0028] S22. Apply Kalman filtering to denoise the sensor data, and use the following recursive formula:

[0029]

[0030] Among them, is the estimated state at the current moment, is the estimated state at the previous moment, and K k is the Kalman gain, z k is the observed value, and H k is the observation matrix;

[0031] S23. Use the quartile method for outlier detection. An outlier is defined as data that exceeds 1.5 times the range between the lower quartile and the upper quartile. By removing these abnormal data, the accuracy of subsequent analysis is ensured. The formula is as follows:

[0032] Outlier = (Data < Q1 - 1.5×(Q3 - Q1)) ∨ (Data > Q3 + 1.5×(Q3 - Q1));

[0033] Among them, Outlier represents the quartile, Data represents the detected data, Q1 is the lower quartile, and Q3 is the upper quartile.

[0034] Optionally, the specific steps of S3 include:

[0035] S31. Discretize and model the warehouse space, divide the warehouse space into multiple optimization areas, and each area can accommodate specific types of goods. By modeling the attributes of various shelves, define the space utilization rate, material access efficiency, and warehousing cost as the main optimization objectives;

[0036] S32. Define the optimization objective function as a weighted sum form of multiple objectives, and the objective function is comprehensively optimized by balancing each objective:

[0037] F(x) = w1·U(x) + w2·E(x) + w3·C(x) + w4·L(x);

[0038]

[0039] Among them, F(x) is the comprehensive objective function, U(x) is the space utilization rate, E(x) is the access efficiency, C(x) is the warehousing cost, L(x) is the load balance, w1, w2, w3, w4 are the weight coefficients of the objective function, x is the decision variable of the layout plan, A i (x) represents the shelf area in the i-th area, O i (x) is the idle area not occupied by goods in this area, A total is the total area of the warehouse, f i (x) represents the access frequency of goods i, T i (x) is the access time of goods i, T totalis the total access time of all goods in the warehouse, C i (x) is the basic cost of the i-th shelf, P i (x) is the material handling cost, D i (x) is the handling distance of the material, L i (x) is the load of the i-th shelf, L max is the maximum load to ensure the load uniformity of the layout;

[0040] S33. Set constraint conditions to ensure that the optimization result meets the actual requirements of the warehouse. The mathematical model of the constraint conditions is as follows:

[0041] g1(x) ≤ 0, g2(x) ≤ 0, g3(x) ≤ 0, g4(x) ≤ 0, g5(x) ≤ 0;

[0042] Among them, g1(x) is the shelf load limit, g2(x) is the aisle width limit, g3(x) is the material access path limit, g4(x) is the shelf spacing limit, and g5(x) is the material classification limit;

[0043] Optionally, the S4 specifically includes:

[0044] S41. Initialize the particle p of the particle swarm algorithm i , each particle represents a layout scheme x i , initialize the position x of the particle i (t) and velocity v i (t), where the position and velocity represent the solution space of the layout scheme and the search speed of the particle respectively. Initialize the particle swarm through the following formula:

[0045] x i (0) = Random Layout within Bounds;

[0046] v i (0) = Initial Speedbased on Layout Space;

[0047] S42. Calculate the objective function value F(x i ) corresponding to the layout scheme of each particle p i . The objective function considers the warehouse space utilization rate U(x i ), access efficiency E(x i ), warehousing cost C(x i ) and load balance L(x i ), and calculate the objective function value through the following formula:

[0048] F(x i ) = w1·U(x i) + w2·E(x i ) + w3·C(x i ) + w4·L(x i );

[0049] Among them, w1, w2, w3, and w4 are the weight coefficients of the objective function, representing the importance of space utilization, access efficiency, and storage cost in optimization;

[0050] S43. According to the update rule of the particle swarm algorithm, update the velocity and position of the particle. The velocity and position update formulas of the particle are:

[0051] v i (t + 1) = w·v i (t) + c1·r1·(pbest i - x i (t)) + c2·r2·(gbest - x i (t));

[0052] x i (t + 1) = x i (t) + v i (t + 1);

[0053] Among them, v i (t + 1) is the velocity of the i-th particle, x i (t) is the position of the i-th particle, pbest i is the individual optimal position of the i-th particle, gbest is the global optimal position, w is the inertia weight, c1, c2 are acceleration constants, r1, r2 are random numbers, and the search direction of the particle is updated;

[0054] S44. Introduce the simulated annealing algorithm to avoid the particle swarm algorithm falling into a local optimal solution. The simulated annealing algorithm controls the search space of the solution by adjusting the temperature. The temperature update formula is:

[0055] T(t + 1) = T(t)·β;

[0056] Among them, T(t) is the current temperature, and β is the cooling factor;

[0057] S45. In each iteration process, the simulated annealing algorithm decides whether to accept the new solution by calculating the energy difference between the current solution and the new solution. The acceptance probability is calculated by the following formula:

[0058]

[0059] Among them, ΔF = F(x new ) - F(x old) is the difference in the objective function between the new solution and the old solution. The probability P of accepting the new solution gradually decreases as the temperature drops, ensuring that the search process can jump out of the local optimal solution.

[0060] S46. After each iteration, update the velocity and position of the particle, and combine the acceptance mechanism of the simulated annealing algorithm to adjust the layout plan.

[0061] Optionally, the S5 specifically includes:

[0062] S51. Adopt the integrated learning characteristics of the support vector machine and the gradient boosting tree to intelligently evaluate the access efficiency E(x) and storage cost C(x) of each layout plan, and optimize the improved kernel function and the integrated learning model to improve the accuracy and robustness of the evaluation;

[0063] S52. In the support vector machine, use a high-order radial basis kernel function for mapping, defined as:

[0064]

[0065] where ∥x - y∥ 2 is the Euclidean distance between layout plans x and y, σ is the kernel width, α is an additional adjustment parameter, γ is the inner product parameter, <x, y> is the inner product in the feature space, and d is the order of the polynomial kernel function;

[0066] S53. In the gradient boosting tree, the output of each decision tree is calculated by the residual method, and the splitting point of the decision tree node is optimized. To further improve the evaluation ability, the weighted residual sum of squares is used to train the output of each tree:

[0067]

[0068] where y i is the true value of the i-th layout plan, f i-1 (x) is the prediction result of the previous i - 1 rounds, λ is the regularization coefficient, represents the second-order gradient under the j-th feature dimension;

[0069] S54. The total output f(x) of the gradient boosting tree is a weighted combination of multiple decision trees, calculated as:

[0070]

[0071] where f i (x) is the output of the i-th decision tree, α i is the weight of each tree, γ i is the offset of the tree output, II is the indicator function, indicating that the feature x of the layout plan x j belongs to the leaf node R of the i-th tree ij, where n is the number of trees and p is the feature dimension;

[0072] S55. Perform weighted fusion on the evaluation results of the support vector machine and the gradient boosting tree to obtain the final weighted score S(x) of each layout plan. The weighting function is:

[0073]

[0074] where is the output of the SVM, w1, w2, w3 are the weighting coefficients, and f i (x) is the output of the GBT, and finally the comprehensive evaluation score of each layout plan is obtained;

[0075] S56. Sort all layout plans according to the weighted score S(x), and select the layout plan with the highest score as the optimal plan. The selection of the optimal plan depends on the comprehensive results of multiple evaluation criteria to ensure that the warehouse layout plan can maximize the utilization rate of the storage space, the material access efficiency, and minimize the storage cost.

[0076] Optionally, the S6 specifically includes:

[0077] S61. According to the optimal layout plan, determine the selection and placement position of each shelf, ensure the reasonable matching of the shelf type and the warehouse space, and adjust the layout of the shelves according to the load and space requirements to improve the space utilization rate and material access efficiency of the warehouse;

[0078] S62. Automatically adjust the storage areas of various materials according to the material storage requirements, and optimize the placement positions of the materials according to the inbound and outbound frequencies of the materials to ensure the shortest access path and improve the overall access efficiency.

[0079] Optionally, the S7 specifically includes:

[0080] S71. In actual applications, according to the dynamic data in the warehouse, real-time monitor the access frequencies, inbound and outbound speeds, and inventory changes of various goods, and compare these data with the optimization plan to adjust the current shelf layout to adapt to the new warehouse operation status;

[0081] S72. Automatically adjust the shelf layout according to the real-time monitored data, update the shelf positions, the goods storage areas, and the access paths to cope with the changes in the warehouse, ensure the continuous optimization and adaptability of the layout plan, and improve the warehouse operation efficiency.

[0082] The beneficial effects of the present invention are:

[0083] (1) By introducing a combinatorial optimization algorithm, the present invention dynamically adjusts the arrangement order and access path of the shelves during the automatic selection and arrangement of the drive-in shelves in the library body, and adaptively optimizes the shelf arrangement plan according to the actual space layout, cargo access efficiency, and warehousing cost of the warehouse. Traditional methods usually use fixed layout methods, which are difficult to balance space utilization and access efficiency. The dynamic optimization mechanism of the present invention significantly improves the space utilization rate and access efficiency of the warehouse.

[0084] (2) The present invention combines multi-dimensional optimization objectives such as space utilization rate, access efficiency, and warehousing cost during the shelf arrangement process, and globally adjusts the shelf layout through a combinatorial optimization algorithm to ensure that the arrangement plan can balance the relationship between different optimization objectives during the optimization process, reducing the deviation and deficiency that may be caused by traditional methods in single-objective optimization, and ensuring the efficiency and feasibility of overall warehousing management.

[0085] (3) By dynamically adjusting and real-time monitoring the warehouse data, and combining the optimized shelf layout parameters, the present invention improves the adaptability of the target project under the warehousing requirements and data environment of the target project. Especially in cross-project scenarios with large data distribution differences, it can effectively align data differences and significantly improve the adaptability of the shelf arrangement plan. Description of the Drawings

[0086] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation to the present invention. In the drawings:

[0087] Figure 1 is a flowchart of a method for automatically selecting and arranging drive-in shelves in a library body based on a combinatorial optimization algorithm proposed by the present invention. Detailed Embodiments

[0088] Now, the present invention will be further described in detail with reference to the drawings. These drawings are all simplified schematic diagrams, only illustrating the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0089] Refer to Figure 1 , a method for automatically selecting and arranging drive-in shelves in a library body based on a combinatorial optimization algorithm, includes the following steps:

[0090] S1. Real-time collect the status data of the shelves and goods in the library body through Internet of Things technology;

[0091] S2. Preprocess the collected real-time data, including noise removal and anomaly detection;

[0092] S3. Establish an initial model of the warehouse layout and set the objective functions for shelf selection and layout, including the utilization rate of warehouse space, material access efficiency, and warehousing costs;

[0093] S4. Combine the particle swarm optimization algorithm and the simulated annealing algorithm to perform multi-objective optimization and solution on the initial model, generate multiple drive-in rack layout plans, and make dynamic adjustments according to the constraint conditions;

[0094] S5. Improve and integrate the kernel function of the support vector machine and the ensemble learning characteristics of the gradient boosting tree, intelligently evaluate the access efficiency and warehousing costs of each layout plan, perform weighted ranking, and select the layout plan that best meets the actual needs;

[0095] S6. Automatically adjust the layout of the drive-in racks according to the optimal plan, including the selection of racks, the placement positions, and the storage areas for various materials;

[0096] S7. In actual applications, update the layout of the drive-in racks in real time according to the dynamically changing data in the warehouse to ensure the continuous optimization and adaptability of the layout plan.

[0097] In this embodiment, the specific content of S1 includes:

[0098] S11. Install RFID tags, temperature and humidity sensors, weight sensors, and laser range sensors in the warehouse body to monitor the position information, temperature and humidity, and load status of the goods in real time. Each piece of goods and the rack communicate with the data acquisition platform through the Zigbee protocol and the Wi-Fi protocol;

[0099] S12. Operate according to the sensor working process: The RFID tag automatically records the goods information and sends it to the server every time the goods enter or leave. The temperature and humidity sensor continuously monitors the environmental changes. The weight sensor measures the load of the rack in real time. The laser range sensor is used to accurately measure the distance between the goods and the rack;

[0100] S13. Use the MQTT protocol for data transmission and utilize the lightweight message queue transmission mechanism to ensure the real-time and stability of the data;

[0101] S14. The sensor data is transmitted to the edge computing unit in real time, that is, the network edge close to the data source, and the burden on the main server is reduced through local data processing for data aggregation and analysis;

[0102] S15. Store and transmit all the collected raw data in the following format:

[0103] D = [RFID, Distance, Weight, Humidity, Temperature, Time, Load];

[0104] Among them, D represents the set of original data collected, RFID represents the unique identifier of the goods, Distance is the distance between the goods and the shelf measured by the laser range finder, Weight represents the weight of the goods, Humidity and Temperature respectively represent the temperature and humidity data, Time is the timestamp of data collection, and Load is the load condition on the shelf.

[0105] In this embodiment, the S2 specifically includes:

[0106] S21. Use the k-nearest neighbor interpolation method to fill in the missing values in the sensor data. The formula is as follows:

[0107]

[0108] Among them, X new is the estimated value of the missing data, and X i is the k data values closest to the missing data;

[0109] S22. Apply the Kalman filter to denoise the sensor data. Use the following recursive formula:

[0110]

[0111] Among them, is the estimated state at the current moment, is the estimated state at the previous moment, K k is the Kalman gain, z k is the observed value, and H k is the observation matrix;

[0112] S23. Use the quartile method for outlier detection. The outlier is defined as the data exceeding 1.5 times the range between the lower quartile and the upper quartile. By removing these abnormal data, the accuracy of subsequent analysis is ensured. The formula is as follows:

[0113] Outlier = (Data < Q1 - 1.5×(Q3 - Q1)) ∨ (Data > Q3 + 1.5×(Q3 - Q1));

[0114] Among them, Outlier represents the quartile, Data represents the data to be detected, Q1 is the lower quartile, and Q3 is the upper quartile.

[0115] In this embodiment, the S3 specifically includes:

[0116] S31. Discretize and model the warehouse space, divide the warehouse space into multiple optimization regions, each region can accommodate specific types of goods, and define the space utilization rate, material access efficiency, and warehousing cost as the main optimization objectives by modeling the attributes of various types of shelves;

[0117] S32. Define the optimization objective function as a weighted sum form of multiple objectives, and the objective function is comprehensively optimized by balancing each objective:

[0118] F(x) = w1·U(x) + w2·E(x) + w3·C(x) + w4·L(x);

[0119]

[0120] Among them, F(x) is the comprehensive objective function, U(x) is the space utilization rate, E(x) is the access efficiency, C(x) is the warehousing cost, L(x) is the load balance, w1, w2, w3, w4 are the weight coefficients of the objective function, x is the decision variable of the layout plan, A i (x) represents the shelf area in the i-th region, O i (x) is the idle area not occupied by goods in this region, A total is the total area of the warehouse, f i (x) represents the access frequency of goods i, T i (x) is the access time of goods i, T total is the total access time of all goods in the warehouse, C i (x) is the basic cost of the i-th shelf, P i (x) is the material handling cost, D i (x) is the handling distance of the material, L i (x) is the load of the i-th shelf, L max is the maximum load to ensure the load uniformity of the layout;

[0121] S33. Set the constraint conditions to ensure that the optimization results meet the actual requirements of the warehouse. The mathematical model of the constraint conditions is as follows:

[0122] g1(x) ≤ 0, g2(x) ≤ 0, g3(x) ≤ 0, g4(x) ≤ 0, g5(x) ≤ 0;

[0123] Among them, g1(x) is the shelf load limit, g2(x) is the aisle width limit, g3(x) is the material access path limit, g4(x) is the shelf spacing limit, and g5(x) is the material classification limit;

[0124] In this embodiment, the specific content of S4 includes:

[0125] S41. Initialize the particle p of the particle swarm algorithmi , each particle represents a layout scheme x i , initialize the position x of the particle i (t) and velocity v i (t), where the position and velocity represent the solution space of the layout scheme and the search speed of the particle respectively. Initialize the particle swarm through the following formula:

[0126] x i (0) = Random Layout within Bounds;

[0127] v i (0) = Initial Speedbased on Layout Space;

[0128] S42. Calculate the objective function value F(x i ) corresponding to the layout scheme of each particle p i ). The objective function considers the warehouse space utilization rate U(x i ), the access efficiency E(x i ), the warehousing cost C(x i ) and the load balance L(x i ), and calculate the objective function value through the following formula:

[0129] F(x i ) = w1·U(x i ) + w2·E(x i ) + w3·C(x i ) + w4·L(x i );

[0130] Among them, w1, w2, w3, w4 are the weight coefficients of the objective function, indicating the importance of space utilization rate, access efficiency and warehousing cost in the optimization;

[0131] S43. According to the update rules of the particle swarm algorithm, update the velocity and position of the particle. The velocity and position update formulas of the particle are:

[0132] v i (t + 1) = w·v i (t) + c1·r1·(pbest i - x i )(t)) + c2·r2·(gbest - x i )(t));

[0133] x i (t + 1) = x i (t) + v i (t + 1);

[0134] Among them, vi (t + 1) is the velocity of the i-th particle, x i (t) is the position of the i-th particle, pbest i is the individual best position of the i-th particle, gbest is the global best position, w is the inertia weight, c1 and c2 are acceleration constants, r1 and r2 are random numbers, which update the search direction of the particle;

[0135] S44. Introduce the simulated annealing algorithm to avoid the particle swarm algorithm falling into the local optimal solution. The simulated annealing algorithm controls the search space of the solution by adjusting the temperature. The temperature update formula is:

[0136] T(t + 1) = T(t)·β;

[0137] where, T(t) is the current temperature, and β is the cooling factor;

[0138] S45. In each iteration process, the simulated annealing algorithm decides whether to accept the new solution by calculating the energy difference between the current solution and the new solution. The acceptance probability is calculated by the following formula:

[0139]

[0140] where, ΔF = F(x new ) - F(x old ) is the difference in the objective function between the new solution and the old solution. The acceptance probability P of the new solution gradually decreases as the temperature decreases, ensuring that the search process can jump out of the local optimal solution.

[0141] S46. After each iteration, update the velocity and position of the particle, and combine the acceptance mechanism of the simulated annealing algorithm to adjust the layout scheme.

[0142] In this embodiment, the specific content of S5 includes:

[0143] S51. Adopt the integrated learning characteristics of the support vector machine and the gradient boosting tree to intelligently evaluate the access efficiency E(x) and storage cost C(x) of each layout scheme, and optimize the improved kernel function and the integrated learning model to improve the accuracy and robustness of the evaluation;

[0144] S52. In the support vector machine, use a high-order radial basis kernel function for mapping, which is defined as:

[0145]

[0146] where, ∥x - y∥ 2 is the Euclidean distance between layout schemes x and y, σ is the kernel width, α is an additional adjustment parameter, γ is the inner product parameter, <x, y> is the inner product in the feature space, and d is the order of the polynomial kernel function;

[0147] S53. In the gradient boosting tree, the output of each decision tree is calculated by the residual method, and the splitting points of the decision tree nodes are optimized. To further improve the evaluation ability, the weighted sum of squared residuals is used to train the output of each tree:

[0148]

[0149] where y i is the true value of the i-th layout plan, f i-1 (x) is the prediction result of the previous i - 1 rounds, λ is the regularization coefficient, represents the second-order gradient under the j-th feature dimension;

[0150] S54. The total output f(x) of the gradient boosting tree is a weighted combination of multiple decision trees, calculated as:

[0151]

[0152] where f i (x) is the output of the i-th decision tree, α i is the weight of each tree, γ i is the offset of the tree output, II is the indicator function, indicating that the feature x j of the layout plan x belongs to the leaf node R ij of the i-th tree, n is the number of trees, and p is the feature dimension;

[0153] S55. The evaluation results of the support vector machine and the gradient boosting tree are weighted and fused to obtain the final weighted score S(x) of each layout plan. The weighting function is:

[0154]

[0155] where, is the output of the SVM, w1, w2, w3 are the weighting coefficients, f i (x) is the output of the GBT, and finally the comprehensive evaluation score of each layout plan is obtained;

[0156] S56. According to the weighted score S(x), all layout plans are sorted, and the layout plan with the highest score is selected as the optimal plan. The selection of the optimal plan depends on the comprehensive results of multiple evaluation criteria to ensure that the warehouse layout plan can maximize the utilization rate of the storage space, the material access efficiency, and minimize the storage cost.

[0157] In this embodiment, the specific content of S6 includes:

[0158] S61. According to the optimal layout plan, determine the selection and placement positions of each shelf, ensure a reasonable match between the shelf type and the warehouse space, and adjust the shelf layout according to the load and space requirements to improve the space utilization rate and material access efficiency of the warehouse;

[0159] S62. According to the material storage requirements, automatically adjust the storage areas of various materials, and optimize the placement positions of materials according to the inbound and outbound frequencies of the materials to ensure the shortest access path and improve the overall access efficiency.

[0160] In this embodiment, the specific steps of S7 are as follows:

[0161] S71. In actual application, according to the dynamic data in the warehouse, real-time monitor the access frequencies, inbound and outbound speeds, and inventory changes of various goods, and compare these data with the optimization plan to adjust the current shelf layout to adapt to the new warehouse operation status;

[0162] S72. According to the real-time monitored data, automatically adjust the shelf layout, update the shelf positions, material storage areas, and access paths to cope with the changes in the warehouse, ensure the continuous optimization and adaptability of the layout plan, and improve the warehouse operation efficiency.

[0163] Embodiment 1:

[0164] To verify the feasibility and superiority of the present invention in actual application, this embodiment selects a large warehouse management system as the experimental object. This warehouse is mainly used for storing and sorting goods from various e-commerce platforms. The warehouse layout adopts a drive-in rack structure, with a wide variety of goods and large differences in access frequencies. In warehouse management, the traditional shelf layout method cannot efficiently handle large-scale goods access tasks, and has low space utilization rate and access efficiency. Therefore, how to optimize the shelf layout through scientific methods, improve the access efficiency, and reduce the operation cost has become an urgent problem to be solved in warehouse management.

[0165] This embodiment adopts an automatic selection and layout method for drive-in racks in the library based on a combinatorial optimization algorithm. Through this method, the warehouse shelf layout is optimized to solve the deficiencies of the traditional layout method in terms of space waste, access efficiency, and dynamic adaptability. The goal of this method is to optimize the shelf layout, improve the space utilization rate, shorten the access time, and reduce the warehousing cost.

[0166] To comprehensively evaluate the effectiveness of the method of the present invention, the data is compared from three aspects: before optimization, after optimization, and the improvement amplitude.

[0167] The method of the present invention comprehensively compares the data of each index, and the experimental results are shown in Table 1:

[0168] Table 1 Comparison data of warehouse management efficiency before and after implementation

[0169] Index Before optimization After optimization Improvement rate Warehouse space utilization rate 70％ 90％ 28.6% improvement Single access time 5 minutes 3 minutes 40% improvement Monthly workload of operators 350 hours 280 hours 20% reduction Warehousing cost 120,000 yuan / month 95,000 yuan / month 20.8% reduction Frequency of goods in and out 6 times / hour 9 times / hour 50% improvement Layout adjustment time 8 days 4 days 50% reduction

[0170] As can be seen from Table 1, the optimized warehouse space utilization rate has increased significantly, from 70% to 90%, and the access time per operation has been reduced by 40%, thus greatly improving the access efficiency of the warehouse. The monthly workload of the operator has been reduced by 20%, and the warehousing cost has decreased by 20.8%. These data fully demonstrate the remarkable effects of the method of the present invention in improving warehouse management efficiency, reducing costs, and enhancing the work efficiency of operators.

[0171] In addition, during the optimization process of this embodiment, dynamic changing requirements are considered, and it can adjust the shelf layout and access paths in real time according to changes in the warehouse environment, goods types, and access frequencies, ensuring the adaptability and flexibility of the warehouse. Compared with traditional methods, the optimized system can not only provide more efficient goods storage and sorting, but also adapt to changes in market demands in real time, enhancing the overall intelligent level of warehousing management.

[0172] The present invention dynamically adjusts the arrangement order and access paths of the shelves during the automatic selection and layout of the drive-in shelves in the library body by introducing a combinatorial optimization algorithm, and adaptively optimizes the shelf layout plan according to the actual space layout of the warehouse, goods access efficiency, and warehousing cost. Traditional methods usually use fixed layout methods and it is difficult to balance space utilization rate and access efficiency. The dynamic optimization mechanism of the present invention significantly improves the space utilization rate and access efficiency of the warehouse.

[0173] The present invention combines multi-dimensional optimization objectives such as space utilization rate, access efficiency, and warehousing cost during the shelf layout process, and globally adjusts the shelf layout through a combinatorial optimization algorithm to ensure that the layout plan can balance the relationships between different optimization objectives during the optimization process, reducing the deviations and deficiencies that may be caused by traditional methods in single-objective optimization, and ensuring the efficiency and feasibility of overall warehousing management.

[0174] The present invention realizes the improvement of the adaptability of the target project by dynamically adjusting and real-time monitoring the warehouse data, combined with the optimized shelf layout parameters, under the warehousing requirements and data environment of the target project. Especially in cross-project scenarios with large differences in data distribution, it can effectively align data differences and significantly improve the adaptability of the shelf layout plan. The adaptability of the method of the present invention in different warehousing environments has been improved by 20%, significantly reducing the operating costs of the warehouse, and it has been successfully applied to multiple actual warehouse management systems.

[0175] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and all should be covered within the protection scope of the present invention.

Claims

1. A method for automatic selection and arrangement of drive-in shelves in a warehouse based on a combined optimization algorithm, characterized in that: The steps include: S1. Collect the status data of shelves and goods in the warehouse in real time through the Internet of Things technology; S2, preprocessing the collected real-time data, including noise removal and anomaly detection; S3. Establish an initial model of warehouse layout and set the objective function of shelf selection and layout, including warehouse space utilization, material access efficiency and storage cost; S4. Combine the particle swarm algorithm and simulated annealing algorithm to solve the multi-objective optimization of the initial model, generate multiple drive-in shelf layout plans, and dynamically adjust them according to the constraints; S5. Improve and integrate the kernel function of the support vector machine and the ensemble learning characteristics of the gradient boosting tree, intelligently evaluate the access efficiency and storage cost of each layout scheme, perform weighted sorting, and select the layout scheme that best meets actual needs; S6. Automatically adjust the layout of the drive-in rack according to the optimal solution, including the selection of racks, placement of racks and storage areas for various types of materials; S7. In actual applications, the layout of the drive-in racks is updated in real time according to the dynamically changing data in the warehouse to ensure the continuous optimization and adaptability of the layout plan.

2. The method for automatic selection and arrangement of drive-in shelves in a warehouse based on a combined optimization algorithm according to claim 1 is characterized in that: The S1 specifically includes: S11. Install RFID tags, temperature and humidity sensors, weight sensors and laser ranging sensors in the warehouse to monitor the location information, temperature and humidity and load status of goods in real time. Each product and shelf communicates with the data collection platform through Zigbee protocol and Wi-Fi protocol; S12. Operate according to the sensor workflow: RFID tags automatically record cargo information and send it to the server every time cargo enters or leaves the warehouse. Temperature and humidity sensors continuously monitor environmental changes. Weight sensors measure shelf loads in real time. Laser distance sensors are used to accurately measure the distance between cargo and shelves. S13. Use MQTT protocol for data transmission and use lightweight message queue transmission mechanism to ensure the real-time and stability of data; S14, sensor data is transmitted in real time to the edge computing unit, i.e., the edge of the network close to the data source, to reduce the burden on the main server through local data processing, and to perform data aggregation and analysis; S15. All collected raw data are stored and transmitted in the following format: D=[RFID,Distance,Weight,Humidity,Temperature,Time,Load]; Where D represents the collected raw data set, RFID represents the unique identifier of the goods, Distance represents the distance between the goods and the shelf measured by the laser ranging sensor, Weight represents the weight of the goods, Humidity and Temperature represent the temperature and humidity data respectively, Time represents the timestamp of data collection, and Load represents the load condition on the shelf.

3. The method for automatic selection and arrangement of drive-in shelves in a warehouse based on a combined optimization algorithm according to claim 1 is characterized in that: The S2 specifically includes: S21. Use the k-nearest neighbor interpolation method to fill the missing values ​​in the sensor data. The formula is as follows: Among them, X new is the estimated value of the missing data, X i are the k data values ​​closest to the missing data; S22. Apply Kalman filtering to denoise the sensor data using the following recursive formula: in, is the estimated state at the current moment, is the estimated state at the previous moment, K k is the Kalman gain, z k is the observed value, H k is the observation matrix; S23. Use the quartile method to detect outliers. Outliers are defined as data that exceeds 1.5 times the range between the lower and upper quartiles. By eliminating these outliers, the accuracy of subsequent analysis is ensured. The formula is as follows: Outlier=(Data<Q1-1.5×(Q3-Q1))∨(Data> Q3+1.5×(Q3-Q1)); Among them, Outlier represents the quartile, Data represents the detected data, Q1 is the lower quartile, and Q3 is the upper quartile.

4. The method for automatic selection and arrangement of drive-in shelves in a warehouse based on a combined optimization algorithm according to claim 1 is characterized in that: The S3 specifically includes: S31. Discrete modeling of warehouse space, dividing the warehouse space into multiple optimization areas, each area can accommodate a specific type of goods, and define space utilization, material access efficiency and storage cost as the main optimization goals by modeling the attributes of various types of shelves; S32. Define the optimization objective function as a weighted sum of multiple objectives. The objective function is comprehensively optimized by balancing various objectives: F(x)=w1·U(x)+w2·E(x)+w3·C(x)+w4·L(x); Among them, F(x) is the comprehensive objective function, U(x) is the space utilization, E(x) is the access efficiency, C(x) is the storage cost, L(x) is the load balance, w1, w2, w3, w4 are the weight coefficients of the objective function, x is the decision variable of the layout plan, A i (x) represents the shelf area in the i-th area, O i (x) is the free area in the area not occupied by goods, A total is the total area of ​​the warehouse, f i (x) represents the access frequency of item i, T i (x) is the access time of item i, T total is the total access time of all goods in the warehouse, C i (x) is the basic cost of the i-th shelf, P i (x) is the material handling cost, D i (x) is the material handling distance, L i (x) is the load of the i-th shelf, L max For maximum load, ensure load uniformity of the layout; S33. Set constraints to ensure that the optimization results meet the actual needs of the warehouse. The mathematical model of the constraints is as follows: g1(x)≤0,g2(x)≤0,g3(x)≤0,g4(x)≤0,g5(x)≤0; Among them, g1(x) is the shelf load limit, g2(x) is the channel width limit, g3(x) is the material access path limit, g4(x) is the shelf spacing limit, and g5(x) is the material classification limit.

5. The method for automatic selection and arrangement of drive-in shelves in a warehouse based on a combined optimization algorithm according to claim 1 is characterized in that: The S4 specifically includes: S41, initialize the particle p of the particle swarm algorithm i , each particle represents a layout solution x i , initialize the particle's position x i (t) and speed v i (t), where position and speed represent the solution space of the layout scheme and the search speed of the particle respectively. The particle swarm is initialized by the following formula: x i (0)=Random Layout within Bounds; v i (0)=Initial Speed based on Layout Space; S42, calculate each particle p i The objective function value F(x i ), the objective function considers the warehouse space utilization U(x i ), access efficiency E(x i ), storage cost C(x i ) and load balancing L(x i ), and calculate the objective function value by the following formula: F(x i )=w1·U(x i )+w2·E(x i )+w3·C(x i )+w4·L(x i ); Among them, w1, w2, w3, and w4 are weight coefficients of the objective function, indicating the importance of space utilization, access efficiency, and storage cost in optimization; S43. According to the updating rule of the particle swarm algorithm, the speed and position of the particles are updated. The updating formula of the speed and position of the particles is: v i (t+1)=w·v i (t)+c1·r1·(pbest i -x i (t))+c2·r2·(gbest-x i (t)); x i (t+1)=x i (t)+v i (t+1); Among them, v i (t+10 is the speed of the ith particle, x i (t0 is the position of the i-th particle, pbest i is the individual optimal position of the i-th particle, gbest is the global optimal position, w is the inertia weight, c1, c2 are acceleration constants, r1, r2 are random numbers, and the search direction of the particle is updated; S44, introduce simulated annealing algorithm to avoid the particle swarm algorithm falling into the local optimal solution. The simulated annealing algorithm controls the search space of the solution by adjusting the temperature. The temperature update formula is: T(t+1)=T(t)·β; Where T(t) is the current temperature and β is the cooling factor; S45. In each iteration, the simulated annealing algorithm determines whether to accept the new solution by calculating the energy difference between the current solution and the new solution. The probability of acceptance is calculated by the following formula: Where, ΔF=F(x new 0-F(x old ) is the objective function difference between the new solution and the old solution. The probability P of accepting the new solution decreases as the temperature decreases, ensuring that the search process can jump out of the local optimal solution. S46. After each iteration, update the speed and position of the particles, and adjust the layout plan in combination with the acceptance mechanism of the simulated annealing algorithm.

6. The method for automatic selection and arrangement of drive-in shelves in a warehouse based on a combined optimization algorithm according to claim 1 is characterized in that: The S5 specifically includes: S51. Adopt the integrated learning characteristics of support vector machine and gradient boosting tree to intelligently evaluate the access efficiency E(x) and storage cost C(x) of each layout scheme, and optimize the improved kernel function and integrated learning model to improve the accuracy and robustness of the evaluation; S52. In the support vector machine, a high-order radial basis kernel function is used for mapping, which is defined as: Among them, ∥xy∥ 2 is the Euclidean distance between layout solutions x and y, σ is the kernel width, α is an additional adjustment parameter, γ is the inner product parameter,<x,y> is the inner product in the feature space, d is the order of the polynomial kernel function; S53. In the gradient boosting tree, the output of each decision tree is calculated by the residual method, and the splitting point of the decision tree node is optimized. In order to further improve the evaluation ability, the weighted residual sum of squares is used to train the output of each tree: Among them, y i is the true value of the i-th layout solution, f i-1 (x) is the prediction result of the first i-1 rounds, λ is the regularization coefficient, Represents the second-order gradient under the j-th feature dimension; S54, the total output f(x) of the gradient boosting tree is a weighted combination of multiple decision trees, calculated as: Among them, f i (x) is the output of the i-th decision tree, α i is the weight of each tree, γ i is the offset of the tree output, and II is the indicator function, which represents the feature x of the layout solution x. j The leaf node R belongs to the i-th tree ij , n is the number of trees, p is the feature dimension; S55, weighted fusion of the evaluation results of the support vector machine and the gradient boosting tree is performed to obtain the final weighted score S(x) of each layout scheme, and the weighting function is: in, is the output of SVM, w1,w2,w3 are weighted coefficients, f i (x) is the output of GBT, and finally the comprehensive evaluation score of each layout scheme is obtained; S56. Sort all layout plans according to the weighted score S(x), and select the layout plan with the highest score as the optimal plan. The selection of the optimal plan depends on the comprehensive results of multiple evaluation criteria to ensure that the warehouse layout plan can maximize the utilization of storage space and material access efficiency, and minimize storage costs.

7. The method for automatic selection and arrangement of drive-in shelves in a warehouse based on a combined optimization algorithm according to claim 1 is characterized in that: The S6 specifically includes: S61. According to the optimal layout plan, determine the selection and placement of each shelf, ensure the reasonable match between shelf type and warehouse space, and adjust the layout of shelves according to load and space requirements to improve the space utilization rate and material access efficiency of the warehouse; S62. According to the material storage requirements, the storage areas of various materials are automatically adjusted, and the placement of materials is optimized according to the frequency of material entry and exit, ensuring the shortest access path and improving the overall access efficiency.

8. The method for automatic selection and arrangement of drive-in shelves in a warehouse based on a combined optimization algorithm according to claim 1 is characterized in that: The S7 specifically includes: S71. In actual application, the access frequency, in-and-out speed and inventory changes of various types of goods are monitored in real time based on the dynamic data in the warehouse, and these data are compared with the optimization plan to adjust the current shelf layout to adapt it to the new warehouse operation status; S72. Automatically adjust shelf layout based on real-time monitoring data, update shelf locations, cargo storage areas, and access paths to respond to changes within the warehouse, ensure continuous optimization and adaptability of layout plans, and improve warehouse operating efficiency.

Images