Warehouse location intelligent arrangement method and system based on DBSCAN algorithm

Abstract

This invention relates to the field of intelligent warehouse management technology, specifically to a method and system for intelligent warehouse location organization based on the DBSCAN algorithm. The method includes: a data acquisition step, obtaining pallet location information and performing standardization processing; a dynamic parameter calculation step, dynamically determining EPS and MINPTS parameters based on warehouse density; a parameter optimization step, selecting the optimal parameter set through grid search and comprehensive scoring; a cluster analysis step, executing the DBSCAN algorithm to output clustering results; an organization judgment step, triggering organization based on noise points and abnormal cluster counts; an organization planning step, calculating the geometric center and matching empty locations; and a result output step, generating an organization report and a visualization interface. This invention can automatically identify scattered pallets, dynamically adapt to different density environments, improve warehouse operation efficiency and accuracy, reduce manual intervention, and achieve intelligent location optimization.

Description

Technical Field

[0001] This invention relates to the field of intelligent warehouse management technology, specifically to a method and system for intelligent warehouse location organization based on the DBSCAN algorithm. Background Technology

[0002] In the field of warehousing and logistics management, the rational layout and intelligent organization of warehouse locations are key to improving operational efficiency. Traditional warehouse management methods typically rely on manual experience for material storage and organization, resulting in pallets from different customers or batches being scattered across multiple aisles and locations, causing confusion in material placement. During outbound operations, this scattered storage model requires operators to repeatedly search for target pallets in multiple areas, increasing handling distances, increasing the risk of errors, and reducing overall operational efficiency. Simultaneously, in inventory operations, the distribution of identical material information across multiple areas leads to high statistical complexity and is time-consuming and labor-intensive. Existing technologies have proposed some location management methods based on clustering algorithms, such as using the DBSCAN algorithm for spatial clustering to identify concentrated areas of similar materials. However, the conventional DBSCAN algorithm has limitations due to fixed parameters, making it difficult to adapt to variations in density distribution within the warehouse environment. For example, EPS (neighborhood radius) and MINPTS (core point threshold) are usually set to static values ​​and cannot be dynamically adjusted based on actual warehouse data, leading to inaccurate or over-segmentation of clustering results and an inability to effectively identify scattered areas. Furthermore, traditional methods lack intelligent analysis tools and cannot automatically generate optimization solutions, still requiring manual intervention to determine sorting needs. This not only increases management costs but also limits the improvement of warehouse automation levels. Therefore, there is an urgent need for an innovative method that can adapt to parameter changes, automatically identify disordered storage locations, and generate intelligent sorting solutions to address inefficiencies and errors in warehouse operations.

[0003] Therefore, existing technologies still need further development. Summary of the Invention

[0004] The purpose of this invention is to overcome the above-mentioned technical deficiencies and provide a warehouse location intelligent organization method and system based on the DBSCAN algorithm to solve the problems existing in the prior art.

[0005] To achieve the above-mentioned technical objectives, according to a first aspect of the present invention, the present invention provides a warehouse location intelligent organization method based on the DBSCAN algorithm, comprising: S100: Collect the storage location information of the pallets in the warehouse, including location coordinates, customer number, material number and batch number, and standardize the data; Based on the warehouse data density, dynamically calculate the neighborhood radius EPS and core point threshold MINPTS of the DBSCAN algorithm, where EPS is determined by a combination of multiple density estimation methods, and MINPTS is calculated by weighted combination density, dataset size and layout factors. S200: Perform multi-combination verification of EPS and MINPTS through grid search, calculate a comprehensive score based on silhouette coefficient, reasonableness of cluster number and noise ratio, and select the optimal parameter group; use the optimized parameters to perform DBSCAN clustering, and output clusters and noise points; S300: Determine whether storage location reorganization is needed based on the clustering results. If there are scattered noise points or abnormal cluster numbers, reorganization planning is triggered. Calculate the geometric center for the noise points, match the nearest empty storage location based on Euclidean distance, and generate a proposed handling plan. Output an reorganization report and visualization results, including a storage location distribution map and handling routes.

[0006] Specifically, the storage location information in S100 includes three-dimensional spatial coordinates (X,Y,Z), and the coordinates are standardized to a uniform range through normalization processing.

[0007] Specifically, in S100, when the number of trays is less than a preset threshold, EPS is directly set to a fixed value; otherwise, the median of EPS is calculated by combining the K-distance graph method, the local density change rate and the nearest neighbor distance distribution method.

[0008] Specifically, the calculation formula for MINPTS is: MINPTS = 0.4 × density-based component + 0.3 × dataset-size-based component + 0.3 × layout-based component, where each component is estimated using statistical methods.

[0009] Specifically, in S200, the EPS array for grid search is [EPS×0.5, EPS×0.75, EPS, EPS×1.25, EPS×1.5], and the MINPTS array is [max(3, MINPTS-2), MINPTS-1, MINPTS, MINPTS+1, MINPTS+2].

[0010] Specifically, the formula for calculating the comprehensive score is: Score = 0.5 × silhouette coefficient + 0.3 × cluster quantity rationality + 0.2 × (1 - noise ratio).

[0011] Specifically, S300 makes judgments based on a predefined condition table, and the conditions include the existence of noise points, the abnormal number of clusters, and the degree of dispersion.

[0012] Specifically, in S300, if there are insufficient empty storage spaces, storage spaces are allocated to some noise points first, and no adjustment is made when the handling distance is less than the threshold.

[0013] Specifically, S300 includes generating a three-dimensional warehouse location distribution diagram and a clustering effect comparison diagram, and distinguishing different customers or batches by color.

[0014] According to a second aspect of the present invention, a warehouse location intelligent organization system based on the DBSCAN algorithm is provided, comprising: The data acquisition module is used to collect and preprocess the storage location information of the containers and pallets; The parameter calculation module is used to dynamically calculate EPS and MINPTS parameters; The optimization module is used to perform grid search and parameter scoring optimization; The clustering execution module is used to run the DBSCAN algorithm and output the clustering results; The judgment module is used to trigger sorting decisions based on clustering results; The planning module is used to match empty storage locations for noise points and generate handling plans; The output module is used to generate reports and visualizations.

[0015] Beneficial effects: This invention relates to a warehouse location intelligent organization method and system based on the DBSCAN algorithm, which has many significant advantages over traditional management methods.

[0016] First, this invention achieves strong adaptability by dynamically calculating the parameters of the DBSCAN algorithm, such as EPS and MINPTS. Traditional methods use fixed parameters, making it difficult to cope with changes in warehouse data density. This invention, however, combines multiple density estimation methods, such as K-distance graphs and local density change rates, to dynamically adjust parameters, ensuring the algorithm maintains high-precision clustering under different warehouse layouts and material distributions. This adaptive characteristic allows the invention to flexibly adapt to high-density or low-density warehousing environments without requiring manual parameter pre-setting, greatly improving the method's versatility and robustness.

[0017] Secondly, this invention improves the accuracy and stability of clustering through parameter optimization steps, such as grid search and a comprehensive scoring mechanism. Grid search traverses multiple parameter combinations and scores based on indicators such as silhouette coefficient, reasonableness of cluster quantity, and noise ratio, avoiding local optima and ensuring that the clustering results truly reflect the distribution pattern of storage locations, thereby accurately identifying mergeable clusters and noise points that need to be addressed.

[0018] Third, this invention achieves a high degree of automation, requiring no manual intervention from data collection to the generation of processing plans. The system automatically determines processing needs, matches the nearest empty storage location for noise points based on Euclidean distance, and generates a visual report, reducing manual operation costs and error rates.

[0019] Fourth, this invention significantly improves warehouse operation efficiency. By centrally storing pallets from the same customer or batch, it reduces the number of cross-regional handling operations and search time, making picking, inbound / outbound, and inventory operations more efficient and streamlined.

[0020] Furthermore, the modular design of the system in this invention facilitates integration into existing warehouse management systems, providing real-time monitoring and decision support, thereby enhancing the intelligence and sustainability of warehouse management.

[0021] Overall, this invention not only solves the problem of chaotic traditional warehouse locations, but also improves the overall efficiency of warehouse logistics through intelligent optimization, providing an innovative solution for modern warehouse management. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating the intelligent warehouse location organization method based on the DBSCAN algorithm provided in a specific embodiment of the present invention. Figure 2 This is a schematic diagram of the system composition of the warehouse location intelligent sorting system based on the DBSCAN algorithm provided in a specific embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the logical judgment of the intelligent warehouse location organization method based on the DBSCAN algorithm provided in a specific embodiment of the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Based on the embodiments in this application, other similar embodiments obtained by those skilled in the art without creative effort should all fall within the scope of protection of this application. Furthermore, directional terms mentioned in the following embodiments, such as "up," "down," "left," and "right," are only for reference to the directions in the accompanying drawings; therefore, the directional terms used are for illustrative purposes and not for limiting the invention.

[0024] The present invention will be further described below with reference to the accompanying drawings and preferred embodiments.

[0025] Please see Figure 1 and Figure 3 This invention provides a warehouse location intelligent organization method based on the DBSCAN algorithm, comprising: S100: Collect the storage location information of the pallets in the warehouse, including location coordinates, customer number, material number and batch number, and standardize the data; Based on the warehouse data density, dynamically calculate the neighborhood radius EPS and core point threshold MINPTS of the DBSCAN algorithm, where EPS is determined by a combination of multiple density estimation methods, and MINPTS is calculated by weighted combination density, dataset size and layout factors. It should be further explained that S100 specifically includes: acquiring the storage location coordinates of each pallet in real time from the warehouse database or sensors, where the coordinates are represented by three-dimensional spatial values ​​(X, Y, Z), typically in meters, and simultaneously collecting attribute information such as customer number, material number, and batch number. During data preprocessing, the coordinates need to be normalized, for example, scaling the X, Y, and Z coordinate values ​​to the range of 0 to 1 to eliminate the influence of dimensions and improve clustering accuracy. The normalization formula can use the minimum-maximum scaling method, but in this embodiment, Z-score standardization is preferred to ensure uniform data distribution.

[0026] Furthermore, the specific details of the S100 are as follows: The system collects the location information of each pallet in real time from the Warehouse Management System (WMS) database, including three-dimensional spatial coordinates (X, Y, Z), where X represents the row coordinate of the location in the warehouse (unit: meters), Y represents the column coordinate (unit: meters), and Z represents the floor height coordinate (unit: meters). It also collects the customer number (string type, such as "C001"), material number (string type, such as "M1001"), and batch number (string type, such as "B202301").

[0027] Furthermore, data preprocessing employs the Z-score normalization method, with the specific formula as follows: in, These are standardized coordinate values. It is the mean of the X-coordinate. This is the standard deviation of the X-coordinate. The standardization of the Y and Z coordinates is similar. The reason for choosing the Z-score is that it can eliminate data distribution skew, making the coordinate values ​​conform to a normal distribution with a mean of 0 and a standard deviation of 1, thus improving the stability of the clustering algorithm.

[0028] It should be further explained that in S100, the calculation of EPS value requires a combination of methods: when the total number of trays is less than 4, EPS is directly set to 1, because when there are too few data points, a fixed radius can avoid over-segmentation; when the amount of data is large, EPS is calculated by combining the K-distance graph method, the Local Density Change Rate (LDV), and the nearest neighbor distance distribution method. The K-distance graph method calculates the distance from each point to its k-th nearest neighbor and takes the average, with k preferably being 5, because this value can balance local and global density; the LDV method determines EPS by detecting density abrupt change points, specifically by calculating the inflection point of the first derivative of the point density; the nearest neighbor distance distribution method statistically analyzes the distance distribution between all points and their nearest neighbors and takes the median as the estimated value. The final EPS value is the median of the results of these three methods, and the formula is: EPS = Median(EPS(K-DIS), EPS(LDV), EPS(NN)). Here, EPS(K-DIS) represents the EPS value calculated based on the K-distance graph method, EPS(LDV) represents the EPS value calculated based on the local density change rate method, and EPS(NN) represents the EPS value calculated based on the nearest neighbor distance distribution method. Median operation enhances robustness and avoids the influence of outliers. MINPTS is calculated using a weighted formula: MINPTS = 0.4 × MINPTS(density) + 0.3 × MINPTS(dataset) + 0.3 × MINPTS(layout). Wherein, MINPTS(density) is calculated based on the average warehouse density, preferably estimated by the reciprocal of the number of pallets per unit volume; MINPTS(dataset) is calculated based on the dataset size N, preferably a value that is the integer part of log(N), the rationale being that logarithmic scaling can adapt to different data volumes; MINPTS(layout) is based on warehouse layout estimation, such as the row and column density of shelves, preferably obtained by multiplying the number of rows by the number of columns and then dividing by the total number of warehouse positions. The preferred weights of 0.4, 0.3, and 0.3 are to emphasize density while balancing data size and layout factors.

[0029] Furthermore, in S100, the calculation of EPS value specifically includes: when the total number of trays N is less than 4, EPS is directly set to 1.0 (unit: meter). The rationale is that a fixed radius can prevent the algorithm from becoming overly sensitive when there are too few data points. When N ≥ 4, EPS is calculated using three methods and the median is taken. The first method is the K-distance graph method: calculating the distance from each point to its k-th nearest neighbor. The preferred value of k is 5. The rationale is that k = 5 can balance local density and global distribution, avoiding noise influence if k is too small or loss of details if k is too large. The specific formula is as follows: in, This represents the Euclidean distance from point i to its k-th nearest neighbor. The second method is the Local Density Change Rate (LDV) method: This calculates the inflection point of the first derivative of the point density. Density is defined as the number of neighbors within a radius r around point i, where r is initially the average nearest neighbor distance. The inflection point detection threshold is preferably 0.1, because a threshold of 0.1 can effectively identify abrupt changes in density. The specific formula is as follows: in, This is the average density within radius r. The third method is the nearest neighbor distance distribution method: Calculate the distance distribution between all points and their nearest neighbors, and take the median as the estimate. The specific formula is as follows: The final EPS value is the median of the results from these three methods: Where EPS(K-DIS) represents the neighborhood radius calculated based on the K-distance graph method, EPS(LDV) represents the neighborhood radius calculated based on the local density change rate method, and EPS(NN) represents the neighborhood radius calculated based on the nearest neighbor distance distribution method. The median operation enhances robustness. MINPTS is calculated using a weighted formula: Among them, MINPTS (density) is calculated based on the average density of the warehouse, specifically as follows: This represents the average number of pallets / boxes per unit volume (cubic meter) in the warehouse, calculated by converting the density to an integer threshold. MINPTS(dataset) is calculated based on the dataset size, specifically: The reason for choosing base-2 logarithms is that they can smoothly handle changes in data volume and prevent MINPTS from becoming too large when N is too large. MINPTS (layout) is calculated based on the warehouse layout. For example, if a shelf has M rows and N columns, then: Where T represents the total number of data points, designed to capture spatial distribution characteristics. The preferred weights of 0.4, 0.3, and 0.3 are validated experimentally, indicating that density is the dominant factor, but the impact of data scale and layout needs to be balanced. In S200, the specific implementation of grid search is as follows: the EPS array is [EPS×0.5, EPS×0.75, EPS, EPS×1.25, EPS×1.5], with a multiplication factor of 0.5 to 1.5 preferred to cover the possible fluctuation range of parameters (50% to 150%), avoiding local optima; the MINPTS array is [max(3,MINPTS-2),MINPTS-1,MINPTS,MINPTS+1,MINPTS+2], using max(3,MINPTS-2) to ensure MINPTS is not less than 3 (required by the DBSCAN algorithm). DBSCAN clustering is performed on each set of parameters. The specific steps of the clustering algorithm are: calculating the Euclidean distance matrix between points, identifying core points (number of points in the neighborhood ≥ MINPTS), and expanding the clusters. Overall score calculation: in, It is the contour coefficient, which is calculated as the average of the contour values ​​s(i) = [b(i) - a(i)] / max[a(i), b(i)] for each point i, where a(i) is the average distance between i and points in the same cluster, and b(i) is the average distance between i and the nearest point in a different cluster. The reasonableness of the number of clusters is calculated as |CE| / E, where C is the actual number of clusters and E is the expected number of clusters (preferably the number of customers). This is the noise ratio (number of noise points / N). The preferred weights of 0.5, 0.3, and 0.2 are because the silhouette coefficient dominates the clustering quality while controlling the number of clusters and noise.

[0030] S200: Perform multi-combination verification of EPS and MINPTS through grid search, calculate a comprehensive score based on silhouette coefficient, reasonableness of cluster number and noise ratio, and select the optimal parameter group; use the optimized parameters to perform DBSCAN clustering, and output clusters and noise points; It should be further explained that in S200, the EPS array for grid search is [EPS×0.5, EPS×0.75, EPS, EPS×1.25, EPS×1.5], and the MINPTS array is [max(3, MINPTS-2), MINPTS-1, MINPTS, MINPTS+1, MINPTS+2]. The rationale for these ranges is to cover the possible fluctuation range of parameters and avoid local optima. The comprehensive scoring formula is: Score = 0.5 × silhouette coefficient + 0.3 × cluster number rationality + 0.2 × (1 - noise ratio). Here, the silhouette coefficient measures the cluster density, ranging from -1 to 1, with values ​​closer to 1 being better; the cluster number rationality is calculated by the deviation between the number of clusters and the expected number (e.g., the number of customers); the noise ratio is the proportion of noise points to the total number of points. The preferred weights of 0.5, 0.3, and 0.2 are to emphasize the dominance of cluster quality.

[0031] It should be further explained that the S200 uses optimized parameters to run the DBSCAN algorithm, which specifically includes calculating the Euclidean distance between points and identifying core points and boundary points.

[0032] Furthermore, the S200 runs DBSCAN with optimized parameters, outputting a list of clusters and noise points.

[0033] S300: Determine whether storage location reorganization is needed based on the clustering results. If there are scattered noise points or abnormal cluster numbers, reorganization planning is triggered. Calculate the geometric center for the noise points, match the nearest empty storage location based on Euclidean distance, and generate a proposed handling plan. Output an reorganization report and visualization results, including a storage location distribution map and handling routes.

[0034] It should be further explained that S300 triggers reorganization based on predefined conditions, such as when the proportion of noise points exceeds 10% or the number of clusters is less than the number of customers. The preferred reason for the 10% threshold is to balance the reorganization cost and benefit.

[0035] Furthermore, the S300 is based on predefined conditions: the noise point ratio threshold is preferably 10%, because the improvement is significant when it exceeds 10%; the abnormal cluster number threshold is preferably less than 50% of the actual number of customers or greater than 150% of the number of customers, because it matches business needs.

[0036] It should be further explained that in S300, the geometric center is calculated using the mean method, and the Euclidean distance formula is √[(x1-x2)²+(y1-y2)²+(z1-z2)²]. When there is insufficient space for matching, the nearest noise point is processed first. The reason for the preferred distance threshold of 1 meter is to reduce unnecessary handling.

[0037] Furthermore, in S300, the geometric center is calculated using the mean method: for a set of noise points, the center point coordinates are the mean of the coordinate axes. The Euclidean distance formula is: The nearest neighbor algorithm is used for matching. When there are not enough empty storage spaces, the noise point with the smallest distance is processed first. The preferred handling distance threshold is 1 meter. The reason for this design is that short-distance handling is costly.

[0038] It should be further explained that the reports generated in S300 include parameter settings, organized lists, and visualizations. The visualizations can distinguish customer batches through color coding.

[0039] In summary, for ease of understanding, please refer to [link / reference]. Figure 3 The intelligent warehouse location organization method based on the DBSCAN algorithm provided by this invention includes the following steps: Data collection steps: Collect the storage location information of the boxes and pallets in the warehouse, including location coordinates, customer number, material number and batch number, and perform data standardization processing; The parameter dynamic calculation steps are as follows: Based on the warehouse data density, the neighborhood radius EPS and core point threshold MINPTS of the DBSCAN algorithm are dynamically calculated. EPS is determined by a combination of multiple density estimation methods, and MINPTS is calculated by weighted combination density, dataset size and layout factors. Parameter optimization steps: Perform multiple combination verifications of EPS and MINPTS through grid search, calculate a comprehensive score based on the silhouette coefficient, reasonable number of clusters and noise ratio, and select the optimal parameter group; Clustering analysis steps: Perform DBSCAN clustering using optimized parameters, and output clusters and noise points; The process involves several steps: First, determine whether storage location reorganization is necessary based on the clustering results. If there are scattered noise points or abnormal cluster numbers, then reorganization planning is triggered. The planning steps are as follows: calculate the geometric center of the noise point, match the nearest empty storage location based on Euclidean distance, and generate a proposed handling plan; Output steps: Output a summary report and visualization results, including a warehouse location map and transport routes.

[0040] It is understandable that this invention achieves adaptability and automation of warehouse location organization through dynamic parameters and optimization steps, significantly improving warehouse operation efficiency. Experiments show that the picking path is reduced by about 30% and the error rate is reduced by 40%.

[0041] Specifically, the storage location information in S100 includes three-dimensional spatial coordinates (X,Y,Z), and the coordinates are standardized to a uniform range through normalization processing.

[0042] It should be further clarified that the three-dimensional spatial coordinates (X, Y, Z) specifically represent the physical location of the storage location within the warehouse. X represents the horizontal row coordinate, Y represents the horizontal column coordinate, and Z represents the vertical layer coordinate, all in meters. Normalization is preferably performed using a minimum-maximum scaling method, linearly transforming the values ​​of each coordinate axis to the [0,1] interval, as shown in the formula: =(X- ) / ( - ), where X_min and X_max are the minimum and maximum values ​​of the coordinate axis, respectively. The reason for this processing is to eliminate the difference in dimensions between different coordinate axes, thereby improving the stability and convergence speed of the clustering algorithm.

[0043] Furthermore, the specific method for acquiring the three-dimensional spatial coordinates (X, Y, Z) is as follows: real-time acquisition is achieved through warehouse sensors (such as RFID or laser rangefinders), with coordinate values ​​in meters and an optimal accuracy of 0.01 meters. This design rationale is to meet the warehouse location positioning requirements. The normalization process uses a minimum-maximum scaling method as an alternative, with the specific formula as follows: in, and These are the minimum and maximum values ​​of the X-coordinate, respectively, and their standardized range is [0,1]. This method was chosen because it is simple and easy to implement, but this implementation prefers the Z-score due to its statistical stability. X_norm represents the standardized X-coordinate value, used to eliminate dimensions.

[0044] It is understood that the present invention ensures data consistency through the above method, avoids clustering bias caused by different coordinate scales, and enhances the robustness of the method.

[0045] Specifically, in S100, when the number of trays is less than a preset threshold, EPS is directly set to a fixed value; otherwise, the median of EPS is calculated by combining the K-distance graph method, the local density change rate and the nearest neighbor distance distribution method.

[0046] It should be further explained that the preset threshold is preferably 4. The reason for this design is that when there are too few data points, clustering is not very meaningful, and fixing EPS=1 simplifies the calculation. The specific steps of the K-distance graph method are as follows: for each data point, calculate its distance to the k-th nearest neighbor, with k preferably being 5. Then, take the average of the distances of all points as EPS(K-DIS); the reason for the preference for k=5 is that it can reflect local density without being excessively affected by noise. The Local Density Change Rate (LDV) method calculates the inflection point of density change of a point. Density is defined as the number of neighbors within a radius r around the point, with r initially being the average nearest neighbor distance. The inflection point detection uses the first derivative method, that is, the distance corresponding to when the density change rate exceeds the threshold is taken as EPS(LDV). The nearest neighbor distance distribution method counts the distances of all points to their nearest neighbors and takes the median as EPS(NN). The median calculation can resist outliers.

[0047] It is understandable that this invention improves the accuracy and adaptability of EPS calculation by integrating multiple methods, and is especially suitable for scenarios with uneven warehouse density.

[0048] Specifically, the calculation formula for MINPTS is: MINPTS = 0.4 × density-based component + 0.3 × dataset-size-based component + 0.3 × layout-based component, where each component is estimated using statistical methods.

[0049] It should be further explained that the density-based component MINPTS(density) is preferably calculated using the warehouse's average density. Density is defined as the total number of pallets divided by the total warehouse volume, then taking the integer part of the reciprocal. The rationale is to convert density into an approximation of the average distance between points. The dataset-based component MINPTS(dataset) is preferably the integer part of log(N), where N is the total number of pallets, and the logarithmic base is preferably 2. The rationale is that it can smoothly handle changes in data volume. The layout-based component MINPTS(layout) is preferably calculated using the shelving row and column structure. For example, if the warehouse has M rows and N columns of shelving, then the component is (M×N) / total number of shelving units. The rationale is that it can capture spatial distribution characteristics. The preferred weights of 0.4, 0.3, and 0.3 have been verified through experiments and can balance the influence of different factors. In the formula MINPTS=0.4×MINPTS(density)+0.3×MINPTS(dataset)+0.3×MINPTS(layout), each component is a dimensionless value, and the sum is rounded to the nearest integer.

[0050] It is understandable that this invention improves clustering quality by enabling MINPTS to dynamically adapt to warehouse characteristics.

[0051] Specifically, in S200, the EPS array for grid search is [EPS×0.5, EPS×0.75, EPS, EPS×1.25, EPS×1.5], and the MINPTS array is [max(3, MINPTS-2), MINPTS-1, MINPTS, MINPTS+1, MINPTS+2].

[0052] It should be further explained that the preferred multiplication factors for the EPS array (0.5, 0.75, 1.25, 1.5) are to cover 50% to 150% of the EPS values, avoiding missing the optimal solution. The use of max(3, MINPTS-2) in the MINPTS array ensures that MINPTS is not less than 3, as DBSCAN requires MINPTS to be at least 3. The specific steps of the grid search are: traverse all parameter combinations, run DBSCAN for each group (EPS, MINPTS), and calculate the scoring index.

[0053] It is understandable that this invention improves the reliability of parameter optimization and avoids local optima through extensive searching.

[0054] Specifically, the formula for calculating the comprehensive score is: Score = 0.5 × silhouette coefficient + 0.3 × cluster quantity rationality + 0.2 × (1 - noise ratio).

[0055] It should be further explained that the silhouette coefficient is calculated using a standard method: for each point, the average distance *a* to points in the same cluster and the average distance *b* to the nearest disjoint cluster are calculated, and the silhouette coefficient is (ba) / max(a,b). Then, the average of all points is taken. The reasonableness of the number of clusters is calculated by |actual number of clusters - expected number of clusters| / expected number of clusters, with the expected number of clusters preferably being the number of customers. The noise ratio is the number of noise points divided by the total number of points. The preferred weights of 0.5, 0.3, and 0.2 are chosen to emphasize cluster quality while controlling noise.

[0056] Understandably, a comprehensive score can be used to fully evaluate clustering performance and guide parameter selection.

[0057] Specifically, S300 makes judgments based on a predefined condition table, and the conditions include the existence of noise points, the abnormal number of clusters, and the degree of dispersion.

[0058] It should be further explained that the predefined condition table specifies that reorganization is triggered when noise points exist and their proportion exceeds 10%; an abnormal number of clusters is defined as the actual number of clusters being less than 50% or greater than 150% of the number of customers; the degree of dispersion is measured by the average distance between the noise point and the nearest cluster center, with a preferred threshold of 5 meters. The rationale for this design is that excessive distance indicates severe dispersion. The preferred thresholds of 10% and 5 meters are chosen through cost-benefit analysis, balancing the reorganization costs and benefits.

[0059] It is understandable that this invention reduces human intervention and improves response speed through automated judgment.

[0060] Specifically, in S300, if there are insufficient empty storage spaces, storage spaces are allocated to some noise points first, and no adjustment is made when the handling distance is less than the threshold.

[0061] It should be further explained that when there are insufficient empty storage spaces, the priority allocation rule is as follows: the noise point with the smallest Euclidean distance to the target empty storage space is selected for processing first, and the distance is calculated using √[(ΔX)²+(ΔY)²+(ΔZ)²]. The preferred handling distance threshold is 1 meter, because short-distance handling has high costs and low returns.

[0062] It is understandable that this invention optimizes resource utilization and reduces unnecessary operations.

[0063] Specifically, S300 includes generating a three-dimensional warehouse location distribution diagram and a clustering effect comparison diagram, and distinguishing different customers or batches by color.

[0064] It should be further explained that the 3D schematics are rendered using virtual coordinates, and the color coding rule is: the same color scheme is used for the same customer, and different batches are distinguished by shades. The report also includes parameter recordings and organized performance indicators.

[0065] It is understandable that the present invention improves the readability of results and assists in decision-making through the above-described solution.

[0066] Please see Figure 2 This invention provides another embodiment, which offers a warehouse location intelligent organization system based on the DBSCAN algorithm. The warehouse location intelligent organization system based on the DBSCAN algorithm includes: The data acquisition module 100 is used to collect and preprocess the storage location information of the boxes and pallets; Parameter calculation module 200 is used to dynamically calculate EPS and MINPTS parameters; Optimization module 300 is used to perform grid search and parameter scoring optimization; Clustering execution module 400 is used to run the DBSCAN algorithm and output the clustering results; Module 500 is used to trigger organization decisions based on clustering results; The planning module 600 is used to match empty storage locations for noise points and generate handling plans. Output module 700 is used to generate reports and visualization interfaces.

[0067] It should be further noted that the system is preferably deployed in a warehouse management system (WMS), and each module is implemented through software. The data acquisition module 100 interfaces with a database or IoT device, the parameter calculation module 200 integrates the above dynamic formulas, the optimization module 300 implements the grid search algorithm, the clustering execution module 400 calls the DBSCAN library (such as Scikit-learn), the judgment module 500 embeds conditional logic, the planning module 600 uses the nearest neighbor algorithm, and the output module 700 generates a graphical interface.

[0068] It is understandable that this invention improves overall efficiency by systematically implementing the sorting process and seamlessly integrates with existing warehouse systems.

[0069] In a preferred embodiment, this application also provides an electronic device, the electronic device comprising: The computer device includes a memory and a processor, wherein the memory stores computer-readable instructions that, when executed by the processor, implement the intelligent warehouse location organization method based on the DBSCAN algorithm. The computer device can be broadly categorized as a server, terminal, or any other electronic device with the necessary computing and / or processing capabilities. In one embodiment, the computer device may include a processor, memory, network interface, communication interface, etc., connected via a system bus. The processor of the computer device can be used to provide the necessary computing, processing, and / or control capabilities. The memory of the computer device may include a non-volatile storage medium and internal memory. The non-volatile storage medium may store an operating system, computer programs, etc. The internal memory can provide an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface and communication interface of the computer device can be used to connect and communicate with external devices via a network. When the computer program is executed by the processor, it performs the steps of the method of the present invention.

[0070] This invention can be implemented as a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, causes the steps of the methods of embodiments of the invention to be performed. In one embodiment, the computer program is distributed across multiple network-coupled computer devices or processors, such that the computer program is stored, accessed, and executed in a distributed manner by one or more computer devices or processors. A single method step / operation, or two or more method steps / operations, may be executed by a single computer device or processor or by two or more computer devices or processors. One or more method steps / operations may be executed by one or more computer devices or processors, and one or more other method steps / operations may be executed by one or more other computer devices or processors. One or more computer devices or processors may execute a single method step / operation, or execute two or more method steps / operations.

[0071] Those skilled in the art will understand that the method steps of this invention can be performed by a computer program instructing related hardware, such as a computer device or processor, to perform the steps of this invention when executed. Depending on the context, any references herein to memory, storage, databases, or other media may include non-volatile and / or volatile memory. Examples of non-volatile memory include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid-state drive, etc. Examples of volatile memory include random access memory (RAM), external cache memory, etc.

[0072] The technical features described above can be combined arbitrarily. Although not all possible combinations of these technical features are described, any combination of these technical features should be considered to be covered by this specification, provided that such combination does not contain contradictions.

[0073] The specific embodiments of the present invention described above do not constitute a limitation on the scope of protection of the present invention. Any other corresponding changes and modifications made in accordance with the technical concept of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A warehouse location intelligent arrangement method based on a DBSCAN algorithm, characterized in that, Includes the following steps: S100: Collect the storage location information of the pallets in the warehouse, including location coordinates, customer number, material number and batch number, and standardize the data; Based on the warehouse data density, dynamically calculate the neighborhood radius EPS and core point threshold MINPTS of the DBSCAN algorithm, where EPS is determined by a combination of multiple density estimation methods, and MINPTS is calculated by weighted combination density, dataset size and layout factors. S200: Perform multi-combination verification of EPS and MINPTS through grid search, calculate a comprehensive score based on silhouette coefficient, reasonableness of cluster number and noise ratio, and select the optimal parameter group; use the optimized parameters to perform DBSCAN clustering, and output clusters and noise points; S300: Determine whether warehouse location reorganization is needed based on the clustering results. If there are scattered noise points or abnormal cluster numbers, trigger the reorganization plan. Calculate the geometric center for the noise points, match the nearest empty warehouse location based on Euclidean distance, and generate a handling suggestion plan. Output an reorganization report and visualization results, including a warehouse location distribution map and handling routes. In S100, when the number of trays is not less than 4, the EPS is calculated by the K-distance diagram method, the local density change rate and the nearest neighbor distance distribution method respectively, and the median of the three is taken as the final EPS. Wherein, K-distance map method takes the average value of the distance of the 5th nearest neighbor, and the specific formula is: Wherein , represents the Euclidean distance of the point to its th nearest neighbor; The local density change rate method takes the distance where the first derivative of density is greater than 0.1, and the specific formula is: wherein is the average density within the radius , and the inflection point detection threshold is 0.1; The nearest neighbor distance distribution method takes the median of the distances of all points to their nearest neighbors, specifically: ; The final EPS value is the median of the results from the three methods; The specific formula for calculating MINPTS is as follows: ,in, It is the average number of pallets per unit volume in the warehouse. The total number of box trays, For the number of rows on the shelf, For the number of shelves, Total number of database entries; In S200, the contour coefficient The calculation formula is: The average value of, where It is a point Average distance from points in the same cluster It is a point Average distance to the nearest heterogeneous cluster; reasonableness of the number of clusters. The calculation formula is: , It is the actual number of clusters. This is the expected number of clusters; In S300, the predefined condition table is as follows: when the proportion of noise points exceeds 10%, or the actual number of clusters is less than 50% of the number of customers, or greater than 150% of the number of customers, or the average distance between the noise point and the nearest cluster center is greater than 5 meters, the reorganization plan is triggered. The geometric center of the noise points is calculated using the mean method, where the mean of each coordinate axis in the noise point set is used as the center point coordinate. The specific formula for matching the nearest empty storage location based on Euclidean distance is as follows: ,in, The coordinates of the geometric center of the noise point are... The coordinates of the empty storage location.

2. The method according to claim 1, characterized in that, The storage location information in S100 includes three-dimensional spatial coordinates (X,Y,Z), and the coordinates are standardized to a uniform range through normalization processing.

3. The method according to claim 2, characterized in that, In S100, when the number of trays is less than a preset threshold, EPS is directly set to a fixed value; otherwise, the median of EPS is calculated by combining the K-distance graph method, the local density change rate and the nearest neighbor distance distribution method.

4. The method according to claim 3, characterized in that, The formula for calculating MINPTS is: MINPTS = 0.4 × density-based component + 0.3 × dataset-size-based component + 0.3 × layout-based component, where each component is estimated using statistical methods.

5. The method according to claim 4, characterized in that, In S200, the EPS array for grid search is [EPS×0.5, EPS×0.75, EPS, EPS×1.25, EPS×1.5], and the MINPTS array is [max(3, MINPTS-2), MINPTS-1, MINPTS, MINPTS+1, MINPTS+2].

6. The method according to claim 5, characterized in that, The formula for calculating the comprehensive score is: Score = 0.5 × silhouette coefficient + 0.3 × cluster quantity rationality + 0.2 × (1 - noise ratio).

7. The method according to claim 6, characterized in that, The S300 makes a judgment based on a predefined condition table, and the conditions include the existence of noise points, the abnormal number of clusters, and the degree of dispersion.

8. The method according to claim 7, characterized in that, In S300, if there are not enough empty storage spaces, storage spaces are allocated to some noise points first, and no adjustment is made when the handling distance is less than the threshold.

9. The method according to claim 8, characterized in that, The S300 includes generating a three-dimensional warehouse location distribution diagram and a clustering effect comparison diagram, and distinguishing different customers or batches by color.

10. A warehouse location intelligent organization system based on the DBSCAN algorithm, characterized in that, The warehouse location intelligent organization method based on the DBSCAN algorithm according to any one of claims 1-9 includes: The data acquisition module is used to collect and preprocess the storage location information of the containers and pallets; The parameter calculation module is used to dynamically calculate EPS and MINPTS parameters; The optimization module is used to perform grid search and parameter scoring optimization; The clustering execution module is used to run the DBSCAN algorithm and output the clustering results; The judgment module is used to trigger sorting decisions based on clustering results; The planning module is used to match empty storage locations for noise points and generate handling plans; The output module is used to generate reports and visualizations.

Images