A warehouse scheduling method based on cooperation of WMS and AGV

Abstract

The present application relates to the technical field of warehouse logistics, and discloses a warehouse scheduling method based on WMS and AGV cooperation. The method comprises the following steps: constructing a dynamic storage location data center, integrating parameters such as storage location occupation state and material turnover rate; using a multi-factor decision algorithm to generate a dynamic storage location allocation scheme; generating an AGV carrying task chain according to a real-time warehouse-in and warehouse-out instruction and a task decomposition rule; constructing a multi-AGV cooperative path planning network using a space-time map; inputting the storage location allocation scheme and the path planning network into a coupling scheduling model to generate an AGV cluster scheduling instruction set; and adjusting an AGV travel trajectory through a collision avoidance optimization algorithm to output a final scheduling strategy. The present application realizes the intelligentization and high efficiency of warehouse scheduling, improves the warehouse space utilization rate, operation efficiency and safety, reduces operation cost, and is suitable for various warehouse logistics scenarios.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of warehouse logistics, in particular to a warehouse scheduling method based on WMS and AGV cooperation. BACKGROUND

[0002] In the modern warehouse logistics industry, with the rapid development of e-commerce, manufacturing and other fields, the warehouse scale is continuously expanding, the types of goods and the frequency of warehouse in and out are significantly increasing, and the traditional warehouse scheduling method is facing many challenges. Early warehouse management relies on manual experience for warehouse allocation and task scheduling, which is low in efficiency and prone to errors. Manual recording of warehouse information makes it difficult to keep track of warehouse occupancy in real time, and it is difficult to find materials, resulting in long time for goods to enter and exit the warehouse and low utilization rate of warehouse space. For example, in a large e-commerce warehouse, the amount of goods increases dramatically during promotional activities, and manual scheduling cannot respond in time, resulting in serious goods backlog.

[0003] With the development of technology, some warehouses have introduced warehouse management systems (WMS) to realize the informatization management of warehouse and inventory, which has improved the efficiency of warehouse management to some extent. However, WMS often only relies on simple rules such as first-in-first-out and fixed partitioning when allocating storage space, without fully considering key factors such as material turnover rate, storage environment requirements, and in and out frequency. This makes it possible for materials to be placed in unreasonable warehouse locations, increasing the distance and time cost of transportation.

[0004] At the same time, the application of automated guided vehicles (AGV) in warehouses is gradually widespread, which can realize automated transportation and reduce labor costs. However, in the scenario of multiple AGV cooperation, the problem of path planning is highlighted. Existing path planning methods cannot accurately predict path conflicts, resulting in frequent collisions and congestion between AGVs, reducing the overall transportation efficiency. For example, in a high-density storage warehouse, the number of AGVs is large, and traditional path planning is difficult to effectively coordinate, often resulting in multiple vehicles "fighting for the road" and increasing the time of work stoppage.

[0005] In addition, there is a lack of effective cooperation mechanism between WMS and AGV. The data exchange between the two is not smooth, and the information is not updated in time, making it impossible to form an organic whole for warehouse scheduling. For example, the task instructions issued by WMS cannot be accurately received and executed by AGV in time; the real-time state information of AGV cannot be fed back to WMS in time, making it impossible for WMS to adjust the scheduling strategy according to the actual situation. SUMMARY

[0006] The present application aims to provide a warehouse scheduling method based on WMS and AGV cooperation to solve the problems raised in the background.

[0007] To achieve the above object, the application provides the following technical scheme: a warehouse scheduling method based on WMS and AGV cooperation, the method comprises:

[0008] A dynamic storage location data center is constructed, the dynamic storage location data center contains storage location occupation state, material turnover rate and material attribute correlation parameters; the material attribute correlation parameters include material classification code, storage environment requirement and storage frequency;

[0009] Based on the dynamic storage location data center, a dynamic storage allocation scheme is generated through a multi-factor decision algorithm; the multi-factor decision algorithm includes turnover rate weight adjustment, correlation matching degree calculation and storage location distance coefficient optimization;

[0010] According to real-time storage and retrieval instructions, an AGV carrying task chain is generated through task decomposition rules; the task decomposition rules include order urgency classification, AGV load balancing degree and task node dependency relationship;

[0011] Based on the AGV carrying task chain, a multi-AGV cooperative path planning network is constructed through a space-time map; the space-time map includes storage area obstacle distribution, AGV real-time position and path conflict prediction area;

[0012] The dynamic storage allocation scheme and the multi-AGV cooperative path planning network are input into a coupling scheduling model to generate an AGV cluster scheduling instruction set;

[0013] Based on the AGV cluster scheduling instruction set, an AGV travel trajectory is adjusted through a collision avoidance optimization algorithm to output a final scheduling strategy.

[0014] Preferably, the dynamic storage location data center is constructed, comprising:

[0015] Historical storage and retrieval data are collected to extract material turnover rate features and classification correlation matrix;

[0016] The storage area level is divided according to the physical parameters of the storage location, and the temperature and humidity sensitive area is marked;

[0017] The material turnover rate features, the classification correlation matrix and the storage area level are spatially mapped to generate a dynamic priority sequence of the storage location;

[0018] The dynamic priority sequence is used to update the storage location occupation state to form a dynamic storage location data center.

[0019] Preferably, the dynamic storage allocation scheme is generated through a multi-factor decision algorithm, comprising:

[0020] The product of the material turnover rate and the storage location distance coefficient is calculated as the first allocation weight;

[0021] A second allocation weight is generated based on the matching degree between the material classification code and the storage environment requirements.

[0022] The first allocation weight and the second allocation weight are integrated by a weighted fusion algorithm to generate a dynamic reserve score;

[0023] Based on the dynamic storage location scoring and ranking results, select the target storage location and update the allocation plan.

[0024] Preferably, the step of generating the AGV handling task chain through task decomposition rules includes:

[0025] Parse the order urgency tags in inbound and outbound instructions to classify task priority levels;

[0026] Calculate the available handling capacity value based on the AGV's current load status and remaining battery level;

[0027] Based on the matching relationship between task priority levels and available transport capacity values, a task node dependency graph is generated;

[0028] Convert the task node dependency graph into an AGV transport task chain and mark the task switching trigger conditions.

[0029] Preferably, the step of constructing a multi-AGV collaborative path planning network through a spatiotemporal map includes:

[0030] Collect 3D point cloud data of the reservoir area and mark the boundaries of fixed obstacles and temporary occupied areas;

[0031] Dynamic passage rights are generated based on the real-time location of the AGV, and path conflict prediction areas are marked.

[0032] Dynamic access rights and path conflict prediction areas are gridded to construct a spatiotemporal map;

[0033] A multi-AGV collaborative path planning network is generated based on a spatiotemporal map, and path priority labels are marked.

[0034] Preferably, the coupled scheduling model includes a path optimization module, which includes:

[0035] Extract the path conflict prediction region from the multi-AGV cooperative path planning network and calculate the congestion penalty factor;

[0036] Dynamically adjust path priority labels based on historical execution data of the AGV cluster scheduling instruction set;

[0037] An optimized path set is generated by balancing the congestion penalty factor and the path priority label through an iterative optimization algorithm.

[0038] The optimized path set is spatiotemporally aligned with the dynamic storage location allocation scheme to output the AGV cluster scheduling instruction set.

[0039] Preferably, the anti-collision optimization algorithm includes:

[0040] Based on the AGV's driving trajectory, predict the position coordinates of the next N time steps and generate a trajectory conflict detection matrix;

[0041] Calculate the safe distance offset between each AGV based on the trajectory conflict detection matrix;

[0042] The AGV speed or path steering angle is adjusted by using a trajectory replanning algorithm to eliminate safety distance offset;

[0043] The adjusted driving trajectory is encoded into the final scheduling strategy, and the spatiotemporal map is updated synchronously.

[0044] Preferably, the method for constructing the classification association matrix includes:

[0045] Calculate the co-occurrence frequency of materials in historical inbound and outbound orders to generate an initial correlation matrix;

[0046] The initial correlation matrix is ​​normalized according to the material classification coding level;

[0047] The classification association matrix is ​​updated based on the normalization results, and the priority sequence of the dynamic storage location data center is associated with it.

[0048] Preferably, the method for generating the dynamic passage right value includes:

[0049] The Euclidean distance between the AGV and the target storage location is calculated based on the AGV's real-time position to generate a basic passage weight value;

[0050] The basic passage weight value is attenuated and compensated based on the remaining time of the AGV task and the remaining battery power.

[0051] By combining the congestion level of the predicted route conflict area, the passage weight value is dynamically adjusted, and the adjusted passage weight value is mapped to the gridded encoding of the spatiotemporal map.

[0052] Preferably, the present invention also includes a warehouse scheduling system based on WMS and AGV collaboration, the system comprising:

[0053] Dynamic storage location modeling module: used to build a dynamic storage location data center, which includes storage location occupancy status, material turnover rate and material attribute association parameters;

[0054] Storage location allocation module: configured to generate dynamic storage location allocation schemes based on a dynamic storage location data center and through a multi-factor decision algorithm;

[0055] Task decomposition module: Used to generate AGV handling task chains based on real-time inbound and outbound instructions and task decomposition rules;

[0056] Path planning module: Used to construct a multi-AGV collaborative path planning network based on the AGV handling task chain using a spatiotemporal map;

[0057] Coupled scheduling module: Inputs the dynamic storage space allocation scheme and the multi-AGV collaborative path planning network into the coupled scheduling model to generate an AGV cluster scheduling instruction set;

[0058] Anti-collision execution module: Based on the AGV cluster scheduling instruction set, it adjusts the AGV driving trajectory through anti-collision optimization algorithm and outputs the final scheduling strategy.

[0059] Compared with the prior art, the beneficial effects of the present invention are:

[0060] This invention provides a warehouse scheduling method and system based on WMS and AGV collaboration. Addressing several key aspects, it effectively solves the problems of traditional warehouse scheduling, bringing significant benefits in multiple areas. In terms of warehouse location management, a dynamic warehouse location data center is constructed to comprehensively integrate information such as warehouse location occupancy status, material turnover rate, and material attribute correlation parameters. By extracting material turnover rate characteristics and classification correlation matrices from historical inbound and outbound data, and combining this with warehouse area hierarchy division, a precise dynamic priority sequence of warehouse locations is generated, enabling dynamic management of warehouse locations. This allows materials to be rationally allocated to suitable warehouse locations based on their characteristics and flow frequency. For example, materials with high turnover rates are preferentially allocated to warehouse locations near entrances and exits to reduce handling distances; highly correlated materials are stored in adjacent warehouse locations for convenient simultaneous retrieval, greatly improving warehouse space utilization and material storage and retrieval efficiency, and avoiding the space waste and search difficulties caused by traditional fixed warehouse location allocation methods.

[0061] In the storage location allocation process, a multi-factor decision-making algorithm comprehensively considers turnover rate weight adjustment, correlation matching degree calculation, and storage location distance coefficient optimization. It generates weights by calculating the product of material turnover rate and storage location distance coefficient, and by calculating the matching degree between material classification codes and storage environment requirements. These weighted fusions yield a dynamic storage location score, which is then used to select the target storage location. Compared to traditional single-factor storage location allocation methods, this scientific decision-making approach can more accurately match materials and storage locations, further improving warehousing efficiency and reducing operating costs.

[0062] Faced with real-time inbound and outbound instructions, a reasonable AGV handling task chain is generated based on task decomposition rules according to order urgency levels, AGV load balancing, and task node dependencies. Order urgency tags are parsed to prioritize tasks, and the available handling capacity is calculated by combining the current AGV load and battery level. A task node dependency graph is constructed and converted into a task chain, while task switching trigger conditions are marked. This ensures that urgent tasks are processed first, AGV load is balanced, and unreasonable task allocation prevents some AGVs from being overworked or idle, thus improving the overall operating efficiency and lifespan of the AGVs.

[0063] The construction of a multi-AGV collaborative path planning network utilizes a spatiotemporal map. It collects 3D point cloud data of the warehouse area, labels obstacles and temporary occupation areas, generates dynamic passage weights based on the real-time AGV positions, and marks predicted path conflict areas. A gridded map is then constructed, generating a planning network and path priority labels. This enables AGVs to avoid obstacles and conflict areas in real time during operation, selecting the optimal path, effectively reducing path conflicts and congestion, and significantly improving the efficiency and safety of multi-AGV collaborative operations.

[0064] The coupled scheduling model organically combines a dynamic storage location allocation scheme with a multi-AGV collaborative path planning network. Its path optimization module generates an optimized path set by calculating congestion penalty factors, dynamically adjusting path priority labels, and using iterative optimization algorithms. This optimized path set is then aligned with the storage location allocation scheme in both time and space, outputting a precise AGV cluster scheduling instruction set. This collaborative scheduling approach ensures smooth connection of warehousing operations, avoids disconnects and conflicts between different stages, and improves the overall accuracy and efficiency of scheduling.

[0065] The collision avoidance optimization algorithm predicts the future position of the AGV based on its travel trajectory, generates a trajectory conflict detection matrix, calculates the safe distance offset, and then uses a trajectory replanning algorithm to adjust the speed or path steering angle to eliminate the offset, updating the travel trajectory and spatiotemporal map. This algorithm provides a solid guarantee for the safe operation of AGVs in the warehousing environment, effectively avoiding collision accidents, reducing equipment damage and cargo loss, and ensuring the stable operation of warehousing operations. Attached Figure Description

[0066] Figure 1 This is a schematic diagram illustrating the working principle of the warehouse scheduling method based on WMS and AGV collaboration described in this invention.

[0067] Figure 2 A flowchart for constructing a dynamic storage location data center;

[0068] Figure 3 A flowchart for generating a dynamic storage allocation scheme;

[0069] Figure 4 This is a flowchart of the path optimization module in the coupled scheduling model. Detailed Implementation

[0070] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0071] Please see Figures 1-4 This invention provides a warehouse scheduling method and system based on the collaboration of WMS (Warehouse Management System) and AGV (Automated Guided Vehicle), to achieve efficient and intelligent warehouse management and improve the overall efficiency and accuracy of warehouse operations. The specific implementation methods of this invention are described in detail below.

[0072] The dynamic warehouse location data center is the foundational data platform for the entire warehouse scheduling process. It includes warehouse location occupancy status, material turnover rate, and material attribute correlation parameters. These parameters cover material classification codes, storage environment requirements, and inbound / outbound frequency. In the actual construction process, historical inbound / outbound data accumulated over a long period is first collected from the warehouse. This data records detailed information such as the inbound / outbound time and quantity for each type of material. Through in-depth analysis of this data, specific data mining algorithms are used to extract material turnover rate characteristics and classification correlation matrices. Simultaneously, warehouse areas are divided into levels based on physical parameters such as the floor where the warehouse is located and the row and column positions of the shelves, and temperature and humidity sensitive areas are marked using temperature and humidity sensors. Then, the material turnover rate characteristics, classification correlation matrix, and warehouse area levels are spatially mapped. A dynamic priority is calculated for each warehouse location, taking into account various factors, thus generating a dynamic priority sequence for warehouse locations. Finally, based on this dynamic priority sequence and combined with real-time material inbound / outbound information, the warehouse location occupancy status is updated, forming the dynamic warehouse location data center.

[0073] The multi-factor decision-making algorithm comprehensively considers multiple key factors such as material turnover rate and storage environment matching degree. First, the product of material turnover rate and storage location distance coefficient is calculated as the first allocation weight. Materials with high turnover rates are tended to be allocated to storage locations closer to the inlet / outlet, reducing handling time and improving warehousing efficiency. The storage location distance coefficient is calculated based on the actual distance between the storage location and the inlet / outlet. A second allocation weight is generated based on the matching degree between material classification codes and storage environment requirements. For example, for temperature-sensitive materials, it is ensured that they are allocated to storage locations that meet their temperature requirements. Next, a weighted fusion algorithm integrates the first and second allocation weights to generate a dynamic storage location score. Based on the ranking results of the dynamic storage location scores, target storage locations are selected from high to low scores, and the allocation scheme is updated in a timely manner to adapt to dynamic changes in warehousing conditions.

[0074] Upon receiving real-time inbound / outbound instructions, the system first parses the order urgency tag in the instruction, classifying tasks into different priority levels. For example, urgent orders are assigned the highest priority. Simultaneously, the AGV's onboard sensors acquire the current load status and remaining battery level, calculating the available handling capacity based on a pre-defined model. Then, based on the matching relationship between task priority levels and available handling capacity, a task node dependency graph is constructed, clearly defining the sequence and dependencies between various handling tasks. Finally, the task node dependency graph is converted into an AGV handling task chain, clearly marking the task switching trigger conditions. For instance, when the AGV completes a handling task and has sufficient battery power and meets load requirements, the next task is triggered.

[0075] Spatiotemporal mapping is crucial for achieving collaborative path planning among multiple AGVs. First, 3D point cloud data of the warehouse area is collected. Advanced 3D scanning technology is used to accurately mark the boundaries of fixed obstacles and temporary occupied areas, such as areas where shelves are under maintenance. Based on the real-time positions of the AGVs, dynamic passage weights are generated using a specific algorithm. Simultaneously, sensor and data analysis techniques are used to mark predicted path conflict areas. The dynamic passage weights and predicted path conflict areas are then gridded to construct an intuitive and accurate spatiotemporal map. Based on this map, an intelligent path planning algorithm is used to generate a collaborative path planning network for multiple AGVs. Each path is labeled with a path priority tag, and higher-priority paths are assigned to AGVs first to improve overall handling efficiency.

[0076] The coupled scheduling model comprehensively considers both storage location allocation and path planning. The path optimization module plays a crucial role. It first extracts predicted path conflict areas from the multi-AGV collaborative path planning network and calculates a congestion penalty factor based on the congestion level of these areas. Simultaneously, it dynamically adjusts path priority labels based on historical execution data from the AGV cluster scheduling instruction set. An iterative optimization algorithm balances the congestion penalty factor and path priority labels to generate an optimized path set. Finally, the optimized path set is spatiotemporally aligned with the dynamic storage location allocation scheme to ensure that AGVs arrive at the correct storage location at the correct time for transport operations, thereby outputting the AGV cluster scheduling instruction set.

[0077] Collision avoidance optimization algorithms are a crucial step in ensuring the safe and efficient operation of AGVs. Based on the AGV's trajectory, a kinematic model is used to predict its position coordinates for the next N time steps, generating a trajectory conflict detection matrix. The safe distance offset between each AGV is calculated based on this matrix; if the offset exceeds the safe range, a collision risk is identified. A trajectory replanning algorithm is then used to adjust the AGV's speed or path steering angle to eliminate the safe distance offset. The adjusted trajectory is encoded into the final scheduling strategy, and the spatiotemporal map is updated synchronously, enabling the system to monitor the latest AGV positions and driving status in real time, achieving continuous and stable warehouse scheduling.

[0078] The technical solution of the present invention will be further described in detail below with reference to specific embodiments:

[0079] Example 1:

[0080] In practical applications, building a dynamic warehouse location data center is the foundation for achieving efficient warehouse scheduling.

[0081] Historical inbound and outbound data is collected, covering records of all materials entering and leaving the warehouse over a period of time, including entry / exit time, material type, quantity, and corresponding storage location information. In-depth analysis of this data extracts material turnover characteristics. Material turnover rate can be calculated as the ratio of the quantity of materials leaving the warehouse to the average inventory quantity over a certain period, reflecting the speed of material flow within the warehouse. Simultaneously, the co-occurrence frequency of materials in historical inbound and outbound orders is statistically analyzed to generate an initial correlation matrix. For example, if material A and material B appear simultaneously in multiple orders, their correlation is considered high. The initial correlation matrix is ​​normalized according to the material classification coding hierarchy to ensure comparability of correlations between materials at different levels, thus obtaining a classification correlation matrix.

[0082] Storage areas are stratified based on physical parameters. For example, based on the warehouse layout and storage conditions, the warehouse is divided into different zones, such as ambient temperature zones and cold storage zones, and temperature and humidity sensitive areas are marked. These zones have specific requirements for the temperature and humidity of the stored materials and require special management.

[0083] Material turnover rate characteristics, classification association matrices, and warehouse area hierarchy are spatially mapped to generate a dynamic priority sequence for warehouse locations. For example, materials with high turnover rates are preferentially assigned to warehouse locations near entrances and exits, while materials with strong correlations are assigned to adjacent warehouse locations as much as possible.

[0084] The system updates storage location occupancy status based on a dynamic priority sequence. When new materials enter or leave the warehouse, the occupancy information of the storage locations is updated in a timely manner, forming a dynamic storage location data center. By continuously updating and improving this data center, real-time and accurate data support is provided for subsequent warehouse scheduling.

[0085] Example 2:

[0086] Multi-factor decision-making algorithms play a crucial role in generating dynamic storage location allocation schemes. The product of material turnover rate and storage location distance coefficient is used as the first allocation weight. Let the material turnover rate be... The storage location distance coefficient is First allocation weight Material turnover rate As mentioned earlier, this reflects the flow rate of the material. A higher value indicates a faster turnover of materials in the warehouse; warehouse location distance coefficient This is determined by the distance between the storage location and the warehouse entrance / exit or key operational area; the closer the distance, the better. The larger the value, the more likely it is to be obtained by calculating the Euclidean distance between the storage location coordinates and the entrance / exit coordinates and then normalizing it.

[0087] A second allocation weight is generated based on the matching degree between material classification codes and warehouse storage environment requirements. Assume the matching degree is... Second allocation weight Match degree The calculation can be performed using pre-set rules. For example, if the required storage environment for a material is room temperature, and a storage location is in the room temperature zone, then the matching degree between the material and the storage location is 1. If the storage location environment partially matches the material requirements, a corresponding score can be assigned based on the degree of matching, such as 0.5.

[0088] A dynamic reserve score is generated by integrating the first and second allocation weights using a weighted fusion algorithm. Let the weighting coefficients be... and Dynamic storage space scoring Weighting coefficients and Adjustments can be made based on actual warehouse management needs and experience. For example, if more emphasis is placed on the impact of material turnover rate on storage location allocation, it can be appropriately increased. The value of .

[0089] Based on the dynamic storage location scoring ranking results, target storage locations are selected and the allocation plan is updated. The dynamic storage location scores of all locations are sorted, and the storage location with the highest score is selected as the target storage location. When new materials need to be stored, the dynamic storage location scores are recalculated according to the above process, and the allocation plan is updated promptly to ensure the rationality and dynamic adaptability of storage location allocation.

[0090] Example 3:

[0091] Upon receiving real-time inbound / outbound instructions, generating AGV handling task chains through task decomposition rules is a crucial step in achieving efficient warehouse handling. This involves parsing the order urgency tags in the inbound / outbound instructions to categorize tasks into priority levels. For example, order urgency can be divided into high, medium, and low levels, with high-urgency orders corresponding to the highest priority tasks, requiring immediate processing.

[0092] Calculate the available handling capacity based on the AGV's current load status and remaining battery level. Assume the AGV's maximum load capacity is... The current load is The remaining battery power is The battery is fully charged. Available transport capacity value This formula takes into account both the AGV's load capacity and battery power to ensure that the AGV has sufficient capacity to complete the handling task and will not be affected by insufficient power during operation.

[0093] Based on the matching relationship between task priority levels and available handling capacity values, a task node dependency graph is generated. For example, high-priority tasks are preferentially assigned to AGVs with high available handling capacity values; for low-priority tasks, after satisfying the allocation of other high-priority tasks, they are then assigned to AGVs with lower remaining available handling capacity values. Through this matching relationship, the dependencies between task nodes are constructed, forming the task node dependency graph.

[0094] The task node dependency graph is converted into an AGV transport task chain, and task switching trigger conditions are labeled. These conditions can include the AGV completing its current task, battery level falling below a certain threshold, or receiving a higher-priority task. For example, after an AGV completes a transport task, it automatically acquires the next task based on the task node dependency graph. If, during task execution, the battery level drops below 20%, a task switch is triggered, and the AGV returns to the charging area to recharge. Once the battery is fully charged, it continues executing the remaining tasks. By rationally constructing the AGV transport task chain and clearly defining the task switching trigger conditions, the operational efficiency and task execution flexibility of the AGV are improved.

[0095] Example 4:

[0096] Constructing a multi-AGV collaborative path planning network is a crucial step in achieving efficient collaborative operations among multiple AGVs in a warehouse, and a spatiotemporal map is the foundation for building this network. The process involves collecting 3D point cloud data of the warehouse area, acquiring 3D spatial information using devices such as LiDAR, and marking the boundaries of fixed obstacles and temporary storage areas. Fixed obstacles include shelves and walls, while temporary storage areas may be areas where loading and unloading operations are underway or where goods are temporarily stored.

[0097] Dynamic passage rights are generated based on the real-time location of the AGV. Let the real-time location of the AGV be... The target storage location is The Euclidean distance between the storage facility and the target storage location is calculated using the Euclidean distance formula. Generate basic passage weights Then, the basic passage weight is attenuated and compensated based on the remaining AGV task time and battery remaining. Let the remaining task time be... The remaining battery power is The battery is fully charged. Attenuation compensation coefficient ( (Total task time), adjusted passage weights Finally, the passage weights are dynamically adjusted based on the congestion level of the predicted path conflict areas. If the congestion level of a predicted path conflict area is high, the passage weight for that area is reduced; if the congestion level is low, the passage weight is appropriately increased. The adjusted passage weights are then mapped to the gridded encoding of the spatiotemporal map.

[0098] Simultaneously, based on the real-time location and direction of travel of the AGVs, path conflict prediction areas are marked. When multiple AGVs are likely to arrive at the same location or intersect paths at a certain time, this is identified as a path conflict prediction area. Dynamic passage weights and path conflict prediction areas are then encoded in a grid to construct a spatiotemporal map. In the spatiotemporal map, each grid contains the passage weights for that area, the presence of obstacles, and path conflict prediction information.

[0099] A multi-AGV collaborative path planning network is generated based on a spatiotemporal map, and path priority labels are labeled. Path priority labels are determined based on factors such as task priority and passage weight. For example, paths for high-priority tasks are assigned higher path priority labels; paths with high passage weights also have their path priority labels increased accordingly. By constructing a multi-AGV collaborative path planning network and labeling path priority, reasonable travel path planning is provided for AGVs, avoiding path conflicts and improving the efficiency of multi-AGV collaborative operations.

[0100] Example 5:

[0101] The path optimization module in the coupled scheduling model is crucial for generating an efficient AGV cluster scheduling instruction set. First, the path conflict prediction region in the multi-AGV cooperative path planning network is extracted, and the congestion penalty factor is calculated. Let the number of AGVs expected to conflict within the path conflict prediction region be... The area of ​​this region is Congestion penalty factor The more conflicting AGVs there are, the smaller the area, and the larger the congestion penalty factor, indicating a higher degree of congestion in that area.

[0102] Based on historical execution data from the AGV cluster scheduling instruction set, path priority labels are dynamically adjusted. For example, if a path frequently experiences congestion or delays during historical execution, its path priority label is lowered; conversely, if a path has high execution efficiency, its path priority label is raised. Through analysis and learning from historical data, path priority labels are continuously optimized to better reflect actual operational conditions.

[0103] An iterative optimization algorithm balances congestion penalty factors and path priority labels to generate a set of optimized paths. Commonly used iterative optimization algorithms include genetic algorithms and simulated annealing. Taking genetic algorithms as an example, paths are treated as individuals, and the population is continuously evolved through selection, crossover, and mutation operations until the optimized paths achieve the best overall performance considering both the congestion penalty factor and path priority labels. In each iteration, the fitness value of each individual is calculated. The fitness value can be determined based on a combination of the congestion penalty factor and path priority labels. ( and These are the weighting coefficients, and Through multiple iterations, an optimized path set is finally generated.

[0104] The optimized path set is spatiotemporally aligned with the dynamic storage location allocation scheme to output an AGV cluster scheduling instruction set. Spatiotemporal alignment refers to matching the optimized paths with the storage locations and inbound / outbound times of materials to ensure that AGVs can execute handling tasks in a reasonable order and at reasonable times. For example, based on the dynamic storage location allocation scheme, the storage locations of materials are determined. Combined with the optimized path set, specific travel paths and task execution times are generated for each AGV, forming an AGV cluster scheduling instruction set to guide the actual operation of the AGVs.

[0105] Example 6:

[0106] Collision avoidance optimization algorithms are a key technology for ensuring the safe operation of AGVs in warehouses. These algorithms predict future collisions based on the AGV's trajectory. The position coordinates of the AGV at each time step. Let the current position coordinates of the AGV be... The driving speed is The driving direction angle is In time step ( The position coordinates of ) can be obtained through the formula Calculation yields ( (Time interval for each time step). Based on these predicted position coordinates, a trajectory collision detection matrix is ​​generated. If the distance between the predicted position coordinates of two AGVs at a future time is less than a safe distance threshold... If the value is positive, the corresponding position in the trajectory conflict detection matrix is ​​marked as 1, indicating that a conflict may exist; otherwise, it is marked as 0.

[0107] The safe distance offset between each AGV is calculated based on the trajectory conflict detection matrix. Let the predicted position coordinates of two AGVs that may conflict be as follows: and Safety distance offset .

[0108] The AGV's speed or path steering angle can be adjusted using a trajectory replanning algorithm to eliminate safety distance offsets. For example, if a potential collision between two AGVs is detected, the speed of one AGV can be reduced, or its path steering angle can be adjusted to avoid the collision zone. The trajectory replanning algorithm can be an improved version of a classic path planning algorithm such as Dijkstra's algorithm, which replans the AGV's travel trajectory while taking into account safety distance offsets and other constraints.

[0109] The adjusted driving trajectory is encoded into the final scheduling strategy, and the spatiotemporal map is updated synchronously. The adjusted driving trajectory information is converted into an instruction format that the AGV can recognize and execute, which is then output as the final scheduling strategy. Simultaneously, the AGV's position information and path conflict prediction areas in the spatiotemporal map are updated so that subsequent path planning and scheduling decisions can be based on the latest status, ensuring the safe and stable operation of warehousing.

[0110] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0111] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A warehouse scheduling method based on WMS and AGV collaboration, characterized in that, include: A dynamic storage location data center is constructed, which includes storage location occupancy status, material turnover rate, and material attribute association parameters; the material attribute association parameters include material classification code, storage environment requirements, and inbound / outbound frequency. Based on the dynamic storage location data center, a dynamic storage location allocation scheme is generated through a multi-factor decision algorithm; the multi-factor decision algorithm includes turnover rate weight adjustment, correlation matching degree calculation, and storage location distance coefficient optimization. Based on real-time inbound and outbound instructions, an AGV handling task chain is generated through task decomposition rules; the task decomposition rules include order urgency classification, AGV load balancing, and task node dependencies. Based on the AGV handling task chain, a multi-AGV collaborative path planning network is constructed using a spatiotemporal map; the spatiotemporal map includes the distribution of obstacles in the warehouse area, the real-time location of AGVs, and the path conflict prediction area. The dynamic storage allocation scheme is coupled with the multi-AGV collaborative path planning network input scheduling model to generate an AGV cluster scheduling instruction set. Based on the AGV cluster scheduling instruction set, the AGV driving trajectory is adjusted through an anti-collision optimization algorithm, and the final scheduling strategy is output. The construction of the dynamic storage location data center includes: Collect historical inbound and outbound data to extract material turnover rate characteristics and classification correlation matrix; The storage area is divided into different levels based on the physical parameters of the storage location, and temperature and humidity sensitive areas are marked. By spatially mapping material turnover rate characteristics, classification association matrix and warehouse area hierarchy, a dynamic priority sequence of warehouse locations is generated. The storage space occupancy status is updated based on the dynamic priority sequence to form a dynamic storage space data center; The coupled scheduling model includes a path optimization module, which includes: Extract the path conflict prediction region from the multi-AGV cooperative path planning network and calculate the congestion penalty factor; Dynamically adjust path priority labels based on historical execution data of the AGV cluster scheduling instruction set; An optimized path set is generated by balancing the congestion penalty factor and the path priority label through an iterative optimization algorithm. The optimized path set is spatiotemporally aligned with the dynamic storage location allocation scheme to output the AGV cluster scheduling instruction set. The anti-collision optimization algorithm includes: Based on the AGV's driving trajectory, predict the position coordinates of the next N time steps and generate a trajectory conflict detection matrix; Calculate the safe distance offset between each AGV based on the trajectory conflict detection matrix; The AGV speed or path steering angle is adjusted by using a trajectory replanning algorithm to eliminate safety distance offset; The adjusted driving trajectory is encoded into the final scheduling strategy, and the spatiotemporal map is updated synchronously. The method for constructing the classification association matrix includes: Calculate the co-occurrence frequency of materials in historical inbound and outbound orders to generate an initial correlation matrix; The initial correlation matrix is ​​normalized according to the material classification coding level; The classification association matrix is ​​updated based on the normalization results, and the priority sequence of the dynamic storage location data center is associated with it.

2. The warehouse scheduling method based on WMS and AGV collaboration according to claim 1, characterized in that, The generation of a dynamic storage allocation scheme through a multi-factor decision-making algorithm includes: The product of material turnover rate and storage location distance coefficient is used as the first allocation weight; A second allocation weight is generated based on the matching degree between the material classification code and the storage environment requirements. The first allocation weight and the second allocation weight are integrated by a weighted fusion algorithm to generate a dynamic reserve score; Based on the dynamic storage location scoring and ranking results, select the target storage location and update the allocation plan.

3. The warehouse scheduling method based on WMS and AGV collaboration according to claim 1, characterized in that, The generation of AGV transport task chains through task decomposition rules includes: Parse the order urgency tags in inbound and outbound instructions to classify task priority levels; Calculate the available handling capacity value based on the AGV's current load status and remaining battery level; Based on the matching relationship between task priority levels and available transport capacity values, a task node dependency graph is generated; Convert the task node dependency graph into an AGV transport task chain and mark the task switching trigger conditions.

4. A warehouse scheduling method based on WMS and AGV collaboration according to claim 1, characterized in that, The construction of a multi-AGV collaborative path planning network using a spatiotemporal map includes: Collect 3D point cloud data of the reservoir area and mark the boundaries of fixed obstacles and temporary occupied areas; Dynamic passage rights are generated based on the real-time location of the AGV, and path conflict prediction areas are marked. Dynamic access rights and path conflict prediction areas are gridded to construct a spatiotemporal map; A multi-AGV collaborative path planning network is generated based on a spatiotemporal map, and path priority labels are marked.

5. A warehouse scheduling method based on WMS and AGV collaboration according to claim 4, characterized in that, The method for generating the dynamic access rights includes: The Euclidean distance between the AGV and the target storage location is calculated based on the AGV's real-time position to generate a basic passage weight value; The basic passage weight value is attenuated and compensated based on the remaining time of the AGV task and the remaining battery power. By combining the congestion level of the predicted route conflict area, the passage weight value is dynamically adjusted, and the adjusted passage weight value is mapped to the gridded encoding of the spatiotemporal map.

6. A warehouse scheduling system based on WMS and AGV collaboration, characterized in that, The device includes a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, when the processor executes the computer program, it implements the steps of the warehouse scheduling method based on WMS and AGV collaboration as described in any one of claims 1 to 5.

Images