Multi-device cooperative warehouse scheduling method and system, and storage medium

Abstract

This application discloses a multi-device collaborative warehouse scheduling method, system, and storage medium. The method includes: decomposing received warehouse tasks; dynamically generating an execution sequence based on the task type, real-time device status, and warehouse layout status; scheduling the execution sequence using a multi-level feedback queue mechanism based on the real-time device status and warehouse layout status to generate an initial scheduling strategy and distribute it to the corresponding execution entities for execution; when any execution entity experiences an anomaly, obtaining the current warehouse scenario status and dynamically adjusting the scheduling weights of each task type; calculating the priority of affected execution stages based on the scheduling weights; generating an updated scheduling strategy and distributing it for execution. This method improves the system's response capability to unexpected events such as device failures and congestion.

Description

Technical Field

[0001] This application relates to the field of warehouse scheduling, and in particular to a warehouse scheduling method, system and storage medium for multi-device collaboration. Background Technology

[0002] With the rapid development of intelligent manufacturing and e-commerce logistics, modern warehousing systems are evolving towards automation and intelligence. The Warehouse Control System (WCS), as the core hub connecting the upper-level Warehouse Management System (WMS) and the lower-level execution equipment, plays a crucial role in warehousing operations. Stacker cranes, automated guided vehicles (AGVs), and conveyors, as core equipment in warehousing operations, directly impact the overall operational efficiency of the warehouse through their coordinated scheduling efficiency.

[0003] Current conventional methods for scheduling and controlling warehouse operations have shortcomings. In terms of system architecture, business, scheduling, and equipment control logic are not decoupled and have a high degree of coupling; in equipment control, each PLC controls a group of devices, lacking unified coordination; task scheduling often uses simple single-dimensional priority determination and lacks multi-task parallel optimization algorithms; multi-device collaborative scheduling is simple and does not fully consider issues such as path conflicts; and the path replanning mechanism is imperfect when dealing with sudden scenarios.

[0004] The aforementioned defects severely impact warehousing operations. High system architecture coupling results in poor scalability, making it difficult to adapt to the collaborative management of multiple types of equipment; insufficient multi-device collaborative scheduling leads to path conflicts, congestion, and deadlocks, hindering global optimization; slow response and high overhead in unexpected scenarios make it difficult to balance global and local considerations; unreasonable task scheduling prevents resource preemption, resulting in long operation cycles; imperfect multi-device collision avoidance and PLC collaboration pose collision risks, and the system's visualization and dynamic configuration capabilities are weak.

[0005] In summary, existing warehouse control systems have many problems in terms of architecture, scheduling, and emergency response, and urgently need to be improved. Summary of the Invention

[0006] Firstly, this application provides a multi-device collaborative warehouse scheduling method, employing the following technical solution: A multi-device collaborative warehouse scheduling method is applied to a warehouse system including stacker cranes, AGVs, and conveyors. The method includes the following steps: The received warehousing tasks are decomposed, and an execution sequence is dynamically generated based on the task type, real-time equipment status, and warehousing layout status. The task types include inbound tasks, outbound tasks, and transfer tasks. Each execution step in the execution sequence is bound to a unique execution entity. Based on the real-time status of the equipment and the warehouse layout status, a multi-level feedback queue mechanism is used to schedule the execution sequence, generate an initial scheduling strategy, and send it to the corresponding execution entity for execution. When any of the execution entities encounters an anomaly, the current warehouse scenario status is reacquired and the scheduling weights of each task type are dynamically adjusted. The priority of the affected execution links is calculated based on the scheduling weights, an updated scheduling strategy is generated, and the strategy is issued for execution.

[0007] By adopting the above technical solutions, a three-layer architecture of dynamically generated execution sequences, multi-level feedback queue scheduling, and scenario-driven anomaly replanning is implemented, achieving a complete closed loop from task decomposition to anomaly recovery. Dynamically generated execution sequences ensure that the scheduling strategy matches the real-time state, multi-level feedback queues enable dynamic priority escalation and de-escalation to avoid task starvation, and anomaly replanning integrates scenario state adjustments to improve the system's responsiveness to sudden events such as equipment failure and congestion.

[0008] Preferably, the step of decomposing the received warehousing task and dynamically generating an execution sequence based on the task type, real-time equipment status, and warehouse layout status specifically includes the following steps: Receive warehousing tasks and break them down to obtain multiple task types for the warehousing tasks. Based on the task types, determine the type sequence of each task corresponding to the executing entity. The system acquires real-time equipment status and warehouse layout status, and dynamically binds the corresponding execution entity to each execution stage in the type sequence based on the real-time equipment status; the real-time equipment status includes real-time equipment location, equipment load rate, equipment fault status, and equipment availability. Based on the warehouse layout status, configure execution parameters for each execution stage; the warehouse layout status includes the distribution of storage locations, the degree of channel congestion, and the occupancy of connection points. An execution sequence is dynamically constructed in real time based on the type sequence, the execution subject, and the execution parameters. Each execution sequence includes at least two of the following execution links: conveyor execution link, AGV execution link, and stacker crane execution link.

[0009] By adopting the above technical solution, the process of generating the execution sequence is refined, the mapping from task type to type sequence is clarified, the execution subject binding driven by real-time device status is clarified, and the parameter configuration driven by warehouse layout status is clarified. This ensures the accurate matching of each execution link with the specific device type and solves the problem of lack of device-level binding in task decomposition in the existing technology.

[0010] Preferably, the process of dynamically constructing the execution sequence in real time further includes dynamically reconstructing the unexecuted execution sequence when the real-time status of the device or the warehouse layout status changes, including at least one of the following situations: When the congestion level of the channel exceeds a preset congestion threshold, a detour path is generated for the AGV execution process; When the equipment load rate of the conveyor exceeds a preset load threshold, a buffer temporary storage step is added before the conveyor execution step; When the stacker crane fails, the execution links bound to the stacker crane will be reassigned to other available stacker cranes.

[0011] By adopting the above technical solution, the mechanism for dynamically reconstructing unexecuted sequences enables proactive responses to scenarios such as channel congestion, equipment overload, and equipment failure. It can instantly adjust the paths of subsequent execution stages, add buffer stages, or reallocate equipment when anomalies occur, thereby improving the robustness and adaptability of the system.

[0012] Preferably, the step of scheduling the execution sequence based on the real-time status of the equipment and the warehouse layout status using a multi-level feedback queue mechanism to generate an initial scheduling strategy and then issuing it to the corresponding execution entity for execution specifically includes the following steps: Obtain the dynamically generated execution sequence, as well as the current real-time status of the device and the warehouse layout status; The execution steps in the execution sequence are input into a multi-level feedback queue mechanism. The initial priority of each execution step is determined according to the device load rate in the real-time status of the device. Based on the initial priority, a scheduling execution sequence containing the scheduling order is output. Based on the scheduling order, the equipment availability, and the channel congestion level, an execution time window is allocated to each execution stage; Based on the scheduling order, the real-time location of the equipment, and the channel congestion level of the warehouse layout, an initial path is planned for the AGV execution process. The initial path is generated globally optimal based on the A* algorithm, and local trajectory optimization is performed using the Dynamic Window Method (DWA). The execution stage, the bound execution entity, the execution time window, and the initial path of the AGV execution stage are encapsulated into scheduling instructions, which are then sequentially sent to the corresponding execution entities according to the scheduling order.

[0013] By adopting the above technical solution, the generation process of the initial scheduling strategy is refined. The scheduling order is output through a multi-level feedback queue, the time window is allocated, the A* global path and DWA local optimization path are planned for the AGV, and the scheduling instructions are encapsulated and issued. This realizes a complete link from queue scheduling to path planning and then to instruction issuance, ensuring the determinism of scheduling and the controllability of execution.

[0014] Preferably, the scheduling execution sequence, which includes the scheduling order, is output according to a multi-level feedback queue mechanism, specifically including the following steps: Based on the device load rate in the real-time status of the device, the initial priority of each execution step is calculated, and the device load rate is negatively correlated with the initial priority; The execution phases are allocated to corresponding priority queues in a multi-level feedback queue according to the initial priority. The multi-level feedback queue contains multiple priority queues, and different priority queues correspond to different time slice lengths and execution priorities. The execution steps in the current priority queue are executed sequentially using a time-slice rotation method. When any execution step is completed within the allocated time slice, it is removed from the queue. When any execution step is not completed within the allocated time slice, it is downgraded to the next priority queue. When the waiting time of any execution step in the priority queue exceeds a first preset time threshold, it is upgraded to a higher priority queue. The output is the scheduling execution sequence after the multi-level feedback queue scheduling, and the execution steps in the scheduling execution sequence are arranged according to the scheduling order.

[0015] By adopting the above technical solution, the core logic of the multi-level feedback queue mechanism is defined in depth. The initial priority is determined by the device load rate, time slice rotation is executed, and unfinished downgrade and waiting timeout upgrade tasks are implemented. This achieves dynamic and fine-grained adjustment of priorities, avoids long-term starvation of low-priority tasks, and improves system throughput efficiency.

[0016] Preferably, during execution, the process further includes establishing a multi-time-scale scheduling framework to dynamically optimize the issued initial scheduling strategy, specifically including the following steps: At a global scale, the actual execution progress of each execution entity and the real-time status of the device are periodically acquired to identify the deviation between the initial scheduling strategy and the actual execution progress. When the deviation exceeds a preset deviation threshold, the optimal task allocation scheme is recalculated based on the unexecuted execution steps and the real-time status of the devices of each execution subject. Based on the re-allocated task allocation scheme, the execution order of the unexecuted execution steps and the binding relationship of the execution subjects are adjusted. At a local scale, the operational status of each of the aforementioned execution entities is monitored in real time; When a temporary obstacle or sudden congestion is detected on the initial path corresponding to the AGV, an obstacle avoidance trajectory is generated in real time based on the dynamic window method, and the original path is locally corrected. When the actual execution time of any of the execution steps exceeds the execution time window, the waiting time of the subsequent dependent steps of the execution step is dynamically adjusted, and the passage priority of the relevant AGV is updated.

[0017] By adopting the above technical solutions, a multi-timescale scheduling framework was established. At the global scale, task allocation and execution sequence are periodically optimized, and at the local scale, dynamic changes such as temporary obstacles and execution timeouts are responded to in real time. Through the arbitration of conflicts by the collaborative mechanism, a balance between global optimization and local real-time is achieved, thereby enhancing the dynamic adaptability of the scheduling strategy.

[0018] Preferably, when any of the execution entities malfunctions, the current warehouse scenario status is reacquired and the scheduling weights of each task type are dynamically adjusted. The priority of the affected execution stages is calculated based on the scheduling weights, an updated scheduling strategy is generated, and execution is initiated. Specifically, this includes the following steps: When an abnormal event is detected in any of the aforementioned execution entities, a replanning process is triggered; Obtain the current warehouse scenario status, which includes at least one of the following: inventory occupancy rate, backlog of outbound tasks, number of available storage locations, and degree of channel congestion. The state of the warehousing scenario is input into a preset weight adjustment model to dynamically adjust the scheduling weights of the inbound task, the outbound task, and the transfer task. Wherein, the inventory occupancy rate and the backlog of outbound tasks are related to the scheduling weight of the outbound tasks, the number of free storage locations is related to the scheduling weight of the inbound tasks, and the degree of channel congestion is related to the scheduling weight of the transfer tasks. Identify the execution steps affected by the abnormal event, and recalculate the priority of the affected execution steps according to the adjusted scheduling weight; Based on the recalculated priorities, the execution entities are reassigned and the execution order is adjusted for the affected execution stages. The paths for the affected AGV execution stages are replanned, and the updated scheduling strategy is generated and issued for execution.

[0019] By adopting the above technical solutions, the complete data flow of anomaly-triggered replanning has been refined. From anomaly detection, scenario status acquisition, dynamic weight adjustment, affected link identification, priority recalculation to update strategy generation and distribution, a closed-loop replanning process has been formed. In particular, scenario statuses such as inventory occupancy rate and outbound backlog are used as the basis for weight adjustment, realizing scenario-driven fine-grained scheduling.

[0020] Preferably, the step of recalculating the priority of the affected execution stage based on the adjusted scheduling weights specifically includes the following steps: Based on the execution sequence, the execution stage dependency graph generated is used to locate the execution stage where the abnormal event occurred using a forward propagation method, and the execution stage and all downstream execution stages are marked as affected execution stages. A multi-factor dynamic weighting mechanism is used to calculate the priority score of each affected execution stage. The multi-factor dynamic weighting mechanism includes a task urgency factor, an auction algorithm factor, a device status factor, and a scheduling weight factor. The scheduling weight factor is the dynamically adjusted scheduling weight of the task type, and the task urgency is determined according to the preset outbound time limit; the auction algorithm factor is determined according to at least one of the remaining power of the AGV, the current path length, and the estimated completion time; and the equipment status factor includes at least one of the remaining power of the AGV, the stacker crane load rate, and the conveyor queue length. The affected execution steps are sorted according to the priority scores to generate an updated execution order.

[0021] By adopting the above technical solution, the specific implementation of priority recalculation is further defined. The affected links and their downstream are located by the dependency graph of the execution links. A multi-factor dynamic weighting mechanism is adopted to calculate the priority score by comprehensively considering the task urgency, auction algorithm factor, equipment status factor and scheduling weight factor. The auction algorithm factor introduces the AGV bidding mechanism to realize the distributed optimization allocation of conflict paths.

[0022] Secondly, this application provides a multi-device collaborative warehouse scheduling system, which adopts the following technical solution: A multi-device collaborative warehouse scheduling system includes the following modules: The execution sequence generation module is used to decompose the received warehousing tasks and dynamically generate an execution sequence based on the task type, real-time equipment status, and warehousing layout status of the warehousing tasks. The task types include inbound tasks, outbound tasks, and transfer tasks. Each execution step in the execution sequence is bound to a unique execution entity. The multi-level feedback queue scheduling module is used to schedule the execution sequence according to the real-time status of the equipment and the warehouse layout status, and generate an initial scheduling strategy and send it to the corresponding execution subject for execution. The exception feedback and replanning module is used to reacquire the current warehouse scenario status and dynamically adjust the scheduling weights of each task type when any of the execution entities encounters an exception. Based on the scheduling weights, the priority of the affected execution links is calculated, an updated scheduling strategy is generated, and the strategy is issued for execution.

[0023] By adopting the above technical solution, a system architecture corresponding to the above method is provided. Through the collaboration of the execution sequence generation module, the multi-level feedback queue scheduling module, and the exception feedback and replanning module, a complete functional closed loop from task decomposition to exception recovery is realized, which facilitates system-level deployment and implementation.

[0024] Thirdly, this application provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the multi-device collaborative warehouse scheduling method as described above.

[0025] In summary, this application includes at least one of the following beneficial technical effects: (1) This application decomposes the tasks of warehousing, warehousing, and transfer into equipment-level execution sequences, thereby realizing the temporal coupling and precise coordination of three heterogeneous equipment: stacker crane, AGV, and conveyor. This solves the problem of lack of equipment-level binding in task decomposition and improves the determinism and efficiency of multi-equipment collaborative operation.

[0026] (2) This application realizes the dynamic upgrading and downgrading of priority and time slice round-robin scheduling through a multi-level feedback queue mechanism, and re-acquires the warehouse scene status when an anomaly occurs, dynamically adjusts the task type weight and recalculates the priority using a multi-factor weighting mechanism, forming a three-layer dynamic scheduling architecture of global optimization, local fine-tuning and scene-driven replanning, which significantly improves the system's response capability to sudden events such as equipment failure, congestion and emergency tasks. (3) This application realizes the fine-grained dynamic management of task priority through a multi-level feedback queue scheduling mechanism; it realizes the distributed resource optimization configuration by executing AGV conflict path allocation through an auction algorithm; and it combines congestion prediction with hierarchical path planning to realize the synergy between A* global planning and DWA local optimization. Attached Figure Description

[0027] Figure 1 This is a flowchart of a multi-device collaborative warehouse scheduling method; Figure 2 This is an architecture diagram of a multi-device collaborative warehouse scheduling system. Detailed Implementation

[0028] This application provides a multi-device collaborative warehouse scheduling method, system, and storage medium. To make the objectives, technical solutions, and advantages of this application clearer, the implementation methods of this application will be further described in detail below.

[0029] The following describes in further detail an embodiment of a multi-device collaborative warehouse scheduling method of this application, with reference to the accompanying drawings.

[0030] A multi-device collaborative warehouse scheduling method is applied to a warehouse system including stacker cranes, AGVs, and conveyors. The process is as follows: Figure 1 As shown, it includes the following steps: S1. Decompose the received warehousing tasks and dynamically generate an execution sequence based on the task type, real-time equipment status, and warehouse layout status. Task types include inbound tasks, outbound tasks, and transfer tasks. Each execution step in the execution sequence is bound to a unique execution entity. The specific steps include the following: S11. Receive warehousing tasks and break them down to obtain multiple task types for warehousing tasks. Based on the task type, determine the type sequence of each task corresponding to the executing entity. In this embodiment, when the task type is warehousing, the type sequence is: conveyor type, AGV type, stacker crane type; When the task type is outbound, the type sequence is: stacker crane type, AGV type, conveyor type; When the task type is database transfer, the type sequence is: stacker crane type, AGV type, stacker crane type.

[0031] S12. Obtain the real-time status of the equipment and the warehouse layout status. Based on the real-time status of the equipment, dynamically bind the corresponding execution subject to each execution stage in the type sequence. The real-time status of the equipment includes the real-time location of the equipment, the equipment load rate, the equipment fault status, and the equipment availability. When multiple devices of the same type are available, select the device with the lowest load rate as the execution subject; When all devices of a certain type fail, the execution stage is marked as pending scheduling, and the replanning process in step S3 is triggered.

[0032] S13. Configure execution parameters for each execution stage based on the warehouse layout status, including time window, route selection, and speed setting. Warehouse layout status includes location distribution, aisle congestion levels, and connection point occupancy. Determine the target cargo location coordinates for the stacker crane's execution phase based on the cargo location distribution; The initial path for the AGV execution process is planned based on the degree of channel congestion; the higher the degree of congestion, the greater the path weight. The connection time window for the conveyor's execution process is determined based on the occupancy status of the connection points; S14. An execution sequence is dynamically constructed in real time based on the type sequence, execution subject, and execution parameters. This enables the dynamic generation of executable scheduling instructions from task type, real-time equipment status, and warehouse layout status. Each execution sequence contains at least two of the following execution stages: conveyor execution stage, AGV execution stage, and stacker crane execution stage.

[0033] S15. When the real-time status of the equipment or the warehouse layout status changes, the unexecuted execution sequence is dynamically reconstructed, including at least one of the following situations: When the congestion level of the channel exceeds the preset congestion threshold, a detour path is generated for the AGV execution process; When the equipment load rate of the conveyor exceeds the preset load threshold, a buffer temporary storage step is added before the conveyor execution step; When a stacker crane malfunctions, the execution links bound to that stacker crane will be reassigned to other available stacker cranes.

[0034] In this embodiment, the execution sequence is dynamically generated with the optimization objective of minimizing at least one of the following: minimizing total completion time, minimizing device energy consumption, and minimizing path congestion. The optimal execution sequence is calculated in real time using a heuristic algorithm. The heuristic algorithm is the A* algorithm, whose input parameters include the real-time status of the equipment, the warehouse layout status, and the task type, and the output is the sequence of execution steps bound to the equipment.

[0035] S2. Based on the real-time status of the equipment and the warehouse layout, a multi-level feedback queue mechanism is used to schedule the execution sequence, generate an initial scheduling strategy, and distribute it to the corresponding execution entities for execution. This includes the following steps: S21. Obtain the dynamically generated execution sequence, as well as the current real-time status of the equipment and the warehouse layout status.

[0036] S22. Input the execution steps in the execution sequence into a multi-level feedback queue mechanism, determine the initial priority of each execution step based on the device load rate in the real-time device status, and output a scheduling execution sequence containing the scheduling order based on the initial priority. Specifically, this includes the following steps: S221. Calculate the initial priority of each execution step based on the equipment load rate in the real-time status of the equipment. The equipment load rate is negatively correlated with the initial priority.

[0037] Specifically, the equipment load rate is defined as the ratio of the number of tasks currently being executed by the executor to its maximum concurrency capacity. The lower the load rate, the more idle the equipment is, and the corresponding execution link should receive a higher initial priority. For example, when the load rate of a conveyor is 20%, its associated execution link is assigned priority 1 (highest); when the load rate of an AGV is 80%, its associated execution link is assigned priority 3 (lower).

[0038] S222. The execution phase is assigned to the corresponding priority queue in the multi-level feedback queue according to the initial priority. The multi-level feedback queue contains multiple priority queues, and different priority queues correspond to different time slice lengths and execution priorities.

[0039] For example, suppose a multi-level feedback queue includes three priority queues: the first priority queue has a time slice of 2 seconds, the second priority queue has a time slice of 3 seconds, and the third priority queue has a time slice of 5 seconds. The higher the priority of the queue, the shorter its time slice, to ensure that high-priority tasks can get a processing opportunity quickly.

[0040] S223. The execution steps in the current priority queue are executed sequentially using a time-slice round-robin method. When any execution step is completed within the allocated time slice, it is removed from the queue. When any execution step is not completed within the allocated time slice, it is downgraded to the next priority queue. When the waiting time of any execution step in the priority queue exceeds the first preset time threshold (e.g., 10 seconds), it is upgraded to a higher priority queue to prevent low-priority tasks from "starving" due to long-term lack of execution.

[0041] In one specific implementation, the multi-level feedback queue mechanism also includes: The execution steps that need to wait for a device response are identified as the first type. The execution steps corresponding to the first type are automatically given a higher priority than the current queue. For example, when an AGV execution step is waiting for a conveyor to connect, this step is an I / O intensive task, and the system automatically raises its priority by one level to reduce the waiting time.

[0042] Execution steps whose execution time exceeds a second preset time threshold are identified as the second type. Execution steps corresponding to the second type are downgraded in the scheduling process. For example, the execution steps of AGVs in long-distance transportation are computationally intensive tasks. The system will downgrade them in the scheduling process to avoid occupying high-priority queues for a long time.

[0043] The first type includes the execution phase of waiting for AGV to connect, and the second type includes the AGV execution phase of long-distance transportation.

[0044] S224. Output the scheduling execution sequence after multi-level feedback queue scheduling. The execution steps in the scheduling execution sequence are arranged in the scheduling order.

[0045] After the above scheduling, the system outputs a scheduled execution sequence with a clear execution order. For example, the original execution sequence is [E1 (conveyor #3), E2 (AGV #5), E3 (stacker crane #2)], and after multi-level feedback queue scheduling, the output scheduled execution sequence may be [E1, E3, E2], where E1 and E3 are given priority for execution due to their lower equipment load rates.

[0046] S23. Based on the scheduling order, equipment availability (i.e., whether each execution entity is currently idle) and channel congestion level, allocate an execution time window for each execution stage; the execution time window includes the expected start time and the expected end time.

[0047] For example, for the first execution stage E1 (conveyor #3) in the scheduling sequence, if conveyor #3 is currently idle and the channel is unobstructed, its start time is set to the current time T0, and its end time is calculated as T1 based on the preset execution duration. For the second execution stage E2 (AGV #5), the execution completion time T1 of E1, as well as the current position of AGV #5 and the degree of channel congestion, need to be considered. If AGV #5 needs to move to the connection point and the channel is slightly congested, its start time is set to T1 + ΔT, and its end time is T2, and so on.

[0048] S24. Based on the scheduling order, real-time equipment location, and channel congestion level of the warehouse layout, plan the initial path for the AGV execution process.

[0049] The initial path is generated using the A* algorithm to create a globally optimal path from the origin to the destination. The heuristic function of the A* algorithm incorporates congestion prediction as a weighting factor in the path cost, guiding the AGV to prioritize routes with lower congestion levels. The congestion prediction is dynamically updated based on historical path data and the current task distribution.

[0050] Based on the globally optimal path, the Dynamic Window Approach (DWA) is used for local trajectory optimization. The DWA algorithm samples the velocity-angular velocity combination within the prediction time window, evaluates the obstacle distance, velocity target, and smoothness of each trajectory, and outputs the optimal control command in real time, enabling the AGV to flexibly avoid sudden obstacles during operation.

[0051] S25. Encapsulate the execution phase, the bound execution entity, the execution time window (expected start and end time), and the initial path (coordinate sequence) of the AGV execution phase into scheduling instructions, and issue them to the corresponding execution entities in the order of scheduling.

[0052] After receiving the instruction, the executing entity performs the corresponding warehousing operations according to the time window and path requirements in the instruction.

[0053] S3. During execution, to address the technical issue of the system failing to respond to dynamic changes after the scheduling strategy is issued, the system monitors the operational status of each execution entity in real time and establishes a multi-time-scale scheduling framework to dynamically optimize the issued initial scheduling strategy. Specifically, task allocation is periodically optimized at the global scale, and the operating parameters of the execution entities are adjusted in real time at the local scale. When path congestion or equipment load fluctuations are detected, the AGV's operating path or execution order is fine-tuned in real time. This includes the following steps: S31. At the global scale, periodically (e.g., every 30 seconds) trigger global scale optimization to obtain the actual execution progress and real-time status of each execution entity, and identify the deviation between the initial scheduling strategy and the actual execution progress; wherein, the actual execution progress includes the number of completed execution steps and the completion percentage of the current execution step.

[0054] The actual execution progress collected is compared with the preset execution time window in the initial scheduling strategy, and the deviation value is calculated.

[0055] S32. When the deviation exceeds the preset deviation threshold (e.g., 15%), the system triggers a global re-optimization.

[0056] Based on the unexecuted execution stages and the real-time status of each execution entity's equipment, the unexecuted execution stages include those that have not yet started and those that have been interrupted. Using dynamic generation logic, the optimal task allocation scheme is recalculated. Specifically, the system uses the Hungarian algorithm or a genetic algorithm to find the optimal match based on the current load rate, fault status, availability of each execution entity, and the capacity requirement tags of each execution stage. The objective function is to minimize the total completion time.

[0057] Based on the reassigned task allocation scheme, the execution order and binding relationship of the execution entities of the unexecuted execution stages are adjusted, and the updated scheduling strategy is temporarily stored until it takes effect in the next execution cycle.

[0058] S33. At the local scale, monitor the operational status of each executing entity in real time and respond quickly to emergencies.

[0059] S331. When a temporary obstacle or sudden congestion is detected on the initial path corresponding to the AGV, a local path correction is triggered. An obstacle avoidance trajectory is generated in real time based on the dynamic window method, and the original path is locally corrected.

[0060] The DWA algorithm samples multiple velocity combinations in the velocity space, with each combination corresponding to a candidate trajectory. The system evaluates each candidate trajectory based on the following three metrics: Obstacle distance: The distance between the trajectory and the nearest obstacle; the greater the distance, the higher the score. Target direction: The angular deviation between the trajectory endpoint and the target point; the smaller the deviation, the higher the score. Velocity smoothness: the rate of change of the trajectory's velocity; the smaller the change, the higher the score.

[0061] The trajectory with the highest overall score is selected as the local correction path, and the corrected path instructions are sent to the AGV for execution in real time. The entire process is completed in milliseconds, ensuring that the AGV can avoid obstacles in time without affecting the overall operation efficiency.

[0062] S332. When the actual execution time of any execution stage exceeds the execution time window, a timeout handling mechanism is triggered to dynamically adjust the waiting time of subsequent dependent stages and update the passage priority of the relevant AGVs. Specifically: First, identify the subsequent dependent processes affected by this execution process. Dependencies include time-series dependencies (such as "AGVs can only connect after the conveyor is completed") and resource dependencies (such as multiple AGVs sharing the same channel).

[0063] Secondly, the system dynamically adjusts the waiting time of subsequent dependent links based on the timeout duration. For example, if the conveyor execution link times out by 10 seconds, the system will automatically extend the waiting time of the AGV connection link that depends on the conveyor by 10 seconds.

[0064] Simultaneously, the system updates the passage priority of the relevant AGVs. Based on the multi-level feedback queue mechanism, the priority of the affected AGVs is temporarily increased by one level to ensure their priority passage in subsequent paths, thereby compensating for the delays caused by timeouts.

[0065] S34. When there is a conflict between the local-scale adjustment and the global-scale task allocation scheme, arbitration shall be conducted according to the following coordination mechanism: Specifically, when there is a discrepancy between the local scale adjustment instruction and the global scale optimization scheme, the system triggers an arbitration mechanism, including but not limited to: Local adjustments require AGVs to detour through a certain channel, while the global optimization scheme requires the AGVs to travel along the original path; local adjustments increase the passage priority of a certain AGV, while the global optimization scheme gives the AGVs a lower priority.

[0066] If the impact of a local adjustment is limited to a single roadway or a single AGV and does not affect the timing constraints of other execution stages, then the local adjustment shall be executed first. Local adjustments may affect the timing constraints of multiple execution stages or lead to resource competition. For example, if multiple AGVs are competing for the same channel at the same time, the local adjustment will be suspended, and the global optimization plan will be used as the basis. The local adjustment requirements will be used as the input for the next cycle of global optimization.

[0067] In addition, the system supports flexible configuration of multiple arbitration principles, including at least one of the following: high-priority task priority principle, resource consumption minimization principle, and task completion time shortest principle.

[0068] S4. When any execution entity encounters an anomaly, the current warehouse scenario status is reacquired and the scheduling weights of each task type are dynamically adjusted. Based on the scheduling weights, the priority of the affected execution stages is calculated, an updated scheduling strategy is generated, and execution is initiated. The specific steps include the following: S41. When an abnormal event is detected in any execution entity, the replanning process is triggered; Examples include: failure of the execution entity (such as AGV motor overheating, stacker crane positioning loss, conveyor communication interruption); path congestion (multiple vehicles forming traffic bottlenecks at intersections or narrow alleys); addition of new high-priority tasks (such as emergency order insertion, production line shutdown for emergency repairs); task execution timeout (task not completed within the specified time).

[0069] S42. Obtain the current warehouse scenario status, which includes at least one of the following: inventory occupancy rate, backlog of outbound tasks, number of available storage locations, and degree of channel congestion.

[0070] S43. Input the warehouse scenario status into the preset weight adjustment model to dynamically adjust the scheduling weights of inbound tasks, outbound tasks, and transfer tasks. Among them, the inventory occupancy rate and the backlog of outbound tasks are related to the scheduling weight of outbound tasks, the number of available storage locations is related to the scheduling weight of inbound tasks, and the degree of channel congestion is related to the scheduling weight of transfer tasks.

[0071] In this embodiment, the weight adjustment follows the following rules: When the inventory occupancy rate exceeds the preset inventory occupancy threshold, it indicates that the warehouse storage space is tight. The system increases the scheduling weight of outbound tasks to speed up the outbound release of goods and free up storage space. When the number of available storage locations is lower than the preset threshold, it indicates that there are not enough available storage locations. The system will increase the scheduling weight of the inbound task, but this needs to be judged in conjunction with the inventory occupancy rate. When the backlog of outbound tasks exceeds the preset backlog threshold, it indicates that the outbound pressure is high, and the system increases the scheduling weight of outbound tasks. When the congestion level of a passage exceeds the preset congestion level threshold, it indicates that the transportation passage is blocked. The system will appropriately reduce the weight of the warehouse transfer task that needs to pass through the congested passage, or increase the weight of the detour route task.

[0072] In one specific implementation, the weight adjustment model has a built-in weight mapping table. When the inventory occupancy rate exceeds 90%, the system increases the weight of the outbound task from the default value of 0.3 to 0.6; when the number of idle storage locations is less than 10% of the total number of storage locations, the system increases the weight of the inbound task from the default value of 0.4 to 0.7.

[0073] In another specific implementation, "increasing scheduling weight" means that, under the same conditions, the task type with increased weight is given priority over other task types in the execution order. For example, when the inventory occupancy rate exceeds 90%, the system marks outbound tasks as "high priority" and inserts all outbound tasks at the head of the task queue, ensuring that outbound tasks are executed before inbound and transfer tasks. "Increasing scheduling weight" is achieved by adjusting the insertion position of tasks in the task queue: the higher the inventory occupancy rate, the earlier the outbound task is inserted in the queue. When the inventory occupancy rate is between 80% and 90%, outbound tasks are inserted in the first 30% of the queue; when the inventory occupancy rate exceeds 90%, outbound tasks are inserted in the first 10% of the queue, ensuring that outbound tasks are executed with priority.

[0074] S44. Identify the execution steps affected by the abnormal event. Based on the adjusted scheduling weights, combined with task urgency, equipment status, and the waiting time of the execution steps, recalculate the priority of the affected execution steps. This includes the following steps: S441. An execution stage dependency graph generated based on the execution sequence, which records the temporal coupling relationship between each execution stage and its upstream and downstream stages: For inbound tasks: the AGV process can only begin after the conveyor process is completed; the stacker crane process can only begin after the AGV process is completed. For outbound tasks: the AGV process can only begin after the stacker crane process is completed; the conveyor process can only begin after the AGV process is completed. For warehouse transfer tasks: the AGV process can only begin after the stacker crane (picking) process is completed; the stacker crane (discharging) process can only begin after the AGV process is completed.

[0075] When an abnormal event occurs, the forward propagation method is used to locate the execution link where the abnormal event occurred, and the execution link and all downstream execution links are marked as affected execution links; for example, when an AGV fails during the execution of an inbound task, the AGV link is marked as failed, and its downstream stacker crane link is also marked as affected.

[0076] S442. A multi-factor dynamic weighting mechanism is used to calculate the priority score of each affected execution link. The multi-factor dynamic weighting mechanism includes task urgency factor, auction algorithm factor, equipment status factor and scheduling weight factor.

[0077] The formula for calculating priority scores is: ; Among them, the scheduling weight factor is the scheduling weight of the task type after dynamic adjustment. When the inventory occupancy rate is high, the outbound task receives a higher scheduling weight; when there are insufficient empty storage spaces, the inbound task receives a higher scheduling weight.

[0078] The urgency of a task is determined based on a preset outbound time limit or order priority; the shorter the outbound time limit or the higher the order priority, the larger the value of the task urgency factor.

[0079] The auction algorithm factor is determined based on at least one of the AGV's remaining power, current path length, and estimated completion time. When multiple AGVs compete for the same path resource, each AGV calculates its bid based on its remaining power, current path length, and estimated completion time. The central scheduler uses the Vickrey auction model to allocate passage rights. The AGV with the higher bid gets passage rights, and the corresponding execution stage gets a higher auction algorithm factor.

[0080] Equipment status factors include at least one of the following: AGV remaining power, stacker crane load rate, and conveyor queue length; the better the equipment status (e.g., sufficient AGV remaining power, low stacker crane load rate, and short conveyor queue), the larger the value of the equipment status factor.

[0081] S443. Sort the affected execution steps according to priority scores and generate an updated execution order.

[0082] S45. Based on the recalculated priority, reallocate execution entities to the affected execution stages and adjust the execution order: When an AGV malfunctions, the unfinished execution steps are transferred to a backup AGV; When a stacker crane malfunctions, the execution links bound to that stacker crane will be reassigned to other available stacker cranes in the same area; When a conveyor malfunctions, a buffer zone is added in front of the malfunctioning conveyor to temporarily store information, and subsequent execution processes are routed to the backup conveyor channel.

[0083] S46. The affected AGV execution stages are replanned using a layered architecture: the global layer generates the optimal path based on the A* algorithm, the local layer performs real-time obstacle avoidance using the Dynamic Window Method (DWA), and incremental path updates are performed using the D Lite algorithm, recalculating only the affected nodes to avoid global recalculation. The updated scheduling strategy is then generated and executed, completing the replanning loop.

[0084] Based on the same inventive concept described above, this application also discloses a multi-device collaborative warehouse scheduling system, with the following architecture: Figure 2 As shown, it includes the following modules: The execution sequence generation module is used to decompose the received warehousing tasks and dynamically generate execution sequences based on the task type, real-time equipment status, and warehousing layout status. The task types include inbound tasks, outbound tasks, and transfer tasks. Each execution step in the execution sequence is bound to a unique execution entity. The multi-level feedback queue scheduling module is used to schedule the execution sequence based on the real-time status of the equipment and the warehouse layout status, generate an initial scheduling strategy, and send it to the corresponding execution subject for execution. The anomaly feedback and replanning module is used to reacquire the current warehouse scenario status and dynamically adjust the scheduling weights of each task type when any execution entity encounters an anomaly. Based on the scheduling weights, the priority of the affected execution links is calculated, an updated scheduling strategy is generated, and the strategy is issued for execution.

[0085] This system adopts a five-layer architecture, from top to bottom: presentation layer, management layer, scheduling layer, control layer, and device layer. Each layer interacts with the others through standardized interfaces to jointly schedule and execute warehousing tasks.

[0086] The equipment layer includes field devices such as stacker cranes, AGVs, conveyors, sensors, and PLCs. Each PLC controls a group of devices, responsible for local logic control and status feedback. Stacker cranes handle vertical storage and retrieval operations on the racks, AGVs handle horizontal transportation operations within the warehouse area, and conveyors handle batch transport of goods along fixed aisles. Each device communicates with the upper control layer via a fieldbus, reporting its real-time status (running, faulty, idle) and receiving control commands.

[0087] The control layer connects to the device layer via an industrial communication protocol to enable command issuance and status acquisition. The control layer includes a device status monitoring module, a command execution and feedback module, and an anomaly handling and alarm module. A heartbeat mechanism maintains stable communication between the control layer and the device layer; when a device communication interruption is detected for more than a preset time threshold (e.g., 500ms), an anomaly report is automatically triggered.

[0088] The scheduling layer receives task instructions from the management layer and performs task decomposition, path planning, and equipment scheduling. The scheduling layer includes a task decomposition and allocation module, a path planning and conflict detection module, a priority determination and task queuing module, and an AGV and conveyor collaborative scheduling module. The scheduling layer uses the multi-level feedback queue mechanism of this embodiment to schedule the execution stages and plans paths for the AGVs based on the A* algorithm and the Dynamic Window Method (DWA).

[0089] The management layer interfaces with the WMS system to receive inbound and outbound tasks and provide feedback on task execution status. The management layer includes an upstream system interface module, a system status visualization module, and a scheduling strategy configuration and optimization module. Based on the task type (inbound, outbound, transfer) issued by the WMS, the management layer invokes the scheduling layer to decompose the task and generate scheduling strategies.

[0090] The presentation layer provides a large-scale scheduling and monitoring dashboard via web or client, supporting real-time viewing of task progress, device status, and alarm information, as well as manual intervention and scheduling strategy adjustments. The presentation layer includes a visual monitoring interface, a task and device status display module, and a user operation and log recording module.

[0091] In this embodiment, the data interaction between different levels follows the following logic: Presentation and Management Layers: Task commands are issued and status is displayed via the HTTP / WebSocket protocol. Users configure scheduling strategies through the presentation layer, and the management layer pushes device status and task progress to the presentation layer in real time.

[0092] Management and scheduling layers: Task decomposition and scheduling strategy distribution are implemented through internal APIs or message queues. The management layer converts WMS tasks into standard format task instructions, and the scheduling layer returns the task execution status.

[0093] Scheduling and Control Layers: These layers issue control commands and acquire equipment status via TCP / IP or industrial protocols. The scheduling layer converts the execution sequence into control commands recognizable by the equipment, while the control layer provides real-time feedback on the equipment status.

[0094] Control Layer and Equipment Layer: Real-time control and status acquisition of equipment are achieved through PLC and fieldbus. The control layer uses a polling method to collect equipment status and a heartbeat mechanism to maintain communication.

[0095] The five-layer decoupled architecture of this invention achieves complete separation of business logic, scheduling decision, equipment control, and physical execution. Each layer only completes data interaction through standardized interfaces, which can quickly adapt to new equipment types and warehousing business scenarios, and greatly reduces the difficulty of system maintenance and upgrades.

[0096] Based on the same inventive concept described above, this application also discloses a computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set. The at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the multi-device collaborative warehouse scheduling method described above.

[0097] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware, or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and other media capable of storing program code.

[0098] The above description is merely an optional embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A multi-device collaborative warehouse scheduling method, characterized in that, Applied to warehousing systems including stacker cranes, AGVs, and conveyors, the method includes the following steps: The received warehousing tasks are decomposed, and an execution sequence is dynamically generated based on the task type, real-time equipment status, and warehousing layout status. The task types include inbound tasks, outbound tasks, and transfer tasks. Each execution step in the execution sequence is bound to a unique execution entity. Based on the real-time status of the equipment and the warehouse layout status, a multi-level feedback queue mechanism is used to schedule the execution sequence, generate an initial scheduling strategy, and send it to the corresponding execution entity for execution. When any of the execution entities malfunctions, the current warehouse scenario status is obtained and the scheduling weights of each task type are dynamically adjusted. The priority of the affected execution links is calculated based on the scheduling weights, an updated scheduling strategy is generated, and the strategy is issued for execution.

2. The multi-device collaborative warehouse scheduling method according to claim 1, characterized in that, The process of decomposing the received warehousing tasks and dynamically generating an execution sequence based on the task type, real-time equipment status, and warehouse layout status includes the following steps: Receive warehousing tasks and break them down to obtain multiple task types for the warehousing tasks. Based on the task types, determine the type sequence of each task corresponding to the executing entity. The system acquires real-time equipment status and warehouse layout status, and dynamically binds the corresponding execution entity to each execution stage in the type sequence based on the real-time equipment status; the real-time equipment status includes real-time equipment location, equipment load rate, equipment fault status, and equipment availability. Based on the warehouse layout status, configure execution parameters for each execution stage; the warehouse layout status includes the distribution of storage locations, the degree of channel congestion, and the occupancy of connection points. An execution sequence is dynamically constructed in real time based on the type sequence, the execution subject, and the execution parameters. Each execution sequence includes at least two of the following execution links: conveyor execution link, AGV execution link, and stacker crane execution link.

3. The multi-device collaborative warehouse scheduling method according to claim 2, characterized in that, The process of dynamically constructing the execution sequence in real time also includes dynamically reconstructing the unexecuted execution sequence when the real-time status of the device or the warehouse layout status changes, including at least one of the following situations: When the congestion level of the channel exceeds a preset congestion threshold, a detour path is generated for the AGV execution process; When the equipment load rate of the conveyor exceeds a preset load threshold, a buffer temporary storage step is added before the conveyor execution step; When the stacker crane fails, the execution links bound to the stacker crane will be reassigned to other available stacker cranes.

4. The multi-device collaborative warehouse scheduling method according to claim 2, characterized in that, The step of scheduling the execution sequence based on the real-time status of the equipment and the warehouse layout status using a multi-level feedback queue mechanism to generate an initial scheduling strategy and then issuing it to the corresponding execution entity for execution includes the following steps: Obtain the dynamically generated execution sequence, as well as the current real-time status of the device and the warehouse layout status; The execution steps in the execution sequence are input into a multi-level feedback queue mechanism. The initial priority of each execution step is determined according to the device load rate in the real-time status of the device. Based on the initial priority, a scheduling execution sequence containing the scheduling order is output. Based on the scheduling order, the equipment availability, and the channel congestion level, an execution time window is allocated to each execution stage; Based on the scheduling order, the real-time location of the equipment, and the channel congestion level of the warehouse layout, an initial path is planned for the AGV execution process. The initial path is generated globally optimal based on the A* algorithm, and local trajectory optimization is performed using the Dynamic Window Method (DWA). The execution stage, the bound execution entity, the execution time window, and the initial path of the AGV execution stage are encapsulated into scheduling instructions, which are then sequentially sent to the corresponding execution entities according to the scheduling order.

5. The multi-device collaborative warehouse scheduling method according to claim 4, characterized in that, The scheduling execution sequence, which includes the scheduling order, is output based on a multi-level feedback queue mechanism, specifically including the following steps: Based on the device load rate in the real-time status of the device, the initial priority of each execution step is calculated, and the device load rate is negatively correlated with the initial priority; The execution phases are allocated to corresponding priority queues in a multi-level feedback queue according to the initial priority. The multi-level feedback queue contains multiple priority queues, and different priority queues correspond to different time slice lengths and execution priorities. The execution steps in the current priority queue are executed sequentially using a time-slice rotation method. When any execution step is completed within the allocated time slice, it is removed from the queue. When any execution step is not completed within the allocated time slice, it is downgraded to the next priority queue. When the waiting time of any execution step in the priority queue exceeds a first preset time threshold, it is upgraded to a higher priority queue. The output is the scheduling execution sequence after the multi-level feedback queue scheduling, and the execution steps in the scheduling execution sequence are arranged according to the scheduling order.

6. The multi-device collaborative warehouse scheduling method according to claim 4, characterized in that, The execution process also includes establishing a multi-time-scale scheduling framework to dynamically optimize the issued initial scheduling strategy, specifically including the following steps: At a global scale, the actual execution progress of each execution entity and the real-time status of the device are periodically acquired to identify the deviation between the initial scheduling strategy and the actual execution progress. When the deviation exceeds a preset deviation threshold, the optimal task allocation scheme is recalculated based on the unexecuted execution steps and the real-time status of the devices of each execution subject. Based on the re-allocated task allocation scheme, the execution order of the unexecuted execution steps and the binding relationship of the execution subjects are adjusted. At a local scale, the operational status of each of the aforementioned execution entities is monitored in real time; When a temporary obstacle or sudden congestion is detected on the initial path corresponding to the AGV, an obstacle avoidance trajectory is generated in real time based on the dynamic window method, and the original path is locally corrected. When the actual execution time of any of the execution steps exceeds the execution time window, the waiting time of the subsequent dependent steps of the execution step is dynamically adjusted, and the passage priority of the relevant AGV is updated.

7. The multi-device collaborative warehouse scheduling method according to claim 1, characterized in that, When any of the execution entities malfunctions, the current warehouse scenario status is obtained and the scheduling weights of each task type are dynamically adjusted. Based on the scheduling weights, the priority of the affected execution stages is calculated, an updated scheduling strategy is generated, and execution is initiated. This process includes the following steps: When an abnormal event is detected in any of the aforementioned execution entities, a replanning process is triggered; Obtain the current warehouse scenario status, which includes at least one of the following: inventory occupancy rate, backlog of outbound tasks, number of available storage locations, and degree of channel congestion. The state of the warehousing scenario is input into a preset weight adjustment model to dynamically adjust the scheduling weights of the inbound task, the outbound task, and the transfer task. Wherein, the inventory occupancy rate and the backlog of outbound tasks are related to the scheduling weight of the outbound tasks, the number of free storage locations is related to the scheduling weight of the inbound tasks, and the degree of channel congestion is related to the scheduling weight of the transfer tasks. Identify the execution steps affected by the abnormal event, and recalculate the priority of the affected execution steps according to the adjusted scheduling weight; Based on the recalculated priorities, the execution entities are reassigned and the execution order is adjusted for the affected execution stages. The paths for the affected AGV execution stages are replanned, and the updated scheduling strategy is generated and issued for execution.

8. The multi-device collaborative warehouse scheduling method according to claim 7, characterized in that, The step of recalculating the priority of the affected execution stages based on the adjusted scheduling weights specifically includes the following steps: Based on the execution sequence, the execution stage dependency graph generated is used to locate the execution stage where the abnormal event occurred using a forward propagation method, and the execution stage and all downstream execution stages are marked as affected execution stages. A multi-factor dynamic weighting mechanism is used to calculate the priority score of each affected execution stage. The multi-factor dynamic weighting mechanism includes a task urgency factor, an auction algorithm factor, a device status factor, and a scheduling weight factor. The scheduling weight factor is the dynamically adjusted scheduling weight of the task type, and the task urgency is determined according to the preset outbound time limit; the auction algorithm factor is determined according to at least one of the remaining power of the AGV, the current path length, and the estimated completion time; and the equipment status factor includes at least one of the remaining power of the AGV, the stacker crane load rate, and the conveyor queue length. The affected execution steps are sorted according to the priority scores to generate an updated execution order.

9. A multi-device collaborative warehouse scheduling system, characterized in that, Includes the following modules: The execution sequence generation module is used to decompose the received warehousing tasks and dynamically generate an execution sequence based on the task type, real-time equipment status, and warehousing layout status of the warehousing tasks. The task types include inbound tasks, outbound tasks, and transfer tasks. Each execution step in the execution sequence is bound to a unique execution entity. The multi-level feedback queue scheduling module is used to schedule the execution sequence according to the real-time status of the equipment and the warehouse layout status, and generate an initial scheduling strategy and send it to the corresponding execution subject for execution. The exception feedback and replanning module is used to reacquire the current warehouse scenario status and dynamically adjust the scheduling weights of each task type when any of the execution entities encounters an exception. Based on the scheduling weights, the priority of the affected execution links is calculated, an updated scheduling strategy is generated, and the strategy is issued for execution.

10. A computer-readable storage medium, characterized in that, The readable storage medium stores at least one instruction, at least one program, code set, or instruction set, which is loaded and executed by a processor to implement the multi-device collaborative warehouse scheduling method as described in any one of claims 1 to 8.

Images