High-density storage intelligent management system, method and application

Abstract

The invention discloses a high-density warehouse intelligent management system and method and application, and relates to the technical field of intelligent warehouse logistics, the system comprises a dynamic collaborative decision framework, the input end of the dynamic collaborative decision framework is connected with a multi-mode sensing module, and the output end of the dynamic collaborative decision framework is connected with a dynamic game scheduling module and a conflict autonomous resolution module. According to the high-density storage intelligent management system and method and the application, the real-time response capability and the resource utilization rate of the high-density storage system are improved through the dynamic collaborative decision framework and the multi-mode sensing mechanism. Based on an incremental state transition model and a local re-calculation strategy, the system can update equipment capability and environment constraint data in time, and calculation power consumption and response delay caused by traditional global optimization are avoided; through a task-equipment bidirectional matching game algorithm and a hierarchical conflict resolution mechanism, autonomous collaborative scheduling of heterogeneous equipment groups is realized, path cross deadlock and goods shelf scrambling conflicts are effectively reduced, and the system throughput fluctuation rate is reduced.

Description

Technical Field

[0001] The present invention relates to the field of intelligent warehousing and logistics technology, and specifically to a high-density warehousing intelligent management system, method and application. Background Art

[0002] With the increasing demand for space utilization and operational efficiency in the modern logistics industry, high-density warehousing systems, through the in-depth application of three-dimensional racking layouts, dense storage strategies, and automated equipment clusters, have become a core direction for industry upgrades. However, when dealing with complex and dynamic scenarios, existing intelligent management systems still rely on pre-set static task allocation rules and fixed path planning algorithms, resulting in insufficient system flexibility. Frequent dynamic variables in warehousing operations, such as order flow fluctuations, equipment failures, and varying cargo specifications, make it difficult for traditional scheduling strategies based on periodic global optimization to capture these changes in real time. For example, when multiple stacker cranes, AGVs, and shuttles perform inbound and outbound tasks simultaneously, real-time differences in equipment status (such as battery level, load capacity, and operating speed) are not dynamically detected. Consequently, task allocation and equipment capabilities are not accurately matched, easily leading to localized equipment overload or idleness. Furthermore, when heterogeneous equipment paths intersect or compete for resources, the system relies solely on pre-set priorities or manual intervention to resolve conflicts, failing to autonomously predict and resolve potential deadlocks, resulting in process interruptions and reduced throughput. Existing patented technologies mostly focus on the path optimization of a single device or the static design of inventory layout, and fail to build a dynamic decision-making framework covering the collaboration of multiple devices, making it difficult to break through the bottleneck of the overall operating efficiency and robustness of high-density warehousing systems. Summary of the Invention

[0003] (1) Technical problems solved In response to the shortcomings of the existing technology, the present invention provides a high-density warehousing intelligent management system, method and application, which solves the problems of real-time collaborative scheduling and autonomous conflict resolution of heterogeneous equipment groups in dynamic warehousing scenarios.

[0004] (2) Technical solution To achieve the above objectives, the present invention is implemented through the following technical solutions: A high-density warehouse intelligent management system, comprising: The dynamic collaborative decision-making framework has its input connected to the multimodal perception module, and its output connected to the dynamic game scheduling module and the autonomous conflict resolution module respectively; The multimodal perception module collects equipment operating parameters, cargo attributes, and storage channel status data in real time through edge computing nodes deployed on stackers, AGVs, and shuttles, and constructs a spatiotemporal state matrix containing equipment capability vectors, task requirement vectors, and environmental constraint vectors. The dynamic game scheduling module generates task allocation instructions through an asymmetric Nash equilibrium algorithm based on the task priority weights and the equipment comprehensive efficiency index in the spatiotemporal state matrix; The asymmetric Nash equilibrium algorithm prioritizes the competitive power of devices and tasks in task allocation: high-priority tasks are prioritized for high-performance devices, while at least one redundant device is reserved as a buffer for unexpected tasks. For example, if a stacker crane's performance index drops due to excessive path overlap, its high-priority tasks are migrated to a backup shuttle with an independent path and sufficient battery power, ensuring uninterrupted critical tasks.

[0005] The autonomous conflict resolution module analyzes the device motion trajectory prediction data in the spatiotemporal state matrix and triggers the virtual track offset technology or the distributed negotiation protocol to resolve the conflict; A dual-closed-loop verification module, whose input is connected to the dynamic collaborative decision-making framework and whose output reversely corrects the algorithm parameters of the dynamic game scheduling module through the physical system feedback loop; The dual closed-loop verification module includes a digital twin verification platform for simulating scheduling strategy verification in extreme scenarios; The spatiotemporal state matrix of the dynamic collaborative decision-making framework is updated through an incremental state migration model, including: when the equipment status or warehouse environment changes, only the equipment capability vector or environmental constraint vector in the affected area is locally recalculated, and millisecond-level updates are achieved through edge computing nodes.

[0006] The technology of the high-density warehousing intelligent management system lies in building a dynamic collaborative scheduling and autonomous conflict resolution system for heterogeneous equipment groups in high-density warehousing scenarios, and realizing the improvement of the overall efficiency and robustness of the system through multi-module linkage, including: in the equipment status perception and data processing links, the system collects equipment operating parameters in real time through edge computing nodes deployed on stackers, AGVs and shuttle vehicles, including but not limited to power, speed, load and robotic arm lifting height.

[0007] Cargo attributes are captured by scanning labels to determine weight, volume, and order urgency. Warehouse environment data relies on fusion sensing technology using LiDAR and cameras to monitor aisle congestion and shelf occupancy in real time. After preprocessing at edge nodes, this sensing data is mapped into device capability vectors, task requirement vectors, and environmental constraint vectors, creating a dynamically updated spatiotemporal state matrix.

[0008] When there are local changes in device status or environmental constraints, the system uses an incremental update mechanism to recalculate only the vectors in the affected areas. For example, when the AGV's battery power drops, only the device capability vectors of its channel and associated shelf area are updated, rather than globally reconstructing the matrix. This reduces computing power consumption and improves response speed.

[0009] Preferably, the task priority weights in the dynamic game scheduling module are calculated based on a base priority calculated based on cargo weight and volume, and the time decay coefficient of the order urgency is added to generate dynamic task priority weights. The dynamic scheduling process utilizes a two-way matching game mechanism between tasks and equipment. Task priority weights are dynamically generated based on cargo weight, volume, and order urgency. Urgency is calculated using a decay function of the remaining order processing time; the more urgent the time, the faster the weight increases.

[0010] The comprehensive device performance index includes an energy efficiency factor, a path overlap factor, and a task completion factor. The energy efficiency factor is dynamically calculated based on the ratio of the device's unit power consumption to its rated power. The path overlap factor is updated in real time based on the overlap length between the device's current path and the globally planned path. The task completion factor is generated based on weighted statistics of historical task completion timeliness and success rates. The comprehensive device performance index is calculated using energy efficiency, path overlap, and historical task completion rates. When a device's power level falls below a safety threshold or its path overlap is excessively high, the system automatically lowers its performance index to mitigate risk.

[0011] Preferably, the virtual track offset technology of the autonomous conflict resolution module specifically includes: While maintaining the global path planning of the devices, the system adjusts the device's operating speed through microsecond-level speed commands, causing the motion trajectories of the conflicting devices to be misaligned in time and space. The conflict resolution module handles device interaction conflicts in dynamic scenarios through a hierarchical response mechanism. For minor conflicts where paths intersect but the time window is adjustable, the system uses virtual track offset technology. While maintaining the global path planning, it adjusts the device's operating rhythm through microsecond-level speed commands, causing the time and space trajectories of the conflicting devices to be misaligned. For example, when an AGV and a stacker are expected to collide at an intersection, the system sends a deceleration command to the AGV and simultaneously increases the stacker's speed, increasing the time difference between the two to exceed the safety threshold.

[0012] Preferably, the execution steps of the distributed negotiation protocol are: Conflicting devices generate at least two alternative paths based on local state exchange and use a lightweight consensus algorithm to select the path with the best weighted average time and energy consumption. For severe conflicts, such as shelf contention or complete path overlap, a distributed negotiation protocol is triggered. Conflicting devices generate alternative path solutions based on local state exchange and use a consensus algorithm to select the solution with the best overall time and energy consumption. During the negotiation process, if the difference in scores between devices is less than a preset threshold, forced arbitration is implemented based on task priority and device performance index, ensuring that the higher-priority task is executed first.

[0013] Preferably, the digital twin verification platform verifies the scheduling strategy in the following ways: Extreme scenario data, including order spikes, multiple device failures, and channel congestion, was injected into the twin environment. Monte Carlo simulations were then used to evaluate system throughput fluctuations and conflict resolution success rates. System reliability was ensured through a dual closed-loop mechanism, integrating the digital twin verification platform with the physical system feedback loop. The digital twin platform simulated extreme scenarios, such as order spikes, device failures, and channel congestion, in a virtual environment. Monte Carlo simulations were then used to evaluate the robustness of the scheduling strategy and optimize algorithm parameters based on the results.

[0014] Preferably, the physical system feedback loop realizes parameter correction in the following manner: Collect the deviation data between the actual task completion time and the predicted time, and dynamically adjust the weight of the path overlap rate factor in the equipment comprehensive effectiveness index; According to the conflict resolution failure records, the trajectory prediction time window of the spatiotemporal conflict probability field model is optimized.

[0015] The physical feedback loop collects device execution data in real time, compares actual and predicted results, and dynamically adjusts model parameters. For example, if an AGV's task execution time consistently exceeds the specified limit due to slippery surfaces, the system reduces the weight of the path overlap factor in its performance index and simultaneously adds a friction coefficient variable to the twin platform to optimize the simulation model. This dual closed-loop mechanism achieves a smooth transition through a gradual parameter migration strategy, preventing system oscillations caused by sudden parameter changes.

[0016] Preferably, the storage channel status data of the multimodal perception module includes: Channel congestion level perceived through LiDAR and camera fusion; Shelf occupancy status detection results based on infrared sensors.

[0017] For dynamic perception of the warehouse environment, lidar and cameras collaborate to determine channel congestion levels: lidar detects obstacle density and distribution, while cameras analyze equipment queue length and movement speed. Weighted fusion of these two data generates the final congestion level. Infrared sensors monitor shelf occupancy, determining whether a shelf layer is fully loaded based on reflected signal strength. Abnormal fluctuations trigger manual review. For example, a sudden drop in a shelf's infrared signal indicates that goods have been removed. The system immediately unfreezes the shelf and prioritizes replenishment tasks. In the event of sensor data conflicts, the system employs a layered independent processing mechanism, restricting operational permissions to only the abnormal area to avoid global misjudgments.

[0018] Preferably, the local recalculation range of the incremental state transition model is: centered on the device where the state change occurs, and radiating to other devices and shelf areas where there is a risk of intersection with its movement path or resource contention.

[0019] During local updates, the incremental state migration model dynamically defines the impact area based on the impact of the device state change. For example, when a shuttle triggers an environmental constraint update due to a fully loaded shelf, the system only reallocates tasks associated with that shelf and adjusts the path planning for adjacent devices. If a mismatch between tasks and device capabilities occurs after the update, the system forcibly reclaims the erroneous tasks and re-matches them to the optimal device. In scenarios with multiple devices operating concurrently, the system prioritizes low-efficiency device areas by efficiency index, while reserving high-efficiency resources to handle unexpected demands.

[0020] A high-density warehousing intelligent management method includes the following steps: S1: Use edge computing nodes to collect equipment operating parameters, cargo attributes, and storage environment data in real time to build a dynamically updated spatiotemporal state matrix; S2: Based on the task priority weight and the comprehensive equipment efficiency index, the asymmetric Nash equilibrium algorithm is used to achieve dynamic matching between tasks and equipment; S3: Predict device trajectory conflicts and trigger virtual track offset technology or distributed negotiation protocols in a hierarchical manner to resolve conflicts; S4: Optimize scheduling parameters through a dual closed-loop mechanism of the digital twin verification platform and the physical system feedback loop; S5: Modify the spatiotemporal state matrix using an incremental local update strategy based on changes in device status or environmental constraints.

[0021] According to the application of the high-density warehousing intelligent management system, the high-density warehousing intelligent management system is suitable for the following scenarios: Automated goods-to-person picking operations in e-commerce warehouses; High-density three-dimensional cold storage management in cold chain logistics; Real-time material distribution and warehouse scheduling in smart manufacturing factories; Multi-equipment collaborative loading and unloading and route planning for cross-border logistics transit warehouses; Dynamic storage and priority scheduling of temperature-sensitive goods in pharmaceutical warehousing.

[0022] (3) Beneficial effects The present invention provides a high-density warehousing intelligent management system, method, and application. It has the following beneficial effects: (1) This high-density warehousing intelligent management system, method, and application improves the real-time response capability and resource utilization of high-density warehousing systems through a dynamic collaborative decision-making framework and a multimodal perception mechanism. Based on an incremental state migration model and a local recalculation strategy, the system can timely update equipment capabilities and environmental constraint data, avoiding the computing power consumption and response delays brought about by traditional global optimization; through a task-equipment two-way matching game algorithm and a hierarchical conflict resolution mechanism, it realizes autonomous collaborative scheduling of heterogeneous equipment groups, effectively reducing path intersection deadlocks and shelf competition conflicts, and reducing the volatility of system throughput. At the same time, the dual closed-loop verification module ensures the robustness of the scheduling strategy in extreme scenarios through dynamic parameter correction through virtual simulation and physical feedback, solving the defect of traditional static scheduling relying on manual intervention.

[0023] (2) This high-density warehousing intelligent management system, method, and application achieves low-latency and high-reliability conflict resolution through a distributed negotiation protocol and edge computing architecture, compressing conflict response time and reducing the frequency of manual intervention. The combination of a digital twin verification platform and a dynamic self-learning mechanism enables the system to adapt to structural changes in the warehousing environment, shortening the policy adaptation cycle in new scenarios. Furthermore, layered exception handling and incremental update mechanisms prevent the global spread of localized failures, ensuring the continuous execution of critical tasks and providing an intelligent management solution that combines efficiency, stability, and scalability for high-density warehousing scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic diagram of the overall framework of the present invention; Figure 2 This is a control logic timing diagram of the present invention. DETAILED DESCRIPTION

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0026] See also Figure 1 and Figure 2 The present invention provides a technical solution: a high-density warehousing intelligent management system, comprising: The dynamic collaborative decision-making framework has its input connected to the multimodal perception module, and its output connected to the dynamic game scheduling module and the autonomous conflict resolution module respectively; The multimodal perception module collects equipment operating parameters, cargo attributes, and storage channel status data in real time through edge computing nodes deployed on stacker cranes, AGVs, and shuttles, and constructs a spatiotemporal state matrix that includes equipment capability vectors, task requirement vectors, and environmental constraint vectors. The dynamic game scheduling module generates task allocation instructions through an asymmetric Nash equilibrium algorithm based on the task priority weights and equipment comprehensive efficiency index in the spatiotemporal state matrix; The autonomous conflict resolution module analyzes the device motion trajectory prediction data in the spatiotemporal state matrix and triggers virtual track offset technology or distributed negotiation protocols to resolve conflicts. A dual-closed-loop verification module, whose input is connected to the dynamic collaborative decision-making framework and whose output reversely corrects the algorithm parameters of the dynamic game scheduling module through the physical system feedback loop; The dual closed-loop verification module includes a digital twin verification platform for simulating scheduling strategy verification in extreme scenarios; The dynamic collaborative decision-making framework's spatiotemporal state matrix is ​​updated via an incremental state migration model. This includes: when equipment status or the warehouse environment changes, only the equipment capability vector or environmental constraint vector in the affected area is locally recalculated, with millisecond-level updates achieved via edge computing nodes. It should be further explained that, in specific implementations, the dynamic collaborative decision-making framework's spatiotemporal state matrix is ​​locally and dynamically updated via an incremental state migration model.

[0027] In specific implementation, when a change in equipment status or a change in the warehouse environment is detected, the system first locates the affected area, including: if the status of a single device is abnormal, the device capability vector is recalculated only for the area covered by the motion path of the device; if it is an environmental constraint change, the environmental constraint vector is updated for all equipment task paths associated with the channel; among them, equipment status changes include AGV power below the safety threshold and stacker load exceeding the limit, and warehouse environment changes include sudden channel congestion and shelf occupancy status update; the motion path coverage area includes the current task target shelf and the passing channel, and environmental constraint changes include increased channel congestion level.

[0028] During the recalculation process, the edge computing node normalizes the parameters in the device capability vector to an efficiency coefficient between 0 and 1 based on the device's real-time data and the local environmental state. This coefficient is then matched with the cargo weight and volume thresholds in the task requirement vector. If the device capability coefficient falls below the task requirement threshold, the device is marked as "restricted" and task migration is triggered. The parameters in the device capability vector include power and speed. For example, when a shuttle vehicle's environmental constraint vector is updated due to a fully loaded shelf, the system only reallocates pending tasks associated with that shelf and adjusts the path planning of neighboring devices to avoid that area. If a stacker crane's lifting height is insufficient to access a high-level shelf, the lifting parameters in its device capability vector are locally updated, and the task is automatically downgraded to a mid- or low-level shelf.

[0029] For scenarios where multiple devices change their status concurrently, the system sorts them by device efficiency index, prioritizes recalculating the coverage areas of low-efficiency devices, and reserves resources for high-efficiency devices to handle emergency tasks. This scenario involves multiple AGVs issuing simultaneous power alarms.

[0030] After the incremental update is completed, the spatiotemporal state matrix only synchronizes the vector data of the changed part to ensure millisecond-level response. At the same time, verification logic is used to prevent local updates from causing global conflicts, including: if there is a contradiction between the updated device capability vector and the global task allocation, such as the device is assigned a task that exceeds its updated capability threshold, a secondary correction mechanism is triggered to force the recovery of the erroneous task and re-match.

[0031] The task priority weight of the dynamic game scheduling module is calculated based on the basic priority according to the weight and volume of the goods, and the time decay coefficient of the order urgency is superimposed to generate the dynamic task priority weight. It should be further explained that in the specific implementation process, the task priority weight of the dynamic game scheduling module is dynamically generated by the weight, volume and urgency of the order. In the specific implementation, the system first calculates the basic weight according to the ratio of the weight of the goods to the preset standard load threshold, including: if the weight of the goods exceeds the standard value, the priority is linearly increased according to the overweight ratio to ensure matching with high-load equipment; for goods with abnormal volume, such as overlong or overwide, the priority correction coefficient is additionally superimposed according to the proportion of shelf space occupied by it. If the volume exceeds the operating threshold of the sorting robot arm, the manual intervention flag is triggered and the task weight is temporarily frozen.

[0032] The order's urgency is calculated using a time decay coefficient based on the remaining time between the order creation time and the preset delivery deadline. The shorter the remaining time, the faster the decay coefficient increases. When the remaining time falls below a critical threshold of 30 minutes, the decay coefficient switches to exponential growth, forcing the task priority to increase. For example, a standard shipment weighing 50 kg has a base weight of 1.0. If the shipment weighs 80 kg, exceeding the limit by 60%, the base weight increases to 1.6. If the shipment is also an urgent order with only 20 minutes remaining to process, the time decay coefficient is calculated exponentially to 2.3, resulting in a dynamic task priority weight of 1.6 × 2.3 = 3.68.

[0033] For cargo with complex attributes, such as high weight and high urgency, the system adopts a weighted upper limit constraint mechanism to prevent a single task from excessively preempting resources. This includes: when the calculated value exceeds the preset upper limit, the buffer queue is automatically enabled, and the overflow priority is converted into the device priority release right after the task is executed.

[0034] When sensor data anomalies are detected, such as missing cargo weight information, the system generates a temporary weight based on the average of similar historical orders. This weight is then added with a checksum when assigning tasks, and recalibrated after data is restored. If multiple high-priority tasks occur within a time period, the system allocates equipment resources based on a composite ranking of "weight-urgency," prioritizing equipment with the largest remaining capacity to the task with the highest weight percentage. Furthermore, the system reserves idle time for the nearest available equipment for the second-highest-urgency task to prevent task backlogs.

[0035] The comprehensive equipment performance index includes an energy efficiency factor, a path overlap factor, and a task completion factor. The energy efficiency factor is dynamically calculated based on the ratio of the equipment's unit task power consumption to its rated power. The path overlap factor is adjusted in real time based on the length of overlap between the equipment's current path and the globally planned path. The task completion factor is generated based on the weighted statistical value of the timeliness and success rate of historical task completions. It should be further explained that, in the specific implementation process, the comprehensive equipment performance index is dynamically calculated using the real-time collected energy efficiency, path overlap, and historical task completion rates. In the specific implementation, the energy efficiency factor is dynamically adjusted based on the ratio of the equipment's current task power consumption to its rated power. This includes: when the equipment is performing a high-load or long-distance task, if the actual power consumption exceeds 80% of the rated power, its energy efficiency factor is reduced, and a task migration alert is triggered. If the equipment is in a low-load state and the power consumption is less than 30% of the rated power, the factor value is increased to prioritize the allocation of lightweight tasks.

[0036] The path overlap factor is calculated by comparing the length of the overlapping segments between the device's current path and the globally planned path. If the overlap ratio exceeds 50%, the path conflict risk is determined to be high, the factor is dynamically lowered, and the path planning module is triggered to re-evaluate the backup routes. If the device paths are completely independent and non-intersecting, the factor baseline value is maintained to maintain scheduling stability.

[0037] The task completion rate factor is generated based on the device's historical task records. This includes: A positive weighting factor is assigned to devices with an on-time completion rate exceeding 95%. If there have been recent task timeouts or failures, a negative penalty factor is added based on the number of timeouts and failure rate. For example, if an AGV's path overlap rate increases to 60%, its path overlap rate factor drops from 0.9 to 0.7. The system automatically removes it from the list of high-priority task candidates and assigns it to a task in a path-independent, low-conflict area. If a stacker crane's energy efficiency factor drops below 0.5 due to low battery, new tasks are suspended, and its efficiency index gradually recovers after it completes recharging.

[0038] When a device experiences a sudden failure, the system immediately freezes its performance index and marks it "unavailable." It also migrates unfinished tasks to the optimal replacement device, sorted by the performance index of neighboring devices. If multiple devices have similar performance indices, the device with the lowest path overlap and highest task completion rate is prioritized for task completion. For complex anomalies, such as a device experiencing both high path overlap and low task completion rates, the system activates a performance index circuit breaker mechanism. This includes forcing the device into a maintenance queue when the combined index falls below a critical threshold and triggering an incremental state migration model to locally update the task allocation logic in the affected area.

[0039] The autonomous conflict resolution module's virtual track shifting technology specifically involves adjusting the device's operating speed through microsecond-level speed commands, while maintaining the global path plan. This shifts the trajectory of the conflicting devices in time and space. It should be further explained that in its implementation, virtual track shifting resolves conflicts by dynamically adjusting the device's operating speed or local path curvature, while maintaining the global path plan.

[0040] In practice, when the system predicts that a device's trajectory will intersect within the next three seconds, it adjusts its strategy based on the collision time difference and the device's current state. For example, if the time difference is greater than 0.5 seconds but less than 1 second, it prioritizes speed fine-tuning, reducing the speed of the leading device or increasing the acceleration of the trailing device to increase the time difference between the two devices passing the collision point to above a safety threshold. If the time difference is less than 0.5 seconds and the device is lightly loaded, it activates local path curvature adjustment, inserting a small arc into the straight segments of the global path to spatially offset the device's trajectory while ensuring that the total path length deviation does not exceed 2%. For example, if the predicted collision time difference between a stacker crane and an AGV in an intersecting aisle is 0.3 seconds, the system issues a microsecond-level instruction to the AGV, reducing its operating speed from 1.5m / s to 1.2m / s and simultaneously increasing the stacker crane's speed by 0.1m / s, increasing the time difference to 0.8 seconds. If a shuttle cannot completely stagger its path due to narrow shelf spacing, a 5-degree turning arc is added at the end of its straight path to bypass the adjacent equipment operating area.

[0041] For high-load equipment, such as fully loaded stackers, the system limits the speed adjustment range to no more than 10% of the rated speed, and prioritizes path curvature correction to avoid the risk of inertial loss of control. For light equipment, such as unloaded AGVs, the speed fluctuation range is allowed to be expanded to 20%, and a rapid acceleration mode is supported.

[0042] If a device fails and becomes unable to respond to speed commands due to a sudden malfunction, the system immediately freezes its path and triggers global avoidance for neighboring devices. This includes generating a sector-shaped warning zone based on the device's real-time position, forcing other devices to circumvent it until the faulty device is out of the conflict zone. If three or more consecutive virtual track deviations occur within the same channel, the area is identified as a high-risk conflict point, automatically increasing its avoidance weight in the path planning algorithm and prioritizing alternate channels for subsequent task allocation.

[0043] The execution steps of the distributed negotiation protocol are: The conflicting devices generate at least two alternative paths based on local state exchange, and select the path with the best weighted time and energy consumption through a lightweight consensus algorithm. It should be further explained that in the specific implementation process, the distributed negotiation protocol generates a conflict resolution plan through local state exchange and consensus decision-making between conflicting devices. In the specific implementation, when the system determines that there is a severe conflict between devices that cannot be misplaced, the negotiation process is triggered, including: the conflicting devices each generate at least two alternative path plans based on their own power acquired in real time, the remaining time of the current task and local environmental data, such as the length of the backup path and the congestion level of adjacent channels, and perform multi-dimensional evaluation through a lightweight consensus algorithm.

[0044] Evaluation dimensions include the total time taken for a route, estimated energy consumption, and the weight of its impact on the global task queue. Time and energy consumption are weighted differently by device type, with AGVs prioritizing energy consumption and stackers prioritizing time. Each device broadcasts its evaluation results to the conflict group, and the path with the highest overall score is selected as the final solution through majority rule. For example, when three shuttle vehicles compete for the same high-density shelf, each vehicle generates three options: going directly to the target shelf, detouring around an adjacent shelf, and waiting in a queue. The direct route takes the shortest time but has the highest risk of conflict; the detour route adds 10% to the route but can be executed immediately; and the waiting route requires a 15-second delay but has the lowest energy consumption.

[0045] Each device scores the solution based on its own remaining power and the urgency of the task. Among them, high-power devices tend to detour, and low-power devices tend to wait. After two rounds of voting, the detour solution is selected as the optimal solution. If the score difference between devices is less than the preset threshold, the priority arbitration mechanism is enabled, including: sorting according to task weight and device performance index, forcing the highest priority device to execute its preferred solution, and the remaining devices to execute the suboptimal solution. For scenarios where the path fails dynamically, such as sudden congestion in the backup channel during negotiation, the device updates the environmental data in real time and regenerates the solution. If two consecutive negotiations fail, degradation processing is triggered, that is, freezing the conflict area tasks and guiding the relevant devices to the safe buffer zone to wait for manual intervention. During the negotiation process, the system limits the duration of a single negotiation to no more than 500 milliseconds. If it times out, the lowest energy consumption solution is executed by default to prevent system congestion.

[0046] The digital twin verification platform validates scheduling strategies by injecting extreme scenario data into the twin environment, including order spikes, multiple device failures, and channel congestion. Monte Carlo simulations are then used to evaluate system throughput fluctuations and conflict resolution success rates. It should be noted that, during implementation, the digital twin verification platform uses Monte Carlo simulations to validate scheduling strategies under these extreme scenarios. In its specific implementation, the platform injects multi-dimensional extreme working condition data based on a real-time synchronized warehouse virtual model. These include: for order peak scenarios, generating intensive task flows exceeding normal loads, simulating a large influx of urgent orders in a short period of time, and randomly allocating abnormal ratios of cargo weight and volume. For example, the proportion of overweight cargo increases to 40%, testing the task allocation efficiency and equipment load balancing capabilities of the dynamic game scheduling module; for multi-device failure scenarios, randomly selecting stackers, AGVs, or shuttles to mark them as sudden downtime states, and simulating the migration process of unfinished tasks of failed equipment to verify the collaborative response efficiency of the remaining equipment in task redistribution, dynamic path adjustment, and conflict resolution; for aisle congestion scenarios, virtually closing the passages in high-traffic shelf areas, and simultaneously simulating the sudden congestion diffusion effects of adjacent aisles to evaluate the system's response speed in enabling backup paths and the stability of the global task queue.

[0047] During the simulation process, the platform uses the Monte Carlo method to randomly combine the probability of occurrence, duration and superposition pattern of extreme conditions, such as the concurrence of order peaks and equipment failures, and calculates indicators such as throughput fluctuation rate, conflict resolution failure rate and equipment utilization rate. If the throughput fluctuation rate exceeds the preset threshold, it is determined that the scheduling strategy needs to be optimized.

[0048] For example, when the simulation shows that the delay rate of high-priority task allocation exceeds 5%, the platform automatically adjusts the urgency attenuation coefficient in the dynamic game algorithm to strengthen the weight of time-sensitive tasks; if multiple device failures cause the task migration failure rate to increase, the recalculation range of the incremental state migration model is optimized to expand the coverage area of ​​the affected devices.

[0049] For scenarios where the path planning conflict rate surges due to channel closures, the platform dynamically adjusts the environmental constraint vector through a pre-trained backup path library and reversely corrects the negotiation protocol trigger threshold of the conflict resolution module. After verification, the optimized parameters are synchronized to the actual scheduling module through the physical system feedback loop. If the deviation between the actual operating data and the simulation results continues to be lower than the tolerance value, the parameters are determined to be effective; if the deviation rebounds, a secondary simulation verification is triggered and the correction coefficient is added. When structural changes occur in the warehouse layout, such as the addition of new shelves or equipment types, the platform reconstructs the virtual model and reruns historical extreme scenario data to preview the adaptability of the scheduling strategy in the new environment and generate parameter adjustment plans in advance to avoid online risks.

[0050] The physical system feedback loop implements parameter correction by collecting data on the deviation between actual task completion time and predicted time, dynamically adjusting the weight of the path overlap rate factor in the device's comprehensive effectiveness index, and optimizing the trajectory prediction time window of the spatiotemporal conflict probability field model based on records of conflict resolution failures. It should be further explained that during the specific implementation process, the physical system feedback loop dynamically corrects the scheduling algorithm parameters by collecting deviations between device execution data and the prediction model in real time. In specific implementation, the system continuously monitors the difference between the actual completion time of device tasks and the time predicted by the digital twin platform. This includes: if the deviation in the task completion time of the same device exceeds 10% of the preset tolerance value for three consecutive times, it is determined that there is a deviation in its effectiveness index calculation model, triggering a parameter correction process. Specifically, the weight of the path overlap rate factor in the device's comprehensive effectiveness index is reduced, and the proportion of historical data backtracking for the task completion rate factor is increased to weaken the impact of path planning and enhance actual execution capabilities.

[0051] For conflict resolution failure scenarios, the system counts the number of failed conflicts and the causes of each type of conflict on an hourly basis. These include path intersections and shelf grabs. This process includes shortening the trajectory prediction window of the spatiotemporal conflict probability field model, such as from 3 seconds to 2 seconds, if the failure rate of a particular conflict type exceeds a threshold, such as a 5% daily failure rate, and increasing the frequency of updating the environmental constraint vector.

[0052] When a sudden change in the warehouse environment is detected, such as the addition of new shelves or the expansion of equipment clusters, the system links the digital twin platform to reconstruct the virtual model and inject historical extreme scenario data for pre-verification. If the simulation results show that the throughput volatility under the new parameters has decreased, the physical system is updated in stages through a progressive parameter migration strategy to avoid oscillation caused by full-scale switching. For example, when the AGV's task execution time continues to be long due to changes in the ground friction coefficient, the feedback loop reduces the weight of its path overlap rate factor and simultaneously increases the task allocation priority of neighboring equipment; if the failure rate of channel congestion conflict resolution increases sharply during a certain period of time, the system automatically shortens the trajectory prediction time window from 3 seconds to 2.5 seconds and enables high-sensitivity conflict detection mode.

[0053] For complex deviation scenarios, such as when the deviation of the device performance index and the conflict failure rate increase simultaneously, the system will prioritize adjusting the conflict prediction parameters to quickly reduce the risk of task interruption, and then gradually correct the performance index model after the conflict rate stabilizes. During the correction process, the system sets an upper limit on the parameter change, such as a single adjustment not exceeding 20% ​​of the original value, and uses double closed-loop cross-validation to ensure that the adjusted parameters pass stability tests in both virtual environments and actual scenarios. If the deviation still cannot converge after three consecutive corrections, a manual intervention alarm will be triggered, freezing the current parameters and rolling back to the most recent stable version. At the same time, the abnormal device will be marked to enter the maintenance and inspection process.

[0054] The storage channel status data of the multimodal perception module includes: the channel congestion level perceived by the fusion of LiDAR and camera; and the shelf occupancy status detection results based on infrared sensors. It should be further explained that, in the specific implementation process, the multimodal perception module perceives the channel congestion level through the fusion of LiDAR and camera, and detects the shelf occupancy status in combination with infrared sensors. In the specific implementation, the LiDAR scans the obstacle distribution density and movement trend in the channel in real time. When it is detected that the obstacle concentration exceeds the critical value, such as the number of stationary obstacles in a single channel is greater than or equal to 3 or the distance between mobile devices is less than the safety threshold, the channel congestion level is marked as high; the camera analyzes the equipment queue length and moving speed through image recognition technology. If the queue length exceeds 50% of the channel capacity or the average speed drops below 60% of the rated value, the congestion level is dynamically increased. The two types of perception data are combined through a weighted fusion algorithm to generate the final congestion level. This includes: when the judgment results of the lidar and the camera are consistent, the highest level is directly adopted; if there is a difference, such as the lidar determines that the congestion is medium and the camera shows low congestion, a comprehensive value is calculated based on the historical accuracy weight; among which, the historical accuracy weight includes 70% for the lidar and 30% for the camera.

[0055] Infrared sensors are deployed on each shelf layer and detect the storage status of goods by the intensity of the reflected signal. This includes: if the reflection intensity of a shelf layer is continuously lower than the empty baseline value for more than 5 seconds, it is judged to be fully loaded and the operating rights of the shelf are frozen; if the reflection intensity fluctuates violently, such as changing more than 3 times within 10 seconds, a manual review process is triggered and the shelf status is temporarily marked as abnormal. For example, when the laser radar detects that the obstacle density in the channel exceeds the standard due to temporary stacking of goods, but the camera shows that the equipment queue is moving at a normal speed, the system determines the congestion level as medium with a weight of 7:3, and only restricts new tasks from entering the channel, but does not trigger global path replanning; if the shelf infrared sensor detects a sudden drop in reflection intensity, such as from 90% of the full load value to 30%, the shelf is immediately unfrozen and replenishment tasks are assigned priority.

[0056] In the event of sensor data conflicts, such as inconsistent infrared sensor detection results on different layers, the system activates a layered state independent processing mechanism, marking only the abnormal layer as restricted while maintaining normal access to the remaining layers. If a sudden temporary obstruction in a channel, such as fallen cargo, causes both the LiDAR and camera to identify high congestion, the system automatically generates detour instructions and leverages the conflict resolution module to dynamically redistribute tasks passing through that area. Simultaneously, a physical feedback loop is used to adjust the channel capacity parameters of the digital twin platform.

[0057] The local recalculation range of the incremental state transition model is: centered on the device whose state has changed, radiating to other devices and shelf areas where there is a risk of intersection with its movement path or resource contention. It should be further explained that in the specific implementation process, the local recalculation range of the incremental state transition model is centered on the device whose state has changed, radiating to other devices and shelf areas that may be affected by it. In the specific implementation, when a device state change or environmental constraint update is detected, the system first dynamically defines the recalculation radius based on the device's current task path and the predicted data of the motion trajectory of adjacent devices. For example, if the device's own state is abnormal, such as a decrease in battery power, the radiation range covers its current task target shelf, the passing aisles, and other devices whose paths intersect within the next 3 seconds; if the environmental constraint changes, such as a fully loaded shelf, then all task paths and associated devices planned to use the shelf are covered. Among them, device state changes include insufficient AGV battery and stacker load exceeding the limit, and environmental constraint updates include increased aisle congestion levels. For example, when a shuttle triggers an environmental constraint vector update due to a fully loaded shelf, the system only reallocates pending tasks destined for that shelf and adjusts the routing of equipment within a 3-meter radius to avoid that shelf area. If a stacker crane is unable to access a high-level shelf due to insufficient lift height, its radius is limited to all high-level shelves on its task path, and tasks are automatically downgraded to mid- and low-level shelves. For scenarios where multiple devices experience concurrent status changes, such as multiple AGVs experiencing simultaneous power alarms, the system sorts them by their efficiency index, prioritizing local recalculation in areas with low-efficiency equipment while reserving resources for high-efficiency equipment to handle unexpected tasks.

[0058] During the recalculation process, the system only updates the device capability vector and the associated environmental constraint vector of the affected device, and achieves millisecond-level synchronization through the edge nodes. When it is detected that the updated local state is inconsistent with the global task allocation, such as when a device is assigned a task that exceeds its updated capability threshold, a secondary correction mechanism is triggered, including: forced recovery of the erroneous task, and re-matching to the optimal device based on the latest device performance index. For example, when the AGV's load capacity vector is updated from 100kg to 60kg due to a sudden drop in power, the system immediately recovers its assigned 80kg task and re-matches it to a nearby stacker with sufficient load margin; if the channel is suddenly congested, resulting in an update of the environmental constraint vector, all task paths scheduled to pass through the channel will be detour-planned, and will be preferentially assigned to the backup channel with the lowest path overlap rate.

[0059] For scenarios where there is a risk of competition for device resources within the radiation range, such as the intersection of multiple device paths, the system immediately triggers the conflict prediction mechanism after a local update. If a new conflict point is predicted, the dynamic game scheduling module will intervene in advance to redistribute tasks to avoid secondary conflicts caused by incremental updates.

[0060] A high-density warehousing intelligent management method includes the following steps: S1: Use edge computing nodes to collect equipment operating parameters, cargo attributes, and storage environment data in real time to build a dynamically updated spatiotemporal state matrix; S2: Based on the task priority weight and the comprehensive equipment efficiency index, the asymmetric Nash equilibrium algorithm is used to achieve dynamic matching between tasks and equipment; S3: Predict device trajectory conflicts and trigger virtual track offset technology or distributed negotiation protocols in a hierarchical manner to resolve conflicts; S4: Optimize scheduling parameters through a dual closed-loop mechanism of the digital twin verification platform and the physical system feedback loop; S5: Modify the spatiotemporal state matrix using an incremental local update strategy based on changes in device status or environmental constraints.

[0061] Based on the application of high-density warehousing intelligent management system, the high-density warehousing intelligent management system is suitable for the following scenarios, including: automated goods-to-person picking operations in e-commerce warehousing centers; high-density three-dimensional cold storage management in cold chain logistics; real-time material distribution and warehousing scheduling in smart manufacturing factories; multi-equipment collaborative loading and unloading and route planning in cross-border logistics transit warehouses; dynamic storage and priority scheduling of temperature-sensitive goods in pharmaceutical warehousing.

[0062] Through the implementation of this technical solution, the system achieves autonomous coordination and conflict resolution among heterogeneous equipment groups in a dynamic warehousing environment, resolving the issues of delayed response, frequent conflicts, and low resource utilization associated with traditional static scheduling strategies. The deep coupling between modules and the closed-loop verification mechanism ensure the technical solution is both creative and highly feasible, providing a systematic solution for intelligent management of high-density warehousing.

[0063] Through a dynamic collaborative decision-making framework and multimodal perception mechanism, the real-time response capability and resource utilization of high-density warehousing systems are improved. Based on an incremental state migration model and a local recalculation strategy, the system can timely update equipment capabilities and environmental constraint data, avoiding the computing power consumption and response delays caused by traditional global optimization. Through a task-equipment two-way matching game algorithm and a hierarchical conflict resolution mechanism, autonomous collaborative scheduling of heterogeneous equipment groups is achieved, effectively reducing path intersection deadlocks and shelf competition conflicts, and reducing system throughput volatility. At the same time, the dual closed-loop verification module ensures the robustness of the scheduling strategy in extreme scenarios through dynamic parameter correction based on virtual simulation and physical feedback, solving the problem of traditional static scheduling relying on human intervention.

[0064] Through a distributed negotiation protocol and edge computing architecture, low-latency and high-reliability conflict resolution is achieved, shortening conflict response times and reducing the frequency of manual intervention. The combination of a digital twin verification platform and a dynamic self-learning mechanism enables the system to adapt to structural changes in the warehousing environment, shortening the policy adaptation cycle in new scenarios. Furthermore, layered exception handling and incremental update mechanisms prevent the global spread of localized failures, ensuring the continuous execution of critical tasks and providing an intelligent management solution that combines efficiency, stability, and scalability for high-density warehousing scenarios.

[0065] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0066] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A high-density warehouse intelligent management system, characterized in that: include: The dynamic collaborative decision-making framework has its input connected to the multimodal perception module, and its output connected to the dynamic game scheduling module and the autonomous conflict resolution module respectively; The multimodal perception module collects equipment operating parameters, cargo attributes, and storage channel status data in real time through edge computing nodes deployed on stackers, AGVs, and shuttles, and constructs a spatiotemporal state matrix containing equipment capability vectors, task requirement vectors, and environmental constraint vectors. The dynamic game scheduling module generates task allocation instructions through an asymmetric Nash equilibrium algorithm based on the task priority weights and the equipment comprehensive efficiency index in the spatiotemporal state matrix; The autonomous conflict resolution module analyzes the device motion trajectory prediction data in the spatiotemporal state matrix and triggers the virtual track offset technology or the distributed negotiation protocol to resolve the conflict; A dual-closed-loop verification module, whose input is connected to the dynamic collaborative decision-making framework and whose output reversely corrects the algorithm parameters of the dynamic game scheduling module through the physical system feedback loop; The dual closed-loop verification module includes a digital twin verification platform for simulating scheduling strategy verification in extreme scenarios; The spatiotemporal state matrix of the dynamic collaborative decision-making framework is updated through an incremental state migration model, including: when the equipment status or warehouse environment changes, only the equipment capability vector or environmental constraint vector in the affected area is locally recalculated, and millisecond-level updates are achieved through edge computing nodes.

2. The high-density warehouse intelligent management system according to claim 1, characterized in that: The task priority weight of the dynamic game scheduling module is calculated based on the basic priority of the cargo weight and volume, and the time decay coefficient of the order urgency is added to generate the dynamic task priority weight; The comprehensive equipment performance index includes an energy efficiency factor, a path overlap factor, and a task completion factor; the energy efficiency factor is dynamically calculated based on the ratio of the equipment's unit task power consumption to its rated power; the path overlap factor is corrected in real time based on the overlapping length between the equipment's current path and the globally planned path; and the task completion factor is generated based on the weighted statistical value of the historical task completion timeliness and success rate.

3. The high-density warehouse intelligent management system according to claim 1, characterized in that: The virtual track offset technology of the autonomous conflict resolution module specifically includes: Under the premise that the global path planning of the equipment remains unchanged, the equipment running speed is adjusted through microsecond-level speed instructions, so that the motion trajectory of the conflicting equipment is dislocated in the time and space dimensions.

4. The high-density warehouse intelligent management system according to claim 3, characterized in that: The execution steps of the distributed negotiation protocol are: The conflicting devices generate at least two alternative paths based on local state exchange, and select the path with the best weighted average time and energy consumption through a lightweight consensus algorithm.

5. The high-density warehouse intelligent management system according to claim 1, characterized in that: The digital twin verification platform verifies the scheduling strategy in the following ways: Extreme scenario data such as order peak impact, multiple device failures and channel congestion are injected into the twin environment, and the system throughput fluctuation rate and conflict resolution success rate are evaluated through Monte Carlo simulation.

6. The high-density warehouse intelligent management system according to claim 1, characterized in that: The physical system feedback loop achieves parameter correction in the following ways: Collect the deviation data between the actual task completion time and the predicted time, and dynamically adjust the weight of the path overlap rate factor in the equipment comprehensive effectiveness index; According to the conflict resolution failure records, the trajectory prediction time window of the spatiotemporal conflict probability field model is optimized.

7. The high-density warehouse intelligent management system according to claim 1, characterized in that: The storage channel status data of the multimodal perception module includes: Channel congestion level perceived through LiDAR and camera fusion; Shelf occupancy status detection results based on infrared sensors.

8. The high-density warehouse intelligent management system according to claim 1, characterized in that: The local recalculation range of the incremental state transition model is: Centered on the equipment experiencing status changes, the system radiates to other equipment and shelf areas where there is a risk of intersection with its movement path or resource competition.

9. A high-density warehousing intelligent management method, characterized in that: The following steps are involved: S1: Use edge computing nodes to collect equipment operating parameters, cargo attributes, and storage environment data in real time to build a dynamically updated spatiotemporal state matrix; S2: Based on the task priority weight and the comprehensive equipment efficiency index, the asymmetric Nash equilibrium algorithm is used to achieve dynamic matching between tasks and equipment; S3: Predict device trajectory conflicts and trigger virtual track offset technology or distributed negotiation protocols in a hierarchical manner to resolve conflicts; S4: Optimize scheduling parameters through a dual closed-loop mechanism of the digital twin verification platform and the physical system feedback loop; S5: Modify the spatiotemporal state matrix using an incremental local update strategy based on changes in device status or environmental constraints.

10. A high-density warehouse intelligent management system according to any one of claims 1-8, characterized in that: The high-density warehousing intelligent management system is applied to the following scenarios: Automated goods-to-person picking operations in e-commerce warehouses; High-density three-dimensional cold storage management in cold chain logistics; Real-time material distribution and warehouse scheduling in smart manufacturing factories; Multi-equipment collaborative loading and unloading and route planning for cross-border logistics transit warehouses; Dynamic storage and priority scheduling of temperature-sensitive goods in pharmaceutical warehousing.

Images