Automatic operation optimization method and system for automated stereoscopic warehouse based on digital twinning

Abstract

The present application relates to the technical field of intelligent warehousing, and more particularly to an automated stereoscopic warehouse operation optimization method and system based on digital twinning. The method comprises the following steps: obtaining a warehouse-out wave task, a warehouse logistics topology and real-time state data of equipment, analyzing a digital twinning relationship, and constructing a task propagation chain; combining the equipment state to recursively propagate the task node occupation interval, and performing occupation interval mapping to form a task space-time load; performing overlap analysis on the task occupation interval to identify node load convergence and congestion propagation chains; based on the congestion propagation chain, pre-elimination deduction is carried out, and scheme simulation is carried out in the digital twinning environment to obtain deduction simulation data. The present application can identify implicit load convergence and congestion propagation paths in advance, realize congestion pre-elimination, reduce task backlog and equipment waiting, improve the operation efficiency and stability of the stereoscopic warehouse, and enhance the forward-looking nature of scheduling.

Description

Technical Field

[0001] This invention relates to the field of intelligent warehousing technology, and in particular to an automated three-dimensional warehouse operation optimization method and system based on digital twins. Background Technology

[0002] Automated storage and retrieval systems (AS / RS) typically consist of racking systems, stacker cranes, shuttle cars, aisle conveyors, elevators, buffer conveyor sections, and inbound / outbound workstations. They function through a warehouse management system and a warehouse control system to handle the receiving, storage, picking, and outbound operations of goods. Compared to a conventional floor-level warehouse, an outbound task in an AS / RS requires traversing multiple spatially connected logistics nodes with sequential constraints. Different tasks may also share stacker cranes, elevators, redirecting conveyors, or buffer nodes. Therefore, the operational status of each piece of equipment and the order in which tasks are executed directly impact the overall outbound cycle time of the warehouse.

[0003] In actual operation, multiple orders are typically issued in batches according to outbound waves. When multiple tasks within the same wave target adjacent storage locations, the same aisles, shared elevators, or the same outbound interface, different tasks will create resource occupancy issues at logistics nodes that are close in time or partially overlap. Existing scheduling methods mostly sort tasks based on task priority, current equipment location, shortest running path, current queue length, or real-time node occupancy status. While these methods can handle existing equipment conflicts or localized congestion, they struggle to anticipate the continuous migration of load across multiple logistics nodes based on tasks that have not yet been executed. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes an automated three-dimensional warehouse operation optimization method and system based on digital twins, thereby solving at least one of the aforementioned technical problems.

[0005] This application provides a method for optimizing the operation of an automated storage and retrieval system based on digital twins, including the following steps: S1: Acquire outbound wave task data, warehousing and logistics topology data, and real-time equipment status data of the automated warehouse; perform digital twin relationship analysis based on the outbound wave task data and warehousing and logistics topology data to obtain task propagation chain data; S2: Based on the task propagation chain data and the real-time status data of the device, perform recursive calculations to obtain the task node occupancy interval data; based on the task node occupancy interval data, perform occupancy interval mapping to obtain the task spatiotemporal load data; S3: Perform overlap analysis on the task node occupancy interval data based on the task spatiotemporal load data to obtain node load aggregation data; perform congestion propagation analysis on the node load aggregation data to obtain congestion propagation chain data. S4: Perform preliminary mitigation simulation based on congestion propagation chain data to obtain preliminary mitigation data; perform digital twin simulation based on preliminary mitigation data to obtain simulation data.

[0006] This invention maps outbound wave tasks, warehouse logistics topology, and real-time equipment status to a unified digital twin environment, constructing a continuous propagation relationship of tasks among storage locations, stacker cranes, conveyor lines, elevators, and outbound exits. By utilizing node occupancy interval recursion and occupancy interval mapping, the originally discrete task execution states are transformed into a spatiotemporal load representation that can be jointly analyzed along the time axis and logistics nodes. Local load convergence states are identified through task occupancy interval overlap analysis, and the propagation direction, propagation delay, and cascading range of congestion between adjacent nodes are further characterized. This allows for the early identification of potential congestion chains before explicit congestion occurs in the physical warehouse. Pre-emptive mitigation simulations of congestion propagation chains are performed, and simulations verify solutions such as task release adjustments, equipment service reordering, path switching, and cache diversion within the digital twin model. This reduces downtime, idle travel, and task backlog caused by relying on on-site trial and error, improving the foresight, coordination, and stability of warehouse operation scheduling.

[0007] Preferably, the digital twin relationship parsing in S1 is specifically as follows: A digital twin graph is constructed based on the warehouse logistics topology data to obtain logistics topology graph data; task tuples are discretized based on the outbound wave task data to obtain task tuple data; the task tuple data and logistics topology graph data are aligned with digital twin entities to obtain twin state synchronization data; and a task propagation chain is constructed based on the twin state synchronization data to obtain task propagation chain data.

[0008] This invention transforms the warehousing and logistics topology into a computable digital twin graph, allowing physical objects such as storage locations, stacker cranes, conveyor lines, elevators, and outbound exits to be represented by unified nodes and connecting edges. Outbound wave tasks are discretized into task tuples containing task identifiers, target storage locations, target outbound exits, load attributes, and priorities, facilitating standardized processing of task information from different sources and with different structures. By aligning task tuples with virtual entities in the twin graph, a unique mapping relationship can be established between actual tasks, logistics nodes, and service equipment, reducing task mismatches caused by inconsistent node naming, unclear equipment affiliation, or missing path relationships. Constructing a task propagation chain fully preserves the order of task flow from the storage location to the outbound end and the resource call relationships.

[0009] Preferably, the recursive calculation in S2 is as follows: The physical flow delay of nodes is calculated based on the task propagation chain data and the real-time status data of the equipment to obtain the action delay data; the timing sequence is recursively deduced based on the action delay data to obtain the timing sequence data; and the occupied interval is explicitly latched based on the timing sequence data to obtain the task node occupied interval data.

[0010] This invention combines the node sequence in the task propagation chain with the real-time location, operating speed, current occupancy status, remaining service time, and inter-node transfer distance to calculate the actual physical flow delay of tasks between logistics nodes. This avoids time prediction deviations caused by relying solely on rated cycle times or fixed empirical values. Through time-series concatenation and recursion, the completion time of the previous node, the inter-node transfer delay, and the available time of the next node are passed step by step, creating a continuous and mutually constrained time sequence relationship between tasks on stacker cranes, conveyor lines, elevators, and outbound nodes. Explicitly latching the occupancy intervals of the time-series concatenation results clarifies the expected entry time, duration of continuous occupancy, and release time of each task for equipment, channels, and buffer resources. This provides accurate time boundaries for subsequent occupancy interval projection, load overlap identification, and congestion propagation prediction, improving the accuracy of warehouse operation status prediction and the reliability of scheduling decisions.

[0011] Preferably, the occupied interval display in S2 is specifically as follows: Based on the task node occupancy interval data, occupancy boundary events are extracted to obtain occupancy boundary event data; based on the occupancy boundary event data, non-equidistant event carrying segments are performed to obtain event carrying interval data; based on the event carrying interval data, task occupancy load maps are generated to obtain task load map data; based on the task load map data, logistics topology coordinate mapping is performed to obtain load spatiotemporal coordinate data; based on the load spatiotemporal coordinate data, cross-node load map continuation is performed to obtain continuous task mapping trajectory data; based on the continuous task mapping trajectory data, multi-dimensional sparse load encapsulation is performed to obtain task spatiotemporal load data.

[0012] This invention extracts events from the entry, state transition, and departure boundaries of task node occupancy intervals and segments them non-equidistantly based on the actual state change times. This avoids the distortion caused by fixed time slices merging short-term occupancy, transient waiting, and handover processes, making the time representation more closely match the actual operating rhythm of stacker cranes, conveyor lines, and hoists. Each event-carrying interval is converted into a task load map and projected onto the corresponding logistics node and topology connection edge, simultaneously preserving the task occupancy intensity, duration, spatial location, and propagation direction. Through cross-node load map continuation, the complete load migration process of a task leaving the previous node, being transferred via the connection edge, and entering the next node can be continuously described, avoiding the fragmentation of load information between nodes. Finally, multi-dimensional sparse load encapsulation is used, recording only the effective spatiotemporal units where occupancy changes occur, reducing redundant data storage and computational load.

[0013] Preferably, the overlap analysis in S3 specifically includes: Downstream stagnation constraints are identified based on task spatiotemporal load data and task node occupancy interval data to obtain downstream stagnation constraint data. Upstream occupancy boundaries are then reverse-expanded based on the downstream stagnation constraint data to obtain upstream reverse-expansion interval data. Occupancy interval incremental differences are then performed based on the upstream reverse-expansion interval data to obtain reverse-expansion incremental data. Secondary overlaps between different tasks on the same node are identified based on the reverse-expansion incremental data to obtain secondary occupancy overlap data. Upstream lock-in backtracking iterations are then performed based on the secondary occupancy overlap data to obtain stable backtracking occupancy data. Finally, node load aggregation is determined based on the stable backtracking occupancy data to obtain node load aggregation data.

[0014] This invention overcomes the limitations of static overlap judgment based solely on the original occupied interval by applying the expected congestion status of downstream nodes to the upstream task occupancy boundary. It accurately characterizes the continuous locking of upstream resources caused by insufficient downstream buffering, unreleased equipment, or handover delays. By incrementally differentiating the reverse expansion interval, only the newly occupied portion is extracted, and secondary overlaps between it and other tasks on the same node are identified, reducing the interference of existing normal occupancy on convergence judgment. Utilizing upstream lock-up backtracking iteration, the newly added congestion is propagated step-by-step along the task propagation chain to the preceding nodes until the occupancy boundary stabilizes, thereby identifying multi-level, backtracking load convergence induced by local congestion. The resulting node load convergence data more realistically reflects the formation range, duration, and associated tasks of hidden congestion in the warehouse.

[0015] Preferably, the upstream locking backtracking iteration specifically involves: Based on the overlapping occupancy data, the overlapping stagnation increment is extracted to obtain the overlapping stagnation increment data; based on the overlapping stagnation increment data, the preceding lock occupancy boundary is back-calculated to obtain the preceding lock occupancy boundary data; based on the preceding lock occupancy boundary data, the backtracking occupancy increment is extracted to obtain the backtracking newly added occupancy data; based on the backtracking newly added occupancy data, the preceding heterogeneous task overlap is re-determined to obtain the backtracking overlapping update data; based on the backtracking overlapping update data, multi-level lock occupancy convergence iteration is performed to obtain stable backtracking occupancy data.

[0016] This invention extracts the new waiting amount caused by mutual task obstruction from the secondary occupancy overlap results and propagates this obstruction increment back to the preceding logistics node. This dynamically corrects the original lock-up end boundary of the preceding node, preventing the downstream delay impact from being limited to a single node. By backtracking the occupancy increment extraction, only the newly added resource occupancy interval after the boundary delay is retained, and the overlap between the newly added interval and the occupancy status of other tasks in the preceding node is re-determined. This continuously detects multi-level chain lock-up induced by a single local blockage. Through multi-level lock-up convergence iteration, the task waiting time, node release time, and newly added occupancy range are updated layer by layer until adjacent iterations no longer generate new occupancy increments. This ensures that the obtained stable backtracking occupancy data can truly reflect the diffusion boundary of the obstruction impact on the task propagation chain.

[0017] Preferably, the congestion propagation analysis in S3 is as follows: Congestion source cores are identified based on node load aggregation data to obtain congestion source core data. Unabsorbed loads are stripped from the congestion source core data to obtain propagation residual load data. Adjacent load injection data is obtained by targeted injection into adjacent logistics nodes based on the propagation residual load data. Node carrying capacity erosion analysis is performed based on the adjacent load injection data to obtain propagation trigger node data. Cross-node time delay cascading recursion is performed based on the propagation trigger node data to obtain cascaded congestion trajectory data. Propagation node time sequence is concatenated based on the cascaded congestion trajectory data to obtain congestion propagation chain data.

[0018] This invention further identifies congestion source cores with persistent overload characteristics and spillover trends from the node load aggregation results, avoiding misjudging instantaneous concurrency or short-term task accumulation as actual congestion. By stripping unabsorbed load, the portion that a node can complete its service within the current time window or that can be absorbed by the remaining buffer is deducted, retaining only the excess load that needs to be passed to adjacent nodes, thereby improving the targeting of propagation analysis. The propagated remaining load is injected into adjacent nodes according to the logistics topology connection relationship and task flow direction, and the degree of erosion of node service capacity, buffer capacity, and available time margin by the additional load is analyzed, which can identify risk nodes that are not yet congested but are about to be triggered in advance. Through cross-node delay cascading recursion and propagation node time sequence concatenation, the path, time, and intensity of congestion spreading step by step from the source core to the elevator, conveyor line, and outlet can be represented, providing a continuous and quantifiable basis for selecting the pre-cutoff location and generating the resolution scheme.

[0019] Preferably, the pre-elimination deduction in S4 is specifically as follows: Based on the congestion propagation chain data, the propagation reverse intervention window is traced back to obtain the pre-intervention window data; based on the pre-intervention window data, the propagation truncation object is filtered to obtain the propagation truncation object data; based on the propagation truncation object data, the resolution action coupling arrangement is performed to obtain the pre-resolution candidate data, where the resolution action includes at least one of the following: task release time adjustment, device service order reordering, task propagation path switching, and temporary cache node allocation; based on the pre-resolution candidate data, the congestion propagation residual is iterated to obtain the pre-resolution data.

[0020] This invention, by tracing back the latest adjustable time of each node along the congestion propagation chain, can identify intervention windows with effective handling capacity before congestion actually forms, avoiding intervention that is too late and loses the cutoff condition. Based on the pre-intervention window, propagation cutoff nodes and related tasks are selected, allowing resources to be concentrated on key locations that significantly reduce downstream cascading loads. By coupling and orchestrating actions such as adjusting task release times, rearranging equipment service sequences, switching task propagation paths, and allocating temporary cache nodes, multiple candidate solutions can be formed that consider task time limits, equipment service ranges, path reachability, and cache capacity constraints. By iterating through the remaining congestion load after implementing each candidate solution, the number of propagation nodes, remaining excess load, congestion duration, and task delay costs can be quantitatively compared. This provides more targeted and executable pre-emptive solutions for subsequent digital twin simulations, improving the foresight and overall synergy of warehouse congestion control.

[0021] Preferably, the digital twin simulation in S4 specifically includes: Based on the pre-resolved data, a twin initial state branch mapping is performed to obtain twin branch data for the resolved scheme; based on the twin branch data for the resolved scheme, task event-driven evolution is performed to obtain twin load evolution data; based on the twin load evolution data, congestion propagation and decay quantification is performed to obtain simulation data.

[0022] This invention uses branch mapping based on the same initial state of a digital twin to represent different pre-congestion mitigation schemes. This ensures that all candidate schemes are compared and simulated under the same conditions of equipment location, task queue, node occupancy status, and buffer availability, avoiding distortion of evaluation results due to differences in initial operating conditions. By using a task event-driven approach to sequentially simulate task release, equipment pickup, node transfer, buffer waiting, and task departure, the evolution of the load on each logistics node over time after the mitigation action is implemented can be dynamically reproduced. Quantifying the time of congestion source core decay, the reduction in the number of propagation nodes, the remaining undigested load, task delays, and additional equipment travel allows for accurate evaluation of the degree to which each scheme interrupts the congestion propagation chain and its operational cost. This enables pre-verification of schemes without interfering with the normal operation of the physical warehouse, providing objective, comparable, and executable simulation data for the selection of the target mitigation scheme and the subsequent issuance of scheduling instructions.

[0023] Preferably, this application also provides a digital twin-based automated warehouse operation optimization system for executing the digital twin-based automated warehouse operation optimization method described above. The digital twin-based automated warehouse operation optimization system includes: The digital twin relationship parsing module is used to acquire outbound wave task data, warehousing and logistics topology data, and real-time equipment status data of automated warehouses; and to perform digital twin relationship parsing based on the outbound wave task data and warehousing and logistics topology data to obtain task propagation chain data. The task spatiotemporal load module is used to perform recursive calculations based on task propagation chain data and real-time device status data to obtain task node occupancy interval data; and to perform occupancy interval mapping based on task node occupancy interval data to obtain task spatiotemporal load data. The congestion propagation analysis module is used to perform overlap analysis on the occupancy interval data of task nodes based on the spatiotemporal load data of the task to obtain the node load aggregation data; and to perform congestion propagation analysis based on the node load aggregation data to obtain the congestion propagation chain data. The pre-congestion mitigation simulation module is used to perform pre-congestion mitigation simulation based on congestion propagation chain data to obtain pre-congestion mitigation data; and to perform digital twin simulation based on the pre-congestion mitigation data to obtain simulation data.

[0024] This invention unifies and correlates outbound wave tasks, warehouse logistics topology, and real-time equipment status. Through digital twin relationship analysis, it forms a task propagation chain that reflects the step-by-step flow of goods from storage locations, stacker cranes, aisle conveyor lines, elevators, and finally to the outbound gate. This establishes a task-equipment-node mapping foundation consistent with the physical warehouse for subsequent operational status simulation. Based on this, by combining the current equipment location, operating speed, remaining service time, and node availability, the arrival, waiting, service, and departure times of tasks at each logistics node are recursively calculated, forming node occupancy intervals with clear time boundaries. By mapping these occupancy intervals, discrete tasks are transformed into spatiotemporal loads continuously distributed along the time dimension and logistics topology. Further, through interval overlap analysis, it not only identifies direct task convergence at the same node but also characterizes upstream occupancy extensions caused by downstream congestion and secondary overlaps of multiple tasks. Based on this, it extracts the remaining load not absorbed by node service capacity and buffer capacity, constructing a propagation chain where congestion cascades and spreads over time between adjacent logistics nodes. Finally, by reverse-engineering the congestion propagation chain, the intervention window is determined, and the task release time, equipment service sequence, propagation path, and temporary buffer nodes are combined and simulated. Branch simulations are carried out under the same twin initial state. Under the condition of not interfering with the operation of the physical warehouse, the congestion reduction speed, remaining excess load, task delay, and additional equipment travel of each scheme can be quantitatively compared, thereby improving the foresight, coordination, and operational stability of the automated warehouse scheduling. Attached Figure Description

[0025] Other features, objects, and advantages of this application will become more apparent from the following detailed description of the non-limiting embodiments, taken with reference to the accompanying drawings: Figure 1 A flowchart illustrating the steps of an automated storage and retrieval system operation optimization method based on digital twins is shown in one embodiment. Figure 2 A flowchart illustrating the steps of a digital twin relationship resolution method according to one embodiment is shown. Figure 3 A flowchart illustrating the steps of a recursive calculation method according to one embodiment is shown. Figure 4 A flowchart illustrating the steps of an overlap analysis method according to one embodiment is shown. Detailed Implementation

[0026] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0027] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. Functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0028] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0029] The system first obtains the starting location, target outbound port, planned release time, required completion time, and task priority for each task, and then constructs a logistics topology map by combining equipment nodes, logistics connection edges, rated passage time, and connection activation status. After digital twin entity alignment, 36 task-level propagation chains are formed, of which 21 task propagation chains pass through hoist No. 2 together, and 17 task propagation chains pass through the buffer node before the outbound port together. This leads to the formation of corresponding wave propagation relationships.

[0030] After calculating the physical flow delay of nodes and performing time-series recursion based on real-time equipment status data, a total of 284 task node occupancy intervals were obtained. The occupancy interval display results show that from 620 seconds to 684 seconds, hoist node No. 2 simultaneously had 3 task occupancy requirements, while its maximum simultaneous capacity was 1; from 658 seconds to 731 seconds, the cache node... The number of occupied tasks reached 4, while its capacity was 2. Through upstream reverse expansion and lockout backtracking iteration, the congestion of the hoist was identified and directed to its preceding transport node. , and lane entrance cache nodes The congestion propagates, forming a congestion propagation chain with four propagation nodes. In the baseline simulation, the cumulative overcapacity duration of this propagation chain is 126 seconds, the maximum cascading delay is 94 seconds, causing nine tasks to wait, of which three tasks exceed their required completion time, with the maximum overdue time being 71 seconds.

[0031] The system traces back along the congestion propagation chain to identify three pre-intervention windows and selects the appropriate tasks. , and As the target of propagation interception. The pre-resolution scheme is as follows: the task The release time will be delayed by 35 seconds; the task will be... The service order in the waiting queue of elevator No. 2 has been adjusted from 2nd to 3rd; the task... From the original propagation path "transport node" —Elevator No. 2—Cache Node Switch to "Conveyor Node" —Elevator No. 1—Cache Node "; at the same time, the task Temporarily allocated to a cache node with 1 slot remaining. The cache stay time is 42 seconds.

[0032] After mapping the aforementioned pre-processing data to the digital twin branch, event-driven simulation was performed. Simulation results show that the overcapacity duration of hoist node 2 decreased from 64 seconds to 0 seconds, and the cache node... The overcapacity duration was reduced from 62 seconds to 8 seconds, and the cumulative overcapacity duration of the entire congestion propagation chain was reduced from 126 seconds to 8 seconds, a reduction of 118 seconds, or 93.65%; the number of propagation nodes was reduced from 4 to 1, a reduction of 75%; the maximum cascading latency was reduced from 94 seconds to 17 seconds, a reduction of 77 seconds, or 81.91%; the number of affected tasks was reduced from 9 to 2, and the cumulative waiting time for tasks was reduced from 428 seconds to 96 seconds, a reduction of 332 seconds, or 77.57%.

[0033] Meanwhile, all three overdue tasks in the original baseline plan were completed as required after the adjustment. The overall completion time of the wave was shortened from 31 minutes and 46 seconds after the original plan was extended to 30 minutes and 18 seconds, a reduction of 88 seconds. The idling time of the No. 2 hoist waiting for upstream release was reduced from 73 seconds to 21 seconds, and the peak number of cache nodes was reduced from 4 to 2. No cache overflow or new path congestion occurred. It can be seen that the present invention can identify upstream locking and cross-node congestion propagation caused by insufficient downstream capacity before the outbound tasks officially form a serious backlog. It can also resolve these issues in advance by adjusting the release time, reordering the service sequence, switching the propagation path, and allocating temporary caches. This reduces the duration of node overcapacity, cascading latency, and the risk of task overdue, and improves the continuity and stability of outbound wave execution in automated warehouses.

[0034] Please see Figures 1 to 4 This application provides a method for optimizing the operation of an automated three-dimensional warehouse based on digital twins, including the following steps: S1: Acquire outbound wave task data, warehousing and logistics topology data, and real-time equipment status data of the automated warehouse; perform digital twin relationship analysis based on the outbound wave task data and warehousing and logistics topology data to obtain task propagation chain data; Specifically, the outbound wave task data of the automated storage and retrieval system includes wave number. Task Number Target cargo number The location of the goods in the warehouse Target outbound port Task Priority Planned release time and required completion time The warehouse logistics topology data includes the set of equipment nodes corresponding to stacker cranes, conveyors, elevators, shuttles, level change points, buffer positions, and outbound exits. And the set E of logistics connection edges that allow cargo transfer between each device node. Each logistics connection edge Record at least the starting node Arrival at the node Conveying direction, permitted cargo types, and rated passage time. and connection enabled status ,in =1 indicates that the current connection can participate in task propagation. =0 indicates that the connection is closed or unreachable.

[0035] The system is based on the location of the goods in the warehouse. Query the first execution device node bound to this storage location to obtain the task start node. Based on the target outbound port Query the corresponding end node to obtain the task termination node. .by Starting from the edge, subsequent nodes are retrieved layer by layer according to the connection activation status, transport direction, and cargo type adaptation rules. satisfy =1. Conveying direction from point to And the target goods The size, weight, and vehicle type are all at the node When within the allowable range, It is considered a valid successor node for the current task; otherwise, the connection edge is discarded.

[0036] For tasks with multiple valid paths, the system filters them sequentially: first, it retains the complete path that reaches the target exit; second, it retains the path with the fewest device nodes; if the number of nodes is the same, it retains the path with the shortest total rated travel time. Any path Represented as: Its rated propagation time is: The system numbers each node in the determined path according to the order of cargo flow, and generates a node sequence number, predecessor node, successor node, entry method, exit method, and associated equipment number for each task node. This is based on the wave numbering. and the moment of mission release This involves aggregating the task paths within the same wave to form task-level and wave-level propagation chains. A task-level propagation chain refers to a chain based on a single outbound task. Taking the task as the object, according to the actual logistics flow sequence of the goods corresponding to the task from the starting storage location to the target exit, a directed link is formed by sequentially connecting the equipment nodes, buffer nodes, logistics connection edges, and the rated passage time between nodes along the way. Buffer nodes are intermediate logistics nodes in the automated warehouse logistics path used for temporary storage of goods or vehicles and connecting the operation rhythm of upstream and downstream equipment, such as temporary storage locations on conveyor lines, buffer locations at aisle entrances, waiting locations in front of elevators, or temporary storage locations in front of exits. Wave-level propagation chains refer to those based on the same outbound wave. The entire task propagation chain is treated as an object, and is a collection of multi-task related links formed by summarizing the propagation chains of each task level according to the planned release time, task priority, and shared node relationships. The obtained task propagation chain data includes the task number. Starting node Termination node Ordered node sequence The set of edges connecting nodes, and the rated passage time for each edge. Task Priority Planned release time and required completion time .

[0037] S2: Based on the task propagation chain data and the real-time status data of the device, perform recursive calculations to obtain the task node occupancy interval data; based on the task node occupancy interval data, perform occupancy interval mapping to obtain the task spatiotemporal load data; Specifically, the task propagation chain data includes tasks ordered node sequence ( Let i be the starting logistics node for the i-th task. For the second logistics node of the i-th task, (The terminating logistics node of the i-th task) and the task release time. Task Priority and the rated passage time between adjacent nodes Real-time device status data includes nodes. Operating status Current occupied tasks and estimated remaining operation time Node waiting queue Real-time running speed and estimated time for fault recovery The system uses the task start node as the recursive starting point. When the start node is idle, the task entry time is taken as... When the starting node is occupied, take... ,in For the current calculation time, This represents the estimated remaining job duration for the tasks currently being executed by the starting node, i.e., the remaining time required from the current computation time until the starting node releases available capacity; when a node fails, it takes... , This refers to the estimated fault recovery time of the starting node, i.e., the expected time point at which the equipment recovers from a faulty state to a state where it can receive and perform tasks. The node processing time is adjusted based on the rated passage time and real-time speed. ,in This is the rated speed recorded in the topology data. The node departure time is... For subsequent nodes Its entry time is recursively calculated as follows: ,in, The earliest available time for a node; when a node is idle, the earliest available time is taken. The node retrieves data while it is executing a task. When a node has queued tasks, it will serve tasks according to the rules of first-come, first-served and prioritizing higher-priority tasks at the same release time. The remaining time of each previous task is summed up sequentially. ; Take when node fails This yields the occupied range for each task node: The system encapsulates task number, node number, entry time, exit time, and interval duration to form task node interval data. The system then organizes the warehouse operation timeline according to a preset time granularity. Divided into time slots , This is the discrete time slot number. If the occupied interval... With time slot If a non-empty intersection exists, then the task... Mapped to node The corresponding time slot; its screening occupancy is expressed as: ,in, This represents the actual overlap duration. The system records this in a three-dimensional structure: task-node-timeslot. This yields the task's spatiotemporal load data.

[0038] S3: Perform overlap analysis on the task node occupancy interval data based on the task spatiotemporal load data to obtain node load aggregation data; perform congestion propagation analysis on the node load aggregation data to obtain congestion propagation chain data. Specifically, the task spatiotemporal load data includes tasks At the node Time slot occupancy The data on the occupied interval of the task node includes the occupied interval. Node number, predecessor node, successor node, and task priority The system is based on nodes. and time slot As an index, summarize the set of tasks that are simultaneously mapped to this node and this time slot: And calculate the node time slot load: ,when When, it indicates that the node is not over-occupied within that time slot; when When this time slot is used, it is designated as the load convergence time slot. For logistics nodes The maximum simultaneous carrying capacity. For consecutive load convergence time slots, they are merged into convergence intervals according to the principle of time adjacency. , For nodes The start time of the q-th load aggregation interval, For nodes The end time of the q-th load aggregation interval, the length of which is the excess duration. The system will... Intersecting task occupancy intervals They are incorporated into the same aggregation unit, and the aggregation task set, number of tasks, earliest entry time, latest exit time, and excess duration are recorded to form node load aggregation data.

[0039] The system performs congestion propagation analysis on each convergence unit based on the predecessor-successor relationship in the task propagation chain. For nodes... convergence range If the task The actual time when a node can leave is later than the originally planned time when its successor node enters, i.e., the following condition is met: The delay is then defined as: , For the task At the end of the occupation at the k-th node, Let this be the planned entry time of the task at node k+1, and let the task... In the successor node The entire occupied area has been moved to: If the shifted interval overlaps with an already occupied interval of the successor node, then... Once a congestion propagation node is identified, it continues to propagate to its subsequent nodes according to the same rules; if no overlap occurs, or if a delayed task reaches the outbound termination node, the propagation of that branch stops. For multiple tasks within the same aggregation unit, they are propagated separately according to the original task propagation chain order. The system will specify the starting aggregation node, aggregation interval, affected tasks (and aggregation interval). The congestion propagation chain data is obtained by sequentially encapsulating the occupied intervals of tasks with overlapping elements, the propagation nodes at each level (obtained by querying the successor nodes in the task propagation chain), the occupied intervals before and after propagation, the node delay, and the propagation termination node.

[0040] S4: Perform preliminary mitigation simulation based on congestion propagation chain data to obtain preliminary mitigation data; perform digital twin simulation based on preliminary mitigation data to obtain simulation data.

[0041] Specifically, the congestion propagation chain data includes the initial convergence node. Convergence Zone Affected task set The propagation node sequence and node occupied range for each task. Node latency And the propagation termination node. The system locates the node in the propagation chain where the intervals first overlap, and traces forward along the task propagation direction, identifying the task nodes before that node with adjustable conditions as candidate resolution nodes. Adjustable conditions include: the task has not yet been released, the task is located in a cache node, the task's corresponding path has a reachable backup branch, or there is a spare time between the task's release time and the required completion time. For each congested propagation chain, the system sequentially performs three types of deductions: release delay, cache pre-positioning, and path switching. For tasks that have not yet been released... Adjust its release time to: ,in, From a time slot It begins to increase sequentially by time slot, and satisfies... , The rated propagation time for the task , For the task The required completion time. For tasks that have entered the propagation chain and whose predecessor nodes have caching capabilities, their departure time from the cache node is postponed until the end of the downstream conflict interval, i.e.: , For the task The end time of occupation or departure time at the j-th logistics node For nodes The end time of the qth load aggregation interval is calculated, and the entry and exit times of subsequent nodes are recursively calculated. For tasks with backup paths, the alternative path is selected from those paths that do not pass through the initial aggregation node and can reach the target outlet, with the fewest number of nodes and the shortest rated propagation time.

[0042] For each resolution scheme generated, the system recalculates the node occupancy interval for the relevant tasks. If, after adjustment, all nodes satisfy the condition of being within the same time slot... And the task completion time is no later than If the load aggregation still exists, the solution is marked as valid; if so, the release delay time slot is increased, the cache dwell position is adjusted, or the alternative path is changed. The system determines the final solution in the order of the fewest modified tasks, the shortest total release delay time, and the fewest path switching times, and obtains the pre-resolution data, which includes the adjusted tasks, adjustment methods, original and adjusted release times, cache dwell intervals, alternative paths, and updated node occupancy intervals.

[0043] The pre-processing data is written into the digital twin model of the automated warehouse, replacing the release time, node sequence, and occupied interval of the corresponding task. The simulation clock is advanced according to the rated speed of the equipment, real-time status, and node connection relationships. During the simulation, the load of each node is recorded time-slot by time slot. The simulation data includes task waiting time, device idle time, task completion time, and whether congestion propagation occurs again. The load convergence range, number of propagation nodes, maximum latency, and task completion status before and after congestion resolution are compared and encapsulated to obtain the simulation data.

[0044] Preferably, the digital twin relationship parsing in S1 is specifically as follows: S11: Construct a digital twin based on the warehouse logistics topology data to obtain logistics topology map data; Specifically, the stacker cranes, shuttle cars, conveyors, elevators, layer-changing entrances, buffer positions, storage locations, and outbound entrances in the warehouse logistics topology data are read, and each entity is mapped to a digital twin node. And establish a set of node attributes for each node: ,in, Number the device or location. For node type, For spatial coordinates or the location of the tunnel-layer-column, To accommodate the number of tasks simultaneously, For the rated operating speed, Enable the node status. Establish directed connection edges based on the physical connectivity between devices and the allowed cargo flow direction. Only when goods can be transported from the node Direct transfer to node Establish connecting edges and record each edge: ,in, For the direction of transport, For the rated passage duration, For the set of permitted cargo specifications, Enable the connection. The logistics topology graph consists of a set of nodes V, a set of connecting edges E, a set of node attributes, and a set of edge attributes. Each edge attribute is read from the equipment interface configuration, conveying direction configuration, equipment rated parameters, cargo passage specifications and connection start / stop status in the warehouse logistics topology data. Abnormal records with no equipment number, no connection direction or non-existent nodes at both ends of the connection are removed to obtain the logistics topology map data.

[0045] S12: Discretize the task tuples based on the outbound wave task data to obtain task tuple data; Specifically, according to wave numbering Outbound tasks are grouped, with a single item to be shipped or a single vehicle handling instruction serving as the smallest task unit. For cases where a single outbound instruction contains items in multiple storage locations, it is further divided into multiple tasks based on the location of the items. And retain the common wave number. Extract the task number for each task. Goods number Goods specifications Starting location Target outbound port Planned release time Required completion time and priority Construct the task tuple: Among them, cargo specifications At a minimum, it includes cargo length, width, height, weight, and vehicle type. If tasks are not explicitly prioritized, they are sorted by required completion time, with tasks requiring earlier completion times having higher priority; if required completion times are the same, they are arranged in the order they were generated by task number. The task tuple dataset consists of all task tuples within the same wave. .

[0046] S13: Align the task tuple data and logistics topology data with digital twin entities to obtain twin state synchronization data; Specifically, for each task tuple According to the starting storage location In the logistics topology map, query the storage location nodes with the same location attributes; if the storage location node does not have the ability to perform handling, then query the first stacker crane, shuttle car, or conveyor node along its unique outbound direction and determine that node as the task starting node. According to the target outbound port Query the outbound node with the same number to determine the task termination node. Regarding the specifications of the goods Match each node and connecting edge to its allowed specifications item by item. For connecting edges... Only if the following conditions are met: , , When this happens, mark the connecting edge as a task. The edges are marked as available; otherwise, they are marked as unavailable. This indicates that the dimensions, weight, and vehicle type of the goods are within the allowable range of the connecting edge and the reaching node. The system will assign a task number. Goods number Task start node Task termination node Available node set Available connection edge set Release time Completion Time and priority Binding is performed to create a synchronized twin state record: If the start node and end node cannot be obtained, or there is no available connection edge between the start node and the end node, the corresponding task will be marked as a non-propagable task and an exception will be output, and it will not enter the subsequent propagation chain construction.

[0047] S14: Construct the task propagation chain based on the twin state synchronization data to obtain the task propagation chain data.

[0048] Specifically, starting from the task node Starting from the search point, within the set of available connecting edges... The process continues by querying successor nodes layer by layer along the directed edges until the task termination node is reached. During the search process, the same path is not allowed to repeatedly pass through the same node to avoid loop propagation caused by transport loops. Each reachable path obtained is recorded as follows: When multiple reachable paths exist, first compare the number of nodes contained in each path and retain the path with the fewest nodes; if the number of nodes is the same, calculate the rated propagation time for each path. The path with the shortest rated propagation time is retained; if the rated propagation times are still the same, a unique path is determined according to the device number order of the first branch node in the path. Nodes in the final path are numbered according to the cargo flow order, and a predecessor node, successor node, corresponding connecting edge, and rated passage time are generated for each node. A task propagation chain is then constructed. ,in, This refers to the wave number or wave identifier of the b-th outbound wave. For the final ordered sequence of nodes, This represents the rated propagation duration for the corresponding path. Within the same wave, each task propagation chain is sorted according to its planned release time; if the release times are the same, they are sorted from highest to lowest priority, thus obtaining the task propagation chain data.

[0049] Preferably, the recursive calculation in S2 is as follows: S21: Calculate the physical flow delay of nodes based on task propagation chain data and real-time device status data to obtain action delay data; Specifically, reading the tasks in the task propagation chain data An ordered sequence of nodes: and the rated passage time between adjacent nodes The real-time status data of the device includes at least the node. Current device status Real-time running speed Rated operating speed Current remaining action duration Waiting task queue Expected time for fault recovery and node capacity The system determines the corresponding physical actions based on the node type. For stacker crane nodes, the actions of picking, horizontal movement, vertical lifting, and unloading are sequentially connected; for conveyor nodes, the time required for goods to pass through the conveyor section is used as the flow action; for elevator nodes, the actions of entering, lifting, and exiting the elevator are sequentially connected; for buffer nodes, the time required for goods to enter and leave the buffer position is used as the action delay. The basic action delay of the node is expressed as follows: , among which, when When the value is 0 and the device is in a fault state, division is not performed directly; instead, the start time of the node's action is restricted to [time missing]. Then; when real-time speed is missing, retrieve If the node is currently occupied by another task, then the remaining action time will be... This is recorded as the waiting delay. If a node has a waiting queue, the priority is determined by the following rules: tasks that entered the queue first take precedence; if entry times are the same, tasks with higher priority take precedence; if priorities are still the same, tasks with smaller task numbers take precedence. Previous task set And calculate the queuing delay: , For the task At the node The basic operation latency, i.e., excluding queuing and fault recovery time, is the time required for the corresponding picking, conveying, lifting, buffer entry / exit, or unloading operation at the task completion node; for buffer nodes with a capacity greater than 1, latency is only considered when the current occupied quantity reaches [a certain threshold]. Calculate queuing delay in real time; retrieve queued items if capacity limit is not reached. =0. Therefore, the task... At the node The corresponding basic action delay, queuing delay, fault recovery constraints, and node numbers are encapsulated to form an action delay record: The action delay records of all nodes constitute the action delay data.

[0050] S22: Perform timing concatenation recursion based on the action delay data to obtain timing concatenation data; Specifically, in terms of tasks The planned release time As the starting point for recursion from the first node. For the first node... If the equipment is operating normally, its operation will begin at: If the equipment is in a faulty state, then take: The action of the first node ends at: For subsequent nodes First, the end time of the previous node's action. This represents the earliest time the goods arrive at the current node. If the current node is functioning normally, then: If the current node fails, then: The current node's action ends at: When there is a handover interlock relationship between adjacent nodes (i.e., a task transfer permission constraint is set between adjacent logistics nodes: the task can only be transferred from the predecessor node to the successor node when the predecessor node has completed the material discharge preparation and the successor node is in an accepting state), the predecessor node's occupation can only end after the predecessor node has completed the material discharge and the successor node is in an accepting state. In this case, the actual end time of the predecessor node is corrected as follows: and with Replace the original end time. The system executes the above recursion sequentially from front to back according to the node sequence to obtain time-series concatenated data containing task number, node number, arrival time, action start time, action end time, and waiting time.

[0051] S23: Explicitly latch the occupied interval based on the time-series concatenated data to obtain the occupied interval data of the task node.

[0052] Specifically, the system converts the start and end times of actions in the time-series concatenated data into explicit occupancy intervals according to the actual resource occupancy rules of the nodes. For stacker cranes, hoists, and shuttle car nodes, the entire period from the start of task execution to the completion of unloading is considered as being occupied by the task, and its occupancy interval is as follows: For conveyor nodes, the period from when goods enter the corresponding conveyor segment to when they completely leave the segment is considered occupied. For buffer nodes, the period from when goods enter the buffer position to when they leave the buffer position is considered occupied. If goods need to wait for downstream equipment, the end time of their buffer occupancy is taken as the start time of the subsequent node. The system latches each occupied interval according to the left-closed, right-open principle, meaning the task... Constantly occupying the node, in Nodes are released at all times to prevent adjacent tasks from being repeatedly flagged as conflicts at the same handover time. For nodes corrected by interlocks in the time-series recursion, the corrected end time is used. As the right endpoint of the interval. (Task number) Wave number Node number The predecessor node, successor node, action start time, action end time, duration of occupation, and node state are encapsulated as follows: ,in: All node occupancy records constitute the task node occupancy range data.

[0053] Preferably, the occupied interval display in S2 is specifically as follows: S24: Extract the occupancy boundary events based on the task node occupancy interval data to obtain the occupancy boundary event data; Specifically, read the occupancy records from the occupancy range data of the task node: ,in, and Tasks For nodes The system specifies the start and end times of occupancy for each left-closed, right-open occupancy interval. It can be broken down into an entry event and a release event, represented as follows: , Here, +1 indicates that a task begins occupying a node, and -1 indicates that a task ends occupying a node. All events are sorted in ascending order of event time; when multiple events have the same time, release events are processed first, followed by entry events, to avoid false overlaps caused by previous tasks not being released at the same handover time. The event time, task number, node number, event type, and wave number are encapsulated to obtain the occupation boundary event data.

[0054] S25: Perform non-equidistant event carrying segmentation based on the occupancy boundary event data to obtain event carrying interval data; Specifically, extract all distinct event moments and form an ordered time set: Instead of segmenting according to a fixed time granularity, the system uses any two adjacent event times as segment boundaries to construct non-equidistant event carrying intervals: Its duration is: Due to the interval No new entry or release events occur within this interval, therefore the task occupancy state on the same node remains unchanged within this interval. The system is at the left endpoint of the interval. The release and entry events occurring at that moment are applied sequentially, and the updated set of occupied tasks is latched as the set of tasks to be carried in that interval. Event-carrying interval data is formed by interval number, start and end times, interval length, and the set of tasks it carries.

[0055] S26: Generate task load map data based on event carrying interval data; Specifically, for each event carrying interval Query the set of tasks item by item The nodes currently occupied by each task. If the task Node occupied range satisfy: Then at node The above generates a video clip showing the workload of a task: The last digit, 1, indicates that the task continuously occupies the corresponding node throughout the entire carrying range. If the node capacity is... >1. Each task is still generated as a separate videotape, without proportionalizing multiple tasks into a single workload. The system merges videotapes from the same task and at the same node that are consecutive in time, resulting in: It also records the number of the carrying area it covers, forming task load videotape data.

[0056] S27: Perform logistics topology coordinate mapping based on task load video data to obtain load spatiotemporal coordinate data; Specifically, read the location attributes of each node in the logistics topology map data. For storage locations, buffer locations, and outbound exits, the locations will be uniformly represented by lane number, layer number, and column number: ,in, Indicates the coordinates of the tunnel or horizontal passage. Indicates column direction coordinates, This represents the floor height coordinates. For moving equipment such as stacker cranes, conveyors, and hoists, the coordinates of their service area center or their respective track are used as the topology node coordinates. The system will map each task load to... Bind the time interval to the node coordinates, construct: This transforms the task occupancy relationship from "task-node-time" to "task-spatial coordinates-time interval". For reversing or connecting nodes that only have logical connections but no physical spatial location, the intersection point of the line connecting the coordinates of its predecessor node and successor node is taken as the projection coordinate to obtain the load spatiotemporal coordinate data.

[0057] S28: Based on the load spatiotemporal coordinate data, perform cross-node load video tape continuation to obtain continuous task projection trajectory data; Specifically, the system processes the same task according to the order of nodes in the task propagation chain. The load spatiotemporal coordinate records are sorted. For adjacent nodes... and If the end time of the previous node's video tape and the start time of the next node's video tape satisfy the following: Then, the two spatiotemporal coordinate segments are directly joined end-to-end. If the following conditions are met: Then query whether the time interval corresponds to a connected edge. The physical transfer process; if so, then generate the transfer connection segment based on the coordinates of both ends of the connecting edge: ,in, This refers to the current simulation moment or trajectory sampling moment during the task's cross-node transfer process. This is used to represent the continuous movement of goods between two nodes. If the time interval corresponds to task waiting rather than physical movement, the task position is maintained at the predecessor cache node until the successor node begins receiving data. The entire task's node imagery and transfer segments are sequentially connected to obtain continuous task projection trajectory data.

[0058] S29: Perform multidimensional sparse load encapsulation based on continuous task projection trajectory data to obtain task spatiotemporal load data.

[0059] Specifically, the system uses task number, wave number, node number, topology coordinates, carrying area number, and event time as indexes to record only data items that actually have occupied or transported data, and does not create zero-value records for unoccupied node-time combinations. A single sparse load record is represented as follows: ,in, =1 indicates that the task occupies a node within this interval. This indicates that the record belongs to node occupancy, cache waiting, or cross-node transfer. For all records within the same node and the same bearer range, the system retains a task detail set: And record the number of tasks: This forms the spatiotemporal load data of tasks, organized by wave, task, node, spatial coordinate, event interval, and load type.

[0060] Preferably, the overlap analysis in S3 specifically includes: S31: Based on the task spatiotemporal load data and the task node occupancy interval data, downstream stagnation constraints are identified to obtain downstream stagnation constraint data. Specifically, reading tasks In adjacent nodes , The occupied range above: , And query the successor node from the task spatiotemporal load data. Number of tasks occupied within the event carrying range and node capacity When the task reaches the expected handover time of the successor node. The carrying capacity of the area satisfies: Or the earliest time when the successor node can receive data. Later than the original release time of the precursor node At this point, the task is determined to be subject to downstream delay constraints. The delay duration is: The system encapsulates the task number, predecessor node, successor node, original release time, successor receiveable time, and dwell time into downstream dwell constraint data. For outbound termination nodes without successor nodes, no downstream dwell constraints are generated.

[0061] S32: Perform reverse expansion of the upstream occupancy boundary based on the downstream retention constraint data to obtain the upstream reverse expansion interval data; Specifically, for tasks with downstream stagnation constraints, the predecessor node is interlocked according to the inter-node handover mechanism. The occupancy end boundary is from the original time. Extended to the actual receivable time of the successor node: Thus, the occupied interval after reverse expansion is obtained: Here, "reverse expansion" refers to the downstream node's reception delay being propagated backward along the task propagation chain to the upstream node, rather than advancing the start time of the interval. If an independent buffer is set between the predecessor and successor nodes, and that buffer has remaining capacity within the corresponding time, then expansion only extends to the buffer node, without further extending the occupied interval of the upstream execution device; if there is no buffer or the buffer is full, then the occupied boundary of the predecessor node is extended as described above. The system records the original occupied interval, the expanded occupied interval, and the expansion source node to obtain the upstream reverse expansion interval data.

[0062] S33: Obtain the incremental data of the reverse expansion interval by performing incremental difference of the occupied interval based on the upstream reverse expansion interval data; Specifically, the system will expand the occupied area in reverse. Compared with the original occupied interval Perform set difference: Since the two intervals have the same initial boundary, therefore when hour: The duration of its increment is: ,when When the value is 0, no incremental record is generated. The system binds the task number, node number, incremental start and end times, incremental duration, and the corresponding downstream source of delay to form reverse expansion incremental data. This increment only represents the resource lock-in time added due to the downstream's inability to receive the data in a timely manner, and does not include the original normal operation time.

[0063] S34: Based on the reverse expansion incremental data, perform secondary overlap identification of different tasks at the same node to obtain secondary occupation overlap data; Specifically, for each incremental interval Query the same node On other tasks The original occupied range or the expanded occupied range ,in When the following conditions are met: At that time, determine the task New lock-in and tasks caused by downstream delays A secondary overlap occurs. The overlapping interval is represented as follows: The system retains the record only when the right end of the overlapping interval is greater than the left end, and records the delayed task causing the lock, the blocked subsequent tasks, the node number, and the overlap duration. If the node capacity... If the number of tasks exceeds 1, then the number of tasks existing simultaneously during the overlapping period will be counted. Only if the number of tasks exceeds 1 will the count be counted. Only then is it determined to be a valid secondary occupation overlap, and secondary occupation overlap data is obtained.

[0064] S35: Perform upstream locking backtracking iteration based on the secondary occupancy overlap data to obtain stable backtracking occupancy data; Specifically, for nodes that cannot enter as originally planned due to a second overlap... Task The actual start time of this node is postponed to the time when the conflict task is released: And maintain its original duration of action. Unchanged, update end time: Then query the task predecessor node If the task is waiting to enter the node If the predecessor node is still occupied during this period, the end time of the predecessor node occupancy will be extended to [the next period]. The system then performs incremental difference analysis of occupied intervals and secondary overlap identification of different tasks at the same node. Following the task propagation chain, the system backtracks upstream from the conflicting node, adding each newly generated expansion interval to the pending set. If no new expansion increment or secondary overlap is generated in a given iteration, or if the backtracking reaches the task's starting node, the iteration stops. To prevent circular propagation, the same "task-node-interval right boundary" combination is only allowed to be updated once, and replacement is only performed if the new end time is later than the previous round's end time. This yields stable backtracking occupancy data containing the original occupied interval, expansion intervals from each round, the source of the lock, affected tasks, and iteration levels.

[0065] S36: Perform node aggregation load determination on the stable backtracking occupancy data to obtain node load aggregation data.

[0066] Specifically, , For nodes The number of concurrent tasks in the m-th decision interval. For the i-th outbound task, For the m-th load determination time interval, For the task At the node The effective occupied interval after reverse expansion or backtracking correction, when: When this interval is reached, it is designated as the node load aggregation interval; if However, if the occupancy of at least one task is formed by reverse expansion and there are subsequent tasks waiting to enter, it is marked as a critical convergence interval. The maximum backtracking level comes from the multi-level lock-up backtracking process. The node where the first and second occupancy overlaps is recorded as level 0. The level is incremented by 1 for each backtracking to the preceding node of the task propagation chain. For all backtracking paths within the same convergence unit, the maximum level reached is taken. The cumulative expansion duration comes from the newly added occupancy intervals in each round of reverse expansion and backtracking. For the same task and the same node, only the length of the newly added non-overlapping intervals in each round is accumulated. The system merges convergence intervals on the same node that are consecutive in time and have the same set of participating tasks, and records the convergence node, convergence start and end time, node capacity, set of simultaneously locked tasks, set of waiting tasks, first delayed source node, maximum backtracking level, and cumulative expansion duration to form node load convergence data.

[0067] Preferably, the upstream locking backtracking iteration specifically involves: S351: Extract the overlapping hindrance increment based on the secondary occupancy overlap data to obtain the overlapping hindrance increment data; Specifically, read the overlapping records in the secondary occupancy overlapping data: Among them, the task Extended occupation of nodes due to downstream congestion The lock task, task To enter the node according to the original timeline The obstructed mission This refers to the effective overlap interval between the two tasks. This represents the start time of the second occupancy of the overlapping interval. This is the end time of the overlapping interval between secondary occupancy. The system queries the locking task. At the node Release time after expansion and the obstructed mission Original start time Computation of hindered tasks New waiting time: , For nodes The earliest release time to restore the node to an acceptable state is when, after any occupied tasks release resources, the current occupied amount first decreases to less than the node's capacity. The system encapsulates the blocked task, conflicting node, start and end times of the blockage, duration of the blockage, and the corresponding lock-originating task to obtain overlapping blockage incremental data. Blocked tasks are those that cannot enter the node as planned due to insufficient capacity.

[0068] S352: Based on the overlapping blocking incremental data, the preceding locking boundary is back-inferred to obtain the preceding locking boundary data; Specifically, the task In the preceding node The actual release time is determined to be the later of the originally planned release time and the actual available entry time of the subsequent node. , For the task At the node The actual start time after the above-mentioned delay correction, If there are independent cache nodes between the preceding node and the conflicting node Then query the cache node in the range The amount of space occupied within the cache node. When a cache node still has remaining capacity, only the end boundary of the cache node's occupancy is extended to... The upstream execution device is not extended; when the cache node is full, the preceding node is selected and the execution boundary is pushed back along the task propagation chain. The system records the blocked task, the current conflicting node, the preceding node that is pushed back, the original release boundary, the release boundary after the push back, and the reason for the push back, thus obtaining the preceding lock boundary data.

[0069] S353: Based on the previous lock boundary data, perform backtracking occupancy increment extraction to obtain the backtracking newly added occupancy data; Specifically, the task In the preceding node The additional lock holding time caused by the successor node's inability to receive the lock in time is the amount of time that the release time of the predecessor node is postponed. If the preceding node has already formed an expanded interval in the previous backtracking round, the difference between the current backtracking end time and the previous round end time is used as the new amount for this round. Only the unprocessed interval portion is retained to avoid the same interval being repeatedly involved in overlapping judgments. The system encapsulates the task number, preceding node, backtracking level, start and end times of new occupation, new duration, and corresponding downstream conflicting nodes to obtain the backtracking new occupation data.

[0070] S354: Based on the newly added back-tracked occupied data, perform a re-judgment of the overlapping of previous heterogeneous tasks to obtain the back-tracked overlapping updated data; Specifically, for each newly occupied interval during backtracking Query the same predecessor node On other tasks Current effective occupied range ,in The currently valid occupied interval includes the original occupied interval, the upstream reverse expansion interval, and the backtracking expansion interval confirmed in previous iterations. Computational task The newly added backtracking interval is related to another task on the same preceding node. The actual overlapping intervals between the currently active occupied intervals In this case, the record is retained only if the right end of the interval is greater than the left end. For the task The backtracking of the newly added occupancy begins at the previous node. For the task In the preceding node The current effective occupancy start time on, For the task At the end of the backtracking addition and occupancy time on the preceding node. For the task In the preceding node The system calculates the current effective occupancy end time. When the same node and the same blocked task are simultaneously affected by multiple locked tasks, the system does not accumulate the repetition duration, but instead takes the latest release time that allowed the task to enter the node as the new blocking boundary. The system encapsulates the new locked task, the new blocked task, the preceding node, the overlapping interval, the node capacity, and the corresponding backtracking level to obtain backtracking overlapping update data, which is then used as input for the next round of overlapping blocking increment extraction.

[0071] S355: Perform multi-level locking convergence iteration based on the backtracking overlapping update data to obtain stable backtracking occupancy data.

[0072] Specifically, the system establishes a set of overlapping objects to be processed. ,in This refers to the iteration round. Initially, the overlapping data from the second occupancy is written... In the r-th round, proceed sequentially to... The overlapping records within the process execute steps S351 to S354, which involve extracting the overlapping delay increment during the upstream locking backtracking iteration and re-judging the overlapping of previous tasks, thus generating the newly added occupancy set for this round of backtracking. and new overlapping sets In the (r+1)th round of backtracking iteration, the task At the node The end time of the occupation is the later of the end time determined in the previous round and the end time newly calculated in this round. , After the r-th round of lock-holding backtracking iteration, the task At the node The current occupancy end time on, The candidate occupancy end time is derived from newly generated overlapping blockages or previous lock boundaries in this iteration. Replacement is only performed when the new boundary is later than the current boundary. Therefore, the occupancy interval in each round only extends and does not shrink. If the newly generated occupancy boundary is the same as the recorded boundary, no duplicate increments are generated. The system terminates the iteration when any of the following conditions are met: no new backtracking occupancy interval is generated in this round; no new valid cross-task overlap is generated in this round; all backtracking paths have reached the task start node; or all pending nodes have available cache capacity, so that the lock does not continue to propagate upstream. After the iteration ends, the system summarizes the original occupancy interval, final lock end boundary, cumulative newly added occupancy interval, direct lock source, indirect lock source, maximum backtracking level, and affected task set for each task and node to form stable backtracking occupancy data. The affected task set is the combination of the triggering task, blocked task, and subsequent indirectly delayed tasks.

[0073] Preferably, the congestion propagation analysis in S3 is as follows: S37: Identify congestion source cores based on node load aggregation data to obtain congestion source core data; Specifically, read the aggregation node from the node load aggregation data. Convergence Zone Node capacity Locked task set Waiting task set The system first identifies the source node of the congestion and the maximum backtracking level. It checks each convergence interval chronologically and identifies the node that first experiences overcapacity occupancy or full capacity lockout during the same congestion association process, accompanied by subsequent task waiting, as the candidate source core. When a node is in the interval... Internal satisfaction: Or satisfy And waiting for the task set When at least one occupied interval is formed by reverse expansion or backtracking locking, the node is identified as a congestion convergence source core. For nodes The number of effectively occupied tasks within the q-th load aggregation interval, i.e., within the aggregation interval. The system calculates the total number of different tasks simultaneously occupying the node. If multiple nodes meet the conditions at the same time, the system traces back along the task propagation chain and selects the node that first generates downstream congestion constraints as the source core. If they cannot be distinguished, the upstream node priority rule in the logistics topology is used to determine the source core. The system encapsulates the source core node, source core formation time, convergence interval, locked tasks, waiting tasks, node capacity, and source core origin to obtain congestion source core data.

[0074] S38: Strip the unabsorbed load from the congestion source core data to obtain the propagated residual load data; Specifically, targeting the source core node Within the convergence zone, the system determines the portion of the tasks that a node can normally handle based on its capacity. Tasks are arranged in ascending order of their originally scheduled entry time; if entry times are the same, they are sorted by task priority and task number, with the first... Tasks that can be accepted by a node are marked as accepted tasks, and the remaining tasks are marked as unaccepted tasks. At the node Unresolved delay duration / Remaining delay duration due to node's temporary inability to accept ,in For the task Original plan to enter node The moment when the residual block begins. The moment when the source core node releases the first capacity slot among the currently occupied tasks is the end time of the remaining congestion. If releasing one slot still cannot accommodate all waiting tasks, the acceptable time corresponding to each unaccepted task is determined one by one according to the release event sequence. The system encapsulates the unaccepted tasks, source core node, start and end times of the remaining congestion, and the duration of the remaining congestion to obtain the propagation of remaining load data.

[0075] S39: Based on the propagation of the remaining load data, perform targeted injection of adjacent logistics nodes to obtain adjacent load injection data; Specifically, for unprocessed tasks Query the source core node based on its task propagation chain. direct predecessor node Since the task failed to reach the source core node as planned, its remaining blocked load was injected backward along the task path into the direct predecessor node, instead of spreading to unrelated adjacent nodes. Unable to access downstream nodes And in the upstream adjacent nodes The newly formed load injection interval ,in These are the adjacent nodes that actually receive the obstruction. This refers to the time when the task was originally occupied by the node and ended. The system records the source core node, injection node, task number, injection direction, injection interval, and source of obstruction, forming adjacent load injection data.

[0076] S310: Perform node load margin erosion analysis based on adjacent load injection data to obtain propagation trigger node data; Specifically, the system queries the injection node. The number of currently active tasks within the injection interval and node capacity And define the node carrying capacity margin: For task injection interval If the entire interval satisfies If the injection can be absorbed by the node's remaining capacity and does not trigger further propagation, then the injection can be absorbed by the node's remaining capacity and will not trigger further propagation; if there exists a time sub-interval that satisfies: This indicates that the remaining load has eroded the entire capacity margin. From the task... For nodes Within the resulting load injection interval, the portion of the time when the extraction node has no remaining capacity is the interval that actually triggers congestion propagation. When the intersection is not empty, the nodes are... The node is identified as the propagation trigger node, and its trigger time, trigger interval, triggering task, node capacity, original set of occupied tasks, and source core node are recorded to obtain propagation trigger node data. For nodes with a capacity greater than 1, the triggering condition is that the number of simultaneously occupied tasks exceeds the capacity. The triggering task is the task that injects new load.

[0077] S311: Based on the data from the propagation trigger node, perform cross-node time delay cascading recursion to obtain cascading congestion trajectory data; Specifically, the task At the node Cascading delay caused by upstream load injection or node capacity being fully utilized , For the task At the node The actual start time after congestion propagation correction. For the task At the node If the mission is scheduled to begin at the original start time... It continues to occupy its predecessor node while waiting. The cascading delay is then converted into a new adjacent injection interval, and adjacent logistics node directional injection and node capacity margin erosion analysis are performed sequentially. If available cache capacity exists, the cascading terminates at the cache node; if it backtracks to the task start node, the task release delay is recorded as the termination result. The reason for propagation termination is determined based on the stopping conditions when the cascading recursion stops, including adjacent nodes having remaining capacity and absorbing the injected load, available cache nodes truncating the propagation, backtracking to the task start node, reaching the outbound termination node, or subsequent nodes no longer forming effective overlap. The system records each level of trigger node, affected task, injection interval, cascading delay, and propagation termination reason in the order of "source core node - first trigger node - subsequent trigger nodes" to obtain cascading congestion trajectory data.

[0078] S312: Based on the cascaded congestion trajectory data, the propagation nodes are sequentially connected to obtain the congestion propagation chain data.

[0079] Specifically, the system uses the source core number and the initial unresolved tasks as indexes to arrange congestion trajectories at all levels belonging to the same propagation process in ascending order of trigger time. For two adjacent propagation records, a propagation connection is established only if the predecessor node of the currently blocked task is the same as the next trigger node, and the next trigger time is not earlier than the current injection start time. A single propagation chain is represented as follows: The corresponding propagation time sequence is represented as: , For the congestion source node of the g-th congestion propagation chain, This is the first-level propagation node triggered by the congestion source core. For the second-level propagation node triggered by the first-level propagation, This is the termination node of the g-th congestion propagation chain. For the source core convergence interval corresponding to the congested source core node, This is the propagation trigger interval for the first-level propagation node. This is the propagation trigger interval for the second-level propagation node. For the propagation trigger interval of the r-th propagation node, the system simultaneously records the triggering task, blocked task, and cascading delay corresponding to each propagation node. The cumulative propagation delay and the reason for propagation termination are calculated. The cumulative propagation delay is accumulated sequentially according to the actual increase in delay at each level, but overlapping time portions are only retained once to avoid double counting. If multiple task branches are generated from the same source core, they are formed into sub-propagation chains and associated through a common source core number. If multiple propagation branches converge at the same node and in the same time interval, they are merged into a single convergence propagation node, but the source of each task is retained. The final congestion propagation chain data includes the source core node, source core interval, propagation node sequence, propagation time interval sequence, set of affected tasks, delay at each level, cumulative delay, and termination node.

[0080] Preferably, the pre-elimination deduction in S4 is specifically as follows: S41: Based on the congestion propagation chain data, backtrack the propagation reverse intervention window to obtain the forward intervention window data; Specifically, read the source core node in the congestion propagation chain data. Timing of source nucleus formation Propagation node sequence The system defines the propagation trigger intervals at each level, triggering tasks, blocked tasks, and cascading delays. The system is based on the source kernel formation time. To trace back to the endpoint, we reverse the task propagation chain of each affected task and query its still adjustable preceding nodes and task states. For tasks that have not yet been released... Its intervention window starts at the current simulation time. The window endpoint is the latest release adjustment time that will not cause source core convergence. If the task passes through nodes sequentially after release... To the source kernel node The cumulative physical transit time before it reaches the source nucleus is: , For the task The cumulative basic turnaround time required from the task release location to the source core node of the g-th congestion propagation chain. For the task The kth logistics node in the propagation path, For the congestion source node of the g-th congestion propagation chain, For the task At the node The basic action delay, corresponding to the latest intervention time, is: , The moment when the source core of the g-th congestion propagation chain is formed determines the intervention window for release: , For tasks that have been released but have not yet reached the source core node, the expected departure time of their current node is used as the starting point of the window, and the time when the task enters the next unadjustable node is used as the ending point. If the current node has waiting capabilities, backup paths, or available temporary caches, then the node is marked as an interventionable node. If the device action has already started and cannot be stopped, no intervention window is generated during the current action, and backtracking begins from the next controllable handover point. The system encapsulates the task number, propagation chain number, interventionable node, window start and end times, allowed action types, and the corresponding source core formation time to obtain the preceding intervention window data.

[0081] S42: Based on the data from the pre-intervention window, filter the propagation truncation objects to obtain propagation truncation object data; Specifically, for each congestion propagation chain, the system filters tasks that can alter the arrival order, arrival time, or propagation path before the source core is formed as candidate truncation targets. Candidate tasks should meet at least one of the following conditions: the task has not yet been released; the task is located on a cache node that allows waiting; the task has not yet entered the irreversible transport segment; there is an alternative path after the current node that does not pass through the source core node; or there is an allocable temporary cache node upstream of the source core. For candidate tasks... The system calculates its direct occupancy relationship with the source core region. If the task's expected occupancy region on the source core node... Convergence region of source and nucleus intersect: If a task does not directly occupy the source core node, but its locking causes any triggering node in the propagation chain to form an injection interval, then it is considered an indirectly truncated object.

[0082] The selection process follows deterministic rules: Prioritize tasks upstream of the source core that have not yet been executed; secondly, select tasks that only require adjusting their release time to avoid the convergence zone; thirdly, select tasks that can reside at the nearest preceding cache node; and finally, select tasks that require switching propagation paths. For multiple tasks within the same level, prioritize tasks whose intervention location is closest to the source core node and whose intervention window is not yet closed; if still the same, select tasks according to their original priority from low to high to avoid delaying urgent tasks. The system encapsulates the truncated task, truncated node, truncated type, corresponding propagation chain, original arrival time at the source core, and set of executable actions to obtain propagation truncated object data.

[0083] S43: Based on the propagation truncation object data, perform the coupling and arrangement of resolution actions to obtain the candidate data for prior resolution, wherein the resolution actions include at least one of the following: task release time adjustment, device service order reordering, task propagation path switching, and temporary cache node allocation; Specifically, for each propagation truncation object, the system generates candidate solutions in the order of "single action first, then combination of actions." Regarding the adjustment of the task release time, the task... The release time is determined by Adjusted to: ,in, This represents the actual time interval or preset minimum control duration between the event carrying interval boundaries, with n increasing from 1. After adjustment, it should satisfy: Furthermore, the adjusted source core occupancy range no longer overlaps with the original convergence range due to overcapacity. For the task The rated total propagation time required to travel from the starting node to the target outbound node along its task propagation chain is obtained by summing the rated travel times of each node or connecting edge on the task path.

[0084] For device service order reordering, only tasks that have not yet started execution and are waiting for the same device are reordered. The system maintains the principle of executing high-priority tasks first; under the same priority, tasks that can be directly received by the next node are executed first, and tasks with no available capacity on downstream nodes are postponed. The reordering result must not interrupt the execution of currently executing device actions, nor change the order of completed tasks.

[0085] For task propagation path switching, the system searches for an alternative path from the current node to the target outbound port from the set of available connection edges obtained from the twin state synchronization data. The alternative path must not pass through the source core node and triggered nodes in the current propagation chain, and should meet the constraints of cargo specifications, equipment activation status, and conveying direction. When multiple paths exist, the path with the fewest nodes and the shortest rated propagation time is selected in sequence.

[0086] For temporary cache node allocation, the system searches along the path from the current task node to the source core node for cache nodes that are not full, have matching cargo specifications, and are directly connected to the original propagation path. The allocated cache space is: ,in For the task Enter temporary cache node At that moment, The earliest time after the source core node or the next node becomes acceptable is taken, i.e., the task. Leave the temporary cache node At that moment.

[0087] When a single action cannot eliminate the propagation, the system generates a combination of solutions in the order of "release adjustment and service order rearrangement", "release adjustment and temporary cache allocation", and "path switching and service order rearrangement". Each candidate solution records the adjustment task, action type, action execution node, original parameters, adjusted parameters, action sequence, and the set of affected tasks, forming preliminary resolution candidate data.

[0088] S44: Based on the candidate data for the previous resolution, perform congestion propagation residual evolution to obtain the data for the previous resolution.

[0089] Specifically, the system writes the release time, service order, propagation path, and cache occupancy range of each candidate solution into the corresponding task propagation chain, and re-executes the node timing concatenation recursion, occupancy range explicitation, and node load convergence determination. For each node in the original congestion propagation chain... Count the number of effective tasks after the candidate solutions are executed. And define the propagation residual interval as: If there is no overcapacity occupancy, but the node is at full capacity and there are still tasks waiting, then the corresponding time period is recorded as the critical propagation residual.

[0090] The system iterates node by node along the original propagation direction. If the propagation residual at the source core node is empty, the original first propagation trigger node is checked again; if the node still has residual, it is used as the new starting point to re-execute the unabsorbed load stripping, adjacent directional injection, and load margin erosion judgment until the propagation terminates. A candidate scheme is considered valid if it meets the following conditions: the original source core convergence interval is eliminated; no new overcapacity nodes are formed; and the completion time of all adjustment tasks is no later than the required completion time. The number of temporary cache nodes does not exceed their capacity; all nodes in the backup path remain available. When multiple candidate schemes are effective, the final scheme is determined in the following order: prioritize the scheme involving the fewest tasks; if the number of tasks is the same, select the scheme with the fewest action types; if the action types are the same, select the scheme with the shortest total task release delay; if still the same, select the scheme with the fewest path switching times; if still indistinguishable, select the scheme with the shortest temporary cache occupation duration. The propagation chain number, truncated task, intervention window, resolution action and its execution order, original and adjusted release time, device service order, propagation path, cache occupation range, adjusted node occupation range, residual evolution result, and propagation termination position are encapsulated to obtain the pre-resolution data.

[0091] Preferably, the digital twin simulation in S4 specifically includes: S45: Perform twin initial state branch mapping based on the previous resolution data to obtain twin branch data of the resolution scheme; Specifically, read the current simulation time. The underlying digital twin state includes device node status, current task location, task execution stage, node occupancy range, device waiting queue, remaining cache node capacity, and logistics connection edge activation status. This underlying state is then replicated into independent twin state branches, with each pre-processing solution corresponding to one twin branch. And retain a baseline branch that does not perform a resolution action. .

[0092] The system writes the adjustment task, task release time, device service order, alternative propagation path, and temporary cache occupancy range from the pre-processing data into the corresponding twin branches. For task release time adjustments, the original release event is... Replace with: For device service order reordering, update nodes Waiting queue: , After the service order is rearranged, the nodes The first task to be executed in the waiting queue. After the service order is rearranged, the nodes The second task to be executed in the waiting queue. After the service order is rearranged, the nodes The nth task in the waiting queue is selected, but the currently executing and completed tasks remain unchanged. For path switching, the original node sequence of the tasks is changed. Replace with candidate paths For temporary cache allocation, entry and exit events are established at the specified cache node.

[0093] To avoid repeatedly simulating historical states, only Events that have not yet occurred are written into the branch event set. If the execution time of a resolution action is earlier than... If the action is applied to a node that has already been completed, the solution is determined to be unmappable and a failure flag is generated. The branch number, corresponding solution number, initial device state, task state, updated task propagation chain, waiting queue, cache allocation relationship, and set of events to be executed are encapsulated to obtain the solution twin branch data.

[0094] S46: Perform task event-driven evolution based on the twin branch data of the resolution scheme to obtain twin load evolution data; Specifically, for each twin branch, an event queue is established, arranged in ascending order of event time: , This is the first pending event in the twin branch event queue, sorted by occurrence time. This is the second pending event in the twin branch event queue, sorted by occurrence time. This is the m-th pending event in the twin branch event queue, sorted by occurrence time. The event includes at least the following: task release event, node entry event, device action start event, device action end event, cache entry event, cache exit event, node release event, and task completion event. When multiple events occur at the same time, they are processed sequentially in the order of "action end, node release, cache exit, task completion, task release, node entry, action start, cache entry" to avoid resources being considered occupied even after being released at the same time. The system advances the twin clock starting from the earliest event. After a task is released, it checks if the head node of its propagation chain has the acceptance conditions; if the node is idle or has a remaining capacity greater than zero, a node entry and action start event are generated; otherwise, the task is added to the node waiting queue. The device action duration is calculated using the action delay obtained by S21 / node physical flow delay. The action ends at: , For the task At the node The action begins at the moment the task is prepared by the node. Transfer to successor node At that time, check the capacity, device status, and connection edge enable status of the successor node. A transfer event is generated only if the successor node is ready to receive; otherwise, the task continues to hold the current node, and the next check time is set to the most recent release event time of the successor node. For tasks with allocated temporary buffers, priority is given to entering the designated buffer node, and the buffer usage must not exceed [a certain limit]. For each event processed, the system updates the set of tasks occupied by the corresponding node: And obtain the instantaneous load of the node: The system latches node load, number of waiting tasks, cache usage, cumulative task waiting time, and device idle status between adjacent event moments, forming non-equidistant load evolution intervals. The simulation continues until all affected tasks are completed, or until a preset simulation termination time is reached, obtaining twin load evolution data.

[0095] S47: Based on the twin load evolution data, the congestion propagation and decay quantification is performed to obtain the simulation data.

[0096] Specifically, the system in the baseline branch Extract the original congestion source core node, propagation node sequence, overcapacity interval, critical full load interval, propagation delay at each level, and the set of affected tasks. In each resolution scheme branch... The corresponding results are extracted according to the same node and the same task scope to ensure that they can be directly compared before and after the resolution. , This represents the duration of the interval. For the nodes in this branch The actual number of tasks occupied at time t. For logistics nodes The maximum simultaneous capacity, that is, the upper limit of the number of tasks that the node is allowed to accommodate, process, or occupy at the same time. For the first In each twin branch, the node The m-th load evolution interval; the corresponding congestion duration reduction is: , , For nodes in the baseline branch The cumulative overcapacity duration, The cumulative overcapacity duration in the h-th solution branch. For the maximum cascading delay decay, The maximum cascade delay in the baseline branch, The maximum cascading delay in the h-th resolution branch; the corresponding propagation node change and waiting time decay are: ,in, The number of nodes used for the baseline branch propagation. For the first Number of branch propagation nodes for each scheme; ,in, and Tasks The cumulative waiting time in the baseline branch and the scheme branch. When the scheme branch satisfies the following conditions: all source core nodes no longer have overcapacity intervals, the number of propagation nodes is zero, the temporary cache is not overcapacity, and the completion time of all tasks is not later than the required completion time, the scheme is marked as completely fading; when there are still propagation nodes, but their number, duration, or cascading delay is less than that of the baseline branch, it is marked as partially fading; when new overcapacity nodes are generated, tasks are overdue, or the cache overflows, it is marked as fading failure. The scheme number, twin branch number, event evolution sequence, node load timing, source core disappearance status, propagation node changes, overcapacity duration fading amount, waiting time fading amount, maximum cascading delay fading amount, task completion time, cache usage, and scheme validity identifier are encapsulated to obtain the simulation data.

[0097] Preferably, this application also provides a digital twin-based automated warehouse operation optimization system for executing the digital twin-based automated warehouse operation optimization method described above. The digital twin-based automated warehouse operation optimization system includes: The digital twin relationship parsing module is used to acquire outbound wave task data, warehousing and logistics topology data, and real-time equipment status data of automated warehouses; and to perform digital twin relationship parsing based on the outbound wave task data and warehousing and logistics topology data to obtain task propagation chain data. The task spatiotemporal load module is used to perform recursive calculations based on task propagation chain data and real-time device status data to obtain task node occupancy interval data; and to perform occupancy interval mapping based on task node occupancy interval data to obtain task spatiotemporal load data. The congestion propagation analysis module is used to perform overlap analysis on the occupancy interval data of task nodes based on the spatiotemporal load data of the task to obtain the node load aggregation data; and to perform congestion propagation analysis based on the node load aggregation data to obtain the congestion propagation chain data. The pre-congestion mitigation simulation module is used to perform pre-congestion mitigation simulation based on congestion propagation chain data to obtain pre-congestion mitigation data; and to perform digital twin simulation based on the pre-congestion mitigation data to obtain simulation data.

[0098] Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended application documents rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of the equivalents of the application documents be incorporated into the invention.

[0099] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for optimizing the operation of an automated three-dimensional warehouse based on digital twins, characterized in that, Includes the following steps: S1: Acquire outbound wave task data, warehousing and logistics topology data, and real-time equipment status data of the automated warehouse; perform digital twin relationship analysis based on the outbound wave task data and warehousing and logistics topology data to obtain task propagation chain data; S2: Based on the task propagation chain data and the real-time status data of the device, perform recursive calculations to obtain the task node occupancy interval data; based on the task node occupancy interval data, perform occupancy interval mapping to obtain the task spatiotemporal load data; S3: Perform overlap analysis on the task node occupancy interval data based on the task spatiotemporal load data to obtain node load aggregation data; perform congestion propagation analysis on the node load aggregation data to obtain congestion propagation chain data. S4: Perform preliminary mitigation simulation based on congestion propagation chain data to obtain preliminary mitigation data; perform digital twin simulation based on preliminary mitigation data to obtain simulation data.

2. The method according to claim 1, characterized in that, The specific analysis of digital twin relationships in S1 is as follows: A digital twin graph is constructed based on the warehouse logistics topology data to obtain logistics topology graph data; task tuples are discretized based on the outbound wave task data to obtain task tuple data; the task tuple data and logistics topology graph data are aligned with digital twin entities to obtain twin state synchronization data; and a task propagation chain is constructed based on the twin state synchronization data to obtain task propagation chain data.

3. The method according to claim 1, characterized in that, The recursive calculation in S2 is as follows: The physical flow delay of nodes is calculated based on the task propagation chain data and the real-time status data of the equipment to obtain the action delay data; the timing sequence is recursively deduced based on the action delay data to obtain the timing sequence data; and the occupied interval is explicitly latched based on the timing sequence data to obtain the task node occupied interval data.

4. The method according to claim 1, characterized in that, The specific details of the interval display in S2 are as follows: Based on the task node occupancy interval data, occupancy boundary events are extracted to obtain occupancy boundary event data; based on the occupancy boundary event data, non-equidistant event carrying segments are performed to obtain event carrying interval data; based on the event carrying interval data, task occupancy load maps are generated to obtain task load map data; based on the task load map data, logistics topology coordinate mapping is performed to obtain load spatiotemporal coordinate data; based on the load spatiotemporal coordinate data, cross-node load map continuation is performed to obtain continuous task mapping trajectory data; based on the continuous task mapping trajectory data, multi-dimensional sparse load encapsulation is performed to obtain task spatiotemporal load data.

5. The method according to claim 1, characterized in that, The overlap analysis in S3 is as follows: Downstream stagnation constraints are identified based on task spatiotemporal load data and task node occupancy interval data to obtain downstream stagnation constraint data; upstream occupancy boundaries are reverse-expanded based on downstream stagnation constraint data to obtain upstream reverse expansion interval data; and occupancy interval incremental difference is performed based on upstream reverse expansion interval data to obtain reverse expansion incremental data. Based on the reverse expansion incremental data, secondary overlap identification of different tasks on the same node is performed to obtain secondary occupancy overlap data; based on the secondary occupancy overlap data, upstream lock-in backtracking iteration is performed to obtain stable backtracking occupancy data; node aggregation load determination is performed on the stable backtracking occupancy data to obtain node load aggregation data.

6. The method according to claim 5, characterized in that, The upstream lock-holding backtracking iteration is specifically as follows: Based on the overlapping occupancy data, the overlapping stagnation increment is extracted to obtain the overlapping stagnation increment data; based on the overlapping stagnation increment data, the preceding lock occupancy boundary is back-calculated to obtain the preceding lock occupancy boundary data; based on the preceding lock occupancy boundary data, the backtracking occupancy increment is extracted to obtain the backtracking newly added occupancy data; based on the backtracking newly added occupancy data, the preceding heterogeneous task overlap is re-determined to obtain the backtracking overlapping update data; based on the backtracking overlapping update data, multi-level lock occupancy convergence iteration is performed to obtain stable backtracking occupancy data.

7. The method according to claim 1, characterized in that, The congestion propagation analysis in S3 is as follows: Congestion source cores are identified based on node load aggregation data to obtain congestion source core data; unabsorbed loads are stripped from the congestion source core data to obtain propagated residual load data. Based on the propagation of remaining load data, targeted injection is performed on adjacent logistics nodes to obtain adjacent load injection data; Based on the adjacent load injection data, node load margin erosion analysis is performed to obtain propagation trigger node data; Based on the propagation trigger node data, cross-node time delay cascading recursion is performed to obtain cascaded congestion trajectory data; based on the cascaded congestion trajectory data, the propagation node time sequence is concatenated to obtain congestion propagation chain data.

8. The method according to claim 1, characterized in that, The pre-solution derivation in S4 is as follows: Based on the congestion propagation chain data, the propagation reverse intervention window is traced back to obtain the pre-intervention window data; based on the pre-intervention window data, the propagation truncation object is filtered to obtain the propagation truncation object data; Based on the propagation truncation object data, the resolution actions are coupled and arranged to obtain the preliminary resolution candidate data. The resolution actions include at least one of the following: task release time adjustment, device service order reordering, task propagation path switching, and temporary cache node allocation. Based on the preliminary resolution candidate data, the congestion propagation residual evolution is performed to obtain the preliminary resolution data.

9. The method according to claim 1, characterized in that, The digital twin simulation in S4 is specifically as follows: Based on the pre-resolved data, a twin initial state branch mapping is performed to obtain twin branch data for the resolved scheme; based on the twin branch data for the resolved scheme, task event-driven evolution is performed to obtain twin load evolution data; based on the twin load evolution data, congestion propagation and decay quantification is performed to obtain simulation data.

10. An automated storage and retrieval system (AS / RS) operation optimization system based on digital twins, characterized in that: For executing the automated storage and retrieval system (AS / RS) operation optimization method based on digital twins as described in claim 1, the automated storage and retrieval system (AS / RS) operation optimization system based on digital twins includes: The digital twin relationship parsing module is used to acquire outbound wave task data, warehousing and logistics topology data, and real-time equipment status data of automated warehouses; and to perform digital twin relationship parsing based on the outbound wave task data and warehousing and logistics topology data to obtain task propagation chain data. The task spatiotemporal load module is used to perform recursive calculations based on task propagation chain data and real-time device status data to obtain task node occupancy interval data; and to perform occupancy interval mapping based on task node occupancy interval data to obtain task spatiotemporal load data. The congestion propagation analysis module is used to perform overlap analysis on the occupancy interval data of task nodes based on the spatiotemporal load data of the task to obtain the node load aggregation data; and to perform congestion propagation analysis based on the node load aggregation data to obtain the congestion propagation chain data. The pre-congestion mitigation simulation module is used to perform pre-congestion mitigation simulation based on congestion propagation chain data to obtain pre-congestion mitigation data; and to perform digital twin simulation based on the pre-congestion mitigation data to obtain simulation data.

Images