A big data-based smart park information management system

Abstract

The present application relates to the technical field of smart park management, in particular to a smart park information management system based on big data, the present application constructs a shared physical path conflict graph, uniformly models various types of operation path relationships and capacity constraints in the park, and forms a time delay distribution library through real-time data calibration. On this basis, the position of the operation unit is tracked in real time, the capacity cascade release prediction table is generated by combining the time delay distribution parameters, the time sequence continuity is checked node by node according to the operation type path chain, and the effective combination window is generated for the tenant to select. Further introduce the deviation trigger judgment and cascade correction mechanism to dynamically correct the actual operation deviation, at the same time, the time delay distribution parameters are updated by the sample library and the conflict graph is written back, the model self-adaptive optimization is realized. The method can improve the path resource utilization rate, reduce the conflict probability, and enhance the accuracy and stability of scheduling in the multi-tenant environment.

Description

Technical Field

[0001] This invention relates to the field of smart park management technology, specifically a smart park information management system based on big data. Background Technology

[0002] As industrial parks evolve into complexes, they often bring together various tenants, including e-commerce, manufacturing, and logistics companies, forming a typical multi-tenant shared warehousing hub operating model. In this model, the park's physical facilities, such as shelving space, loading and unloading platforms, cold chain equipment, and transportation resources, are shared concurrently by multiple tenants. Simultaneously, each tenant's inventory data, inbound and outbound records, and cargo information constitute their own core business data, requiring strict information segregation and compliance requirements.

[0003] In existing technologies, smart park information management systems typically adopt two types of architectures: one is a strong isolation scheme where each tenant deploys its own subsystems. While this scheme can ensure data security, the systems of each tenant are fragmented, and shared physical resources cannot be coordinated and scheduled across tenants. During peak periods of concurrent inbound and outbound operations by multiple tenants, resource mismatch problems frequently occur, such as platform congestion and equipment idleness. The other is a centralized management scheme that unifies data aggregation. This scheme can achieve centralized scheduling of resources, but the data boundaries of each tenant are blurred, posing a security risk of cross-access to sensitive information. It is difficult to meet the data isolation requirements and compliance requirements in multi-tenant scenarios in parks.

[0004] The fundamental flaw of the two types of solutions mentioned above is that neither can perform dynamic collaborative scheduling of multi-tenant competitively shared warehousing resources based on big data under the premise of strict data isolation. This results in low overall resource utilization of the shared warehousing hub in the park, and there is an irreconcilable structural contradiction between data security and resource sharing efficiency.

[0005] To address this, a smart park information management system based on big data is proposed. Summary of the Invention

[0006] The purpose of this invention is to provide a smart park information management system based on big data. By constructing a shared physical path conflict map and a time delay distribution library, and combining capacity cascading release prediction, combined window generation and deviation cascading correction mechanism, it realizes the time-series collaborative scheduling and dynamic optimization management of multi-tenant operations in the shared path network.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A smart park information management system based on big data includes: The path status distribution module is used to construct a shared physical path conflict map based on the physical path layout of the park's storage area, and to mark the status delay distribution of each path segment by actual measurement to form a delay distribution library. The path cascading release module is used to track the location of each work unit's path node in real time based on a shared physical path conflict map, and calculate the expected release time of each path segment's capacity by combining the delay distribution library, and generate a capacity cascading release prediction table. The timing combination window distribution module is used to receive the job types reported by tenants through the isolation channel, determine the corresponding path node sequence based on the shared physical path conflict diagram, check the full chain timing coherence availability node by node on the capacity cascade release prediction table, and distribute the combination windows that meet the conditions through the tenant isolation channel. The path delay correction module is used to compare the actual dwell time of each path segment with the corresponding average value in the delay distribution library after the tenant confirms the combined window. When the deviation exceeds the distribution range, the predicted release time of the affected subsequent nodes is delayed and corrected, and adjustment notifications are pushed through the isolation channels of each tenant. The measured duration is written into the sample library to continuously update the delay distribution library, and the correction parameters are written back to the shared physical path conflict map.

[0008] Preferably, the construction process of the shared physical path conflict map is as follows: the passable paths of the work units in the storage area are divided into continuous path segments. The maximum parallel capacity of each path segment is determined based on the net width of the path segment, the safety margins on both sides, the standard vehicle width of a single work unit, and the safety clearance on one side, forming the capacity parameters of each path segment; a directed connection relationship is established between adjacent path segments based on the planar layout connectivity; the sequence of path segments that the four types of operations—receiving, shipping, cold chain, and cross-docking—must pass through in sequence is written into the shared physical path conflict map in the form of a directed path chain; the capacity parameters, the directed connectivity relationships, and the operation type path chains together constitute the complete topological structure of the shared physical path conflict map.

[0009] Preferably, the calibration process of the time delay distribution library is as follows: multiple rounds of on-site measurements are performed on each path segment under both load and no-load conditions, and the actual travel time of each measurement is recorded; the sample mean and sample dispersion of the travel time of each path segment under both load and no-load conditions are statistically analyzed to form the time delay distribution parameters of the path segment under the corresponding load condition; the time delay distribution parameters of each path segment are stored in the time delay distribution library with the path segment number and load condition as dual indexes; each hatch segment of the cold chain corridor is calibrated separately according to the same process, and the time delay distribution parameters are stored independently with the hatch number and load condition as indexes.

[0010] Preferably, the process of generating the capacity cascading release prediction table is as follows: when each work unit passes through a path node, the work unit number, the current path node number, and the time of passing the node are recorded. Based on the records of two consecutive node passes for the same work unit, the current path segment and the time of entering the path segment are determined. Based on the average delay distribution of the path segment under the current load state, the expected time of the work unit leaving the current path segment is calculated, which is used as the expected release time of the currently occupied capacity of the path segment. The calculation is performed for all work units in transit, and the expected release times of each path segment are summarized by path segment number and time dimension. Combined with the committed occupancy status recorded for each path segment, a capacity cascading release prediction table covering a preset time domain is generated.

[0011] Preferably, the process of generating a valid combination window is as follows: taking the candidate start time as the starting point for verification, based on the path node sequence determined in the shared physical path conflict graph according to the job type, verify whether the predicted available capacity corresponding to the candidate start time of the first node meets the requirements of this job; if it does, the mean of the delay distribution of the first node is accumulated to the candidate start time as the verification time of the second node, and the same verification is performed on the second node; this process is continued node by node along the path node sequence, and the mean of the delay distribution of the preceding node is accumulated on the verification time of the preceding node for each verification time; only when the predicted available capacity of all nodes in the path node sequence meets the requirements at their respective verification times is the candidate start time marked as a valid combination window; all valid combination windows are arranged in chronological order, and the one that is most recently issued through the tenant isolation channel is selected, with each window including the predicted available capacity status of each path node at the corresponding verification time.

[0012] Preferably, the data flow rules for the tenant isolation channel are as follows: When each tenant submits a job request, it only needs to report the job type identifier through the isolation channel; based on the job type identifier, the corresponding path node sequence in the shared physical path conflict graph is retrieved, and after completing the full-chain time-series coherence availability verification, the combined window information sent to the tenant isolation channel only includes the predicted available capacity status value and corresponding time of each path node; when the combined window is extended, the adjustment notification pushed to the affected tenant isolation channel only includes the affected path node number and the adjusted predicted available start time, without disclosing the attribution information of the job unit from which the deviation originates; no horizontal data flow occurs between the tenant isolation channels.

[0013] Preferably, the deviation triggering judgment and cascading correction process specifically involves: continuously acquiring the actual dwell time of each work unit on each path segment, comparing the actual dwell time with the mean of the delay distribution of the corresponding path segment and the corresponding load state in the delay distribution library; when the actual dwell time exceeds the sum of the mean and dispersion of the delay distribution of the corresponding state of the path segment, it is determined that a deviation has occurred in this passage, and the excess amount is extracted as the deviation increment; the registered expected release time of each node in the subsequent path node sequence of the work unit is sequentially updated with the deviation increment, and the predicted available capacity of each time slot of the corresponding path segment in the capacity cascading release prediction table is synchronously corrected; other tenant commitment occupancy records existing on the path nodes affected by the postponement correction are retrieved, and adjustment notifications are pushed through the corresponding tenant isolation channel for records with overlapping time sequences, and the notification content only includes the affected path node number and the suggested adjusted start time.

[0014] Preferably, the process of rolling updates to the latency distribution library and writing back to the shared physical path conflict map is as follows: After each operation chain is completed, the actual dwell time of each path segment in this operation is extracted from the node records of each path node, and written into the sample library of the corresponding path segment according to the path segment number and load status; the sample library of each path segment maintains a fixed-size set of the most recently measured samples in a first-in-first-out manner. After each new sample is written, the sample mean and dispersion are recalculated based on the current sample set, and the updated latency distribution parameters are written over the corresponding index entries in the latency distribution library; after the latency distribution library is updated, the updated latency distribution mean of each path segment is synchronously written back to the latency parameters of the corresponding path segment in the shared physical path conflict map to ensure that the latency benchmark of each path segment in the conflict map is consistent with the actual operating status; the updated conflict map and the latency distribution library together serve as the input benchmark for the generation of the capacity cascade release prediction table, the combined window verification, and the deviation judgment in the next cycle.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention constructs a dual-core model of "shared physical path conflict graph + latency distribution library" to uniformly express the originally discrete path resources, job type path chains, and capacity constraints within the park. This enables refined modeling and explicit conflict management for multi-tenant, multi-type jobs within the same physical path network. Compared to existing scheduling methods that rely on empirical rules or simple queuing mechanisms, this solution can accurately characterize capacity limits and travel latency characteristics at the path segment granularity, transforming resource conflict avoidance from post-event avoidance to pre-event computability and prediction. This significantly reduces path congestion and resource contention probability, improving overall traffic efficiency and system controllability.

[0016] 2. This invention proposes a capacity cascading release prediction and combined window generation mechanism. By dynamically tracking and calculating the average latency of in-transit work units, a capacity release prediction table covering the future time domain is formed. Based on this, the temporal coherence and availability of the path chain are verified node by node, and effective combined windows that meet the constraints of the entire link are selected. This mechanism breaks through the limitations of traditional scheduling that only schedules a single node or a single resource, and realizes full-chain collaborative optimization scheduling across path segments. It can effectively avoid the problem of "local feasibility, global conflict", significantly improve the foresight and overall optimality of scheduling decisions, and reduce duplicate queuing and invalid waiting.

[0017] 3. This invention achieves adaptive closed-loop optimization of actual operating conditions by introducing a deviation-triggered judgment and cascading correction mechanism, combined with a rolling update strategy for the sample library. When the actual dwell time deviates from the distribution range, the system can automatically adjust the release time of subsequent nodes and synchronously update the prediction table and tenant window to ensure consistency between scheduling results and actual execution. Simultaneously, by continuously introducing the latest sample update delay distribution parameters, the model can dynamically reflect equipment status, environmental changes, and load fluctuations, exhibiting stronger robustness and long-term adaptability compared to static parameter models, thereby improving system prediction accuracy and operational stability. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the smart park information management method provided by the present invention; Figure 2 A schematic diagram of the structure of a smart park information management system based on big data provided by the present invention; Figure 3 This is a schematic diagram of the data flow process of the multi-tenant isolation channel provided by the present invention; Figure 4 This is a schematic diagram of the deviation triggering determination and cascade correction process provided by the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0020] Example 1: Please see Figure 1 and Figure 2This invention provides a smart park information management system based on big data. The technical solution is as follows: a path status distribution module, used to construct a shared physical path conflict map based on the physical path layout of the park's storage area, and to calibrate the status delay distribution of each path segment using actual measurements, forming a delay distribution database; a path cascading release module, used to track the path node positions of each work unit in real time based on the shared physical path conflict map, and to calculate the expected release time of the capacity of each path segment in conjunction with the delay distribution database, generating a capacity cascading release prediction table; and a time-series combination window distribution module, used to receive the work types reported by tenants through the isolation channel. Based on the shared physical path conflict map, the corresponding path node sequence is determined. The availability of the entire chain's temporal continuity is checked node by node on the capacity cascade release prediction table. Combined windows that meet the conditions are issued through the tenant isolation channel. The path delay correction module is used to compare the actual dwell time of each path segment with the corresponding mean in the delay distribution library after the tenant confirms the combined window. When the deviation exceeds the distribution range, the predicted release time of the affected subsequent nodes is delayed and adjusted, and adjustment notifications are pushed to each tenant isolation channel. The measured duration is written into the sample library to continuously update the delay distribution library, and the correction parameters are written back to the shared physical path conflict map.

[0021] Furthermore, the construction process of the shared physical path conflict map is as follows: the passable paths of the work units within the storage area are divided into continuous path segments. The maximum parallel capacity of each path segment is determined based on the net width of the path segment, the safety margins on both sides, the standard vehicle width of a single work unit, and the safety clearance on one side, forming the capacity parameters of each path segment; a directed connection relationship is established between adjacent path segments based on the planar layout connectivity; the sequence of path segments that each of the four types of operations—receiving, shipping, cold chain, and cross-docking—must pass through in sequence is written into the shared physical path conflict map in the form of a directed path chain; the capacity parameters, the directed connectivity relationships, and the operation type path chains together constitute the complete topological structure of the shared physical path conflict map.

[0022] Specifically, during the system deployment phase, the administrator measures the clear width of the passageways for forklifts along each segment of the park's warehouse area based on the actual measured floor plan. The maximum parallel capacity of each path segment is determined based on the clear width of the passageway, the safety margins on both sides, the standard width of a single work unit, and the safety clearance on one side. The specific calculation process is as follows: First, subtract the safety margins on both sides from the clear width of the passageway to obtain the effective clear width; second, add twice the safety clearance on one side to the standard width of a single work unit to obtain the minimum passage width required by a single work unit; finally, divide the effective clear width by the minimum passage width and round down to determine the maximum parallel capacity of the path segment. When the effective clear width is less than the minimum passage width, the path segment is marked as prohibited, and its capacity parameter is recorded as zero in the conflict diagram. The specific values ​​for the side safety margins and single-sided safety clearance are determined based on the minimum safe operating clearance specified in the forklift model's parameter manual. The safety margin on the side closest to a fixed wall or rack structure is the minimum safe clearance between the forklift forks in their maximum extended state and the aisle boundary. When two adjacent forklifts are running parallel, the single-sided safety clearance between them is the minimum clearance width required to ensure safe passage for personnel at the specified operating speed. These parameters are entered into the system by the administrator during the deployment phase after consulting the corresponding model's parameter manual, without relying on estimations or empirical values. If the remaining width of the aisle after subtracting the side safety margins does not meet the minimum passage width requirement for accommodating a single forklift, the path segment is marked as prohibited and not included in the conflict diagram node.

[0023] The directed connectivity between adjacent path segments is determined based on the warehouse layout. The sequence of path segments that each of the four operation types—receiving, shipping, cold chain transshipment, and cross-warehouse transfer—must traverse is entered by the administrator according to the actual process routes of each operation, forming directed path chains for each operation type, which are then written into the conflict graph. The path segment capacity parameters, directed connectivity, and the directed path chains of the four operation types together constitute the complete topology of the shared physical path conflict graph, serving as the structural foundation for all subsequent scheduling calculations.

[0024] Furthermore, the calibration process of the time delay distribution library is as follows: multiple rounds of on-site measurements are conducted on each path segment under both load and no-load conditions, and the actual travel time of each measurement is recorded; the sample mean and sample dispersion of the travel time of each path segment under both load and no-load conditions are statistically analyzed to form the time delay distribution parameters of the path segment under the corresponding load condition; the time delay distribution parameters of each path segment are stored in the time delay distribution library with the path segment number and load condition as dual indexes; each hatch segment of the cold chain corridor is calibrated separately according to the same process, and the time delay distribution parameters are stored independently with the hatch number and load condition as indexes.

[0025] Specifically, the calibration of the time delay distribution library was completed before the system officially went live. Calibration personnel used forklifts actually in use within the park to conduct multiple rounds of on-site measurements on each path segment under both loaded and unloaded conditions. These measurements were taken with forklifts carrying typical single pallet weights of common goods types in the warehouse (specific weight values ​​were entered into the machine parameter table by the administrator based on the average pallet weight of the main goods categories). The actual travel time from the starting point of the path segment to the exit point was recorded each time. The number of measurement rounds was determined by the convergence of the statistical characteristics of the measured data: when, after adding a new round of measurement, the change in the sample mean under the current state of the corresponding path segment was less than 3% of the existing mean, and the change in the sample standard deviation was less than 5% of the existing standard deviation, the distribution parameters under that state were considered to have converged, and the number of rounds could be stopped.

[0026] After calibration, the system calculates the mean and standard deviation of all samples for each path segment under each load condition, storing them in the delay distribution database using both path segment number and load condition as dual indexes. The mean reflects the average throughput efficiency under that condition, while the standard deviation reflects the natural fluctuation range of throughput time. The system performs a distribution fit test on the sample data for each path segment. If the samples significantly deviate from a normal distribution and exhibit right skewness, the threshold for triggering the deviation judgment is adjusted from the basic one standard deviation upwards to 1.5 to 2 times to reduce invalid alarms caused by long-tail delays. Each hatch segment of the cold chain corridor is calibrated separately according to the same process and stored independently using hatch number and load condition as indexes. Each record in the delay distribution database contains two parameters: mean and standard deviation, which serve as the baseline input for subsequent capacity release time estimation, full-chain timing verification, and deviation judgment threshold calculation.

[0027] Furthermore, the process of generating the capacity cascading release prediction table is as follows: When each work unit passes through a path node, the work unit number, the current path node number, and the node passing time are recorded. Based on the records of two consecutive node passing times for the same work unit, the current path segment and the time of entering the path segment are determined. Based on the average delay distribution of the path segment under the current load state, the expected time when the work unit leaves the current path segment is calculated, which is used as the expected release time of the currently occupied capacity of the path segment. The calculation is performed for all work units in transit, and the expected release times of each path segment are summarized by path segment number and time dimension. Combined with the committed occupancy status recorded for each path segment, a capacity cascading release prediction table covering a preset time domain is generated. The capacity cascading release prediction table is generated during system initialization. Whenever a work unit passes through a path node, the incremental update of the relevant time slot is triggered. To prevent concurrent read and write conflicts, the system adopts a transaction locking mechanism for read and write operations on the prediction table. After any tenant requests to enter the verification process, the corresponding time slot is locked until the committed occupancy registration is completed.

[0028] Specifically, after the system is officially launched, it continuously acquires path node crossing events for each forklift in transit within the park. The system maintains a tracking record for each forklift in transit, including its most recent crossing node number and corresponding time. Whenever a forklift crosses a new node, the system updates the tracking record with the current node record, thereby determining the forklift's current path segment and the time it entered that segment. The system reads the current task identifier carried by the forklift, determines its load status based on the task stage (the travel stage before unloading is considered loaded, and the return to the platform after unloading is considered empty), retrieves the average value of the path segment under the current load status from the time delay distribution database, and calculates the expected time for the forklift to leave the path segment by adding the entry time to the average value, which is used as the expected release time of the currently occupied capacity share of the path segment.

[0029] The system performs the above calculations on all forklifts in transit, summarizing the estimated release times of all forklifts in transit for each path segment with the registered committed occupancy records, and generating a capacity cascading release prediction table with path segment number and time axis as dimensions. The time domain covered by the prediction table (i.e., the length of time projected into the future) is configured during the deployment phase based on the historical maximum completion time of the longest operation type (cross-docking), ensuring that the table always covers the remaining execution time of all committed operations that may still be in progress. The predicted available capacity of each path segment within each time slot is determined by subtracting the sum of the in-transit occupancy and committed occupancy in the current time slot from the maximum parallel capacity of that path segment; a negative value is truncated to zero. Each forklift passing a node triggers an update of the relevant time slot for the corresponding path segment.

[0030] Furthermore, the process of generating effective combination windows is as follows: taking the candidate start time as the starting point for verification, based on the path node sequence determined in the shared physical path conflict graph according to the job type, the predicted available capacity corresponding to the candidate start time of the first node is verified to meet the requirements of this job; if it meets the requirements, the mean of the delay distribution of the first node is accumulated to the candidate start time as the verification time of the second node, and the same verification is performed on the second node; this process is continued node by node along the path node sequence, and the mean of the delay distribution of the preceding node is accumulated on the verification time of the preceding node for each verification time; only when the predicted available capacity of all nodes in the path node sequence meets the requirements at their respective verification times is the candidate start time marked as an effective combination window; all effective combination windows are arranged in chronological order, and the one that is most recently issued through the tenant isolation channel is selected, with each window including the predicted available capacity status of each path node at the corresponding verification time.

[0031] Furthermore, the data flow rules for the tenant isolation channel refer to... Figure 3Specifically: When submitting a job request, each tenant only needs to report the job type identifier through the isolation channel, without requiring information on the product category, quantity, or commercial attributes of the consignor and consignee; based on the job type identifier, the corresponding path node sequence in the shared physical path conflict graph is retrieved, and after completing the full-chain temporal coherence and availability verification, the combined window information sent to the tenant's isolation channel only includes the predicted available capacity status value and corresponding time of each path node, without including the job information, identity information, or committed occupancy source information of other tenants; when the combined window is extended, the adjustment notification pushed to the affected tenant's isolation channel only includes the affected path node number and the adjusted predicted available start time, without disclosing the attribution information of the job unit from which the deviation originates; no horizontal data flow occurs between the tenants' isolation channels.

[0032] Specifically, when tenants submit work requests through their dedicated isolation channel, they only need to provide a work type identifier (one of four categories). For receiving and picking operations, tenants also need to provide the expected vehicle type arriving at the port. The system will then retrieve the estimated standard pallet loading capacity for that vehicle type from the vehicle model parameter table. This estimate is used to calculate the required pallet space in the temporary storage area, and is considered a physical parameter input, not related to the commercial attributes of the goods. No other commercial data is required.

[0033] Upon receiving a request, the system retrieves the directed path chain for that job type from the conflict graph to determine the path node sequence. Then, starting from the current time, it examines candidate start times one by one in five-minute increments: for the first node in the path node sequence, it checks whether the predicted available capacity within the corresponding time slot of the candidate time meets the requirements of this job; if so, it adds the mean of the delay distribution of the first node to the candidate start time to obtain the second node's examination time, and performs the same examination on the second node; this process continues along the node sequence, with each node's examination time based on the sum of the mean of the delay distribution of the preceding node; only when the predicted available capacity of all nodes in the path node sequence at their respective examination times is not less than one is the candidate start time marked as a valid combination window.

[0034] The system arranges the effective combination windows in chronological order and pushes the most recent effective windows to the tenant's isolation channel. The number of pushed windows is typically three (fewer than three may result in tenants lacking reasonable choices, while more than five increases the decision-making burden). System administrators can configure this within a range of one to five based on the park's scheduling habits. Each pushed window includes the predicted available capacity value for each path node at the corresponding inspection time, described as a physical state (e.g., predicted available capacity for a platform node: two), and does not contain information from any other tenants. No lateral data flow occurs between tenant isolation channels; when responding to each tenant's request, the system only reads the conflict graph, latency distribution library, and prediction table, without accessing any other tenant's data.

[0035] After a tenant confirms the selected window, the system deducts one portion of the predicted available capacity of all path nodes involved in the window within the corresponding time slot, marks them as committed occupancy, and immediately reflects this in the prediction table, affecting the window verification results of other tenants. This implicitly achieves cross-tenant resource conflict avoidance without exposing any tenant information.

[0036] Furthermore, the process of deviation triggering determination and cascade correction refers to... Figure 4 Specifically, the process involves: continuously acquiring the actual dwell time of each work unit on each path segment, comparing the actual dwell time with the mean of the delay distribution of the corresponding path segment and the corresponding load state in the delay distribution database; when the actual dwell time exceeds the sum of the mean and dispersion of the delay distribution of the corresponding state of the path segment, it is determined that a deviation has occurred in this passage, and the excess amount is extracted as the deviation increment; using the deviation increment, the registered expected release time of each node in the subsequent path node sequence of the work unit is sequentially updated with a delay, and the predicted available capacity of each time slot of the corresponding path segment in the capacity cascade release prediction table is simultaneously corrected; retrieving other tenant commitment occupancy records that already exist on the path nodes affected by the delay correction, and pushing adjustment notifications through the corresponding tenant isolation channel for records with overlapping time sequences, the notification content only includes the affected path node number and the suggested adjusted start time.

[0037] Specifically, during the execution of the tenant-confirmed combination window, the system continuously acquires the actual dwell time of the forklift on each path segment (determined by the difference between two consecutive node-passing event records). For the actual dwell time of each path segment, the system compares it with the mean and standard deviation of the corresponding load state in the latency distribution database: when the actual dwell time exceeds the mean plus one standard deviation, the system determines that there is a deviation in the passage of this path segment. The basis for selecting the mean plus one standard deviation as the judgment threshold is as follows: under the premise that the distribution of warehouse forklift path passage time is a single-peak distribution, passage within the mean is considered normal; passage between the mean and the mean plus one standard deviation is considered high but within the natural fluctuation range; and passage exceeding the mean plus one standard deviation is highly likely to be an abnormal delay caused by physical interference such as temporary avoidance, abnormal cargo handling, or equipment deceleration, which requires triggering correction. If the park management wishes to adjust the sensitivity, the multiplier can be set to between 0.5 and 2 times in the system configuration. The specific value should be determined after a trial operation period of no less than one month after the initial launch, and the correspondence between the deviation trigger frequency and the actual cascading effect occurrence rate should be statistically analyzed to determine the optimal configuration value.

[0038] After the system determines the deviation, it extracts the portion of the actual dwell time exceeding the threshold as the deviation increment. This increment is then used to sequentially update the registered expected release times of each node in the subsequent path node sequence for the forklift, simultaneously correcting the predicted available capacity of each time slot in the affected path segment in the prediction table. Subsequently, the system scans for records of committed occupancy by other parties within the corrected time period for the affected path nodes. For records with overlapping time sequences, an adjustment notification is pushed through the corresponding tenant's isolation channel. The notification only includes the affected path node number and the suggested adjusted available start time, without disclosing any information about the source of the deviation. The notified tenant independently decides within their isolation channel whether to accept the extension or reapply for a combined window, and the confirmation result is written back to the prediction table.

[0039] The expected release times of subsequent nodes are updated using the deviation increment. The update process includes conflict resolution steps: if the postponement results in the available capacity of a subsequent node being less than zero at the corrected time, the release time of that node is further postponed to the next time when available capacity is available, and the additional postponement amount is added to the deviation increment of all subsequent nodes. After the deviation determination is completed, the system synchronously updates the sample database and performs parameter write-back.

[0040] Furthermore, the process of continuously updating the latency distribution library and writing back the shared physical path conflict map is as follows: After each operation chain is completed, the actual dwell time of each path segment in this operation is extracted from the node records of each path node, and written into the sample library of the corresponding path segment according to the path segment number and load status; the sample library of each path segment maintains a fixed-size set of the most recently measured samples in a first-in-first-out manner. After each new sample is written, the sample mean and dispersion are recalculated based on the current sample set, and the updated latency distribution parameters are written over the corresponding index entries in the latency distribution library; after the latency distribution library is updated, the updated latency distribution mean of each path segment is synchronously written back to the latency parameters of the corresponding path segment in the shared physical path conflict map to ensure that the latency benchmark of each path segment in the conflict map is consistent with the actual operating status; the updated conflict map and the latency distribution library together serve as the input benchmark for the generation of the capacity cascade release prediction table, the combined window verification, and the deviation judgment in the next cycle.

[0041] Specifically, after each work chain is completed (triggered by the forklift performing the operation passing a node for the last time and returning to the standby area), the system extracts the actual dwell time of each path segment from all node-passing records for this operation, and adds it to the corresponding path segment's sample library according to the path segment number and load status. Each path segment's sample library maintains a fixed-size set of recently measured samples using a first-in, first-out (FIFO) approach. The sample library capacity is typically set to cover the number of operations under the same load status for that path segment within the past three months—a three-month window is sufficient to capture seasonal operational pattern changes (such as peak and off-peak freight times during holidays) while ensuring that old samples are replaced by new samples after three months, reflecting the evolution of the park's physical condition; system administrators can adjust the capacity within a range of one to six months based on the park's annual freight volume and seasonal peak and off-peak characteristics.

[0042] Each time a new sample is added, the system recalculates the mean and standard deviation based on the current sample set, overwrites the corresponding index entries in the time delay distribution library, and synchronously writes the updated mean of each path segment back to the time delay parameters of the corresponding path segment in the shared physical path conflict map. The updated conflict map and time delay distribution library take effect immediately, serving as the input benchmark for the next round of prediction table generation, combined window verification, and bias determination, completing a data-driven closed loop consisting of four innovative points: conflict map and time delay distribution construction → real-time tracking and prediction table generation → combined window verification and commitment registration → bias detection and cascade correction → sample addition and parameter write-back → next cycle conflict map and time delay distribution update.

[0043] This embodiment focuses on a multi-tenant shared warehousing hub scenario, constructing a data-driven smart park information management closed-loop system. The system first divides the actual physical paths of the warehousing area into path segments, determining the parallel capacity of each path segment by combining aisle width, safety margin, and forklift parameters. A shared physical path conflict graph is then constructed using path connectivity relationships and path chains of different work types, serving as the basis for scheduling. Subsequently, multiple rounds of field tests are conducted on each path segment under load and idle conditions to statistically analyze the mean and dispersion of travel time, forming a latency distribution database to provide a basis for subsequent predictions. During system operation, the location and status of each work unit are tracked in real time, and the capacity release time of the path segment is calculated based on the mean latency, generating a capacity cascading release prediction table. Based on this, for work requests submitted by tenants, the timing and capacity availability are verified node by node along the path chain, and a combination window that meets the constraints of the entire chain is selected and issued, achieving coordinated resource allocation. Simultaneously, the system uses an isolation channel mechanism to transmit only necessary physical status information, preventing data leakage between tenants. During execution, if the actual travel time exceeds the statistical distribution range, a deviation determination is triggered, and subsequent nodes are cascaded and adjusted accordingly. The prediction table is updated synchronously, and relevant tenants are notified, achieving dynamic adjustment. Finally, the system continuously writes actual operating data into the sample database, updates the latency distribution parameters on a rolling basis, and writes them back to the conflict graph, achieving adaptive optimization of the model. Overall, a closed-loop mechanism of "modeling—prediction—scheduling—correction—feedback" is formed, achieving efficient collaboration and dynamic optimization of park resources while ensuring multi-tenant isolation.

[0044] Example 2: Example 1 has fully described the entire process of constructing a shared physical path conflict graph and calibrating the latency distribution library, generating a capacity cascading release prediction table in real time, verifying the time sequence consistency of the entire chain and allocating combined windows, as well as the deviation-triggered cascading correction and rolling updates of the latency distribution library. This example adds the following functional modules to the above process.

[0045] The steps for verifying the temporal coherence and availability of the entire chain node by node also include: Based on the directed path chains of each job type, the system reads the dispersion of all path segments under the corresponding load state from the time delay distribution library, and sums up the dispersion of all path segments to obtain the total chain cumulative uncertainty of the directed path chain of that job type. Using the total chain cumulative uncertainty of the receiving and picking job path chain as the benchmark reference value, the ratio of the total chain cumulative uncertainty of other job types to the benchmark reference value is used as the search time domain expansion coefficient. When generating effective combination windows for each job type, the system sets the search range of the candidate start time to the base time domain multiplied by the search time domain expansion coefficient corresponding to the job type, so that job types with longer path chains and higher cumulative uncertainties obtain a relatively longer candidate window search range. The total chain cumulative uncertainty and the corresponding search time domain expansion coefficient are stored with the job type as the index, and are automatically recalculated and updated synchronously after each update of the time delay distribution library.

[0046] In the step of verifying the temporal coherence and availability of the entire chain node by node, the system sets the search range of candidate start times to the base time domain multiplied by the search time domain expansion coefficient. The base time domain is set during the deployment phase, with a value range of 2 to 8 hours; the search time domain expansion coefficient is limited to a value range of 0.6 to 2.0. If the calculated coefficient exceeds this range, it is truncated according to the boundary value to prevent the search range from being too large and thus wasting computing resources.

[0047] Specifically, after initial calibration or each rolling update of the time delay distribution database, the system performs full-chain cumulative uncertainty calculations for four types of operations: receiving, shipping, cold chain, and cross-warehouse. Taking cross-warehouse transfer as an example, its directed path chain sequentially includes several nodes such as the platform access section, the first section of the forklift aisle, the entrance section of the temporary storage transfer area, the second section of the forklift aisle, and the exit docking section. The system reads the standard deviation of each of the above path segments under load conditions from the time delay distribution database one by one, and adds up the values ​​sequentially. The total obtained is the full-chain cumulative uncertainty for this type of operation. The same calculation result for receiving operations is used as the benchmark reference value. The system divides the full-chain cumulative uncertainty of each type of operation by the benchmark reference value to obtain the search time-domain expansion coefficient for each, and stores it in the conflict graph auxiliary parameter table with the operation type as the index. Cross-warehouse transshipment routes typically contain more path segments than receiving routes, with an expansion coefficient greater than one, corresponding to a search time domain longer than the base time domain. Outbound shipping routes are relatively shorter, with an expansion coefficient close to or even less than one, and a search time domain comparable to or slightly shorter than the base time domain. When a tenant submits a cross-warehouse transshipment request, the system reads the search time domain expansion coefficient corresponding to that type, multiplies the base time domain by this coefficient, and uses it as the upper bound of the search range for the candidate start time. This ensures that long-chain operation types always have sufficient effective combination window candidates to push within the search range, eliminating the situation where long-chain operation types have zero pushable windows due to insufficient fixed search time domains. The search time domain expansion coefficient is automatically recalculated and synchronously written after each update of the time-delay distribution database, maintaining consistency with physical parameters.

[0048] The steps of continuously comparing the actual dwell time of the path segments and detecting deviations also include: Each time a deviation trigger event occurs, the system writes the trigger path segment number and the corresponding time period identifier into the deviation pattern library. The time period identifier is determined by the combination of the hourly time period and the day of the week in which the trigger time occurs. The deviation pattern library uses the path segment number and the time period identifier as a composite index to cumulatively record the number of deviation triggers for each time period and the total number of operations passing through during the same time period. The system periodically calculates the deviation trigger rate for each path segment and each time period. When the deviation trigger rate for a certain path segment and a certain time period exceeds a preset multiple of the average deviation trigger rate for the entire path segment, the time period is marked as a high-risk time period for that path segment and stored in the high-risk time period marker library. The preset multiple threshold is determined based on historical distribution after the system has accumulated at least three months of deviation records, with the constraint that the number of marked high-risk time periods does not exceed one-quarter of the total number of time periods in the day. When generating an effective combination window, if the node verification time of any path segment in the path node sequence falls into the candidate start time of the high-risk time period of that path segment, the system directly skips it without verification and pre-removes historical high-risk time periods from the candidate set.

[0049] Specifically, based on the deviation detection mechanism, the system simultaneously writes the combined identifier of the triggering path segment number, the hour (in whole hours), and the day of the week (Monday to Sunday) into the deviation pattern library each time a deviation trigger event occurs. The deviation pattern library uses the path segment number and time period identifier as a composite key, and two counters: the cumulative number of deviation triggers and the total number of operations passing through during the same time period. The total number of operations passing through is determined by the count of forklift passing through nodes within the corresponding time period, while the deviation trigger count only includes events that pass the deviation judgment in that specific instance. At a fixed time each day (such as the midnight off-peak period), the system calculates the deviation trigger rate for each path segment and each time period based on the latest cumulative data in the deviation pattern library, and calculates the average deviation trigger rate for all time periods for each path segment. The preset multiple threshold is determined after accumulating three months of deviation records following the initial system launch, using the following method: The deviation trigger rates for each path segment and time period are sorted from high to low. The multiple value corresponding to the time period in the top quarter that is precisely marked as a high-risk time period is determined, and this value is written into the system configuration as the preset multiple threshold. If the required multiples for different path segments are inconsistent, the minimum value of the multiples obtained for each path segment is taken, and the highest-risk mode of the most stringent path segment is used to protect global consistency. After the threshold is determined, the system marks time periods with deviation trigger rates exceeding the average of the path segment multiplied by the preset multiple as high-risk, and stores them in the high-risk time period marker library. When generating an effective combination window, the system adds an exclusion step: For each node in the path node sequence, the time period identifier of its test time is calculated, and the high-risk time period marker library is searched. If the node is marked as high-risk in the time period corresponding to its test time, the current candidate start time is skipped directly, and the next longer candidate time test is performed without performing capacity availability testing, thus preventing high-deviation time periods from being included in the effective window from the allocation source.

[0050] The steps for generating the capacity cascade release prediction table also include: While maintaining the capacity cascading release prediction table, the system counts the number of currently valid commitments occupied in each time slot of each path segment in real time to obtain the concurrent commitment density of that path segment and time slot. When the concurrent commitment density of a certain time slot in a certain path segment reaches a preset proportion of the maximum parallel capacity of that path segment, the system deducts a portion of the predicted available capacity of that time slot as buffer capacity, so that the time slot presents a lower available capacity value for subsequent combination window checks. The preset proportion threshold is determined based on the statistical comparison relationship of deviation trigger rate under different commitment density intervals after the system has accumulated no less than two months of commitment density and cascading correction event corresponding records. When the concurrent commitment density in a certain time slot drops below the preset proportion due to commitment completion or cancellation, the system automatically releases the corresponding capacity reservation and restores that portion of capacity to an available state.

[0051] In the step of generating the capacity cascading release prediction table, when the concurrent commitment density of a certain path segment reaches a preset proportion of the maximum parallel capacity, the system deducts one unit of buffer capacity. The preset proportion threshold ranges from 0.5 to 0.9. By monitoring the deviation trigger rate under different density intervals, when the trigger rate is significantly higher than twice that of the low-density benchmark, the proportion corresponding to that density is determined as the effective threshold.

[0052] Specifically, while maintaining the capacity cascading release prediction table, the system maintains a concurrent commitment counter for each time slot of each path segment. Whenever a commitment is added to a time slot of a path segment due to a tenant confirming a valid combination window, the concurrent commitment counter for that time slot of that path segment is incremented by one; it is decremented when a commitment is fulfilled or canceled. After each update of the concurrent commitment counter, the system compares the current count with a preset ratio of the maximum parallel capacity of that path segment: if the count value is greater than or equal to the integer of the maximum parallel capacity multiplied by the preset ratio (rounded up), the system subtracts one unit from the predicted available capacity of that time slot of that path segment in the prediction table and marks it as buffer capacity. This buffer capacity does not respond to any capacity demand queries for subsequent window checks and is automatically released and restored to available capacity only when the concurrent commitment count drops below a threshold. The preset ratio threshold is determined after the system has accumulated more than two months of records corresponding to the commitment density and cascaded correction events: the deviation trigger rate of each path segment at each concurrent commitment density level is extracted, the concurrent commitment density value corresponding to the point where the deviation trigger rate begins to be significantly higher than the low density benchmark (specifically, it is judged as exceeding the low density benchmark by more than twice) is found, and the density value is divided by the maximum parallel capacity of the path segment to obtain the preset ratio threshold, which is written into the system configuration.

[0053] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A smart park information management system based on big data, characterized in that, include: The path status distribution module is used to construct a shared physical path conflict map based on the physical path layout of the park's storage area, and to mark the status delay distribution of each path segment by actual measurement to form a delay distribution library. The path cascading release module is used to track the location of each work unit's path node in real time based on a shared physical path conflict map, and calculate the expected release time of each path segment's capacity by combining the delay distribution library, and generate a capacity cascading release prediction table. The timing combination window distribution module is used to receive the job types reported by tenants through the isolation channel, determine the corresponding path node sequence based on the shared physical path conflict diagram, check the full chain timing coherence availability node by node on the capacity cascade release prediction table, and distribute the combination windows that meet the conditions through the tenant isolation channel. The specific data flow rules for the tenant isolation channel are as follows: When each tenant submits a job request, it only needs to report the job type identifier through the isolation channel; based on the job type identifier, the corresponding path node sequence in the shared physical path conflict graph is retrieved, and after completing the full-chain time-series coherence availability verification, the combined window information sent to the tenant isolation channel only includes the predicted available capacity status value and corresponding time of each path node; when the combined window is extended, the adjustment notification pushed to the affected tenant isolation channel only includes the affected path node number and the adjusted predicted available start time, without disclosing the attribution information of the job unit from which the deviation originates; no horizontal data flow occurs between the tenant isolation channels. The path delay correction module is used to compare the actual dwell time of each path segment with the corresponding average value in the delay distribution library after the tenant confirms the combined window. When the deviation exceeds the distribution range, the predicted release time of the affected subsequent nodes is delayed and corrected, and adjustment notifications are pushed through the isolation channels of each tenant. The measured duration is written into the sample library to continuously update the delay distribution library, and the correction parameters are written back to the shared physical path conflict map.

2. The smart park information management system based on big data according to claim 1, characterized in that: The construction process of the shared physical path conflict map is as follows: the passable paths of the work units in the storage area are divided into continuous path segments. The maximum parallel capacity of each path segment is determined based on the net width of the path segment, the safety margins on both sides, the standard vehicle width of a single work unit, and the safety clearance on one side, forming the capacity parameters of each path segment; based on the connectivity relationship of the planar layout, directed connectivity relationships are established between adjacent path segments; the sequence of path segments that the four types of operations—receiving, shipping, cold chain, and cross-docking—must pass through in sequence are written into the shared physical path conflict map in the form of directed path chains; the capacity parameters, directed connectivity relationships, and operation type path chains together constitute the complete topology of the shared physical path conflict map.

3. The smart park information management system based on big data according to claim 1, characterized in that: The calibration process of the time delay distribution library is as follows: multiple rounds of on-site measurements are conducted on each path segment under both load and no-load conditions, and the actual travel time of each measurement is recorded; the sample mean and sample dispersion of the travel time of each path segment under both load and no-load conditions are statistically analyzed to form the time delay distribution parameters of the path segment under the corresponding load condition; the time delay distribution parameters of each path segment are stored in the time delay distribution library with the path segment number and load condition as dual indexes; each hatch segment of the cold chain corridor is calibrated separately according to the same process, and the time delay distribution parameters are stored independently with the hatch number and load condition as indexes.

4. The smart park information management system based on big data according to claim 1, characterized in that: The process of generating the capacity cascading release prediction table is as follows: When each work unit passes through a path node, the work unit number, the current path node number, and the time of passing the node are recorded. Based on the records of two consecutive node passes for the same work unit, the current path segment and the time of entering the path segment are determined. Based on the average delay distribution of the path segment under the current load state, the expected time of the work unit leaving the current path segment is calculated, which is used as the expected release time of the currently occupied capacity of the path segment. The calculation is performed for all work units in transit, and the expected release times of each path segment are summarized by path segment number and time dimension. Combined with the committed occupancy information recorded for each path segment, a capacity cascading release prediction table covering the preset time domain is generated.

5. The smart park information management system based on big data according to claim 1, characterized in that: The process of generating an effective combination window is as follows: taking the candidate start time as the starting point for testing, based on the sequence of path nodes determined in the shared physical path conflict graph according to the job type, the predicted available capacity corresponding to the candidate start time of the first node is tested to see if it meets the requirements of this job; if it does, the mean of the delay distribution of the first node is accumulated to the candidate start time as the test time of the second node, and the same test is performed on the second node. The process proceeds node by node along the path node sequence in this manner, with each test time being based on the test time of the preceding node plus the mean of the delay distribution of the preceding node; only when the predicted available capacity of all nodes in the path node sequence meets the requirements at their respective test times is the candidate starting time marked as a valid combination window. All valid combination windows are arranged in chronological order, and the one that was most recently distributed through the tenant isolation channel is selected. Each window includes the predicted available capacity status of each path node at the corresponding verification time.

6. The smart park information management system based on big data according to claim 1, characterized in that: The process of deviation triggering judgment and cascading correction is as follows: continuously acquire the actual dwell time of each work unit in each path segment, compare the actual dwell time with the mean of the delay distribution of the corresponding path segment and the corresponding load state in the delay distribution library; when the actual dwell time exceeds the sum of the mean of the delay distribution of the corresponding state of the path segment and the degree of dispersion, it is determined that a deviation has occurred in this passage, and the excess amount is extracted as the deviation increment; use the deviation increment to sequentially update the registered expected release time of each node in the subsequent path node sequence of the work unit, and synchronously correct the predicted available capacity of each time slot of the corresponding path segment in the capacity cascading release prediction table. Retrieve existing tenant commitment occupancy records on path nodes affected by the extension correction. For records with overlapping time sequences, push adjustment notifications through the corresponding tenant isolation channel. The notification content only includes the affected path node number and the suggested start time after adjustment.

7. The smart park information management system based on big data according to claim 1, characterized in that: The process of rolling updates to the latency distribution library and writing back to the shared physical path conflict map is as follows: After each operation chain is completed, the actual dwell time of each path segment in this operation is extracted from the node records of each path node, and written into the sample library of the corresponding path segment according to the path segment number and load status; Each path segment sample library maintains a fixed-size set of the most recently measured samples in a first-in-first-out manner. After each new sample is written, the sample mean and dispersion are recalculated based on the current sample set, and the updated latency distribution parameters are written over the corresponding index entries in the latency distribution library. After the latency distribution library is updated, the updated latency distribution average of each path segment will be synchronously written back to the latency parameters of the corresponding path segment in the shared physical path conflict map to ensure that the latency benchmark of each path segment in the conflict map is consistent with the actual operating status. The updated conflict graph and the time delay distribution library together serve as the input benchmark for generating the next cycle capacity cascade release prediction table, verifying the combined window, and determining the deviation.

Images