A Smart Flower Operation Management System and Method Based on MES Scheduling

Abstract

This invention relates to the field of flower transaction management technology, providing a smart flower operation management system and method based on MES scheduling. The MES-based smart flower operation management system includes a subsystem for barrelled flowers, a subsystem for boxed flowers, a staggered scheduling and warehousing collaboration system, and an outbound scheduling controller. Barrelled flowers are processed entirely on the first floor, with a state machine forcibly regulating the batch state sequence changes, supporting both physical and virtual warehouse modes and three business modes. Barrelled flowers undergo full inspection and warehousing on the first floor, while boxed flowers undergo random inspection and grading on the second floor. Local AI inference for grading is achieved through a 5G edge gateway, and the stacker crane edge controller achieves distributed collision avoidance via 5G-D2D. The staggered scheduling system is located at the barrelled flower warehousing end, eliminating congestion through dynamic time window calculation and a dual scheduling strategy. This invention solves three major problems: congestion during flower warehousing warehousing warehousing, grading delays, and mismatch between auction and logistics times, thereby improving warehousing throughput and outbound efficiency.

Description

Technical Field

[0001] This invention relates to the field of flower transaction management technology, and in particular to a smart operation management system and method for flowers based on MES scheduling. Background Technology

[0002] Flower auction transactions have strict time requirements. From arrival at the warehouse, grading, warehousing, and arrangement to the auction, the entire preparation process must be completed before a fixed time on the same day. However, existing flower operation and management systems have technical deficiencies in the following three core aspects: In the fresh flower (especially barrelled flowers) warehousing process, when multiple distribution lines converge on the main conveyor line simultaneously, existing technology uses a passive triggering method for merging. This means that after the flower barrels reach the buffer zone, they are released randomly or according to a simple first-come, first-served principle. This leads to alternating periods of congestion and idleness at the merging points. Congestion causes the main line to overload and stop, while idleness results in energy waste as the main line idles, leading to low overall warehousing efficiency. Furthermore, it cannot guarantee that flower barrels merge into the main line at a constant interval, affecting the smooth reception of the downstream warehousing system.

[0003] The current grading process for boxed or barrelled flowers relies on uploading images to a central server for AI inference, resulting in significant network transmission latency. Simultaneously, the stacker crane's path planning and collision avoidance depend entirely on a central scheduling system; when the network fluctuates or the central node fails, the entire system is prone to paralysis. Furthermore, after grading, flowers from the same batch cannot be centrally stored in the warehouse, forcing the stacker crane to traverse multiple aisles and retrieve goods multiple times during outbound processing, leading to low retrieval efficiency.

[0004] Flower auctions typically end at a fixed time each day, but there's an inherent conflict between logistics and the auction window. After a flower is sold and shipped from the central warehouse, loading often cannot be completed on the same day or delivered the next day, resulting in long waiting times for buyers. Current technology lacks a mechanism to handle logistics and auctions in parallel, and the outbound process cannot consolidate and deliver multiple orders from the same buyer, leading to low efficiency. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a smart operation management system and method for flowers based on MES scheduling.

[0006] The technical solution adopted by this invention to solve its technical problem is: a smart flower operation and management system based on MES scheduling, comprising: MES serves as a unified central dispatch platform; The barrel flower operation subsystem includes a vertical warehouse, a filling machine, a visual grading system, a transaction system, and a closed-loop cleaning line. In the barrel flower operation subsystem, the entire process of empty barrel delivery, filling, height adjustment, flower placement, grading, warehousing, auction, delivery, picking up, empty barrel return, cleaning, and return to the warehouse is completed on the first floor. The boxed flower operation subsystem includes an inlet, an outlet, a sampling and sorting line, intelligent grading equipment, a return line, a straight line, an automated warehouse, and a stacker crane. In the boxed flower operation subsystem, the boxed flowers complete sampling and grading, warehousing, and pickup on the second floor. The staggered scheduling and warehousing coordination system is set at the inbound end of the bottled flower operation subsystem, including multiple distribution lines, one main roller distribution line, buffer zones and actuators set at the end of each distribution line, and a central scheduling controller; The outbound dispatch controller communicates with the MES and the central dispatch controller. In the barrel flower operation subsystem, MES uses a state machine to drive the state transition of the data model centered on the flower collection batch, and supports both physical library mode and virtual library mode.

[0007] By defining separate subsystems for barrelled flowers, boxed flowers, off-peak scheduling and warehousing collaboration, and outbound scheduling controllers, an operational management framework covering all categories of fresh flowers is established, providing a foundation for the collaborative work of subsequent functional modules and ensuring the unified scheduling and synchronization of information flow and physical flow.

[0008] Furthermore, in the staggered scheduling and warehouse collaboration system, sensors are installed at both the inlet and outlet of the buffer zone. The sensor at the inlet is used to detect the entry of the flower barrel and trigger a reporting event, while the sensor at the outlet is used to detect that the flower barrel has been successfully released into the main roller flow line. The central scheduling controller acquires the status of all buffer zones and the speed of the main roller streamline in real time, dynamically calculates the release time window, and issues release commands to the actuators according to the priority scheduling strategy or the batch collaborative scheduling strategy; the length of the release time window is calculated based on the ratio of the preset safety distance to the current running speed of the main roller streamline.

[0009] By using buffer sensors to detect the arrival and release status of flower buckets in real time, the central dispatch controller dynamically calculates the release time window and issues release instructions according to priority or batch coordination strategies, so that flower buckets merge into the main line at a constant interval, eliminating clustering and congestion, and improving warehousing efficiency and main line utilization.

[0010] Furthermore, the boxed flower operation subsystem also includes a 5G edge gateway located next to the intelligent grading device and an edge controller located in the stacker crane control cabinet. The 5G edge gateway is used for local AI inference to complete the grading of the boxed flowers and only uploads the grading results to the MES. The edge controllers of multiple stackers achieve distributed path planning and real-time collision avoidance through 5G-D2D direct communication. When the 5G network is interrupted, the 5G edge gateway and edge controller can still independently complete the grading and collision avoidance control, and automatically synchronize with the MES after the network is restored.

[0011] By deploying 5G edge gateways to complete AI classification locally and only uploading the classification results, the response time is greatly shortened. At the same time, the stacker crane edge controller achieves distributed collision avoidance through 5G-D2D direct communication. When the network is interrupted, each node can still operate independently and synchronize afterwards.

[0012] Furthermore, in the barrel flower operation subsystem, the state machine mandates that the status of the flower collection batches change unidirectionally in the following order: pending collection, collected, graded, grouped, auctioned, auction in progress, and completed, and changes sequentially in this order. In physical warehouse mode, graded flower barrels enter the automated storage system and then participate in the auction; in virtual warehouse mode, graded flower barrels are pre-shipped and enter the distribution and transportation system, and are simultaneously marked as virtual warehouse status in MES to participate in the auction in parallel, realizing the principle of first shipment, auction in transit, and immediate pickup upon arrival.

[0013] The state machine mandates that batch states change unidirectionally in a fixed order, ensuring process traceability. The physical warehouse mode is suitable for short-distance transactions, while the virtual warehouse mode enables first-out, in-transit auction, and immediate pickup upon arrival, effectively shortening the buyer's waiting time.

[0014] Furthermore, the outbound scheduling controller receives buyer orders, aggregates all batches ordered by the buyer by buyer ID, queries the batch-aisle mapping table, and if all batches are located in the same aisle, generates a merged pickup instruction for the stacker crane to pick up the goods at once; dynamically selects an available pickup port and directs all batches of flower barrels or boxed flowers to the same pickup port for centralized loading; when too many of the same batches cause congestion at a pickup port, dynamically allocates 2 to 3 pickup ports as dedicated pickup ports for the same buyer.

[0015] The outbound dispatch controller aggregates orders by buyer ID and merges pickup instructions for batches within the same aisle, avoiding multiple cross-aisle operations by the stacker crane. At the same time, it dynamically allocates pickup ports, and can allocate 2-3 dedicated port groups when congestion occurs, enabling the same buyer to load the truck in one go, greatly reducing the number of manual round trips and pickup port switching time.

[0016] A method for intelligent flower operation management based on MES scheduling, applied to the intelligent flower operation management system based on MES scheduling described above, includes the following steps: MES issues an empty barrel outbound instruction. After the empty flower barrels are filled, height adjusted, flowers placed, and visually graded, they are stored in the warehouse. After the auction is completed, they are outbound and picked up. The empty barrels are cleaned in a closed-loop cleaning line and then returned to the warehouse. Boxed flowers enter through the warehouse entrance. The boxed flowers that are sampled are transported to the intelligent grading equipment to complete the AI ​​grading and then converge through the return line. The boxed flowers that are not sampled pass directly through the straight line. Finally, they are sent to the vertical warehouse. The staggered scheduling of flower barrels before they enter the warehouse is executed by the central scheduling controller. It obtains the status of all branch line buffer zones and the speed of the main roller streamline in real time, dynamically calculates the release time window, and releases flower barrels to the main roller streamline in sequence according to the scheduling strategy.

[0017] The processing paths for bucket-packaged and box-packaged flowers are clearly separated. At the same time, staggered scheduling is embedded in the front end of the bucket-packaged flower warehousing, realizing the fully automated management of the entire process from empty buckets leaving the warehouse to cleaning and returning them to the warehouse, and from box-packaged flowers entering the warehouse to storage in the automated warehouse.

[0018] Furthermore, the scheduling strategy includes a priority scheduling strategy and a batch cooperative scheduling strategy: In the priority scheduling strategy, the central scheduling controller calculates the priority of each branch line according to the formula: priority = buffer occupancy rate weight × buffer occupancy rate + waiting time coefficient weight × waiting time coefficient + historical allocation balance coefficient weight × historical allocation balance coefficient. In the batch collaborative scheduling strategy, when multiple flower barrels with the same batch identifier are detected at the head of the buffer queue of the same diversion line, a continuous release time window is allocated to the batch, so that it can continuously merge into the main roller flow line with the minimum safety interval and be stored in the continuous storage location of the same lane.

[0019] The priority scheduling strategy dynamically calculates the priority of each diversion line through three factors: buffer occupancy rate, waiting time, and historical allocation balance coefficient, ensuring fairness and efficiency during peak periods. The batch collaborative scheduling strategy allocates continuous release time windows to the same batch of flower buckets, allowing them to merge into the main line with the minimum safe interval and be stored in continuous storage locations in the same lane, realizing centralized storage of batches and improving outbound retrieval efficiency.

[0020] Furthermore, the automatic grading process for boxed flowers includes: The 5G edge gateway runs an AI model to complete the level determination locally and generates structured results that include batch, level and confidence level; When the confidence level is lower than the preset threshold, the edge gateway requests the MES to perform a second review or manual intervention; otherwise, it directly adopts the local result and uploads the level data to the MES. The grade data will be automatically synchronized and updated to other batches of the same variety and grade under the same supplier that have not been sampled. The MES central cloud periodically retrains the AI ​​model based on the data reported by the edge nodes and pushes the updated model to each edge gateway.

[0021] The 5G edge gateway completes image inference and rating determination locally, and only requests MES review when the confidence level is below the threshold, which greatly reduces network dependence. The rating data is automatically synchronized to all unrated batches of the same type from the same supplier to avoid duplicate rating. The MES central cloud regularly retrains the model based on edge node data and pushes updates to achieve edge-cloud collaborative evolution.

[0022] Furthermore, the closed-loop cleaning process for the flower barrels includes: after delivery, the empty flower barrels first pass through the tipping and water-pouring station to pour out the residual liquid, then pass through the tipping internal cleaning station for high-pressure water and atomized cleaning agent spraying cleaning, then pass through the tipping preliminary drying station for the first round of drying, and finally pass through the straightening deep drying station for the second round of deep drying; after drying, the qualified empty flower barrels are returned to the vertical warehouse to await the next dispatch by MES.

[0023] Empty flower barrels pass through four stations in sequence: tipping and water pouring, internal high-pressure spray cleaning, preliminary drying, and deep drying. Each station is accompanied by a tipping action to change the posture of the flower barrel. After the flower barrels pass the cleaning, they are returned to the vertical warehouse to await the next dispatch, forming a complete closed loop cycle to ensure the reliability of the returned flower barrels for reuse.

[0024] Furthermore, the barrel-type flower operation subsystem supports three business models: In the standard mode, after the classification is completed, the goods are stored in the physical warehouse, then grouped, auctioned, distributed and transported, and picked up. In the pre-out mode, after classification, the virtual warehouse is marked and pre-out shipment is carried out. During transportation, batches in the virtual warehouse are grouped and auctioned in parallel. After the auction, the virtual warehouse is released. After the physical goods arrive at the destination pick-up point, they are collected and picked up. In the mixed mode, some flower barrels are processed according to the standard mode, while others are processed according to the pre-shipment mode, with both batches of goods being executed in parallel.

[0025] The standard mode is suitable for flowers that arrive on time before the auction. After grading, they are put into storage before the auction. The pre-shipment mode is suitable for flowers that cannot arrive on time before the auction. After grading, they are pre-shipped immediately, and the virtual warehouse participates in the auction in parallel to achieve immediate pickup upon arrival. The hybrid mode can be flexibly allocated according to the distribution of buyers. Some are processed in the standard mode and some in the pre-shipment mode in parallel to meet the complex scenario requirements of large-volume, multi-destination transactions.

[0026] The beneficial effects of this invention are: This invention utilizes a staggered scheduling and warehousing collaboration system. The central scheduling controller monitors the status of all branch line buffer zones and the speed of the main line in real time, dynamically calculates release time windows, and allocates release permissions according to priority or batch collaboration strategies, enabling the flower buckets to orderly merge into the main roller flow line at a constant interval. Compared with the existing passive triggering paralleling technology, it eliminates the "crowding" congestion and "idle window" phenomenon at the paralleling port, reduces the idling rate and conveying downtime, and significantly improves the inbound processing volume per unit time. This invention enables AI inference and classification to be completed locally via a 5G edge gateway located next to the intelligent classification device, significantly reducing classification response time. Distributed path planning and real-time collision avoidance are achieved through 5G-D2D direct communication between stacker crane edge controllers, significantly reducing stacker crane command latency. Even when the 5G network is interrupted, each edge node can still independently complete classification and collision avoidance control, automatically synchronizing upon network recovery. Simultaneously, through a batch collaborative scheduling strategy, the same batch of flowers converges into the main line and is stored in consecutive locations within the same aisle within a continuous time window, achieving centralized storage of batches within the aisle. This eliminates the need for multiple retrievals across aisles during outbound operations, greatly improving retrieval efficiency.

[0027] This invention enforces the standardized flow sequence of flower collection batches through a MES state machine, ensuring full traceability. Using a virtual warehouse model, graded flower barrels can be pre-shipped and enter distribution transportation, while simultaneously being marked as virtual warehouses in the MES to participate in auctions in parallel. This achieves first-out, in-transit auction, and immediate pickup upon arrival, effectively resolving the conflict between logistics time and fixed auction times, and shortening buyer waiting time. The outbound scheduling controller aggregates orders by buyer ID, merges pickup instructions from the same lane batch, and dynamically allocates pickup points, enabling centralized loading of multiple orders from the same buyer, significantly reducing manual back-and-forth trips and pickup point switching time. Furthermore, the system supports standard, pre-outbound, and hybrid business modes, flexibly adapting to local, intercity, and large-volume transaction scenarios. Attached Figure Description

[0028] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0029] Figure 1 This is a schematic diagram of the existing parallel connection timing technology.

[0030] Figure 2 This is a schematic diagram of the overall staggered scheduling and warehouse collaboration system of the present invention.

[0031] Figure 3 This is a diagram showing the overall layout of the off-peak parallel scheduling and warehousing coordination system of the present invention.

[0032] Figure 4 This is a schematic diagram of the parallel scheduling and warehousing coordination system of the present invention.

[0033] Figure 5 This is a block diagram illustrating the principle of the central dispatch controller in the staggered peak scheduling and warehouse collaboration system of this invention.

[0034] Figure 6 This is a flowchart of the operation of the staggered scheduling and warehouse collaboration system of the present invention.

[0035] Figure 7 This is a timing diagram of the near-term allocation of inbound batches in the staggered peak scheduling and warehousing collaboration system of this invention.

[0036] Figure 8 This is a flowchart of the outbound aggregation and same-port scheduling process of the staggered peak scheduling and warehouse collaboration system of the present invention.

[0037] Figure 9 This is a hierarchical architecture diagram of the barrel-packaged flower operation subsystem of the present invention.

[0038] Figure 10 This is a schematic diagram illustrating the synchronization of information flow and physical flow in the barrel-packaged flower operation subsystem of this invention.

[0039] Figure 11 This is a flow chart of the operation subsystem of the barrel-packaged flower business of the present invention.

[0040] Figure 12 This is a distribution diagram of the barrel-packaged flower operation subsystem of the present invention.

[0041] Figure 13 This is a schematic diagram of the structure of the flower bucket in the flower bucket operation subsystem of the present invention.

[0042] Figure 14 This is a schematic diagram of the filling machine in the barrel-type flower operation subsystem of the present invention.

[0043] Figure 15 This is a flowchart of the operation subsystem for the barrel-packaged flower business of the present invention.

[0044] Figure 16 This is a flow chart of the flower collection batches in the barrel flower operation subsystem of this invention.

[0045] Figure 17 This is a diagram showing the status of auction orders in the barrel-packaged flower operation subsystem of this invention.

[0046] Figure 18 This is a flow status diagram of the collection and distribution transportation link in the barrel flower operation subsystem of the present invention.

[0047] Figure 19 This is a flowchart of one form of the process from rating to pickup in the barrel-packaged flower operation subsystem of the present invention.

[0048] Figure 20 This is a flowchart of another form of the process from rating to pickup in the barrel-packaged flower operation subsystem of the present invention.

[0049] Figure 21 This is the overall architecture diagram of the boxed flower operation subsystem of the present invention.

[0050] Figure 22 This is an overall layout diagram of the boxed flower operation subsystem of the present invention.

[0051] Figure 23 This is a schematic diagram of the internal structure of the intelligent grading device in the boxed flower operation subsystem of the present invention.

[0052] Figure 24 This is a flowchart of the operation subsystem for boxed flower packaging of the present invention.

[0053] Figure 25 This is a flowchart of step one in the operation of the boxed flower operation subsystem of the present invention.

[0054] Figure 26 This is a flowchart showing the specific steps two to four in the operation of the boxed flower operation subsystem of the present invention.

[0055] Figure 27 This is a schematic diagram of step four in the operation of the boxed flower operation subsystem of the present invention.

[0056] Figure 28 This is a flowchart of step five in the operation of the boxed flower operation subsystem of the present invention.

[0057] Figure 29 This is a flowchart of step six in the operation of the boxed flower operation subsystem of the present invention. Detailed Implementation

[0058] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0059] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, features defined with "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0060] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0061] like Figure 1 As shown, existing passively triggered paralleling technology suffers from alternating periods of congestion and idleness. To address these issues, a smart flower operation management system and method based on MES scheduling is proposed. Figure 2 , Figure 11 and Figure 20 As shown, the intelligent flower operation management system based on MES scheduling includes a Manufacturing Execution System (MES) as a unified central scheduling platform, a subsystem for operating barrel-packaged flowers, a subsystem for operating box-packaged flowers, a staggered scheduling and warehousing collaboration system, and an outbound scheduling controller. The MES communicates with the barrel-packaged flower operation subsystem, the box-packaged flower operation subsystem, the staggered scheduling and warehousing collaboration system, and the outbound scheduling controller. The MES establishes a data model centered on flower collection batches, drives the state transitions of each stage through a state machine, and supports both physical and virtual warehouse modes.

[0062] like Figure 3 As shown, the off-peak scheduling and warehousing coordination system is set at the inbound end of the bottled flower operation subsystem. It includes multiple distribution lines (five in this embodiment), a main roller conveyor line, buffer zones and actuators at the end of each distribution line, and a central scheduling controller. The buffer zone is located at the end of each distribution line, before the merging point, and has a capacity of 6-8 flower barrels, allowing continuously arriving barrels from upstream to form an orderly queue for release. Sensors are installed at both the buffer zone entrance and exit. The entrance sensor detects the entry of a flower barrel and triggers a reporting event, including at least the batch identifier and distribution line number of the barrel. The exit sensor detects that the barrel has been successfully released to the main roller conveyor line. The actuator is a conical electric roller installed at the buffer zone exit, used to release the barrel at a specified time in response to instructions from the central scheduling controller. When the conical electric roller is not rotating, the flower barrels are blocked at the outlet of the buffer zone; when the central dispatch controller issues a release command, the conical electric roller rotates at a set speed, relying on the friction of the conical surface and the tilt angle to send the flower barrels located at the outlet one by one to the main roller streamline, which has the characteristics of smooth action and high release accuracy.

[0063] like Figure 5As shown, the central dispatch controller is an independent industrial PC or high-performance PLC, internally running a peak-shaving scheduler. It can monitor the status of all branch lines in real time, calculate the priority of each branch line, and dynamically allocate time windows to avoid peak conflicts. The central dispatch controller acquires three types of parameters from the input layer in real time: production rhythm, status of each buffer zone, and main line speed. Speed ​​sensors installed on the main roller conveyor line acquire the main line running speed in real time and feed it back to the central dispatch controller.

[0064] The central dispatch controller dynamically calculates the length of the release time window based on the current running speed of the main drum and the preset ideal safety distance between the drums. (The safety distance between the drums is set as follows.) Let the current running speed of the main roller streamline be v (unit: m / s), then the length of the release time window is... The calculation formula is: , in, The unit is s.

[0065] For example, if the principal linear velocity v = 0.5 m / s, then =0.6 / 0.5=1.2s. The central dispatch controller uses this to determine the start and end times of the next available time window. For example, if the current time is... The window is [ , + ].

[0066] like Figure 4 As shown, after the above time window calculation and scheduling strategy allocation, the release time points of each diversion line are evenly distributed, eliminating the phenomenon of alternating "clustering" congestion and "idle" periods at the merging points.

[0067] The central scheduling controller determines the current stream line with release permission based on the status information of each buffer and the scheduling policy arbitration. This scheme supports two scheduling policies: First scheduling strategy: General priority scheduling strategy. The central scheduling controller dynamically adjusts the priority based on the backlog of each bypass buffer, calculated according to the following formula: Priority = Buffer occupancy rate weight × Buffer occupancy rate + Waiting time coefficient weight × Waiting time coefficient + Historical allocation balancing coefficient weight × Historical allocation balancing coefficient Among them, the buffer occupancy rate reflects the current backlog level of the diversion line buffer; when it is nearly full, the priority is increased. The waiting time coefficient reflects the length of time that the buckets in the buffer have been waiting; it increases with the waiting time. The historical allocation balance coefficient reflects the proportion of windows that the diversion line has obtained; the higher the proportion, the smaller the coefficient. Each weight coefficient can be dynamically configured by the user through the HMI interface. For example: Priority = 0.6 × Buffer Occupancy Rate + 0.3 × Waiting Time Coefficient + 0.1 × Historical Allocation Balance Coefficient.

[0068] The central dispatch controller uses a normalized multi-factor weighted model to calculate priorities, ensuring that the values ​​of each factor are all within the range of [0,1]. The formula is as follows: , Among them, the weighting coefficient , and satisfy Users can configure it through the HMI interface.

[0069] The buffer occupancy rate is calculated using the following formula: The value ranges from [0,1]. When the capacity is 6 barrels, 3 barrels are... .

[0070] The normalized waiting time coefficient is calculated as follows: , The maximum tolerable waiting time set for the system (e.g., 10 seconds, configurable). Exceeding... The coefficient is always 1 to avoid linear divergence.

[0071] The historical distribution equilibrium coefficient is calculated using the following formula: , The number of time windows that the branch line i has acquired. This represents the total number of allocation windows; the more windows already allocated, the more... The smaller the value, the better the long-term equilibrium of rotation.

[0072] Example: , , A certain line , , ,but =0.6×0.8+0.3×0.5+0.1×0.9=0.72.

[0073] The central dispatch controller sorts the tasks from highest to lowest priority according to the calculated priority and allocates the current time window to the highest priority feeder. To facilitate understanding of the priority formula's response to different scenarios, Table 1 shows typical allocation behavior examples under three dominant factors. In actual operation, priority is calculated jointly by the three factors, and is not limited to a single factor.

[0074] Table 1. Example of time window allocation (illustrating the effect of dynamic priority)

[0075] The mechanism for achieving balanced rotation is as follows: historical distribution equilibrium coefficient. Continuously accumulate the number of windows acquired by each branch line. When a branch line acquires a large number of windows, its... Automatically reduce its priority, thus lowering it; conversely, lines that haven't received a window for a long time... As the value approaches 1, its priority increases. This negative feedback mechanism ensures fair release in a long-term statistical sense, macroscopically manifesting as a rotation effect (e.g., Figure 4 (As shown). Even in the short term due to or Dominance leads to a certain line continuously gaining windows, but in the long run, it will also be... Level it.

[0076] Second scheduling strategy: Batch coordinated scheduling strategy, such as Figure 7 As shown, three barrels from the same batch (e.g., batch A) enter the buffer of the same distribution line (e.g., distribution line 1) consecutively after passing through the same grading device. The central scheduling controller detects multiple barrels with the same batch identifier at the head of the distribution line buffer queue and immediately triggers the batch coordinated scheduling strategy, allocating consecutive release time windows (e.g., T+0, T+2, T+4 seconds) for this batch and issuing release instructions for each barrel to the actuators of distribution line 1. The actuators release the barrels sequentially according to the time sequence, ensuring they continuously merge into the main roller flow line with the minimum safe interval. Simultaneously, priority queues for other distribution lines are maintained, and the subsequent window allocation order is calculated based on a general priority formula.

[0077] Since each branch line has a buffer capacity of ≥6 buckets, the central scheduling controller can scan the continuous bucket sequence from the outlet inward in the buffer queue in real time. When three or more consecutive buckets belonging to the same batch are detected (e.g., the three buckets of batch A occupy the head position of the buffer), the batch collaborative scheduling strategy is triggered. The system reserves a continuous release time window for this batch (e.g., T+0, T+2, T+4) and issues a sequence release command to the actuator of the branch line.

[0078] When multiple routing lines simultaneously trigger batch coordination requests (e.g., multiple batches simultaneously form a continuous sequence at the head of their respective buffers), the central scheduling controller performs batch queuing arbitration: (1) Add all batch requests to the “Batch Collaboration Pending Queue” and sort them by trigger timestamp or comprehensive priority (calculated by the general priority formula).

[0079] (2) Assign a continuous window sequence to each batch in turn, and only one batch is allowed to occupy the main continuous window at the same time.

[0080] (3) After the current batch is released, the system checks the queue to be processed. If there are still batches waiting, the system immediately starts the continuous window allocation of the next batch.

[0081] (4) During the batch release, the ordinary single-bucket release requests of other diversion lines are temporarily suspended, but their waiting time coefficient Wi will continue to accumulate to ensure that no starvation occurs after the batch ends.

[0082] This mechanism avoids "localized congestion" caused by multiple batches simultaneously vying for the main line. For example... Figure 7 As shown, after batch A is released continuously, the system resumes general priority scheduling or processes the next batch.

[0083] After batch A is continuously released (after T+4), the system allocates windows T+6, T+8, and T+10 to distribution lines 2, 3, 4, and 5 according to priority, to avoid long waiting times on any distribution line (capacity balancing). The three barrels of batch A converge into the main line with a safe distance and are finally stored in continuous storage locations 01-03 in aisle A, achieving centralized storage of the same batch within the aisle. The batch-aisle-storage location mapping is shown in Table 2. Table 2

[0084] like Figure 8 As shown, the outbound batch collaboration process is as follows: Step 1: The outbound dispatch controller receives buyer orders from the upper-level system, aggregates all batches ordered by the buyer by "Buyer ID", and splits the orders into batch groups.

[0085] Step 2: The outbound dispatch controller queries the batch-lane mapping table (as shown in Table 2) to obtain the lane and location information of each batch and determine whether each batch is located in the same lane.

[0086] Step 3: If the batches are located in the same aisle, a combined picking instruction is generated, and the stacker crane in that aisle is sent a single instruction to pick up each batch of flower barrels in sequence; if they are located in different aisles, a separate picking instruction is generated, and the stacker crane picks up the barrels in multiple strokes. The picked-up flower barrels are then transported downstream via the main roller conveyor.

[0087] Step 4: The outbound dispatch controller checks the current status (idle / busy) of each pickup point, dynamically selects an idle pickup point, and notifies the PLC to direct all flower buckets to that pickup point. The buyer's vehicle only needs to stop at that pickup point to receive flowers from all batches of flower buckets at once and load them onto the vehicle.

[0088] like Figure 6 As shown, taking peak-hour operation as an example, the complete workflow is as follows: Input layer: The central dispatch controller receives production rhythm parameters, the status of each buffer (including the batch identifier of each buffer's flower barrel), and the speed values ​​fed back by the main line speed sensor in real time.

[0089] Processing Layer: The central dispatch controller performs data preprocessing (cleaning sensor data and updating the buffer queue), demand calculation (calculating the number of flower buckets to be released on each branch line), conflict detection (determining whether multiple buffers are requesting release at the same time), timing optimization (calculating the optimal release order and time window based on priority formulas or batch coordination strategies), and decision generation (determining the recipient of the current time window and the release time).

[0090] Output layer: The central dispatch controller sends a release command to the selected branch line actuator, and the actuator moves at the specified time, and the flower bucket merges into the main roller flow line.

[0091] The aforementioned processing layer operates continuously in a loop, forming a dynamic closed-loop control system that adapts to different production cycles and peak intensity.

[0092] Other implementation methods for staggered peak scheduling and warehouse collaboration systems are as follows: Implementation Method Two: The difference from the above implementation method lies in that, instead of locking specific storage locations in advance, a "time window" and a "lane segment" are locked, allowing the stacker crane to dynamically select available locations within that segment during the filling process, achieving "soft proximity" at the software level. Specifically, while allocating continuous release time windows for the same batch of flower barrels, the central dispatch controller does not specify specific storage location numbers, but rather a target aisle and a range of storage locations (e.g., storage locations 01-10 in aisle A). When performing the storage task, the stacker crane dynamically selects available storage locations within that range based on the actual occupancy of each location during that time period. This implementation method improves the flexibility of storage location allocation while ensuring centralized batch storage, avoiding the storage fragmentation problem caused by locking specific storage locations in advance.

[0093] Implementation Method 3: The difference from the above implementation is that when too many flower buckets from the same batch from the same buyer cause congestion at a single pick-up point, the system can dynamically allocate 2-3 pick-up points as a "dedicated pick-up point group for the same buyer" and automatically guide the buyer's vehicle to move sequentially, achieving semi-centralized delivery. Specifically, the outbound dispatch controller calculates the required number of pick-up points (e.g., one pick-up point for every 50 buckets) based on the number of flower buckets in the buyer's order and the current queuing status at each pick-up point. The system temporarily binds the selected pick-up points as a dedicated pick-up point group for the buyer, and the buyer's vehicle stops at each pick-up point sequentially to complete loading. This implementation alleviates the congestion pressure at a single pick-up point while ensuring centralized delivery.

[0094] Implementation Method Four: The difference from the above implementation is that, for buyers with high-frequency repeat purchases, the system can predict in advance the possible batch combinations they might buy, and intentionally store these batches in the same or adjacent aisles upon receipt, further shortening the outbound retrieval journey. Specifically, the MES analyzes the purchase frequency and batch combination preferences of each buyer based on historical transaction data to generate a predictive model. When a new batch arrives in the warehouse, the MES, based on the prediction results, allocates multiple batches that the buyer might combine to consecutive storage areas in the same or adjacent aisles. After the auction is completed, the outbound dispatch controller can retrieve multiple batches of flowers from the same area at once, reducing the cross-aisle movement distance of the stacker crane.

[0095] Implementation Method 5: The difference from the above implementation is that a movable shuttle car is used as a dynamic pick-up point. After order aggregation, the shuttle car automatically moves to the end closest to the target aisle, loads all batches, and then proceeds to the loading area. Specifically, after the outbound dispatch controller completes order aggregation, it does not guide the flower barrels to a fixed pick-up point, but instead dispatches a shuttle car to the exit of the target aisle. The stacker crane sequentially removes each batch of flower barrels and directly loads them into the shuttle car. After the shuttle car has collected all batches, it automatically proceeds to the loading area. This implementation further reduces the flow distance of the flower barrels on the conveyor line and the secondary handling before loading.

[0096] like Figure 9 As shown, the barrelled flower operation subsystem comprises a management layer, an execution layer, and an equipment layer. The management layer consists of a Management System Execution System (MES), which serves as the central scheduling platform for the entire system, responsible for unified scheduling, data aggregation, status management, and information synchronization. Specifically, the MES sends scheduling instructions to the vertical warehouse, filling machine, visual grading system, auction trading system, and closed-loop cleaning line, and receives event data such as grading results, auction results, and outbound confirmations. The MES connects to the various subsystems in the execution layer via the RESTful API standard protocol, supporting both synchronous calls and asynchronous callback modes, and has a built-in retry queue mechanism to ensure eventual data consistency.

[0097] Each subsystem in the execution layer receives MES instructions and reports the execution status: the vertical warehouse is responsible for the automated storage and retrieval of flower barrels; the filling machine is responsible for filling a preset number of flower barrels with nutrient solution according to a preset amount; the visual grading system is responsible for collecting flower characteristic data and automatically completing the grading; the auction trading system is responsible for displaying flower data and completing the auction trading; and the closed-loop cleaning line is responsible for the standardized cleaning and drying of the recycled empty flower barrels.

[0098] The equipment layer includes stacker cranes, conveyor lines, barcode readers, industrial cameras, flipping mechanisms, high-pressure spray nozzles, drying fans, detection sensors, and PDA handheld terminals, etc.

[0099] like Figure 12 As shown, based on the actual physical flow and information coordination, the bottled flower operation subsystem is divided into six functional areas: the automated storage area, the filling area, the grading area, the anomaly handling area, the auction and outbound area, and the return and cleaning area. The automated storage area serves as the central scheduling node, equipped with 12 aisles, 12 stacker cranes, and 24 rows of shelving to uniformly manage all fresh flowers and empty bottle inventory. The system operates with a dual-thread parallel architecture: thread one handles the entire auction and outbound process (fresh flowers outbound → pickup → empty bottle cleaning and return to storage); thread two handles the entire flower collection and warehousing process (fresh flower arrival and empty bottle supply occur in parallel, converging for grading before warehousing).

[0100] Auction Outbound Process: After goods with auction tasks are dispatched from the automated warehouse, they are grouped and arranged to enter the auction stage. After the auction is completed, the MES synchronizes the auction information to the WMS to generate an outbound task. The MES and WMS then distribute the outbound task to the outbound flow line. Flower barrels are transported to the branch barcode reader (marked by a dotted line as "Flower Bridge," used to connect the flower receiving and inbound lines for data exchange) via the outbound flow line barcode reader, which then identifies and assigns them to the corresponding 20 pickup points. Upon arrival at the pickup point, a pickup slip is automatically printed. Logistics personnel collect the flowers, place them in the turnover box, and attach the slip, while simultaneously clicking "Return Empty Barrel to Warehouse." The empty barrels enter the return flow line, where they undergo a washing machine process including tipping over to drain water, internal and external washing, preliminary drying, and further drying after being straightened. The stacker crane then places the empty barrels back on the shelf and updates the inventory, awaiting the next round of allocation. When there are no auction tasks, the flower barrels remain in the automated warehouse awaiting the next round of allocation.

[0101] The entire flower collection and warehousing process is as follows: Logistics personnel place full frames of flowers into the collection flow line, which then sends them to the collection platform in front of the five grading machines via the third-level flow line. Simultaneously, after the number of buckets is entered into the APP, the MES notifies the WMS to retrieve empty buckets from the automated warehouse. These empty buckets are then sent to the filling area via the outbound flow line and the branch line barcode reader (automatic filling every 8 buckets). After filling, the flowers enter the main pre-grading flow line and are assigned to one of the five grading lines. The grading machines read the barcodes of the buckets, logistics personnel insert the flowers and place batch slips, and counting personnel use PDAs to scan the batch codes to bind the batch to the flower buckets. Grading information is synchronized between the MES and WMS. After the flower buckets flow into the grading buffer zone, they enter the main line in sequence. The WCS reads the barcodes to obtain grading information from the WMS: if information is available, the flowers enter the warehousing flow line (same batch, same aisle); otherwise, they are transferred to the abnormal flow line for manual handling. After warehousing, the WMS updates the inventory, and the flower bucket groups are arranged for auction.

[0102] Global Data Closed Loop: The barcode reader at the junction enables data exchange between the inbound and outbound production lines via a "flower bridge." All inbound, outbound, grading, and empty barrel data are synchronized in real time by MES, WMS, and WCS. The automated warehouse centrally manages all flower and empty barrel inventory. The empty barrel circulation system operates in tandem with the main flower inbound and auction outbound lines, forming a complete closed loop from empty barrel supply, flower collection, AI grading, automated warehouse storage to auction outbound and empty barrel cleaning and return to the warehouse. When there are no auction orders, goods await the next round of allocation in the automated warehouse.

[0103] like Figure 10 and Figure 11 As shown, the system adopts a dual-lane structure. The MES is located in the middle of the upper lane, while the lower lane represents the physical flow path. The lower lane represents a complete flow path from empty barrel outbound to return to the warehouse: Empty barrel outbound → Filling → Height adjustment → Flower placement → Grading → Warehousing → Auction → Outbound → Transportation → Pickup → Empty barrel return → Cleaning → Return to warehouse. A bidirectional arrow connects the upper and lower lanes, indicating the mutual driving relationship between information flow and physical flow. Information flow guides physical flow, and physical flow provides feedback for information updates.

[0104] As the global brain, MES connects downwards to automated production lines, storage and auction systems, closed-loop cleaning lines, and other workstations to achieve task distribution and data binding. The specific workflow is as follows: MES interacts bidirectionally with the front-end system, which supports bucket adjustment and query operations via the APP. MES controls the automated production line, which includes automatic warehousing and filling stations, and manual adjustment and grading stations. After the automatic warehousing and filling station completes the warehousing of empty flower buckets and the filling of nutrient solution, it transports the empty flower buckets containing nutrient solution to the manual adjustment and grading station. After the manual adjustment and grading station completes the height adjustment and visual grading of the flower buckets, it transports the graded flower buckets to the vertical warehouse for storage.

[0105] The vertical warehouse synchronizes data bidirectionally with the MES (Manufacturing Execution System), synchronizing its storage status, location information, and inventory data to the MES. The MES, in turn, synchronizes scheduling instructions and auction results to the vertical warehouse. The vertical warehouse then transports the auctioned flower buckets to the post-auction logistics channel, while the MES simultaneously sends outbound scheduling instructions to this channel.

[0106] like Figure 13 As shown, the flower bucket includes a one-piece molded bucket body, which has an outer shell with a trapezoidal structure that is smaller at the top and larger at the bottom, and an inner cavity with a trapezoidal structure that is larger at the top and smaller at the bottom. Vertical mounting holes are opened at the four corners of the bucket body, and telescopic rods are installed in the mounting holes. The top of the telescopic rods is connected to a barrier. Notches are symmetrically opened on two opposite sides of the barrier. A QR code loading part is set on the upper part of the four walls of the bucket body.

[0107] like Figure 14 As shown, the nutrient solution injection machine can inject nutrient solution into 8 flower pots at one time, and the amount of nutrient solution added is preset by controlling the pump.

[0108] like Figure 16 As shown, MES establishes a data model centered on flower collection batches and drives state transitions through a state machine. The state machine enforces the flow path of the flower collection batch, changing unidirectionally in the following order without skipping steps: Pending collection → Collected → Graded → Grouped → Auctioned → Auction in progress → Completed. The correspondence between each state is as follows: Pending Flower Collection: This corresponds to the batch creation stage, which is entered when the supplier schedules or the system creates the batch.

[0109] Flowers Received: This corresponds to the warehousing process. When the supplier delivers the flowers, the operator scans the container code with a PDA to bind the batch, and the flowers enter the warehousing flow line, the MES receives the warehousing confirmation and the status changes.

[0110] Rating Completed: In the rating process, the status changes after the flower bucket undergoes AI analysis by the visual rating system, the rating information is bound to the flower bucket identifier, and the data is uploaded to the MES.

[0111] Grouped: This refers to the grouping process. After the operator creates a column label on the PC, the MES automatically associates the graded batches with the trolley and generates grouping data, the status changes.

[0112] Already up for auction: This corresponds to the pre-auction preparation stage. After the MES pushes the group data to the transaction system, and the transaction system confirms receipt and prepares to display it, the status changes.

[0113] During the auction process: In the corresponding auction stage, when the auction begins and the buyer bids, MES notifies WMS to release the goods based on the transaction information, and the status of the flower bucket changes.

[0114] Completed: This corresponds to the outbound or pickup process. The status changes after the auction is completed, the transaction system pushes the auction results to MES, MES creates the auction order and associates it with the buyer's information, and the flower bucket is transferred to the buyer's pickup point or enters the outbound distribution process.

[0115] like Figure 17 As shown, auction orders also have corresponding status transitions, which change in the following order: Pending Push, Pushed, Auction in Progress, Sold, Unsold, Cancelled. When the MES pushes the group data to the transaction system, the order status changes from Pending Push to Pushed; when the transaction system starts the auction, the status changes to Auction in Progress; when the buyer completes the bid, the status changes to Sold or Unsold; if an abnormal cancellation occurs during the auction, the status changes to Cancelled.

[0116] The state machine responds to various events, automatically driving state changes and triggering downstream operations: When the PDA scans the code and confirms receipt of flowers, the status changes from pending receipt to received flowers, and at the same time, the MES is triggered to notify the WMS to prepare for data entry.

[0117] When the visual grading system completes the grading and uploads the results, the status changes from "flowers received" to "graded," and at the same time, the MES system notifies the WMS system to update the batch information and put it into the warehouse.

[0118] When an operator creates a column on the PC, the status changes from "classified" to "grouped", and the MES is triggered to prepare the grouped data for push.

[0119] When MES pushes the grouped data to the trading system, the status changes from "Grouped" to "Listed for Auction".

[0120] When the transaction system starts the auction, the status changes from "listed for auction" to "auction in progress," and at the same time, the MES is triggered to notify the WMS to execute the outbound process.

[0121] When the transaction system completes the auction and pushes the results, the status changes from "in the auction process" to "completed". At the same time, it triggers MES to create an auction order, notifies WMS to execute the outbound delivery, and notifies the buyer to prepare for pickup.

[0122] This solution supports two inventory modes: (1) Physical Warehouse Mode: After grading, the graded flower barrels are stored in the automated storage and retrieval system (AS / RS). The WCS allocates aisles based on the flower barrel information, and the stacker crane performs the warehousing. The flower barrels wait for auction instructions in the AS / RS. In this mode, the statuses are sequentially: graded, physical warehouse entry, stored in warehouse, and awaiting auction. This mode is suitable for same-city or short-distance transactions with short logistics times.

[0123] (2) Virtual Warehouse Mode: Suitable for transactions with long logistics times. In this mode, the physical flower buckets that have been graded are shipped out of the physical warehouse and enter the distribution and transportation process in advance. However, the MES logically marks the batch as a virtual warehouse, so that it can still participate in the auction. The status transition is as follows: After being graded, the batch enters the physical warehouse for storage, and then the pre-shipment is executed. The batch type changes from physical warehouse to virtual warehouse. The physical batch enters the distribution and transportation process. At the same time, the batch in the virtual warehouse participates in the grouping and auction normally. The two processes are executed in parallel. After the auction is completed, the MES logically executes virtual shipment and marks the batch in the virtual warehouse as shipped out. At this time, the physical flower buckets that have been pre-shipped are expected to have arrived at the distribution station. The buyer scans the code to pick up the goods and completes the transaction. The core difference between the physical warehouse mode and the virtual warehouse mode is that the physical warehouse mode is to store first, then auction, and then ship out, while the virtual warehouse mode is to ship out first, auction in transit, and pick up upon arrival.

[0124] Information on batches of goods that failed to be pre-shipped will be locked. The current status of the goods needs to be confirmed. The pre-shipment interface allows filtering by batch status to identify any issues. Re-shipping or deleting the batch will directly create a virtual batch and bind it to the failed goods (e.g., in cases of mechanical problems where goods are manually moved out).

[0125] Handling of failed auctions: (1) Re-auction; (2) Destruction (flower auction suppliers generally destroy by default); (3) Return to base via logistics. Goods scheduled for shipment have already arrived at the distribution station and are laid flat on the ground with no pick-up point.

[0126] like Figure 19 and Figure 20 As shown, this solution supports three business models, each adapted to different transaction scenarios: Standard mode: After the classification is completed, the goods are stored in the physical warehouse, then grouped and auctioned. After the auction, the goods are transported to the distribution center and finally picked up. This mode is suitable for same-city or short-distance transactions with short logistics time.

[0127] Pre-shipment mode: After classification, goods are stored in the physical warehouse. Then, batches requiring pre-shipment are selected from the physical warehouse, and pre-shipment is executed to move the physical goods out of the warehouse in advance (the storage type of these batches will change to virtual warehouse after pre-shipment). Distribution and transportation then begin. During the distribution process, batches in the virtual warehouse are grouped and auctioned in parallel. After the auction, virtual shipment is executed, and the physical goods are picked up directly at the distribution site. This mode is suitable for transactions with long logistics times. By shipping in advance, goods are essentially ready for pickup at the distribution station by the end of the auction, effectively shortening customer waiting time.

[0128] Hybrid Mode: After grading, the batches in the "flower bucket" are stored in the physical warehouse (currently, all batches are in the physical warehouse). A subset of batches (the smallest unit is the batch, not the "flower bucket," so there's no splitting of the same batch) are pre-shipped (converted to a virtual warehouse type). The two batches are processed in parallel according to the standard mode and pre-shipment mode paths, respectively. This mode is suitable for high-volume transaction scenarios and can be flexibly allocated based on buyer distribution and logistics timeliness requirements.

[0129] In hybrid mode, auctionable batches are divided into virtual warehouse batches and physical warehouse batches, but they are not distinguished during the auction process and are auctioned together. When a transaction is completed, if the batch belongs to the virtual warehouse, no outbound process will be initiated; instead, the goods will wait for virtual warehouse outbound processing. If the batch belongs to the physical warehouse, the WMS will be notified to process the outbound shipment. Transactions are based on a unique batch ID; each transaction is conducted within a specified batch, meaning the inventory type is determined at the time of the transaction, and there is no concept of assigning a batch only after the transaction is completed.

[0130] Different parks can choose their own outbound mode according to their own needs. Even if all modes are used together, there will be no confusion. Each mode has its own process and software limitations.

[0131] like Figure 18 As shown, the status of the collection and distribution transportation process is forced to change unidirectionally in the following order: Pending processing → Pending pickup → Pending outbound → Pending loading → Loaded → Dispatched → Pending unloading → Unloaded → Pickup completed. The triggering events for each status are as follows: Pending Disposal: Triggered when the auction is completed or the outbound task is generated.

[0132] Pending pickup: Triggered when outbound task is assigned.

[0133] Pending Outbound: Triggered when WMS receives an outbound task.

[0134] Pending Loading: Triggered when the stacker crane finishes leaving the warehouse and the flower barrels are transported to the loading area.

[0135] Loaded: Triggered when the operator uses a PDA to scan the code to confirm loading.

[0136] Departure Completed: Triggered when the driver confirms departure or the system updates the status.

[0137] Pending unloading: Triggered when the vehicle arrives at the pickup location and reports.

[0138] Unloaded: Triggered when the operator uses a PDA to scan the barcode to confirm unloading.

[0139] Goods pickup completed: Triggered when the buyer confirms pickup by scanning the code with a PDA.

[0140] like Figure 11 and Figure 12As shown, the empty flower buckets enter the closed-loop cleaning line after being picked up. The closed-loop cleaning line is configured with the following steps in sequence: a tipping and water-pouring station, a tipping and internal cleaning station, a tipping and preliminary drying station, and a return-to-center and deep drying station. The specific cleaning process is as follows: The empty flower bucket is first flipped 90 degrees by the first flipping mechanism to pour out the remaining liquid.

[0141] Then, it is flipped over by the second flipping mechanism until the bottom of the barrel is facing upwards (90 degrees), and the high-pressure rotating nozzle extends into the barrel to spray high-pressure water and atomized cleaning agent for all-round cleaning.

[0142] After cleaning, the water is rotated 90 degrees by the third flipping mechanism, and the strong air nozzles on the side perform the first purging to remove large water droplets.

[0143] Finally, the flower bucket is flipped back to its upright position and enters the tunnel-style drying channel for multi-directional air drying, undergoing a second round of deep drying.

[0144] After being cleaned and dried, the flower barrels are transported back to the empty barrel area of ​​the vertical warehouse via a conveyor line, awaiting the next MES scheduling, forming a complete closed-loop cycle. Only empty flower barrels that have passed the detection sensor after drying can be returned to the vertical warehouse.

[0145] like Figure 15 As shown, the operation flow of the bucket flower operation subsystem includes: Step 1: The MES issues a flower bucket outbound command. The WMS receives the command and sends it to the WCS. The WCS sends a command to the PLC and instructs the stacker crane to pick up the bucket. The empty flower bucket enters the conveyor line. Upon entering the conveyor line, the empty flower bucket is automatically transferred to the filling machine after being identified by the WCS barcode reader. After accumulating 8 empty flower buckets in the filling machine, nutrient solution is filled into the empty flower buckets. After filling, the solution flows into the grading area and is distributed to the various vision grading devices as needed by the swing wheel machine.

[0146] Step Two: The visual grading system automatically reads the flower bucket identification code. Operators adjust the height of the flower bucket to the appropriate position based on the flower stem length (an automatic telescopic flower bucket driven by a servo motor or electric push rod can also be used). The batch information of the bouquet is then pushed to the visual grading system to achieve automatic batch code binding. The visual grading system collects information on the variety, grade, stem length, and maturity of the flowers. The video module automatically focuses and adjusts the exposure value for image capture. The images are analyzed for defects using an AI model and combined with the data collected from the flowers to achieve automatic grading. The grading information is then bound to the flower bucket identification code and uploaded to the MES, while simultaneously updating the batch grading information. After grading is complete, parallel robots can be added to automatically pick up the flower buckets according to their grade and place them in different buffer conveyor lines or different areas of the automated storage and retrieval system.

[0147] Step 3: After grading, the flower barrels enter the warehousing flow. The MES synchronizes the batch information of the flower barrels to the WMS system. The barcode reader identifies the re-graded information of the flower barrels. The WCS generates an warehousing task based on the flower barrel information and assigns the flower barrels to the aisles according to the identified information. The stacker crane executes the warehousing process, and after completion, updates the information to the WMS. The WMS then synchronizes the information to the MES. Operators create lists for all batches awaiting sale. The MES automatically associates the graded batches and generates group list data.

[0148] Step 4: MES pushes the grouping information to the transaction system. After the transaction system completes the auction, it pushes the results back to MES. MES receives the results, creates an auction order, and associates the buyer information. Based on the auction results, it automatically generates an outbound task, calls the WMS interface to initiate the outbound process, and the WMS plans the outbound sequence according to the aisle location. It creates a collection and distribution task, associates the outbound flowers, places the flowers from the flower buckets into flower frames, and the PDA scans the code to confirm loading and updates the transportation status. When the flower frames arrive at the pickup location, the PDA scans the code to confirm pickup, completing the transaction. To resolve the conflict between logistics time and the fixed auction time, outbound operations can be performed in advance before the auction. After the goods in the physical warehouse are outbound, they are transferred to virtual warehouse management and dispatched (virtual warehouse mode).

[0149] Step 5: After the empty flower buckets are picked up, they are returned to the closed-loop cleaning line and cleaned and dried according to the cleaning process described above.

[0150] Step 6: After drying and passing inspection, the flower buckets are returned to the vertical warehouse to await further dispatch by the MES.

[0151] Other implementation methods for the barrel-type flower operation subsystem are as follows: Implementation Method Two: The difference from the above implementation is that, based on MES scheduling, a UWB (Ultra-Wideband) or RFID active positioning tag is added to each flower barrel, and positioning base stations are deployed in the work area. The system can obtain the precise location of each flower barrel in real time (e.g., in the storage tank, on the filling line, in the cleaning line, etc.). MES no longer relies solely on process-triggered scheduling, but performs dynamic rescheduling by combining location information.

[0152] The specific implementation of dynamic rescheduling is as follows: (1) Real-time acquisition of location information and status mapping Each flower barrel's UWB / RFID tag transmits signals to surrounding positioning base stations at a fixed frequency (e.g., 1Hz). The MES calculates the two-dimensional coordinates of the flower barrel using Time Difference of Arrival (TDOA) or Angle of Arrival (AOA) algorithms, with an accuracy of ±10cm. The MES matches these coordinates with predefined electronic fences of the work area (e.g., pouring area, waiting area, cleaning inlet, cleaning outlet, etc.) to determine the current status of each flower barrel. For example: The coordinates are located within the fence of the injection area → Status = "Injection in progress"; If the coordinates are located within the fenced area of ​​the infusion outlet waiting area and the stay exceeds 5 seconds, the status will be "Infusion completed, awaiting transfer". The coordinates are located inside the cleaning entrance fence → Status = "Pending Cleaning".

[0153] (2) Delay detection and dynamic priority adjustment mechanism MES maintains a task queue, where each task contains the following fields: bucket ID, current status, planned next workstation, planned arrival time, and dynamic priority (initial value 100). Simultaneously, MES monitors the progress of each bucket in real time according to preset standard process cycles (e.g., pouring → cleaning: standard time 30 seconds, allowable deviation ±5 seconds).

[0154] When the MES detects through positioning data that a certain workbench (referred to as a "delayed workbench") has stayed at the current workstation for a longer than the upper limit of the standard cycle time (e.g., staying for 40 seconds, exceeding the 35-second limit), it triggers dynamic rescheduling. It should be noted that "reducing the priority of the delayed workbench" in this step is not a punitive measure, but rather based on the following system optimization logic: (1) The delayed task bucket no longer possesses the "urgent" attribute. The standard tick limit (35 seconds) is the "maximum tolerable waiting time" set by the system. Once the waiting time of task A exceeds this threshold, it means that A has been actually delayed. In scheduling theory, the "marginal urgency" of a delayed task approaches zero—because no matter how the system prioritizes A, the delay time (40 seconds) of A is irrecoverable. Conversely, tasks that have not yet been delayed (such as B and C) are still within the "salvage window," and at this time, the system benefit of protecting the undelayed batches from the impact is far greater than prioritizing the delayed batches.

[0155] (2) Prioritizing delayed tasks will create a "congestion propagation" effect. If task A is sent to the cleaning line first, the delay of A will cause its time window for occupying cleaning line resources to be postponed, which in turn will squeeze the normal processing windows of subsequent tasks B and C, ultimately causing B and C to be forced to delay as well, i.e., creating a "domino effect of delays". The strategy of this invention actively cuts off the delay propagation path by downgrading delayed tasks and upgrading normal tasks, so as to minimize the overall number of delayed tasks in the system.

[0156] (3) Mathematical meaning of delay penalty. Let the system objective function be to minimize the total delay loss: Min Σ (number of delayed tasks × unit loss). When A is delayed, its delay loss is locked at a fixed value, and continuing to prioritize A will not reduce this loss. However, if B and C are delayed, new loss terms will be added. Therefore, downgrading A and upgrading B / C is the only optimal solution under the objective function. Its essence is an online scheduling strategy with the goals of "minimizing maximum delay" and "minimizing the number of delayed tasks".

[0157] Identify the affected task chain: MES queries the subsequent task sequence of the delayed flower bucket (e.g., waiting for transfer → entering the cleaning line → cleaning → exiting the cleaning line → entering the next workstation) and marks these subsequent tasks as "delayed associated tasks".

[0158] Reduce the priority of delayed batch-related tasks: MES reduces the dynamic priority of the delayed bucket and all its subsequent related tasks by one penalty value. (For example The new priority is calculated as follows: new priority = original priority - 30. If the priority drops below 0, it is set to 0. Demoted tasks are automatically moved to the next task in the task queue.

[0159] Increase the priority of normal batch tasks: MES scans the task queue for all tasks not marked as "delay-related" and currently in the "pending" state, and increases their dynamic priority by a compensation value. (For example (and with a priority limit of 100), which moves it forward in the sorting.

[0160] (III) Task Queue Reordering and Execution MES performs a task queue reordering every time it receives a location data update (e.g., every second) or every time a task completion event is triggered. The sorting rule is as follows: tasks are arranged from high to low dynamic priority, and those with the same priority are sorted according to the time they entered the queue.

[0161] After sorting, MES will send the instructions of the bucket and target workstation corresponding to the task at the top of the queue to the corresponding execution equipment (such as AGV, conveyor belt diverter, robotic arm, etc.), and the equipment will execute according to the instructions.

[0162] (iv) Specific examples Assume there are three flower buckets, A, B, and C, all of which have been filled and are waiting to enter the cleaning line, as shown in Table 3: Table 3

[0163] After reordering, the task queue order is: B, C, A. MES prioritizes sending B and C to the cleaning line, while A waits until B and C are released before entering. This example clearly demonstrates the core scheduling logic of this invention: even though A has the longest waiting time, because its delay loss is already locked, the system prioritizes protecting B and C, which have not yet been delayed. This reduces the total number of delayed tasks from a potential 3 (all A, B, and C delayed) to 1 (only A delayed), resulting in optimal overall system efficiency. This implementation can further improve the system's response capability to abnormal events, and is particularly suitable for complex scenarios involving large-scale, multi-threaded parallel processing.

[0164] Implementation Method 3: The difference from the above implementation lies in the use of an automatically telescopic flower barrel driven by a servo motor or electric actuator, and the addition of a laser ranging and automatic height matching module at the vision grading station. After the vision system identifies the length of the flowers, it automatically sends a command to adjust the flower barrel to the optimal height, requiring no manual intervention. Specifically, while acquiring images of the flowers, the vision grading system measures the actual length of the flower stems using a laser ranging module. Based on the difference between the flower stem length and the current height of the flower barrel, the MES sends a telescopic command to the servo motor at the bottom of the flower barrel, automatically adjusting the enclosure to the appropriate position. This implementation further improves the level of automation, reduces manpower, and achieves more precise height matching (accurate to the millimeter level), making it particularly suitable for highly standardized flower factories.

[0165] Implementation Method Four: The difference from the above implementation method lies in the addition of a parallel robot (Delta manipulator) and its vision guidance system after manual adjustment and grading of the workstations. Once the vision grading is completed and the grade is bound to the waste bin ID, the parallel robot can automatically pick up the waste bins according to their grade (A / B / C / D) and place them in different buffer conveyor lines or different areas of the automated storage and retrieval system (AS / RS), achieving automatic sorting and warehousing by grade. Specifically, based on the grade information issued by the MES, the parallel robot picks up grade A waste bins to the grade A buffer line, grade B waste bins to the grade B buffer line, and so on. Wastes of different grades are then stored in different areas of the AS / RS. This implementation method can greatly improve sorting efficiency, reduce the handling work after manual re-judgment, and transform the entire system from semi-automatic to fully automatic.

[0166] Implementation Method 5: The difference from the above implementation is that the closed-loop cleaning line uses a rotary multi-station turntable cleaning machine. The flower barrels sequentially pass through stations such as liquid pouring, brushing, high-pressure spraying, and hot air drying. The turntable rotates continuously with a fixed cycle time, suitable for high-volume, stable production scenarios. Specifically, 8 or 12 flower barrel stations are evenly arranged on the turntable, each station corresponding to one cleaning process. Each time the turntable rotates one station angle, the flower barrel enters the next process, and the cycle time for completing the entire cleaning process is fixed (e.g., 60 seconds / cycle). Compared to a linear rotary cleaning line, this implementation has a smaller footprint and higher throughput.

[0167] For high-end or extremely hygienic fresh flowers (such as export-grade flowers), ultrasonic cleaning can be used in the internal cleaning station to remove microorganisms; the deep drying station is replaced with microwave drying, which is faster and effectively sterilizes. Specifically, in the internal cleaning station, the flower buckets are immersed in an ultrasonic cleaning tank, where the ultrasonic generator produces high-frequency vibrations to remove microorganisms and stubborn stains from the bucket walls. In the deep drying station, the flower buckets enter a microwave drying chamber, where microwave energy rapidly evaporates residual moisture and kills bacteria inside the bucket. This solution is more expensive, but it raises cleaning standards and meets specific market demands.

[0168] Implementation Method Six: The difference from the above implementation is that the unique identifier of the flower bucket in the MES is synchronously written into the blockchain (such as a consortium blockchain) along with the information such as flower grade, batch, and grower. Auctioneers, buyers, and regulators can all verify the on-chain information of the flower bucket by scanning the code, ensuring the data is tamper-proof. Specifically, at each key node such as grading, warehousing, delivery, and auction completion, the MES automatically packages the operation records and status change information into a transaction and submits it to the blockchain network. Users can scan the QR code on the flower bucket to query the complete lifecycle record on the chain, including grower, grading time, grade, auction price, and delivery time. This implementation completely solves the traceability trust problem and is suitable for high-value flowers or export trade scenarios.

[0169] Implementation Method Seven: The difference from the above implementation is that instead of a centralized closed-loop cleaning line, a small integrated cleaning-drying-inspection workstation is set up next to each pick-up point. After the flower barrels are picked up, they are immediately placed into this workstation by staff (or AGVs). After passing cleaning, they are directly dispatched back to the automated warehouse by the MES system. Specifically, each pick-up point is equipped with a small integrated cleaning device that integrates functions such as tipping and emptying, high-pressure spraying, hot air drying, and residue detection. After the buyer picks up the goods, the empty flower barrels are immediately placed into the workstation, where cleaning and drying are completed within 5 minutes. After passing inspection, they are automatically transported to the return-to-warehouse line. This implementation reduces long-distance transportation and waiting time, and is suitable for scenarios where pick-up points are dispersed, space is limited, or where it is desirable to distribute cleaning pressure.

[0170] Implementation Method Eight: The difference from the above implementation is that the core MES service is deployed in the cloud (public / private cloud), while each workstation (filling machine, vision system, stacker crane, washing line) retains only edge controllers. Through 5G or the Industrial Internet, the cloud-based MES and edge controllers synchronize in real time, enabling global scheduling. Specifically, multiple remote flower processing centers share the same cloud-based MES platform, with each center's edge controller responsible for local equipment control and data acquisition. The cloud-based MES centrally manages inventory, batches, orders, and auction information across all processing centers, enabling cross-center inventory transfers and order allocation. This implementation allows multiple remote flower processing centers to share the same MES platform, reducing IT investment at individual sites and is suitable for group-based, multi-park operation models.

[0171] Implementation Method Nine: The difference from the above implementation is that fixed roller conveyor lines are replaced with Automated Guided Vehicles (AGVs) or Autonomous Mobile Robots (AMRs). The MES issues tasks to the AGVs, such as "transport empty flower barrels from the vertical warehouse to the filling station" or "send graded flower barrels into the vertical warehouse." Specifically, the MES communicates with the AGV fleet through a scheduling system, and the AGVs autonomously plan their routes based on task priorities and current locations. When flower barrels need to be moved, the AGVs automatically travel to the pickup point, load the flower barrels by lifting or towing, and then transport them to the target station. This implementation improves the flexibility of the system layout, facilitating future expansion or changes in the process flow, but it requires higher initial investment and has a more complex scheduling algorithm.

[0172] like Figure 21 and Figure 22 As shown, the boxed flower operation subsystem includes a front-end interaction layer, a core scheduling layer, an edge computing layer, a resource and equipment control layer, a hardware execution layer, a communication layer, and a transaction decision layer. The front-end interaction layer includes apps or mini-programs used by suppliers and buyers. The core scheduling layer includes a Management Execution System (MES) for business process orchestration, data integration, and distribution. The edge computing layer includes 5G edge gateways located next to intelligent grading devices and edge controllers located in the stacker crane control cabinet for local AI inference, path planning, and distributed collaboration. The resource and equipment control layer includes a Resource Management System (WMS) and a Resource Control System (WCS). WMS is used for refined inventory management of goods within the automated warehouse, while WCS is used for macro-level task allocation and edge node status monitoring. The hardware execution layer includes automated logistics lines, automatic barcode readers, intelligent grading devices, automated warehouse stacker cranes, and automatic printing equipment. The communication layer includes 5G base stations and a core network to provide low-latency, reliable communication between edge nodes and the MES, as well as 5G-D2D direct communication between stacker crane edge controllers. The transaction decision layer includes an auction platform for facilitating transactions.

[0173] In the boxed flower operation subsystem, boxed flowers complete warehousing, random inspection and grading, and pickup on the second floor. The automated logistics line includes an inbound roller conveyor and a conveyor line, with 5 inbound and outbound ports (lower-level inbound, upper-level outbound). The inbound roller conveyor includes a random inspection diversion line and a straight-through line, used to automatically separate randomly inspected and uninspected boxed flowers. In this embodiment, the vertical warehouse has 8 aisles, each with 1 stacker crane, and each stacker crane is equipped with an edge controller to achieve distributed path planning and collision avoidance coordination.

[0174] like Figure 22 As shown, the boxed flower operation subsystem adopts a six-layer collaborative architecture, forming a complete data and logistics closed loop from front-end business initiation to back-end equipment execution. The front-end interaction layer includes an APP or mini-program used by buyers and an APP or mini-program used by suppliers. Suppliers use the front-end to reserve flower delivery or enter virtual inventory, while buyers use the front-end to participate in bidding and receive transaction feedback. The front-end interaction layer and the auction platform in the transaction decision layer have bidirectional data interaction. The auction platform pushes the matching results of the bidding transactions to the flower auction MES system in the core scheduling layer, while receiving reserved flower delivery data and virtual inventory data from the MES system.

[0175] The MES system, acting as the central control hub of the entire subsystem, is responsible for business process orchestration, data integration, and distribution. The MES operates through four branch links: the first flows to the warehouse management module (including WMS and WCS); the second flows to the 5G communication link at the communication layer; the third flows to hardware execution devices such as intelligent grading equipment and stacker cranes; and the fourth flows to automated printing equipment. Within the warehouse scheduling sub-link, the MES issues inventory management instructions to the WMS. The WMS is responsible for refined inventory management in automated warehouses and also has built-in task allocation functions for the resource and equipment control layer. The WMS assigns macro-level task allocation and edge node status monitoring tasks to the WCS. The WCS interacts bidirectionally with the hardware execution layer: the WCS obtains code reading information from the automatic code reader, receives grading results from the intelligent grading equipment, and issues control instructions to the automated logistics line.

[0176] Boxed flowers enter through the inlet and are conveyed by a roller conveyor to the sampling and sorting point, where they split into two branches. Branch A contains boxed flowers to be sampled: these flowers enter the intelligent grading equipment via a sorting line, where an AI vision system determines their grade. The AI ​​grading result is transmitted back to the WCS and MES systems. After grading, they merge into the straight-through line via a return line. Branch B contains boxed flowers not sampled: these flowers go directly into the straight-through line via a sorting line. The two branches merge on the straight-through line and are finally delivered to the automated warehouse. The automated logistics line (including the inlet roller conveyor, sorting line, straight-through line, return line, and outlet main line) receives control commands from the WCS and drives the entire inlet movement. Automatic barcode readers on the logistics line collect barcode information and transmit it back to the WCS.

[0177] MES sends instructions to the communication layer via 5G communication. The communication layer includes 5G base stations and core network (providing low-latency communication) and 5G D2D direct communication (enabling collaboration between stacker cranes). The communication layer connects downward to the edge computing layer.

[0178] The edge computing layer comprises two types of units: the first is the edge controller deployed within the stacker crane control cabinet, responsible for path planning and collision avoidance coordination; the second is the 5G edge gateway deployed next to the intelligent grading equipment, responsible for local AI inference. The edge computing layer outputs collaborative control commands to the automated warehouse stacker crane and the intelligent grading equipment, enabling real-time local collaboration between the devices.

[0179] Boxed flowers from the inbound straight line are stored in the automated storage and retrieval system (AS / RS), where stacker cranes complete the storage and retrieval. The AS / RS has 8 aisles, and each stacker crane is equipped with an edge controller. The stacker cranes receive collaborative control commands from the edge computing layer and control commands from the MES (Manufacturing Execution System).

[0180] The outbound process is triggered by the MES: the MES sends control commands to the automatic printing equipment, and the automatic printing equipment sends a "printing complete" signal back to the MES after printing. The physical outbound flow is as follows: boxed flowers are taken out of the automated warehouse and transported to the outbound exit area via a separate main outbound line. The outbound endpoint is the automatic printing of the delivery note, completing the outbound delivery.

[0181] Hardware-collected data is transmitted back upwards in reverse, forming a complete closed loop: the reading information collected by the automatic barcode reader is transmitted back to WCS and MES sequentially; the AI ​​grading results generated by the intelligent grading device are transmitted back to WCS and MES sequentially; the automatic printing device transmits the printing completion signal directly back to MES; and the device status of the edge controller and edge gateway is transmitted back to MES via the 5G communication layer. After MES aggregates all warehouse data, transaction data, and device data, it synchronizes them in reverse to the front-end auction platform, as well as the supplier's mini-program and the buyer's mini-program, forming a complete business closed loop.

[0182] In the front-end transaction process, the forward command flow is the auction platform pushing matching results to the MES, while the reverse data flow is the MES pushing pre-arranged delivery and virtual warehouse data to the auction platform. In the warehouse scheduling process, the forward command flow is the MES issuing commands to devices level by level via WMS and WCS, while the reverse data flow is the devices transmitting data back to the MES level by level via WCS and WMS. In the inbound production line process, the forward command flow is the WCS issuing control commands to the logistics line, while the reverse data flow is the barcode readers and grading devices transmitting data back to the WCS. In the edge computing process, the forward command flow is the MES issuing commands to devices via the 5G communication layer and the edge computing layer, while the reverse data flow is the edge layer transmitting device status back to the MES via the 5G communication layer. In the outbound delivery process, the forward command flow is the MES issuing control commands to the automatic printing equipment, while the reverse data flow is the printing equipment transmitting a printing completion signal back to the MES. In the data closed-loop process, the MES aggregates the data from the entire chain and then synchronizes it back to the front-end auction platform and the mini-program.

[0183] This architecture enables digital management of the entire process from order to delivery, with logistics and information flow operating in a closed loop, ensuring the efficiency and traceability of boxed flower operations.

[0184] like Figure 23 As shown, the intelligent grading equipment includes equipment components and control components. The control components include control buttons and a data interface. The equipment components include a conveyor belt, a lid-opening assistance assembly, an industrial camera array, a light source, an image processing unit, and a 5G edge gateway with an embedded GPU. The conveyor belt delivers the boxed flowers to the workstation. After manual assistance in opening the lid, the camera array acquires images under supplementary lighting. The image processing unit preprocesses the images, and the 5G edge gateway with the embedded GPU runs an AI model to determine the grading level (A / B / C / D). The determination result is uploaded to the MES (Manufacturing Execution System) via the data interface. The grading response time is less than 50ms.

[0185] In this embodiment, the AI ​​model built into the intelligent grading device uses a deep convolutional neural network (CNN) for flower grade classification. The output layer is a softmax function, mapping the features of the last layer of the network to the probability distribution of each category. Let the set of flower grade categories be C={A,B,C,D}. For an input image x, the model outputs a four-dimensional probability vector: ,in, Output the probability of each level for the input image. , , The probability is of any level; Confidence score is defined as the highest probability of the class in the model's predictions, i.e. ; For example, if the model outputs p=(0.92,0.05,0.02,0.01), then the confidence level is 0.92 (i.e., 92%), and the model classifies the class as Grade A. This confidence level reflects the model's certainty about the current judgment result. When the probabilities of multiple categories are close (e.g., p=(0.45,0.40,0.10,0.05), the confidence level is only 0.45, indicating that the model has difficulty distinguishing between them, and manual verification is required.

[0186] In this embodiment, the preset threshold is 85%, which was determined based on the following experiments: On a test set of 3000 labeled boxed flower images, confidence-accuracy curves were plotted. Statistical analysis revealed that when the confidence level was ≥85%, the consistency between the model prediction and the manually labeled data reached 98.7%; when the confidence level was <85%, the consistency dropped to 72.3%. Therefore, 85% can be used as the dividing point between the high and low confidence intervals to effectively balance the efficiency and accuracy of automatic judgment.

[0187] After obtaining the output probability vector of the AI ​​model, the edge gateway performs the following judgment logic: calculate ; If Confidence ≥ 85%, the highest probability category output by the model will be directly adopted as the final classification result, and the result (class + confidence) will be uploaded to MES via 5G uplink; If Confidence < 85%, the edge gateway will not adopt the model results for the time being, but instead: Generate a "low confidence warning" message, which includes the original image URL, the model output probability vector, and the confidence value; The MES is requested via the 5G uplink for secondary review (e.g., by calling the backup rule model or the manual review interface). After receiving the request, MES pushes the task to the quality inspector's terminal, where a person can review and enter the results within 10 seconds.

[0188] This mechanism prevents low-confidence judgments from directly entering the production system, reducing the risk of misjudgment.

[0189] The confidence level is calculated as follows: The AI ​​model uses a deep convolutional neural network with a softmax function as the output layer. For the input image, it outputs the probability of each level (A / B / C / D). ,satisfy Confidence level is defined as the probability of the highest class, i.e. For example, when the output is (0.92, 0.05, 0.02, 0.01), the confidence level is 92%, and the model is classified as Grade A. When the confidence level is lower than a preset threshold (85% in this embodiment, which is determined based on the confidence-accuracy curve on the test set; above 85%, the accuracy reaches 98.7%), the edge gateway requests the MES for secondary review or manual intervention; otherwise (i.e., confidence level ≥ 85%), the local result is directly adopted, and the grade data is sent to the MES via the 5G uplink. The MES updates the current batch data and automatically synchronizes the grade information to all other batches of the same type (self-assessed at the same grade) and ungraded batches under the same supplier. At the same time, the edge gateway caches the most recent relevant records (e.g., 100 records) for model self-learning.

[0190] The MES central cloud periodically (e.g., every early morning) retrains the AI ​​model based on data reported by all edge nodes and pushes the updated model to each edge gateway, achieving edge-cloud collaborative evolution. To ensure model version consistency, the system adopts the following mechanism: Each model is assigned a globally unique version number upon release. MES ensures that all grading operations for the same batch of flowers use the same stable version model, and delays task allocation if the versions do not match. A blue-green deployment strategy is adopted to push the new model in batches, and then switch to it uniformly after all edge gateways have been updated, so as to ensure that only one stable version of the model is running in the entire system at any given time. During network outages, the same batch of flowers is logistics-constrained to fixed equipment for processing to avoid cross-device version differences.

[0191] When the 5G network is temporarily interrupted, each edge node independently completes the rating and stacker crane control, and temporarily stores the results locally. After the network is restored, it automatically synchronizes with the MES, and the system does not crash. At the same time, the MES periodically checks the model version consistency of all edge gateways. When an anomaly is detected, it automatically forces synchronization and issues an alarm, thereby preventing rating standard deviations caused by version inconsistencies.

[0192] like Figure 27 As shown, the WCS first breaks down the day's in-warehouse tasks by aisle and pre-distributes them to the edge controllers of each stacker crane. The edge controllers autonomously calculate the optimal picking sequence and travel path, and during execution, exchange real-time positions with the edge controllers of stacker cranes in adjacent aisles via 5G-D2D to achieve distributed collision avoidance. After completing each task, the edge controller only reports its status and time to the WCS, which is responsible for macro-level task balancing. For inbound boxed flowers, the WCS schedules the stacker cranes to automatically shelv the goods according to the storage locations allocated by the WMS, and the WMS updates the physical inventory and synchronizes it with the MES.

[0193] like Figure 25 As shown, the supplier appointment and data entry process includes: (1) Suppliers submit the flower delivery reservation information to MES through APP or mini-program. MES generates virtual inventory for use in subsequent grouping and auction.

[0194] (2) The supplier delivers the turnover box to the receiving port, scans the barcode for verification, enters the flower information into MES, and MES increases the actual physical inventory.

[0195] Setting up virtual inventory allows flowers that cannot arrive in the warehouse before the auction deadline to still participate in the auction that day, avoiding missing the trading window; the auction platform has a sufficient supply of flowers for buyers to bid on every day, unaffected by fluctuations in logistics time; after a successful bid, the supplier ships the flowers themselves, eliminating the transit link of the central warehouse. Real physical inventory is shipped centrally; flowers already in the warehouse are dispatched after the auction via WMS / WCS. Boxed flowers that have undergone random inspection and re-grading have standardized grade data, increasing buyer trust, and the entire process from warehousing to shipment is traceable. Suppliers can choose to have flowers stored in advance or pre-order and then ship directly, depending on their own logistics situation.

[0196] like Figure 26 As shown, when boxed flowers arrive at the warehouse entrance, the barcode is read by a barcode reader, and the information is uploaded to the MES (Manufacturing Execution System). The MES then sends instructions to the WMS (Warehouse Management System) / WCS (Warehouse Control System), and the WCS controls the inbound roller conveyor to send the boxed flowers to the next stage. The MES performs random checks according to the sampling ratio rules. Some boxed flowers are introduced into the intelligent grading equipment, while the rest flow to the automated warehouse via the direct inbound line.

[0197] The boxed flowers for random inspection arrive at the entrance of the intelligent grading equipment. Manual assistance is required to open the lids, and the equipment is activated to enter, where the grading process is completed automatically as described above. The MES dynamically decides whether to conduct random inspections or allow direct passage based on rules, while the WCS controls the physical flow.

[0198] like Figure 28 As shown, the MES (Manufacturing Execution System) aggregates the physical inventory in the WMS (Workshop Management System) and the virtual inventory on the supplier's side, and automatically performs intelligent grouping according to the auction platform's rules, combining flower boxes of the same grade and variety from different suppliers to generate an auction sequence. After grouping, the MES pushes a standardized auction data package to the auction platform. Buyers bid on the auction platform as needed, and transaction information is fed back to the MES in real time.

[0199] like Figure 29As shown, for auctioned goods, MES generates an outbound instruction and sends it to WMS / WCS. WCS then dispatches a stacker crane to retrieve the corresponding boxed flowers from the automated warehouse and transports them to the respective buyers' pickup points via a separate outbound main line. At the pickup point, the barcode reader identifies the boxed flowers, triggering an automatic printing device to print a purchase order. The pickup personnel affix the order, collect the flower boxes, and verify the information by scanning the barcodes using the buyer's app or mini-program, completing the loading and delivery process. The outbound and inbound main lines are independent of each other. Auctioned boxed flowers are directly transported to the outbound area via the independent outbound main line after being retrieved from the automated warehouse, bypassing the grading equipment and avoiding unnecessary duplicate checks, thus improving outbound efficiency.

[0200] For goods from the virtual warehouse, the supplier directly loads and ships the goods at the production site by scanning the code through the APP or mini-program, and updates the logistics status in real time.

[0201] like Figure 24 As shown, the complete operation process of the boxed flower operation subsystem includes the following steps: Step 1: Reservation and Data Entry: Suppliers reserve flower delivery, scan the barcode on the delivery box to enter flower information, and synchronize the data to MES (e.g., ...). Figure 25 (As shown).

[0202] Step 2, Warehousing and Sampling Inspection: When boxed flowers are put into the warehouse, the barcode reader identifies the barcode, the MES notifies the WMS / WCS, and the WCS controls the transmission; the MES performs sampling inspection according to the rules, and the sampled flowers are sent to the intelligent grading equipment.

[0203] Step 3, Automatic Classification: With manual assistance, the cover is opened, the vision component collects images, and the 5G edge gateway AI engine completes the classification within a preset time and sends the data back to MES.

[0204] Step 4: Shelving and Warehousing: After grading, the flowers and those not sampled are combined and enter the automated warehouse. The WCS dispatches the stacker crane to shelve the flowers according to the WMS locations, and the inventory is synchronized to the MES (e.g., Figure 27 (As shown).

[0205] Step 5: Grouping and Auctioning: MES automatically groups items according to the auction platform rules and pushes standard auction data packages; buyers bid, and transaction information is transmitted back to MES in real time (e.g., ...). Figure 28 (As shown).

[0206] Step Six: Outbound and Shipment: MES issues instructions, WCS dispatches stacker cranes to retrieve boxed flowers to each pickup point, and pickup personnel scan the barcodes for verification before loading and shipping (e.g., Figure 29 (As shown).

[0207] Other implementation methods for the boxed flower operation subsystem are as follows: Example 2: The difference from the above implementation method lies in replacing the barcodes on the turnover boxes with RFID electronic tags. RFID readers are installed at locations such as the warehouse entrance, grading equipment, and pick-up points to achieve contactless, batch reading, further shortening the identification time and improving the throughput. Specifically, each turnover box or boxed flower is affixed with an ultra-high frequency RFID tag storing a unique identification code. RFID reader antennas are installed at key nodes (warehouse entrance, sampling and sorting line, grading equipment entrance, and pick-up point). When a turnover box passes by, the tag information is automatically read, eliminating the need for manual alignment and scanning. When multiple turnover boxes pass by simultaneously, the reader can identify all tags in batches. This implementation method can solve the identification problem in the logistics process more efficiently, but the cost is higher than barcodes, and interference may occur in metal or liquid environments.

[0208] Example 3: The difference from the above implementation is that Automated Guided Vehicles (AGVs) replace some of the fixed roller conveyors and stacker cranes. Following WCS instructions, the AGVs autonomously transport flower boxes to designated quality inspection points, automated storage and retrieval system (AS / RS) buffer areas, or pickup points. Specifically, at the flower box receiving end, the AGV picks up the flower boxes from the receiving port and, based on the MES's sampling decision, transports the sampled flower boxes to the intelligent grading equipment area on the second floor, while the unsampled flower boxes are directly transported to the AS / RS buffer area. At the receiving end, the AGV picks up the flower boxes from the AS / RS exit and transports them to the designated pickup point. This implementation increases the system's flexibility and scalability, allowing for more flexible layout, but also correspondingly increases initial investment and maintenance costs.

[0209] Example 4: The difference from the above implementation is that the grading data and circulation process data generated by the intelligent grading equipment are uploaded to the blockchain, forming an immutable blockchain ledger. Suppliers, buyers, auction centers, and other parties can verify the complete lifecycle information of the flower boxes through the on-chain data. Specifically, after each random inspection and grading, the grading result, grading time, grading equipment number, operator, and other information are packaged into a transaction and submitted to the blockchain. Each warehousing, warehousing, and pickup operation is also simultaneously recorded on the blockchain. When picking up the goods, buyers can scan the barcode of the flower box through an app to query the trusted data of the entire chain from planting and grading to auction and warehousing for that batch. This implementation can solve the data trust and grading traceability problems to the greatest extent, but it increases system complexity and data processing burden.

[0210] Example 5: The difference from the above implementation lies in using Wi-Fi 6 or TSN (Time-Sensitive Networking) instead of 5G as the edge communication base. Specifically, in the Wi-Fi 6 solution, the edge gateway accesses the MES via Wi-Fi 6, offering advantages in coverage and cost. In the TSN solution, time synchronization and traffic scheduling ensure deterministic, low-latency transmission of grading data, meeting the stringent requirements of grading devices for network jitter. This implementation can be flexibly chosen based on actual scenario cost and performance requirements, all while remaining true to the core concept of "edge-cloud collaboration."

[0211] By deploying 5G edge gateways next to the intelligent classification devices, AI inference and classification are completed locally, reducing the classification response time from over 200ms in traditional architectures to less than 50ms, and bandwidth consumption is reduced by over 90%. Distributed path planning and real-time collision avoidance are achieved through 5G-D2D direct communication between stacker crane edge controllers, reducing stacker crane command latency from 100ms to less than 10ms. Even when the 5G network is interrupted, each edge node can still independently complete classification and collision avoidance control, achieving a system availability of 99.99%. Through a batch collaborative scheduling strategy, the same batch of flowers is integrated into the main line and stored in consecutive locations within the same aisle within a continuous time window, achieving centralized batch storage and significantly improving outbound retrieval efficiency.

[0212] The MES state machine enforces and standardizes the batch flow sequence of flower collection, ensuring full traceability. A virtual warehouse model enables "first-out, in-transit auction, and immediate pickup upon arrival," effectively resolving the conflict between logistics time and fixed auction times. The outbound scheduling controller aggregates orders by buyer ID, merges pickup instructions from the same lane, and dynamically allocates pickup points, enabling centralized loading of multiple orders from the same buyer. The system supports standard, pre-outbound, and hybrid business modes, flexibly adapting to local, intercity, and large-volume transaction scenarios.

[0213] The barrel / boxed flower operation subsystem of this invention is primarily aimed at large-scale regional flower auction centers (such as trading hubs covering a province or state). These centers process tens of thousands of barrels / boxes of flowers daily, possess professional operation and maintenance teams, and have sufficient financial investment to support the deployment and maintenance of advanced technologies such as edge-cloud AI models, UWB positioning, and AGV fleets. For small and medium-sized flower enterprises and planting bases, this invention also designs a lightweight access solution—a virtual library function—allowing them to participate in centralized trading and logistics coordination at auction centers without having to build their own complex systems.

[0214] After completing the training on the grading rules stipulated in this auction system, small and medium-sized flower enterprises and planting bases can access the system through the following steps: (1) Registration and certification: Enterprises register an account on the auction platform (front-end interaction layer), submit the qualifications of the planting base, and obtain access permission after the platform reviews them.

[0215] (2) Entering goods on mobile or PC: Enterprises use the supplier's APP or mini-program to take pictures of flower samples at the planting base. The system calls the cloud AI grading model (which shares the same model version with the auction center) to remotely assist in grading, or the enterprise can check the goods information according to the auction system's grading rules (A / B / C / D level) after self-inspection, including variety, grade, quantity, etc.

[0216] (3) Virtual warehouse tag: The goods entered by the enterprise are not actually physically stored in the warehouse, but a corresponding electronic voucher is generated in the virtual warehouse of the auction platform. The voucher contains fields such as the unique ID of the goods, grade, enterprise to which it belongs, and storage location (actually still in the cold storage of the enterprise's base).

[0217] (4) Participation in centralized auctions: The auction platform will list the goods in the virtual warehouse together with the goods in the physical warehouse. When the buyer bids, they cannot distinguish whether the goods are in the physical warehouse or the virtual warehouse, thus ensuring the fairness of bidding for small and medium-sized enterprises.

[0218] (5) Logistics coordination after auction: After the auction is completed, the platform generates a shipping notification based on the transaction results: For goods in the virtual warehouse, the platform will push the order to the corresponding enterprise. The enterprise will deliver the flowers to the centralized collection point designated by the auction center at the agreed time, or the platform's third-party logistics will pick up the flowers and ship them together.

[0219] (6) Unified settlement: The auction proceeds are collected by the platform and settled to small and medium-sized enterprises after deducting commissions; buyers only need to pay the platform once and do not need to settle with multiple bases separately.

[0220] The above description is only a specific embodiment of the present invention. Various examples and illustrations do not constitute a limitation on the substantive content of the present invention. Those skilled in the art can make modifications or variations to the above-described specific embodiments after reading the specification without departing from the substance and scope of the invention.

Claims

1. A smart flower operation and management system based on MES scheduling, characterized in that, include: MES serves as a unified central dispatch platform; The barrel flower operation subsystem includes a vertical warehouse, a filling machine, a visual grading system, a transaction system, and a closed-loop cleaning line. In the barrel flower operation subsystem, the entire process of empty barrel delivery, filling, height adjustment, flower placement, grading, warehousing, auction, delivery, picking up, empty barrel return, cleaning, and return to the warehouse is completed on the first floor. The boxed flower operation subsystem includes an inlet, an outlet, a sampling and sorting line, intelligent grading equipment, a return line, a straight line, an automated warehouse, and a stacker crane. In the boxed flower operation subsystem, the boxed flowers complete sampling and grading, warehousing, and pickup on the second floor. The staggered peak scheduling and warehousing coordination system is set at the inbound end of the bottled flower operation subsystem, including multiple distribution lines, one main roller distribution line, buffer zones and actuators set at the end of each distribution line, and a central scheduling controller; The outbound dispatch controller communicates with the MES and the central dispatch controller. In the barrel flower operation subsystem, MES uses a state machine to drive the state transition of the data model centered on the flower collection batch, and supports both physical library mode and virtual library mode.

2. The intelligent flower operation management system based on MES scheduling according to claim 1, characterized in that: In the staggered scheduling and warehouse collaboration system, sensors are installed at both the inlet and outlet of the buffer zone. The sensor at the inlet is used to detect the entry of the flower barrel and trigger a reporting event, while the sensor at the outlet is used to detect that the flower barrel has been successfully released into the main roller flow line. The central scheduling controller acquires the status of all buffer zones and the speed of the main roller streamline in real time, dynamically calculates the release time window, and issues release commands to the actuators according to the priority scheduling strategy or the batch collaborative scheduling strategy; the length of the release time window is calculated based on the ratio of the preset safety distance to the current running speed of the main roller streamline.

3. The intelligent flower operation management system based on MES scheduling according to claim 1, characterized in that, The boxed flower operation subsystem also includes a 5G edge gateway located next to the intelligent grading device and an edge controller located in the stacker crane control cabinet. The 5G edge gateway is used for local AI inference to complete the grading of the boxed flowers and only uploads the grading results to the MES. The edge controllers of multiple stackers achieve distributed path planning and real-time collision avoidance through 5G-D2D direct communication. When the 5G network is interrupted, the 5G edge gateway and edge controller can still independently complete the grading and collision avoidance control, and automatically synchronize with the MES after the network is restored.

4. The intelligent flower operation management system based on MES scheduling according to claim 1, characterized in that, In the barrel flower operation subsystem, the state machine mandates that the status of the flower collection batch changes unidirectionally in the following order: pending collection, collected, graded, grouped, auctioned, auction in progress, completed, and so on. In physical warehouse mode, graded flower barrels enter the automated storage system and then participate in the auction; in virtual warehouse mode, graded flower barrels are pre-shipped and enter the distribution and transportation system, and are simultaneously marked as virtual warehouse status in MES to participate in the auction in parallel, realizing the principle of first shipment, auction in transit, and immediate pickup upon arrival.

5. The intelligent flower operation management system based on MES scheduling according to claim 1, characterized in that, The outbound dispatch controller receives buyer orders, aggregates all batches ordered by the buyer by buyer ID, queries the batch-aisle mapping table, and if all batches are located in the same aisle, generates a merged pickup instruction for the stacker crane to pick up the goods at once; dynamically selects an available pickup port and directs all batches of flower barrels or boxed flowers to the same pickup port for centralized loading; when too many of the same batches cause congestion at a pickup port, it dynamically allocates 2 to 3 pickup ports as dedicated pickup ports for the same buyer.

6. A method for intelligent operation and management of flowers based on MES scheduling, applied to the intelligent operation and management system for flowers based on MES scheduling as described in any one of claims 1 to 5, characterized in that, Includes the following steps: MES issues an empty barrel outbound instruction. After the empty flower barrels are filled, height adjusted, flowers placed, and visually graded, they are stored in the warehouse. After the auction is completed, they are outbound and picked up. The empty barrels are cleaned in a closed-loop cleaning line and then returned to the warehouse. Boxed flowers enter through the warehouse entrance. The boxed flowers that are sampled are transported to the intelligent grading equipment to complete the AI ​​grading and then converge via the return line. The boxed flowers that are not sampled pass directly through the straight line. Finally, they are all sent to the vertical warehouse. Among them, the staggered scheduling of the flower barrels before they are put into storage is executed by the central scheduling controller. It obtains the status of all the branch line buffer zones and the speed of the main roller streamline in real time, dynamically calculates the release time window, and releases the flower barrels to the main roller streamline in sequence according to the scheduling strategy.

7. The intelligent operation and management method for flowers based on MES scheduling according to claim 6, characterized in that, The scheduling strategy includes a priority scheduling strategy and a batch collaborative scheduling strategy: In the priority scheduling strategy, the central scheduling controller calculates the priority of each branch line according to the formula: priority = buffer occupancy rate weight × buffer occupancy rate + waiting time coefficient weight × waiting time coefficient + historical allocation balance coefficient weight × historical allocation balance coefficient. In the batch collaborative scheduling strategy, when multiple flower barrels with the same batch identifier are detected at the head of the buffer queue of the same diversion line, a continuous release time window is allocated to the batch, so that it can continuously merge into the main roller flow line with the minimum safety interval and be stored in the continuous storage location of the same lane.

8. The intelligent operation and management method for flowers based on MES scheduling according to claim 6, characterized in that, The automatic grading process for boxed flowers includes: The 5G edge gateway runs an AI model to complete the level determination locally and generates structured results that include batch, level and confidence level; When the confidence level is lower than the preset threshold, the edge gateway requests the MES to perform a second review or manual intervention; otherwise, it directly adopts the local result and uploads the level data to the MES. The grade data will be automatically synchronized and updated to other batches of the same variety and grade under the same supplier that have not been sampled. The MES central cloud periodically retrains the AI ​​model based on the data reported by the edge nodes and pushes the updated model to each edge gateway.

9. The intelligent operation and management method for flowers based on MES scheduling according to claim 6, characterized in that, The closed-loop cleaning process for the flower barrels includes: after delivery, the empty flower barrels first pass through the tipping and water-pouring station to pour out the residual liquid, then pass through the tipping internal cleaning station for high-pressure water and atomized cleaning agent spraying, then pass through the tipping preliminary drying station for the first round of drying, and finally pass through the straightening deep drying station for the second round of deep drying; after drying, the qualified empty flower barrels are returned to the vertical warehouse to await the next dispatch by MES.

10. The intelligent operation and management method for flowers based on MES scheduling according to claim 6, characterized in that, The barrel-type flower operation subsystem supports three business models: In the standard mode, after the classification is completed, the goods are stored in the physical warehouse, then grouped, auctioned, distributed and transported, and picked up. In the pre-out mode, after classification, the virtual warehouse is marked and pre-out shipment is carried out. During transportation, batches in the virtual warehouse are grouped and auctioned in parallel. After the auction, the virtual warehouse is released. After the physical goods arrive at the destination pick-up point, they are collected and picked up. In the mixed mode, some flower barrels are processed according to the standard mode, while others are processed according to the pre-shipment mode, with both batches of goods being executed in parallel.

Images