Robot mobile shelf system order and shelf online collaborative selection and sorting method

Abstract

The application discloses a kind of robot mobile shelf system order and shelf online collaborative selection and sequencing method, belong to intelligent warehousing technical field.The method faces the online running scene of order continuous arrival and shelf dynamic call, triggers online redetermination when order arrives, order is completed, shelf is released or reaches the rolling decision point;Candidate order set and candidate shelf set are constructed to each workstation;According to the satisfaction ratio of shelf to order remaining demand, order urgency, shelf estimated arrival time and workstation load condition, dynamically assess the real-time adaptation value of order and shelf, real-time service value of shelf to workstation and insertion benefit value of order to workstation;Accordingly, real-time selection target shelf and execute conflict resolution and call, select target order and execute order rearrangement and insertion;According to real-time picking progress, update system state.The method can reduce the number of shelf handling, shorten order response time, improve multi-workstation collaborative picking efficiency.

Description

Technical Field

[0001] This invention belongs to the field of intelligent warehousing technology, specifically relating to a method for online collaborative selection and sorting of orders and shelves in a robotic mobile shelving system. Background Technology

[0002] In recent years, the rapid development of e-commerce and the increasing demands for timely order fulfillment have driven the transformation of warehousing systems towards automation and intelligence. Robotic mobile shelving systems use mobile robots to transport shelves to picking workstations, allowing pickers to complete picking at fixed workstations, after which the robots remove the shelves from the workstation. This model significantly reduces employee walking and searching time, improves order picking efficiency, and has become an important direction for the development of e-commerce warehousing automation and smart logistics.

[0003] However, while robotic mobile shelving systems improve picking efficiency, they also bring complex online scheduling challenges. These systems are characterized by continuous order arrivals, dynamic shelving allocation, real-time inventory changes, multi-location storage of individual items, and parallel operation across multiple workstations. Simultaneously, a single workstation can handle multiple orders, an order may be fulfilled by multiple shelving units, and a single shelving unit may cover multiple order demands, creating a dynamic many-to-many matching relationship between order demand and shelving supply. During the picking process, the continuous entry and exit of orders into the workstation buffer forms the order processing flow, while the sequential arrival and picking of shelving units forms the shelving processing flow. The former determines the shelving handling needs, and the latter determines the order fulfillment status; the two are highly coupled. In multi-workstation scenarios, shelving allocation conflicts and uneven workload distribution may also occur. Therefore, online collaborative determination of order selection, order sorting, shelving selection, and shelving sorting is crucial for improving the system's picking efficiency and reducing operating costs.

[0004] Existing patents related to order picking in robotic mobile shelving systems have studied specific aspects such as wave division, order release, order allocation, and shelving selection. However, these studies often focus on single decisions within specific scenarios and do not fully characterize the dynamic coupling relationships between resources such as orders, shelving, workstations, and robots during online operation. Related journal research also largely bases its work on given orders or fixed wave orders, typically assuming that the system state and operational parameters are known before decision-making, and generating processing solutions through mathematical programming or metaheuristic algorithms. While these offline optimization methods can achieve good results in static scenarios, they struggle to meet the online operational requirements of real-world e-commerce warehousing systems with continuous order arrivals and real-time system state changes. When multiple workstations operate in parallel, shelving call conflicts and uneven workloads can easily arise, affecting system stability. Commonly used methods in practice, such as first-come-first-served, random selection, proximity-based calling, or locally greedy rules, while offering fast responses, fail to fully utilize the dynamic relationships between order demand, shelving supply, and workstation states, easily leading to excessive shelving handling, frequent workstation waiting, and decreased system throughput. Therefore, there is an urgent need for a collaborative selection and sorting method for orders and shelves in the online picking environment of robotic mobile shelving systems. This method should be able to generate order insertion, target shelf selection, and shelf arrival sorting schemes in real time based on the dynamic arrival of orders and the online picking status of workstations. This would reduce shelf handling and workstation waiting time, improve order picking efficiency, and provide scientific and efficient decision support for online order picking in robotic mobile shelving systems. Summary of the Invention

[0005] This invention aims to overcome the shortcomings of existing order picking methods in robotic mobile shelving systems, which are mostly applicable to fixed order pools or local decision-making scenarios and are difficult to adapt to the collaborative optimization of orders, shelves, and workstations in online dynamic operating environments. It provides a method for online collaborative selection and sorting of orders and shelves in robotic mobile shelving systems. This method, based on an event-triggered mechanism, implements rolling decisions in an online operating environment where orders continuously arrive, product inventory and shelf status change in real time, and the load of multiple workstations fluctuates dynamically. It dynamically evaluates the matching relationship between orders and shelves, the real-time service value of shelves to workstations, and the benefits of inserting orders into workstations. It collaboratively determines inserted orders, target shelves, and shelf arrival order, thereby improving the matching efficiency between order demand and shelf supply, reducing shelf handling and workstation waiting times, shortening order response time, and improving the collaborative picking efficiency of multiple workstations.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A method for online collaborative selection and sorting of orders and shelves in a robotic mobile shelving system includes the following steps: Step S1: Build the real-time status of the system.

[0007] At this moment of decision-making The system collects and updates the real-time status of the robotic mobile shelving system. This real-time status includes the dynamic order pool status, workstation status, shelving status, and operation time status.

[0008] Among them, the workstation set is set as The shelving is a collection of The product collection is .

[0009] For any order Record it at that moment The remaining demand vector is ,in, Indicates order At any moment For goods The remaining demand.

[0010] For any shelf Record it at that moment The inventory vector is ,in, Indicates shelf At any moment For goods The available supply quantity.

[0011] For any workstation Record it at that moment The set of orders to be picked is The current processing shelf is The queue of shelves that have been reserved but not yet picked is The workstation has a remaining cache capacity of .

[0012] The dynamic order pool status includes: a collection of newly arrived orders. There is already a set of orders to be assigned. The system includes a collection of orders that are being picked but not yet completed, the arrival time of each order, the waiting time, the urgency parameter, and the remaining order requirements for the goods to be picked.

[0013] The workstation status includes: the current set of orders being picked at each workstation, the currently processed shelves, the queue of shelves that have been reserved but not yet arrived, the remaining cache capacity, the cumulative number of allocated orders, and the cumulative number of shelves that have been called.

[0014] The shelf status includes: the current position of each shelf, the available quantity of each product in the shelf, the current service workstation identifier, and the shelf status marker; the shelf status marker includes one of the following: idle, reserved, in transit, being picked upon arrival, and awaiting release.

[0015] The operational time status includes: the estimated release time of the shelves at the workstation, the estimated arrival time of the candidate shelves at each workstation, and the estimated arrival sequence of the reserved shelves.

[0016] Step S2: Determine whether online re-decision is triggered.

[0017] Detect the current decision moment Has a preset re-decision trigger event occurred? If a preset re-decision trigger event occurs, then while keeping the freeze execution plan unchanged, steps S3 to S10 are executed for unfrozen orders, unfrozen shelves, and unfrozen workstation resources; if no preset re-decision trigger event occurs, then the current freeze execution plan continues to be executed.

[0018] The preset re-decision triggering events include any of the following events: (1) new order arrival event; (2) order completion event; (3) shelf release event; (4) reaching the rolling decision point event, wherein the rolling decision time window is When multiple re-decision events are triggered at the same decision moment, they are merged into a single online re-decision.

[0019] The freeze execution scheme refers to maintaining the allocation relationship, arrival order, and corresponding execution status of the currently processed shelves, called shelves, reserved shelves, and assigned orders that have entered the execution window unchanged within a preset freeze duration.

[0020] Step S3: Construct the candidate set.

[0021] Based on the real-time status of the system, for each workstation Construct a candidate order set and candidate shelf collection .

[0022] First, for each workstation From the newly arrived order collection and existing unassigned order sets Construct a candidate order set The candidate order set includes at least one of the following orders: (1) orders that can be processed by the current shelf. (2) Orders that can be fulfilled immediately at least partially; (3) Orders that can be picked by pre-booked but not yet arrived rack queues. Subsequent orders that meet at least some of the requirements; (3) with workstations Currently picking order collection Orders with a high degree of subsequent matching, the degree of subsequent matching can be determined based on the overlap between the candidate order and the current order being picked, the available shelf set or the reserved shelf set; (4) If all the above order sets are empty, the newly arrived order set and the existing unassigned order set are set as the candidate order set of the workstation.

[0023] Next, a candidate shelf set is constructed from the shelves that are currently idle and have not been reserved, shipped, frozen, or locked by any workstation. Includes at least one of the following shelves: (1) capable of meeting the needs of workstations Currently picking order collection (1) Shelves that can meet at least part of the remaining demand; (2) Shelves that can meet at least part of the demand for candidate orders.

[0024] Step S4: Calculate the real-time fit value (A value) between the order and the shelf.

[0025] For each workstation ,make This represents the set of orders that are currently being picked and still have remaining demand for goods. For any order... This refers to orders currently being picked that still have remaining stock and candidate orders, along with any shelf. This refers to the current picking shelves, the shelves that have been reserved but not yet arrived, and the candidate shelves, which are used to calculate orders. Shelves With workstation At the present moment Real-time adaptation value , is called The value is used to represent the value at the current decision moment. Shelves At the workstation fulfilling orders The overall degree of adaptation to remaining demand. The higher the real-time adaptation value, the better the shelf life. The more suitable it is for fulfilling orders At the workstation The remaining demand for goods.

[0026] The real-time adaptation value Calculate according to the following formula: in, , indicating shelves At any moment For orders The proportion of immediate satisfaction of remaining demand for goods; Indicates order At any moment The urgency, and The urgency level can be determined based on at least one of the following: order waiting time, promised completion time, and order priority. This indicates the shelf after normalization. Arrival at the workstation The estimated arrival time is calculated as follows: in, Indicates shelf At any moment Expected to be transferred to workstation The original estimated arrival time; Indicates the length of the rolling decision window. For workstations Current processing shelves ,make , When an idle mobile robot is available, for shelves that are currently idle and not locked by other workstations... Its original estimated arrival time is calculated according to the following formula: in, Indicates shelf Reassigned to workstation The planned call start time: when the shelf At any moment When a mobile robot is idle, not locked by other workstations, and has available space, then... . Indicates that it is assigned to be used for shelving Move to workstation Mobile robots; Represents robots From current location to shelf The estimated driving distance from the current location; Indicates from the shelf Location to workstation The expected transport distance; This indicates the robot's average speed. This indicates the preparation time required for lifting, placing, and docking the shelves.

[0027] For those who have already made an appointment at the workstation During transport or expected to arrive at the work station The shelves The determination is made by the warehouse control system based on its current transportation status, reservation relationships, expected arrival sequence, and route conflict resolution results.

[0028] in, This is a non-negative weighting parameter, which can be preset based on historical operating data, simulation results, or human experience, or it can be updated periodically based on the system's operating performance. Preferably, let: .

[0029] Step S5: Perform shelf selection, reservation, and conflict resolution.

[0030] make Indicates workstation At any moment Whether new shelves are needed; a value of 1 indicates they are needed, and a value of 0 indicates they are not. And when there are idle robots, the workstation Proceed to the shelf selection step.

[0031] Let (1) Workstation Currently, there are no processing shelves, and the orders that are being picked still have remaining items to be ordered. or candidate order set (2) The current processing racks are insufficient to meet the needs of the workstation. (3) Any remaining demand from orders with remaining merchandise; (4) Subsequent merchandise demand at the workstation that is not yet covered by the current shelf or a reserved shelf; Candidate shelf collection It is not empty, and at least part of the demand for the candidate order cannot be covered by the currently processed shelf or the reserved shelf.

[0032] (5.1) Calculate the real-time service value (B value) of the rack to the workstation. Candidate Shelves and workstation Define the real-time service value of the shelf to the workstation. , is called The value is used to represent the value at the current decision moment. Shelves Called to workstation The overall benefits generated. The higher the real-time service value, the greater the contribution of the rack to the workstation.

[0033] The real-time service value Calculate according to the following formula: in, Indicates time Assigned to workstation A collection of orders that have been picked and still have remaining goods to be ordered; Indicates time workstation The set of candidate orders; Indicates order Assigned to workstation The feasibility coefficient; if the workstation There is remaining cache capacity and orders If it can be inserted, then ,otherwise ; It is to prevent extremely small positive numbers with a denominator of 0; This indicates the shelf after normalization. Arrival at the workstation The estimated arrival time; Let be a non-negative weighting parameter. Preferably, let: .

[0034] (5.2) Perform shelf reservation, conflict resolution and dispatch. For workstations that need to select shelves, select from the candidate shelf set. Select real-time service value The largest shelf is designated as the target shelf. If the target shelf is not locked by other workstations during the expected usage period, a reservation is made directly for it.

[0035] If multiple workstations compete for the same shelf, conflict resolution is performed, specifically including: (1) calculating the real-time service value of each competing workstation using the shelf; (2) retaining the reservation request of the workstation with the higher real-time service value for the shelf; (3) selecting the second-best shelf from the corresponding candidate shelf set for the remaining workstations; (4) if the second-best shelf cannot be directly called at the current time, generating a delayed call plan and updating the estimated arrival time; (5) if there is no second-best shelf in the corresponding candidate shelf set, the workstation is placed in a waiting state, and the candidate shelf set is reconstructed in the next round of decision-making.

[0036] When multiple target shelves with the same real-time service value exist at the same workstation, the shelf closest to the workstation is selected first; if the distance is still the same, the shelf with the smaller shelf number is selected.

[0037] When multiple workstations do not need to select shelves, or when all robots have been assigned, the system will call up each reserved shelf and update the workstation status, shelf status, and operation time status.

[0038] Step S6: Perform order selection, reservation, and insertion.

[0039] When workstation This step is executed if there is remaining cache capacity and the candidate order set is not empty.

[0040] (6.1) Calculate the insert revenue value (C value) of the order to the workstation. For candidate orders and workstation Define the insert revenue value of an order to a workstation as... , is called The value is used to represent the value at the current decision moment. ,Order Insert workstation The overall benefit generated. A higher insertion benefit value indicates that the insertion decision is more conducive to improving order picking efficiency.

[0041] The inserted revenue value Calculate according to the following formula: in, Indicates order With workstation Current processing shelves Real-time adaptation values ​​between them; Indicates workstation At any moment Shelf queues that have been reserved but not yet arrived for picking; Indicates order With workstation The maximum fit value in the reserved shelf queue; Indicates order At any moment The urgency; Indicates order Insert workstation The resulting load imbalance penalty; Let be a non-negative weighting parameter. Preferably, let: .

[0042] In this invention, load balancing metrics are set for each workstation, and a negative penalty is imposed on the insertion scheme if the load on a workstation exceeds a preset allowable deviation after an order is inserted. This method avoids excessive concentration of orders on a few workstations, achieving multi-workstation load balancing control. Specifically, for calculation... Set up workstations At any moment The current load is The load It can be determined based on at least one of the following: the occupancy level of the workstation buffer area, the number of remaining items in the picking order, and the number of shelves that have been reserved but not yet arrived.

[0043] In a preferred embodiment, the workstation The load index can be expressed as: in, Indicates time workstation The collection of orders being picked; Indicates workstation Fixed cache size; Indicates order For goods The remaining demand; Indicates workstation Shelf queues that have been reserved but not yet arrived for picking; and These are the maximum remaining demand and the maximum number of reserved shelves used for normalization, respectively. Let be a non-negative weighting parameter. Preferably, let: .

[0044] Set order At any moment The incremental load brought by inserting the workstation is In a preferred embodiment, Available according to orders The remaining demand for goods is determined, for example: When orders Inserted into workstation After that, workstation The update load is: Correspondingly, orders Insert workstation The average load of the system after that is: Order Insert workstation After that, workstation The deviation relative to the system average load is: Further, define the order Insert workstation The resulting load imbalance penalty is: in, This indicates the allowable load deviation threshold. When This indicates that inserting the order will not cause issues on the workstation. If the load deviation exceeds the allowable range, no load imbalance penalty will be applied; when... When this occurs, it indicates that inserting this order will cause the workstation to... If the system average load is too high, a load imbalance penalty is applied to the portion exceeding the threshold.

[0045] (6.2) Perform order reservation, rearrangement and insertion. For each workstation with remaining cache capacity and a non-empty candidate order set From the candidate order set Select Insert Earnings Value The largest order is designated as the target order, and reservations are made for the target order.

[0046] To further ensure a balanced workload across all workstations, a cyclical, rotating appointment system will be adopted: appointments will be made according to the workstation's schedule. Orders are reserved sequentially from workstation 1. After a single round of reservations is completed, a new round of reservations begins from workstation 1. During the reservation process, the remaining cache capacity of each workstation and whether the order has already been reserved by another workstation are checked simultaneously to complete the orderly reservation of orders in a loop. The loop reservation ends when the remaining cache capacity of all workstations is 0, or when all candidate orders have been reserved.

[0047] If multiple workstations compete for the same order, the order is rearranged, specifically including: (1) calculating the benefit value of each competing workstation inserting the order; (2) retaining the reservation request for the order from the workstation with the higher insertion benefit value; (3) selecting the second-best available order from the corresponding candidate order set for the remaining workstations; (4) if there is no second-best order in the corresponding candidate order set, skip the workstation and reconstruct the candidate order set in the next round of decision-making.

[0048] When there are multiple target orders with the same insertion benefit value, the order with the earlier arrival time is selected first; if the arrival times are still the same, the order with the smaller order number is selected first.

[0049] After all round-robin appointments and order conflict resolution processes are completed, the final appointment order is inserted into the order cache of the corresponding workstation, and the following information is updated: (1) Workstation In the order collection A collection of orders that are being picked and still have remaining goods to be ordered. (2) Dynamic order pool status; (3) Remaining cache capacity of workstation; (4) Cumulative number of allocated orders for workstation; (5) Mapping relationship between orders and workstations.

[0050] Step S7: Calculate the job time and update the status.

[0051] (7.1) Calculate the estimated release time of the shelving Based on the current assigned orders and shelves, calculate the estimated release time of the current shelves, and update the dynamic order pool status, workstation status, shelf status, and operation time status.

[0052] workstation Current Shelf The expected release time is calculated using the following formula: in, Indicates shelf At the workstation The expected release time; Indicates shelf At the workstation The estimated start time for picking is calculated as follows: If the reserved shelf arrives at the workstation and there is no shelf being processed in front of it, the shelf will be set as the current processing shelf upon arrival and picking will begin immediately. If there is a shelf being processed in front of it, the estimated start time for picking is the estimated release time of the previous shelf plus a preparation time. ; Indicates time workstation A collection of orders that are being picked and still have remaining goods to be purchased; Indicates order At any moment For goods The remaining demand; Indicates workstation Shelves At any moment Available products quantity; Indicates workstation At any moment Effective picking rate.

[0053] (7.2) Update status Given a fixed order and shelf space, anticipate changes in shelf inventory and remaining demand for each order in the workstation buffer area after the current picking cycle. (Workstation) Current Shelf Domestic goods The available supply margin is updated according to the following formula: Step S8: Continue to assign shelves to idle robots.

[0054] Determine whether there are still idle robots in the system, and whether there are still unmet order demands after each workstation has completed its expected picking from the currently reserved shelves. If there are idle robots in the system, and at least one workstation still has unmet demands, return to step S3, and continue with steps such as candidate set update, shelf selection, reservation, and call-up based on the updated remaining order demands, shelf status, robot status, and workstation status; otherwise, proceed to step S9.

[0055] Step S9: Perform order picking.

[0056] Based on the aforementioned calculation results, the system executes the actual order picking as planned.

[0057] When a workstation completes a round of picking on the current shelf, the following judgment is made: if there is a completed order, the completed order is removed from the workstation and the corresponding position in the buffer is released; if there is no completed order and the current shelf cannot continue to meet the remaining demand of the picking orders, the current shelf is set to the pending release state.

[0058] During the picking process, if events such as new orders arriving, orders being completed, shelf releases, or the rolling decision point being reached occur, the system's real-time status is updated, and the next round of online re-decision is triggered.

[0059] Step S10: Repeat the execution.

[0060] Repeat steps S2 to S9, continuously update the real-time status of the system based on events such as new order arrival, order completion, shelf release, and reaching the rolling decision point, and perform online collaborative selection and dynamic sorting of orders and shelves.

[0061] When the set of new orders that have arrived up to the current decision cycle, the set of existing orders to be assigned, and the set of picking orders at each workstation are all empty, and there are no reserved but not completed shelves or shelves in transit, the current round of online collaborative selection and sorting process ends.

[0062] In one optional implementation, after the current round of online collaborative selection and sorting process ends, the following are output for each workstation: the dynamic insertion sequence of orders, the dynamic call sequence of shelves, the arrival time of each shelf, the release time of each shelf, and the completion time of each order.

[0063] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) Based on events such as order arrival, order completion, shelf release and arrival rolling decision point, the present invention triggers online re-decision. Under the condition of continuous order arrival and real-time changes in system status, it can dynamically update the order pool, shelf status and workstation status, thereby adapting to the online operation scenario of continuous order arrival and real-time changes in product inventory and shelf status, avoiding the problem that traditional fixed order pool or offline wave method is difficult to respond to real-time changes in the system in a timely manner.

[0064] (2) This invention evaluates the shelf's ability to meet order demands, the shelf's service value to workstations, and the benefits of inserting orders into workstations by using real-time adaptation value, real-time service value, and insertion benefit value, respectively. Based on the above evaluation results, the system can assign shelves to workstations with the highest service value and dynamically insert new orders into the most suitable workstations, thereby improving the matching efficiency between order demand and shelf supply, shelf utilization, reducing shelf handling frequency, and improving order picking efficiency.

[0065] (3) The present invention uses a shelf reservation, locking and conflict resolution mechanism to determine the shelf ownership based on real-time service value when multiple workstations compete for the same shelf, and selects a suboptimal shelf or generates a delayed call plan for workstations that have not obtained the shelf, thereby effectively solving the shelf call conflict problem under the parallel operation of multiple workstations, reducing the repeated rearrangement of shelves and workstation waiting.

[0066] (4) This invention introduces a load imbalance penalty during the order insertion process at the workstation to avoid excessive concentration of orders at some workstations and improve the collaborative picking performance of multiple workstations. At the same time, it sorts and updates the status based on the estimated arrival time and release time of the shelf, making the scheduling scheme more consistent with the actual execution situation, thereby reducing the waiting time of workstations and improving the order response speed and system throughput. Attached Figure Description

[0067] Figure 1 This is a schematic diagram of the overall process of the method of the present invention. Detailed Implementation

[0068] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings and technical solutions.

[0069] like Figure 1 As shown, this invention provides a method for online collaborative selection and sorting of orders and shelves in a robotic mobile shelving system. The system includes... workstations A shelf, The system consists of various products and a continuous flow of arriving orders. Each order contains a demand for several items to be picked, and each shelf stores a certain number of items and their corresponding inventory levels.

[0070] At the initial moment, all workstations have an empty set of picking orders, an empty currently processed shelf, and an empty queue of reserved but not yet arrived shelves. Each shelf is in one of the following states: idle, reserved, in transit, in the process of picking, or waiting to be released. The dynamic order pool can be empty or it can contain some orders that have arrived but have not yet been assigned.

[0071] During system operation, this method applies at each decision-making moment. Perform the following operations: Step 1: Build the system's real-time status. Collect and update the dynamic order pool status, workstation status, shelf status, and operation time status, including order arrival time, waiting time, urgency parameters, remaining order demand, current shelf location, remaining shelf inventory, workstation's picking order set, current shelf status, reserved shelf set, and various estimated times.

[0072] Step 2: Determine if an online re-decision has been triggered. If not, maintain the current frozen execution plan; if triggered, continue with the subsequent steps while keeping the frozen execution plan unchanged.

[0073] Step 3: Construct candidate sets. Based on the currently processed shelves, reserved shelves, remaining product demand for orders in pick-up, and the dynamic order pool, construct candidate order sets and candidate shelf sets for each workstation.

[0074] Step 4: Calculate the real-time fit value (A value) between the order and the shelving. Based on the immediate fulfillment ratio of the remaining order demand by the shelving, the order urgency, and the estimated arrival time of the shelving at the workstation, calculate the real-time fit value between the order and the shelving. Used to indicate shelves For orders At the workstation The degree of overall compatibility.

[0075] Step 5: Perform shelf selection, reservation, and conflict resolution. When a workstation needs a new shelf and there are available robots, the real-time service value of the shelf for the workstation is calculated based on the shelf's compatibility with the workstation's currently picking orders with remaining goods and candidate orders, as well as the estimated time for the shelf to arrive at the workstation. (B value). Select the shelf with the highest real-time service value from the candidate shelf set as the target shelf for reservation; if multiple workstations compete for the same shelf, conflict resolution is performed, and a suboptimal shelf is selected for the workstation that does not obtain the shelf, or a delayed call plan is generated. When multiple workstations do not need to select a shelf or all robots have been assigned, the reserved shelves are called, and the workstation status, shelf status, and operation time status are updated.

[0076] Step 6: Perform order selection, reservation, and insertion. When a workstation has remaining cache capacity and the candidate order set is not empty, calculate the order's benefit to the workstation based on the real-time fit between the order and the workstation's currently processed shelf, the maximum fit between the order and the reserved shelf, the order's urgency, and the load imbalance penalty caused by order insertion. (C value). Select the order with the highest insertion benefit from the candidate order set as the target order for reservation; if multiple workstations compete for the same order, perform order rearrangement and select the second-best order for the workstations that do not receive the order; after all round-robin reservations and order conflict rearrangement processes are completed, insert the final reservation order into the order cache of the corresponding workstation, and update the workstation's picking order set, the order set with remaining product demand, the dynamic order pool status, the workstation's remaining cache capacity, the workstation's cumulative number of allocated orders, and the mapping relationship between orders and workstations.

[0077] Step 7: Calculate the operation time and update the status. Based on the current assigned orders and shelves, calculate the estimated release time of the current shelves, and predict the changes in shelf inventory and remaining demand for each order in the workstation buffer area after the current round of picking.

[0078] Step 8: Reassign shelves to idle robots. Determine if there are still idle robots in the system, and whether there are still unmet order demands after each workstation has completed its expected picking from the currently reserved shelves. If there are idle robots in the system, and at least one workstation still has unmet demands, return to Step 3, and continue updating the candidate set and selecting shelves based on the updated remaining order demands, shelf status, robot status, and workstation status; otherwise, proceed to Step 9.

[0079] Step 9: Execute order picking. Based on the aforementioned calculations, execute actual order picking as planned. If a completed order exists, remove it from the workstation and release the corresponding location in the buffer; if no completed orders exist and the current shelf cannot meet the remaining demand for picking orders, set the current shelf to a pending release state. During the picking process, if events such as order arrival, order completion, shelf release, or rolling decision point occur, update the system status and trigger the next round of online re-decision.

[0080] Step 10, repeat. Repeat steps 2 through 9 until the newly arrived order set, the existing unassigned order set, and the picking order set of each workstation are all empty, and there are no reserved but unfinished processing shelves or shelves in transit. Then, end the current round of online collaborative selection and sorting process.

[0081] The following example, using 2 workstations, 4 shelves, 4 robots, 5 orders, and 6 types of goods, illustrates the online collaborative selection and sorting method for orders and shelves in the robot mobile shelf system described in this invention. At the initial moment, 3 orders have arrived and are awaiting allocation; the remaining 2 orders will arrive online at a subsequent decision-making time. The specific steps are as follows: (1) First round of decision-making Step S1: Build the real-time status of the system.

[0082] At this moment of decision-making It collects and updates the real-time status of the system. The workstation set is... The shelving is a collection of The product collection is Each workstation Fixed cache size .

[0083] At the current decision-making moment, all four robots are idle, located at their initial positions, and have not yet been assigned any handling tasks. The system already has three pending orders, with the remaining demand vector... As shown in Table 1 below: Table 1

[0084] For example, order 1 requires 3 items of product 1 and 2 items of product 2, with no remaining demand for other products.

[0085] At the current decision-making moment, the inventory vector for each shelf... As shown in Table 2 below: Table 2

[0086] For example, shelf 1 can provide 4 items of product 1, 5 items of product 2, and 4 items of product 3 at the current decision moment, but not other items.

[0087] For any workstation Record the workstation status at the current decision-making moment, including the current order collection being picked. Current processing shelves Reservation queue of shelves that have not yet arrived Workstation remaining cache capacity The total number of orders assigned to the workstation is 0, and the total number of shelves called up by the workstation is 0.

[0088] The dynamic order pool status records the set of newly arrived orders. There is already a set of orders to be assigned. The set of orders already picked but not yet completed is empty; the arrival time of each order is the current decision time; and the waiting time is 0. To simplify the calculation in this embodiment, we assume that the initial orders have the same priority and that the urgency parameter is uniformly set. Information on remaining unpicked items for each order As mentioned above.

[0089] In the shelf status, the current position of each shelf is its current storage location, and the available quantity of each product in the shelf is displayed. As mentioned earlier, the current service workstation identifier is empty, and the status markers for all shelves are all idle.

[0090] In the current work schedule, since there are no shelves being processed or reserved but not yet arrived at any of the workstations, the estimated release time for the current shelves is empty. The estimated arrival time of the candidate shelves to each workstation will be calculated in subsequent steps based on the shelf location, robot status, and workstation location.

[0091] Step S2: Determine whether online re-decision is triggered.

[0092] Detect whether a pre-defined re-decision trigger event has occurred. At this point, the system is at the initial decision-making point, which is also the first round of rolling decision-making point, thus triggering online re-decision and executing steps S3 to S10. The rolling decision time window is set... The next rolling decision point is... Second.

[0093] Step S3: Construct the candidate set.

[0094] Based on the current real-time status of the system, for each workstation Construct a candidate order set and candidate shelf collection .

[0095] First, for each workstation Since neither workstation currently has any picking orders or processing shelves, and both have remaining buffer capacity, any existing orders awaiting assignment can be considered as candidate orders. For either workstation... ,have .

[0096] After that, for each workstation Since all shelves are currently idle and not booked, shipped, frozen, or locked by any workstation, the shelf that can meet at least part of the candidate order's demand will be selected from all shelves. Based on order demand and shelf inventory, shelves 1, 2, 3, and 4 can all meet a portion of the remaining demand for at least one candidate order. Therefore, for any workstation... ,have .

[0097] Step S4: Calculate the real-time fit value (A value) between the order and the shelf.

[0098] For each workstation Calculate the orders currently being picked that still have remaining goods to be ordered, as well as the candidate orders. Compared to currently picking shelves, reserved but not yet arrived shelves, and candidate shelves Real-time adaptation value at the current moment Among them, shelves At time 0, orders Immediate satisfaction rate of remaining demand As shown in Table 3 below: Table 3

[0099] For example, The denominator 5 indicates that the total remaining demand for goods 1 and 2 in order 1 is 5, and the numerator 3+2 indicates that shelf 1 can provide 3 goods 1 and 2 goods 2 to order 1.

[0100] Next, the estimated arrival time of each candidate shelf is calculated. Since all shelves are currently idle, they are expected to be available for immediate use. Rolling decision window The robot's average speed meters per second, the preparation time required for lifting, placing, and docking of the shelving. Second.

[0101] The estimated travel distances from each robot to its corresponding shelf are as follows: Robot 1 to shelf 1: 20 meters; Robot 2 to shelf 2: 36 meters; Robot 3 to shelf 3: 30 meters; Robot 4 to shelf 4: 20 meters. The estimated handling distances from each shelf to the workstation are shown in Table 4 below. Table 4

[0102] Calculated according to the formula As shown in Table 5 below: Table 5

[0103] After normalization, we get As shown in Table 6 below: Table 6

[0104] For example, the estimated arrival time of shelf 1 at workstation 1 at the current moment. .

[0105] Weight parameters are pre-set based on human experience. , , The real-time adaptation value under workstation 1 is calculated. As shown in Table 7 below: Table 7

[0106] For example, at workstation 1, the fit value between order 1 and shelf 1 is... .

[0107] Real-time adaptation value under workstation 2 As shown in Table 8 below: Table 8

[0108] Step S5: Shelf selection, reservation and conflict resolution.

[0109] Since none of the workstations currently have processing shelves and the candidate order set is not empty, for any workstation... ,have Meanwhile, there are currently idle robots, and Therefore, each workstation proceeds to the shelf selection step.

[0110] (5.1) Calculate the real-time service value (B value) of the rack to the workstation. For candidate shelves and workstation Calculate the real-time service value of the shelf to the workstation at the current moment. Among them, let Due to the current situation Since it is empty, the first term is 0; Each workstation has remaining cache capacity, therefore the feasibility coefficient for inserting candidate orders into each workstation is 0.5. Let non-negative weight parameters be defined. .

[0111] The real-time service value of each candidate shelf to each workstation is calculated. As shown in Table 9 below: Table 9

[0112] For example, the service value of shelf 1 to workstation 1 is .

[0113] (5.2) Perform shelf reservation, conflict resolution and dispatch. For workstations that need to select shelves, select from the candidate shelf set. Select real-time service value The largest shelf is the target shelf.

[0114] As shown in the table above, both workstation 1 and workstation 2 have shelf 1 as their preferred target shelf, resulting in a scheduling conflict. According to the conflict resolution rules, the real-time service values ​​of the two workstations for shelf 1 are compared. Because... Workstation 1 retains its reservation request for shelf 1. Workstation 2 then selects shelf 2, which has the highest real-time service value from the remaining candidate shelves, as the alternative target shelf and makes a reservation for shelf 2.

[0115] Once multiple workstations no longer require shelf selection, the reserved shelves are invoked, and the system status is updated. The queue of reserved but not yet arrived shelves for workstation 1 is updated. Workstation 1 has accumulated 1 shelf invocation. The current service workstation identifier for shelf 1 is workstation 1, and the status of shelf 1 is updated to "in transit". Update the queue of reserved but not yet arrived shelves for workstation 2. Workstation 2 has accumulated 1 shelf call, the current service workstation identifier for shelf 2 is workstation 2, and the status of shelf 2 is updated to "in transit".

[0116] Step S6: Perform order selection, reservation, and insertion.

[0117] (6.1) Calculate the insert revenue value (C value) of the order to the workstation. After the workstation completes the shelf reservation, it processes the candidate orders. and workstation Calculate the insert revenue value of the order for each workstation. Before calculating the C value, first calculate the load imbalance penalty caused by inserting orders into the workstation. Let... , , .because Since all values ​​are currently empty, the current load of each workstation, according to the formula, is... .

[0118] according to The incremental loads resulting from inserting orders 1, 2, and 3 into the workstation at the current moment are respectively... , , .

[0119] according to ,get As shown in Table 10 below: Table 10

[0120] For example, when order 1 is inserted into workstation 1, workstation The update load is .

[0121] according to Received the order Insert workstation The average system load after that is shown in Table 11 below: Table 11

[0122] For example, the average system load after order 1 is inserted into workstation 1 is: .

[0123] according to Received the order Insert workstation After that, workstation The deviations relative to the system average load are shown in Table 12 below: Table 12

[0124] Set the system's allowable load deviation threshold. According to Received the order Insert workstation The resulting load imbalance penalty is shown in Table 13 below: Table 13

[0125] For example, the load imbalance caused by inserting order 1 into workstation 1 is .

[0126] After that, The insertion benefit value of each candidate order for each workstation is calculated according to the formula. As shown in Table 14 below: Table 14

[0127] For example, the insertion benefit value of order 1 to workstation 1 is .

[0128] (6.2) Perform order reservation, rearrangement and insertion. In this round of order reservation, the reservation and conflict resolution are carried out in a cyclical manner based on the C value calculated at the current decision moment. After the reservation is completed, the order insertion and status update are performed in a unified manner.

[0129] In the first round of reservations, for workstation 1, the order with the highest inserted revenue value is order 1, so workstation 1 reserves order 1; for workstation 2, the order with the highest inserted revenue value is order 2, so workstation 2 reserves order 2. After the first round of reservations is completed, each workstation still has remaining cache capacity, so the second round of reservations is initiated.

[0130] In the second round of booking, for workstation 1, after removing its already booked order 1, the order with the highest insertion benefit value among the remaining candidate orders is order 2. However, order 2 has already been booked by workstation 2, thus triggering an order booking conflict. According to the conflict resolution rules, the insertion benefit values ​​of order 2 for the two competing workstations are compared. Since... Workstation 1 retains workstation 2's reservation request for order 2. Subsequently, workstation 1 selects the second-best available order 3 from the remaining candidate orders and makes a reservation for order 3.

[0131] For workstation 2, after removing its already booked order 2, the remaining candidate orders are order 1 and order 3. However, both order 1 and order 3 have been booked by workstation 1, and the insertion benefit of both for workstation 1 is higher than that for workstation 2. Therefore, workstation 2 will not book any other orders, but will only retain order 2, leaving a cache capacity of 1.

[0132] After the order reservation and conflict resolution are completed, the final reservation order is inserted into the order cache area of ​​the corresponding workstation, and the following information is updated: (1) Workstation In the order collection , A collection of orders that are being picked and still have remaining goods to be ordered. , (2) There is already a set of orders to be assigned. The collection of incomplete orders is already being picked. (3) The remaining cache capacity of workstation 1 is updated to 0, and the cumulative number of allocated orders is updated to 2; the remaining cache capacity of workstation 2 is updated to 1, and the cumulative number of allocated orders is updated to 1. The mapping relationship between orders and workstations is updated as follows: Order 1 and Order 3 are allocated to workstation 1, and Order 2 is allocated to workstation 2.

[0133] Step S7: Calculate the job time and update the status.

[0134] (7.1) Calculate the estimated release time of the shelving Based on the current assigned orders and the shelves, calculate the estimated release time of the current shelf. When the reserved shelf is expected to arrive at the workstation, since there are no shelves being processed in front of the corresponding workstation, the shelf can be set as the current processing shelf upon arrival, and picking is expected to begin immediately. Based on the aforementioned estimated arrival time, the time for shelf 1 to arrive at workstation 1 is 20 seconds, therefore, let Seconds; the time for shelf 2 to reach workstation 2 is 26 seconds, therefore... Seconds. Then, based on the remaining demand of current workstation buffer orders and the available inventory of current processing shelves, calculate the number of items that can be picked in this round and the estimated shelf release time.

[0135] make The estimated release time of the current shelf is calculated based on the number of items picked per second (i.e., 1 item picked every 3 seconds). , .

[0136] (7.2) Update status With orders and shelf availability fixed, changes in shelf inventory and remaining order demand can be anticipated after the current picking cycle. The expected remaining supply of goods for shelves 1 and 2 after this picking cycle is shown in Table 15 below, while shelves 3 and 4 remain unchanged. Table 15

[0137] Accordingly, the remaining demand vector updates for each order within the workstation buffer are expected to be shown in Table 16 below: Table 16

[0138] Based on the expected picking results, order 1 in workstation 1 is expected to be completed, and order 3 still has 2 items of item 5 remaining. Therefore, after the picking is completed, shelf 1 will have Order 2 is expected to be completed in workstation 2, therefore, after the picking is completed on shelf 2, there will be... .

[0139] Step S8: Continue to assign shelves to idle robots.

[0140] Determine whether there are still idle robots in the system, and whether there are workstations with unmet order demands after the currently reserved shelves have completed their expected picking. If there are idle robots in the system, and at least one workstation still has unmet demands, return to step S3, and continue to update the candidate set and select shelves based on the expected updated remaining order demands, shelf status, robot status, and workstation status; otherwise, proceed to step S9.

[0141] In this embodiment, the system has a total of 4 robots. Robot 1 has been assigned to move shelf 1 to workstation 1, and robot 2 has moved shelf 2 to workstation 2. Therefore, robots 3 and 4 are still idle. According to the expected picking result in step S7, order 2 in workstation 2 can be fully satisfied by shelf 2, while order 3 in workstation 1 still has 2 unfulfilled items 5. Therefore, the system still has unmet order demands, and there are still idle robots available to pre-select subsequent shelves. Therefore, the system returns to step S3 to continue selecting shelves for workstation 1 that can satisfy the remaining items 5 of order 3.

[0142] Step S3: Construct the candidate set.

[0143] After returning to step S3, the candidate order set and candidate shelf set are updated based on the currently reserved shelves and the expected picking results. Since orders 1, 2, and 3 in the order pool have all been assigned to workstations, and no new orders are currently arriving, therefore, for any workstation... The candidate order set is It should be noted that an empty candidate order set only indicates that there are currently no new orders to be assigned, and does not mean that the workstation cache is empty.

[0144] Next, the candidate shelf set is updated. Shelves 1 and 2 are currently reserved and locked by workstations 1 and 2 respectively, and cannot be accessed by other workstations. Shelves 3 and 4 remain idle and can both provide item 5. Based on the expected picking results from step S7, order 3 in workstation 1 still has 2 items of item 5 remaining. Therefore, the candidate shelf set for workstation 1 is updated to... .

[0145] Order 2 in workstation 2 can be fully satisfied by the reserved shelf 2, and no remaining picking demand is expected. Therefore, there is no need to continue selecting shelves, and its candidate shelf set is: .

[0146] Step S4: Calculate the real-time fit value (A value) between the order and the shelf.

[0147] For each workstation Calculate the orders currently being picked that still have remaining goods to be ordered, as well as the candidate orders. Compared to currently picking shelves, reserved but not yet arrived shelves, and candidate shelves Real-time adaptation value at the current moment For orders with zero remaining demand, the fit value between them and candidate shelves is no longer calculated. Based on the expected picking results from step S7, the picks for workstation 1 are obtained. As shown in Table 17 below: Table 17

[0148] Next, the estimated arrival time of each candidate shelf is calculated. Since shelves 3 and 4 are currently idle, each shelf is expected to be available for dispatch to the workstation at the current time. The original time is Rolling decision-making time window The robot's average speed meters per second; preparation time for lifting, placing, and docking of shelves. Second.

[0149] The estimated travel distance from robot 3 to shelf 3 is 30 meters, and the estimated transport distance from shelf 3 to workstation 1 is 20 meters; the estimated travel distance from robot 4 to shelf 4 is 20 meters, and the estimated transport distance from shelf 4 to workstation 1 is 14 meters. The original estimated arrival time is calculated from this. As shown in Table 18 below: Table 18

[0150] After normalization, we get As shown in Table 19 below: Table 19

[0151] Set weight parameters , , The calculation results are obtained under workstation 1. As shown in Table 20 below: Table 20

[0152] For example, the real-time adaptation value of order 3 and shelf 4 under workstation 1 is... .

[0153] Step S5: Shelf selection, reservation and conflict resolution.

[0154] Since there are still idle robots in the system, and based on the expected picking results, there will still be a remaining demand for item 5 in workstation 1 after order 3 has been picked from the currently reserved shelves. Workstation 1 proceeds to the shelf selection step. Order 2 in workstation 2 can be fully satisfied by the reserved shelf 2, therefore... Workstation 2 no longer selects the shelf.

[0155] (5.1) Calculate the real-time service value (B value) of the rack to the workstation. For candidate shelves Together with workstation 1, calculate the real-time service value of the shelf to the workstation at the current moment. At this point, based on the expected picking results, the set of valid unfinished picking orders for workstation 1 is... ,and Let non-negative weight parameters be defined. Calculations yielded As shown in Table 21 below: Table 21

[0156] (5.2) Perform shelf reservation, conflict resolution and dispatch. Workstation 1 selects shelf 4, which has a higher real-time service value, as the next target shelf. Since shelf 4 is not locked by other workstations, it can be reserved, and an idle robot can be called to perform the handling task.

[0157] It should be noted that although the earliest estimated arrival time of shelf 4, calculated based on independent handling time, is 19 seconds, which is 20 seconds earlier than the estimated arrival time of shelf 1 at workstation 1, shelf 1 has already been reserved and locked as the first processing shelf for workstation 1 in previous steps. Therefore, shelf 4 only serves as a subsequent reserved shelf for workstation 1, and its actual picking start time is constrained by the shelf processing queue of workstation 1, and will not be earlier than the time when shelf 1 completes its current round of picking and is released. Therefore, the queue of reserved but not yet arrived shelves at workstation 1 is updated according to the planned processing order. .

[0158] Workstation 1 has accumulated 2 racks called. The current service workstation for rack 4 is identified as workstation 1. The status of rack 4 has been updated to "in transit".

[0159] Step S6: Perform order selection, reservation, and insertion.

[0160] (6.1) Calculate the insert revenue value (C value) of the order to the workstation. Since there are no new arriving orders or pending orders in the current order pool, therefore for any workstation The candidate order set is empty. In this case, there is no need to calculate the new order insertion benefit value, nor is it necessary to perform order scheduling, rearrangement, and insertion operations. The existing set of orders being picked remains unchanged for now. , Based on the expected picking results, the expected remaining demand for Orders 1 and 2 is 0, while Order 3 still has 2 items (item 5) remaining. Therefore... , Then, proceed directly to step S7.

[0161] Step S7: Calculate the job time and update the status.

[0162] (7.1) Calculate the estimated release time of the shelving Based on the current assigned orders and the shelves, calculate the estimated release time for the current shelf. When shelf 4 arrives at workstation 1, since the shelf queue at workstation 1 is... Shelf 4 needs to wait for shelf 1 to complete picking before it can begin processing. The estimated release time for shelf 1 is known to be... Seconds, and the preparation time for changing shelves at the workstation for adjacent shelves is... If the time is seconds, then the earliest start time for picking on shelf 4 is... Second.

[0163] Shelf 4 can accommodate the remaining 2 items from order 3, with a picking speed of [missing information]. Items / second, therefore the expected release time for shelf 4 is (7.2) Update status Given a fixed order and shelf allocation, the estimated supply capacity and remaining order demand for each shelf are calculated after the current picking round. After picking is completed on shelf 4, the inventory of item 5 on shelf 4 decreases by 2 units; shelves 1 and 2 maintain their inventory status as projected after the previous picking round; shelf 3 is not used and its inventory remains at its initial state. The projected supply capacity for each shelf is shown in Table 22 below: Table 22

[0164] Accordingly, the remaining demand for each order in the workstation cache is expected to be updated as shown in Table 23 below: Table 23

[0165] Note that after this round of planning is completed, the projected remaining demand for orders 1, 2, and 3 is all 0. , .

[0166] Step S8: Continue to assign shelves to idle robots.

[0167] At this point, the system has a total of 4 robots. Robots 1, 2, and 4 have been assigned to the corresponding workstations for handling shelves 1, 2, and 4, respectively, while robot 3 remains idle. However, based on the aforementioned expected picking results, the remaining demands of orders 1, 2, and 3 will all be met in this round of picking, and each workstation will have no unfulfilled order demands after this round of picking is completed. Therefore, there is no need to continue assigning shelves to the idle robot; proceed directly to step S9.

[0168] Step S9: Perform order picking.

[0169] Based on the aforementioned calculations, the system executes actual order picking as planned. During the picking process, if events such as new orders arriving, orders being completed, shelf releases, or the rolling decision point being reached occur, the system status is updated and the next round of online re-decision is triggered.

[0170] In this example, shelf 1 is expected to arrive at workstation 1 in 20 seconds and begin picking, shelf 2 is expected to arrive at workstation 2 in 26 seconds and begin picking, and shelf 4 will queue up after shelf 1 after arriving at workstation 1.

[0171] At time 30, two new orders arrive. Since this event precedes the next order completion event (order 2 completes in 32 seconds), the system first updates the dynamic order pool status at time 30 and triggers the next round of online re-decision. At this point, orders 1, 2, and 3 are not yet completed, therefore the set of incomplete orders being picked is... The newly arrived order set is There is already a set of orders to be assigned. Orders 4 and 5 both arrive in 30 seconds, with a waiting time of 0. To distinguish between newly arriving orders and initial orders, this embodiment sets the initial urgency of new orders to be [value missing]. The remaining demand for goods to be picked is shown in Table 24 below: Table 24

[0172] Step S10: Repeat.

[0173] Returning to step S2, the online re-decision continues based on the system's real-time status at time 30. In this round of re-decision, the existing shelf 1, shelf 2, and shelf 4, which are already in progress or locked, maintain their established shelf-to-workstation allocation relationships and processing order; the system mainly continues to conduct online collaborative selection for newly arrived orders 4 and 5, as well as still available unfrozen resources.

[0174] If the existing order processing plan is not changed in subsequent decisions, the order completion events are expected to be as follows: Order 2 in workstation 2 will be completed in 32 seconds; Order 1 in workstation 1 will be completed in 38 seconds, at which time Order 3 will still not be completed; Order 3 in workstation 1 will be completed in 46 seconds.

[0175] (2) Second round of decision-making Step S2: Determine whether online re-decision is triggered.

[0176] Because a new order arrived at time 30, the system triggered an online re-decision at time 30 and set the current decision time to [time value missing]. At this time, shelf 1 and shelf 2 are being picked at workstations 1 and 2 respectively; shelf 4 has arrived at workstation 1 and is awaiting processing as a subsequent reserved shelf for workstation 1, with its shelf allocation relationship with workstations and processing order remaining frozen; shelf 3 is still idle. Orders 1, 2, and 3 have not yet triggered completion events, while orders 4 and 5 have just arrived. , Since each workstation has a fixed cache capacity of 2, the remaining cache capacity of the workstation is... , .

[0177] Step S3: Construct the candidate set.

[0178] Based on the real-time system status at time 30, for each workstation Update the candidate order set. Workstation 1 is currently processing shelf 1, which can satisfy item 3 in order 4 and item 1 in order 5. Therefore, the candidate order set for workstation 1 is... Workstation 2 is currently processing shelf 2, which can satisfy item 4 in order 4 and item 1 in order 5. Therefore, the candidate order set for workstation 2 is also [value missing]. .

[0179] Next, update the candidate shelf set. Shelves 1, 2, and 4 are currently locked, so only the unlocked shelf 3 is considered. Shelf 3 can partially meet the requirements of order 5 for products 5 and 6; therefore, from a product matching perspective, shelf 3 can be considered as a candidate shelf. Thus, we have... , .

[0180] Step S4: Calculate the real-time fit value (A value) between the order and the shelf.

[0181] First, calculate the remaining shelving capacity based on the estimated shelving supply after the completion of the previous round of planned operations. For newly arrived orders Instant satisfaction ratio As shown in Table 25 below: Table 25

[0182] Next, calculate the estimated arrival time of the currently picking shelves and candidate shelves. For each workstation's currently picking shelves (shelves 1 and 2) and the reserved and awaiting processing shelf (shelf 4), since they are currently occupied and cannot be used by other workstations, shelves 1 and 4 are unavailable for workstation 2, and their A value is not calculated. Shelf 2 is unavailable for workstation 1, and its A value is not calculated.

[0183] For candidate shelf 3, since it is currently idle, it is expected to be available for allocation to the workstation at this moment. The original time is .

[0184] Assume the estimated travel distance from the robot to shelf 3 is 30 meters, the estimated handling distances from shelf 3 to workstations 1 and 2 are 20 meters and 18 meters respectively, the robot's average travel speed is 2 meters per second, and the preparation time for shelf lifting and placement is... Seconds. Calculated using the formula. As shown in Table 26 below: Table 26

[0185] For each workstation's currently being picked shelves, set their normalized estimated arrival time to 0. For shelves 4 that have been reserved and are waiting to be processed at workstation 1, also set their normalized estimated arrival time to 0. This yields the normalized estimated arrival time. As shown in Table 27 below: Table 27

[0186] Where “-” indicates that the shelf has been locked by another workstation or is currently unavailable to the workstation, so the corresponding A value is not calculated.

[0187] Orders 4 and 5 both have an urgency level of 0.5, and the weighting parameter is... , , At that time, the real-time adaptation value is calculated according to the formula. For example, at workstation 1, the fit value between order 4 and shelf 1 is... . Obtained from workstation 1 As shown in Table 28 below: Table 28

[0188] Workstation 2 As shown in Table 29 below: Table 29

[0189] Step S5: Shelf selection, reservation and conflict resolution.

[0190] The system currently has idle robots, and shelf 3 is an available candidate shelf. For workstation 1, although its candidate shelf set includes shelf 3, since workstation 1 currently has 0 remaining cache capacity, and the remaining demand for order 1 and order 3 has been covered by the current shelf 1 and the subsequent reserved shelf 4, respectively, workstation 1 currently does not have any valid uncovered demand that needs to be met by a new shelf, so no valid shelf reservation request is generated.

[0191] For workstation 2, the current remaining cache capacity is 1, and part of the demand for candidate order 5 can be met by candidate shelf 3, therefore Workstation 2 will then proceed to the shelf selection step.

[0192] (5.1) Calculate the real-time service value (B value) of the rack to the workstation. Calculate the real-time service value of candidate shelf 3 for workstations 1 and 2 respectively. At this time, orders 1 and 3 of workstation 1 are covered by the current shelf 1 and the reserved queue shelf 4 respectively, and order 2 of workstation 2 is covered by the current shelf 2. Therefore, it is possible to... , Workstation 1's cache has no remaining cache capacity, therefore... Workstation 2's buffer has a remaining capacity of 1, therefore .

[0193] according to The value formula and calculation results are shown in Table 30 below: Table 30

[0194] (5.2) Perform shelf reservation, conflict resolution and dispatch. The calculation results show that shelf 3 has a negative service value for workstation 1, indicating that since workstation 1 has no remaining cache capacity and its original order demand has been covered by the current and reserved shelves, continuing to call shelf 3 for workstation 1 will not generate effective benefits. Shelf 3 has a positive service value for workstation 2, and workstation 2 has remaining cache capacity, therefore shelf 3 will be reserved for workstation 2 in this round.

[0195] Schedule a pickup for shelf 3 and call robot 3 to perform the handling task. Update the queue of scheduled but not yet arrived shelves at workstation 2. The cumulative number of shelves called by workstation 2 is updated to 2, the current service workstation identifier of shelf 3 is updated to workstation 2, and the status of shelf 3 is updated to transportation.

[0196] Step S6: Perform order selection, reservation, and insertion.

[0197] At time 30, since there is no remaining capacity in the order cache of workstation 1, no new order insertion is performed; the remaining capacity in the cache of workstation 2 is 1, so a new order needs to be inserted between order 4 and order 5.

[0198] (6.1) Calculate the insert revenue value (C value) of the order to the workstation. For candidate orders For workstation 2, calculate the insert revenue value of the order for that workstation. Before calculating the C value, first calculate the load imbalance penalty caused by inserting orders into the workstation. Let... , The weight parameters are ,current , At this moment, , Therefore, according to the formula, the current load of workstation 1 is The current load on workstation 2 is... .

[0199] according to The incremental load resulting from inserting candidate order 4 into the workstation at the current moment is... The incremental load for order 5 is .

[0200] If order 4 is inserted into workstation 2, the update load of that workstation will be: Correspondingly, the average system load after order 4 is inserted into workstation 2 is: After order 4 is inserted into workstation 2, the deviation of workstation 2 from the system average load is: Furthermore, let the system allowable load deviation threshold be set. Then, according to the load imbalance penalty formula caused by inserting order 4 into workstation 2, we get... Similarly, the load imbalance penalty caused by inserting order 5 into workstation 2 can be calculated as follows: .

[0201] Subsequently, the insertion benefit value of each candidate order for workstation 2 was calculated. As shown in Table 31 below: Table 31

[0202] For example, the insertion gain of order 4 to job 2 is... .

[0203] (6.2) Perform order reservation, rearrangement and insertion. Since workstation 1 currently has no remaining cache capacity, no new order insertion will be performed. Workstation 2 has 1 unit of remaining cache capacity, therefore, the order with the highest profit value will be selected from orders 4 and 5 for insertion. Based on the calculation results... Therefore, workstation 2 made a reservation and inserted order 5, and then updated the system status as follows: (1) workstation In the order collection , (2) Dynamic order pool status, in which order 4, which has not been inserted, enters the set of orders to be allocated. (3) The remaining cache capacity of workstation 2 is updated to 0; (4) The cumulative number of orders allocated to the workstation; (5) The mapping relationship between orders and workstations.

[0204] Step S7: Calculate the job time and update the status.

[0205] (7.1) Calculate the estimated release time of the shelving Based on the current assigned orders and shelves, calculate the estimated release time of the current shelves. When order 5 is inserted into workstation 2, the shelf currently being processed by workstation 2 is shelf 2, and the shelf that has been reserved but not yet arrived is shelf 3. The remaining demand for order 5 is... .

[0206] According to the previous plan, the estimated remaining stock of goods available for shelf 2 after completing the picking of order 2 is [amount missing]. Therefore, shelf 2 can continue to serve the demand for product 1 from order 5. Based on the aforementioned calculations, shelf 2 takes 32 seconds to complete the picking of order 2; therefore, the earliest time to start picking order 5 on shelf 2 is [time to be inserted]. Seconds. Shelf 2 can provide 1 item 1 for order 5. This allows workstation 2 to have a picking speed of [number missing]. The estimated release time for shelf 2 after continuing to serve order 5 is updated to... Second.

[0207] At this point, the remaining demand for order 5 is updated to... Since shelf 3 has already been reserved for workstation 2 in step S5, its expected arrival time at workstation 2 is... .

[0208] Shelf 3 needs to serve the remaining 2 items (item 5) and 1 item (item 6) from order 5, for a total of 3 items. Since shelf 3 arrives later than the estimated release time of shelf 2, the earliest picking start time for shelf 3 is [date / time]. The expected release time for shelf 3 is... .

[0209] (7.2) Update status With orders and shelf availability confirmed, the expected remaining stock of goods for each shelf after this round of planning is completed is shown in Table 32 below: Table 32

[0210] If the current plan continues, orders 1 and 3 at workstation 1 will be completed in 38 seconds and 46 seconds respectively; order 2 at workstation 2 will be completed in 32 seconds, and order 5 will be completed in 65 seconds. Therefore, after this round of planning is completed, the expected remaining demand for each picking order is shown in Table 33 below: Table 33

[0211] Note that order 4 was not inserted into the workstation in this round and remains in the set of orders to be assigned. This will be addressed in subsequent major decisions.

[0212] Step S8: Continue to assign shelves to idle robots.

[0213] At this point, the robots have been assigned to move shelves 1, 2, 4, and 3 respectively, and there are no idle robots in the system. Therefore, even though there is still an order 4 to be assigned in the system, it is temporarily impossible to call up new shelves, so there is no need to return to step S3, and step S9 can be executed directly.

[0214] Step S9: Perform order picking.

[0215] Based on the aforementioned calculation results, the system executes the actual order picking as planned.

[0216] After the second round of decision-making, the earliest triggering event is the completion event of order 2 in workstation 2. Order 2 is completed by shelf 2 and is expected to be completed at time 32. Therefore, at time 32, order 2 is removed from the workstation, triggering the order completion event. Workstation 2 releases one cache location, and the remaining cache capacity of workstation 2 is updated to... At this point, order 5 has not yet been completed and is still held in the order cache of workstation 2. Therefore, the picking order set of workstation 2 is updated to... .

[0217] Meanwhile, orders 1 and 3 in workstation 1 have not yet been completed and still have... Order 4, which has not yet been inserted, remains in a pending allocation state. .

[0218] It should be noted that before the online re-decision is triggered at time 32, according to the previous plan, after order 2 is completed, shelf 2 will continue to pick 1 item 1 for order 5. Therefore, shelf 2 remains the current processing shelf for workstation 2, and its expected release time is 35 seconds. Since the order completion event is a preset re-decision trigger event, the system triggers the next round of online re-decision at time 32.

[0219] If subsequent decisions do not change the existing work plan, the following events are expected: Shelf 2 will be released in 35 seconds; Order 1 in workstation 1 will be completed in 38 seconds; Order 3 in workstation 1 will be completed in 46 seconds; and Order 5 in workstation 2 will be completed in 65 seconds.

[0220] Step S10: Repeat.

[0221] Return to step S2 and continue to perform online re-decision based on the system state at time 32.

[0222] During subsequent online re-decision-making, the system continuously updates its real-time status based on events such as new order arrivals, order completions, and shelf releases. While maintaining the already frozen execution plan, it dynamically adjusts subsequent order insertion and shelf allocation plans. Specifically, when order 2 completes in 32 seconds and releases a cache location on workstation 2, the system inserts the pending order 4 into workstation 2 and utilizes the currently processing shelf 2 on workstation 2 to first fulfill the product 4 requirement of order 4.

[0223] Subsequently, after shelf 1 completes its service and is released at workstation 1 in 38 seconds, the system re-reserves shelf 1 to workstation 2 to fulfill the remaining demand for product 3 in order 4. At the same time, the system maintains the established processing order of reserved shelves, ensuring that shelf 3 first fulfills the remaining demand for order 5, and then shelf 1, which was re-reserved after being released, fulfills the remaining demand for order 4.

[0224] The timeline of the entire process is shown in Table 34 below: Table 34

[0225] Ultimately, orders 2, 1, 3, 5, and 4 were completed in 32 seconds, 38 seconds, 46 seconds, 65 seconds, and 70 seconds, respectively. Specifically, workstation 1 processed shelf 1 and shelf 4 sequentially to complete orders 1 and 3; workstation 2 processed shelf 2, shelf 3, and the reactivated shelf 1 sequentially to complete orders 2, 5, and 4. This result demonstrates that this method can dynamically coordinate order insertion, shelf reuse, and processing order, thereby improving shelf utilization, reducing shelf handling frequency, shortening order response time, and increasing the efficiency of multi-workstation collaborative picking, even with continuous online order arrivals and dynamic system status changes.

Claims

1. A method for online collaborative selection and sorting of orders and shelves in a robotic mobile shelving system, characterized in that, Includes the following steps: Step S1: Construct the real-time system status; At this moment of decision-making The system collects and updates the real-time status of the robotic mobile shelf system; the real-time status of the system includes the dynamic order pool status, workstation status, shelf status, and operation time status. Step S2: Determine if online re-decision is triggered; Detect the current decision moment Has a pre-set re-decision trigger event occurred? If a preset re-decision trigger event occurs, steps S3 to S10 will be executed for unfrozen orders, unfrozen shelves, and unfrozen workstation resources while keeping the freeze execution plan unchanged; if no preset re-decision trigger event occurs, the current freeze execution plan will continue to be executed. Step S3: Construct the candidate set; Based on the real-time status of the system, for each workstation Construct a candidate order set and candidate shelf collection ; Step S4: Calculate the real-time fit value between the order and the shelf, i.e., the A value; For any order that is currently being picked and still has remaining demand for goods, and for any order in the candidate order set. It is associated with any shelf that belongs to the currently picking shelf, the pre-booked but not yet arrived picking shelf, or the candidate shelf set. Calculate orders Shelves With workstation At the present moment Real-time adaptation value , is called The value is used to represent the value at the current decision moment. Shelves At the workstation fulfilling orders The degree of comprehensive adaptation to remaining demand; Step S5: Perform shelf selection, reservation, and conflict resolution; make Indicates workstation At any moment Whether new shelves are needed is indicated by a value of 1 (yes) and 0 (no). And when there are idle robots, the workstation Proceed to the shelf selection step; Candidate Shelves and workstation Calculate the real-time service value of the shelving to the workstation. , is called The value is used to represent the value at the current decision moment. Shelves Called to workstation The resulting overall benefits; for workstations that need to select shelving, from the candidate shelving set Select real-time service value The largest shelf is the target shelf; If the target shelf is not locked by other workstations within the expected call period, the reservation is executed directly; if multiple workstations compete for the same shelf, conflict resolution is performed; when multiple workstations do not need to select a shelf, or the robots have all been assigned, the reserved shelves are called and the workstation status, shelf status, and operation time status are updated. Step S6: Perform order selection, reservation, and insertion; When workstation This step is executed if there is remaining cache capacity and the candidate order set is not empty; For candidate orders and workstation Calculate the insert revenue value of the order to the workstation. , is called The value is used to represent the value at the current decision moment. ,Order Insert workstation The resulting overall benefits; For each workstation with remaining cache capacity and a non-empty candidate order set From the candidate order set Select Insert Earnings Value The largest order is designated as the target order, and reservations are made for the target order; If multiple workstations compete for the same order, then order rearrangement will be performed; Once all round-robin appointments and order conflict resolution processes are complete, the final appointment order is inserted into the order cache area of ​​the corresponding workstation, and the system status is updated. Step S7: Calculate the job time and update the status; Based on the current assigned orders and shelves, calculate the estimated release time of the current shelves, and predict the changes in shelf inventory and remaining demand for each order in the workstation buffer area after the current round of picking is completed; Step S8: Continue to assign shelves to idle robots; Determine whether there are still idle robots in the system, and whether there are still unmet order demands after each workstation has completed the expected picking of the currently reserved shelves; if there are idle robots in the system, and at least one workstation still has unmet demands, return to step S3, and continue to perform steps such as candidate set update, shelf selection, reservation and call based on the updated remaining order demand, shelf status, robot status and workstation status; Otherwise, proceed to step S9; Step S9: Perform order picking; Based on the aforementioned calculation results, the system will execute the actual order picking according to the plan; When a workstation completes a round of picking on the current shelf, the following judgment is made: if there is a completed order, the completed order is removed from the workstation and the corresponding position in the buffer is released; if there is no completed order and the current shelf cannot continue to meet the remaining demand of the picking orders, the current shelf is set to the pending release state. During the picking process, if events such as new order arrival, order completion, shelf release, or reaching the rolling decision point occur, the system's real-time status is updated, and the next round of online re-decision is triggered. Step S10: Repeat the execution; Repeat steps S2 to S9, continuously update the real-time status of the system based on events such as new order arrival, order completion, shelf release, and reaching the rolling decision point, and perform online collaborative selection and dynamic sorting of orders and shelves; When the set of new orders that have arrived up to the current decision cycle, the set of existing orders to be assigned, and the set of picking orders at each workstation are all empty, and there are no reserved but not completed shelves or shelves in transit, the current round of online collaborative selection and sorting process ends.

2. The method for online collaborative selection and sorting of orders and shelves in a robotic mobile shelving system according to claim 1, characterized in that, In step S1: Let the workstation set be The shelving is a collection of The product collection is ; For any order Record it at that moment The remaining demand vector is ,in, Indicates order At any moment For goods The remaining demand; For any shelf Record it at that moment The inventory vector is ,in, Indicates shelf At any moment For goods The available supply quantity; For any workstation Record it at that moment The set of orders to be picked is The current processing shelf is The queue of shelves that have been reserved but not yet picked is The workstation has a remaining cache capacity of ; The dynamic order pool status includes: a set of newly arrived orders. There is already a set of orders to be assigned. The system includes a collection of orders that are being picked but not yet completed, the arrival time of each order, the waiting time, the urgency parameters, and the remaining order requirements for the goods to be picked. The workstation status includes: the current set of orders being picked at each workstation, the currently processed shelves, the queue of shelves that have been reserved but not yet arrived for picking, the remaining cache capacity, the cumulative number of allocated orders, and the cumulative number of shelves that have been called. The shelf status includes: the current position of each shelf, the available quantity of each product in the shelf, the current service workstation identifier, and the shelf status marker; the shelf status marker includes one of the following: idle, reserved, in transit, being picked upon arrival, and awaiting release; The operational time status includes: the estimated release time of the shelves at the workstation, the estimated arrival time of the candidate shelves at each workstation, and the estimated arrival sequence of the reserved shelves.

3. The method for online collaborative selection and sorting of orders and shelves in a robotic mobile shelving system according to claim 1, characterized in that, In step S2: The preset re-decision triggering events include any of the following events: (1) new order arrival event; (2) order completion event; (3) shelf release event; (4) reaching the rolling decision point event, wherein the rolling decision time window is When multiple re-decision events are triggered at the same decision moment, they are merged into a single online re-decision. The freeze execution scheme refers to maintaining the allocation relationship, arrival order, and corresponding execution status of the currently processed shelves, called shelves, reserved shelves, and assigned orders that have entered the execution window unchanged within a preset freeze duration.

4. The method for online collaborative selection and sorting of orders and shelves in a robotic mobile shelving system according to claim 1, characterized in that, The specific process of constructing the candidate order set and candidate shelf set in step S3 is as follows: First, for each workstation From the newly arrived order collection and existing unassigned order sets Construct a candidate order set The candidate order set includes at least one of the following: (1) orders that can be processed by the current shelf. (2) Orders that can be fulfilled immediately at least partially; (3) Orders that can be picked by pre-booked but not yet arrived rack queues. Subsequent orders that meet at least some of the requirements; (3) with workstations Currently picking order collection Orders with a high degree of subsequent matching, the degree of subsequent matching is determined based on the degree of overlap between the candidate order and the current order being picked, the available shelf set or the reserved shelf set; (4) if all the above order sets are empty, the newly arrived order set and the existing unassigned order set are set as the candidate order set of the workstation; Next, a candidate shelf set is constructed from the shelves that are currently idle and have not been reserved, shipped, frozen, or locked by any workstation. Includes at least one of the following shelves: (1) capable of meeting the needs of workstations Currently picking order collection (1) Shelves that can meet at least part of the remaining demand; (2) Shelves that can meet at least part of the demand for candidate orders.

5. The method for online collaborative selection and sorting of orders and shelves in a robotic mobile shelving system according to claim 1, characterized in that, Step S4 is as follows: For each workstation ,make This represents the set of orders currently being picked and still having remaining demand for goods; for any order This refers to orders currently being picked that still have remaining stock and candidate orders, along with any shelf. This refers to the current picking shelves, the shelves that have been reserved but not yet arrived, and the candidate shelves, which are used to calculate orders. Shelves With workstation At the present moment Real-time adaptation value , is called The value is used to represent the value at the current decision moment. Shelves At the workstation fulfilling orders The degree of comprehensive adaptation to remaining demand; The real-time adaptation value Calculate according to the following formula: in, , indicating shelves At any moment For orders The proportion of immediate satisfaction of remaining demand for goods; Indicates order At any moment The urgency, and The urgency level is determined based on at least one of the following: order waiting time, promised completion time, and order priority. This indicates the shelf after normalization. Arrival at the workstation The estimated arrival time is calculated as follows: in, Indicates shelf At any moment Expected to be transferred to workstation The original estimated arrival time; Indicates the length of the rolling decision window; for workstations Current processing shelves ,make , When an idle mobile robot is available, for shelves that are currently idle and not locked by other workstations... Its original estimated arrival time is calculated according to the following formula: in, Indicates shelf Reassigned to workstation The planned call start time: when the shelf At any moment When a mobile robot is idle, not locked by other workstations, and has available space, then... Indicates that it is assigned to be used for shelving Move to workstation Mobile robots; Represents robots From current location to shelf The estimated driving distance from the current location; Indicates from the shelf Location to workstation The expected transport distance; This indicates the robot's average speed. This indicates the preparation time required for lifting, placing, and docking the shelves; These are non-negative weight parameters.

6. The method for online collaborative selection and sorting of orders and shelves in a robotic mobile shelving system according to claim 1, characterized in that, Step S5 is as follows: Let (1) Workstation Currently, there are no processing shelves, and the orders that are being picked still have remaining items to be ordered. or candidate order set (2) The current processing racks are insufficient to meet the needs of the workstation. (3) Any remaining demand from orders with remaining merchandise; (4) Subsequent merchandise demand at the workstation that is not yet covered by the current shelf or a reserved shelf; Candidate shelf collection It is not empty, and at least part of the demand for the candidate order cannot be covered by the currently processed shelf or the reserved shelf; when And when there are idle robots, the workstation Proceed to the following shelf selection steps; (5.1) Calculate the real-time service value of the rack to the workstation, i.e., the B value. Candidate Shelves and workstation Define the real-time service value of the shelf to the workstation. , is called The value is used to represent the value at the current decision moment. Shelves Called to workstation The resulting overall benefits; The real-time service value Calculate according to the following formula: in, Indicates time Assigned to workstation A collection of orders that have been picked and still have remaining goods to be ordered; Indicates time workstation The set of candidate orders; Indicates order Assigned to workstation The feasibility coefficient; if the workstation There is remaining cache capacity and orders If it can be inserted, then ,otherwise ; It is to prevent extremely small positive numbers with a denominator of 0; This indicates the shelf after normalization. Arrival at the workstation The estimated arrival time; These are non-negative weight parameters; (5.2) Perform shelf reservation, conflict resolution and dispatch. For workstations that need to select shelves, select from the candidate shelf set. Select real-time service value The largest shelf is designated as the target shelf; if the target shelf is not locked by other workstations during the expected usage period, a reservation is made directly for it. If multiple workstations compete for the same shelf, conflict resolution is performed, including: (1) calculating the real-time service value of each competing workstation using the shelf; (2) retaining the reservation request of the workstation with the higher real-time service value for the shelf; (3) selecting the second-best shelf from the corresponding candidate shelf set for the remaining workstations; (4) if the second-best shelf cannot be directly called at the current time, generating a delayed call plan and updating the estimated arrival time; (5) if there is no second-best shelf in the corresponding candidate shelf set, the workstation is placed in a waiting state, and the candidate shelf set is reconstructed in the next round of decision-making. When multiple target shelves with the same real-time service value exist at the same workstation, the shelf closest to the workstation is selected first; if the distances are still the same, the shelf with the smaller shelf number is selected. When multiple workstations do not need to select shelves, or when all robots have been assigned, the system will call up each reserved shelf and update the workstation status, shelf status, and operation time status.

7. The method for online collaborative selection and sorting of orders and shelves in a robotic mobile shelving system according to claim 1, characterized in that, Step S6 is as follows: (6.1) Calculate the insert revenue value of the order to the workstation, i.e., the C value. For candidate orders and workstation Define the insert revenue value of an order to a workstation as... , is called The value is used to represent the value at the current decision moment. ,Order Insert workstation The resulting overall benefits; The inserted revenue value Calculate according to the following formula: in, Indicates order With workstation Current processing shelves Real-time adaptation values ​​between them; Indicates workstation At any moment Shelf queues that have been reserved but not yet arrived for picking; Indicates order With workstation The maximum fit value in the reserved shelf queue; Indicates order At any moment The urgency; Indicates order Insert workstation The resulting load imbalance penalty; These are non-negative weight parameters; (6.2) Perform order reservation, rearrangement and insertion. For each workstation with remaining cache capacity and a non-empty candidate order set From the candidate order set Select Insert Earnings Value The largest order is designated as the target order, and reservations are made for the target order; To further ensure a balanced workload across all workstations, a cyclical, rotating appointment system will be adopted: appointments will be made according to the workstation's schedule. Orders are reserved sequentially in order. After a single round of reservations is completed, a new round of reservations is started from workstation 1. During the reservation process, the remaining cache capacity of each workstation and whether the order has been reserved by other workstations are checked simultaneously. This is done in a loop to complete the orderly reservation of orders. When the remaining cache capacity of all workstations is 0, or all candidate orders have been reserved, the round-robin reservation ends. If multiple workstations compete for the same order, then the order is rearranged, specifically including: (1) calculating the benefit value of each competing workstation inserting the order; (2) retaining the reservation request for the order from the workstation with the higher insertion benefit value; (3) selecting the second-best available order from the corresponding candidate order set for the remaining workstations; (4) if there is no second-best order in the corresponding candidate order set, then skip the workstation and reconstruct the candidate order set in the next round of decision-making. When there are multiple target orders with the same insertion benefit value, the order with the earlier arrival time is selected first; if the arrival times are still the same, the order with the smaller order number is selected first. After all round-robin appointments and order conflict resolution processes are completed, the final appointment order is inserted into the order cache of the corresponding workstation, and the following information is updated: (1) Workstation In the order collection A collection of orders that are being picked and still have remaining goods to be ordered. (2) Dynamic order pool status; (3) Remaining cache capacity of workstation; (4) Cumulative number of allocated orders for workstation; (5) Mapping relationship between orders and workstations.

8. The method for online collaborative selection and sorting of orders and shelves in a robotic mobile shelving system according to claim 7, characterized in that, Load balancing metrics are set for each workstation, and a negative penalty is imposed on the insertion scheme if the load on a workstation exceeds a preset allowable deviation after an order is inserted; this is to calculate... Set up workstations At any moment The current load is The load The determination is based on at least one of the following: the occupancy level of the workstation buffer area, the number of remaining items in picking orders, and the number of reserved but not yet arrived shelves. workstation The load metric is expressed as: in, Indicates time workstation The collection of orders being picked; Indicates workstation Fixed cache size; Indicates order For goods The remaining demand; Indicates workstation Shelf queues that have been reserved but not yet arrived for picking; and These are the maximum remaining demand and the maximum number of reserved shelves used for normalization, respectively. These are non-negative weight parameters; Set order At any moment The incremental load brought by inserting the workstation is According to the order The remaining demand for goods is determined. When orders Inserted into workstation After that, workstation The update load is: Correspondingly, orders Insert workstation The average load of the system after that is: Order Insert workstation After that, workstation The deviation relative to the system average load is: Further, define the order Insert workstation The resulting load imbalance penalty is: in, Indicates the allowable load deviation threshold; when This indicates that inserting the order will not cause issues on the workstation. If the load deviation exceeds the allowable range, no load imbalance penalty will be applied; when... When this occurs, it indicates that inserting this order will cause the workstation to... If the system average load is too high, a load imbalance penalty is applied to the portion exceeding the threshold.

9. A method for online collaborative selection and sorting of orders and shelves in a robotic mobile shelving system according to claim 1, characterized in that, Step S7 is as follows: (7.1) Calculate the estimated release time of the shelving Based on the current assigned orders and shelves, calculate the estimated release time of the current shelves, and update the dynamic order pool status, workstation status, shelf status, and operation time status. workstation Current Shelf The expected release time is calculated using the following formula: in, Indicates shelf At the workstation The expected release time; Indicates shelf At the workstation The estimated start time for picking is calculated as follows: If the reserved shelf arrives at the workstation and there is no shelf being processed in front of it, the shelf will be set as the current processing shelf upon arrival and picking will begin immediately; if there is a shelf being processed in front of it, the estimated start time for that shelf will be the estimated release time of the previous shelf plus a preparation time. ; Indicates time workstation A collection of orders that are being picked and still have remaining goods to be purchased; Indicates order At any moment For goods The remaining demand; Indicates workstation Shelves At any moment Available products quantity; Indicates workstation At any moment Effective picking rate; (7.2) Update status Given a fixed order and shelf space, anticipate changes in shelf inventory and remaining demand for each order in the workstation buffer area after the current picking cycle is completed; workstation Current Shelf Domestic goods The available supply margin is updated according to the following formula: 。

Images