Control method and apparatus for product conveyance, and computing device and medium

Abstract

Provided are a control method and apparatus for product conveyance, and a computing device and a medium. The method comprises: acquiring a product occupancy of each data collection position at a current moment, a reservation quantity used for indicating the number of conveyance tasks to be conveyed to each data collection position at the current moment, and a reservation completion time cost corresponding to the reservation quantity; and on the basis of the product occupancy of each data collection position at the current moment, the reservation quantity, and the reservation completion time cost corresponding to the reservation quantity, using a machine learning model to predict a conveyance state corresponding to each data collection position.

Description

Control methods and devices, computing equipment and media for product handling Technical Field

[0001] This disclosure relates to the field of industrial manufacturing technology, and in particular to a control method and apparatus, computing device and computer-readable storage medium for product handling in a factory. Background Technology

[0002] In current manufacturing plants (such as display panel factories), the production of products (e.g., display panels) involves multiple processes. After one process is completed, the product needs to be transferred to the processing equipment for the next process. Finally, the finished product is unloaded from the processing equipment. Therefore, there are numerous (e.g., tens of thousands of times per day) and various types of transfers in manufacturing plants. The number of transfer vehicles (also referred to as trolleys in this text) is limited, and only one unit of product can be transferred at a time (one transfer task). Furthermore, the shelf space of each processing equipment is limited, such as only holding 4-6 units of product. Space must be freed up after the finished products are moved to the next processing equipment before new products can arrive. Therefore, when the number of transfer tasks is large, long waiting times can lead to congestion, idle equipment, and severely impact factory capacity.

[0003] Therefore, how to schedule the transportation of these products to improve the efficiency of the factory and thus increase the factory's capacity is a research topic in this field. Summary of the Invention

[0004] According to one aspect of this disclosure, a control method for product transport in a factory is provided. The control method includes: acquiring, for a current time, the product occupancy at each data acquisition location in the factory, a reservation quantity representing the number of transport tasks to be transported to each data acquisition location at the current time, and a completion reservation time cost corresponding to the reservation quantity; and using a machine learning model, and based on the product occupancy, reservation quantity, and completion reservation time cost corresponding to each data acquisition location at the current time, predicting the transport status corresponding to each data acquisition location. Optionally, the method may further include scheduling product transport in the factory based on the predicted transport status corresponding to each data acquisition location. Optionally, the machine learning model may be trained based on historical product occupancy, historical reservation quantity, and historical completion reservation time cost corresponding to each historical reservation quantity at each data acquisition location for multiple historical sampling times.

[0005] According to another aspect of this disclosure, a control device for product transport in a factory is also provided. The control device may include: an acquisition module for acquiring the product occupancy at each data acquisition location in the factory at a current time, a reservation quantity representing the number of transport tasks to be transported to each data acquisition location at the current time, and a corresponding completion reservation time cost; and a prediction module for predicting the transport status corresponding to each data acquisition location using a trained machine learning model, based on the product occupancy, reservation quantity, and corresponding completion reservation time cost at each data acquisition location at the current time. Optionally, the control device may further include a scheduling module for scheduling product transport in the factory based on the predicted transport status corresponding to each data acquisition location. Optionally, the machine learning model may be trained based on historical product occupancy, historical reservation quantity, and historical completion reservation time cost corresponding to each historical reservation quantity at each data acquisition location for multiple historical sampling times.

[0006] According to another aspect of this disclosure, a computing device is also provided, comprising: a processor; and a memory having a computer-executable program stored thereon, which, when executed by the processor, causes the processor to perform the control method for product handling in a factory as described above.

[0007] According to another aspect of this application, a computer-readable storage medium is also provided, storing a computer program that, when executed by a processor, causes the processor to perform various operations of the control method for product transport in a factory as described above.

[0008] According to another aspect of this application, a computer program product is also provided, comprising a computer program that, when executed by a processor, implements various operations of the control method for product transport in a factory as described above.

[0009] According to the product handling scheduling scheme disclosed herein, a dataset (training dataset and / or test dataset) is constructed for training a machine learning model based on the product occupancy at each preset location over a historical period (multiple historical sampling times) (corresponding to completed handling tasks), the number of scheduled handling orders (i.e., handling tasks scheduled to begin at that time), and the estimated completion time of each scheduled handling order (i.e., the time cost of completing the scheduled order). The trained machine learning model then uses this data to predict the product occupancy at each preset location, the number of scheduled handling orders, and the estimated completion time (i.e., determining the handling status) for a future period, considering the product occupancy, scheduled handling orders, and the corresponding time cost of completing the scheduled handling orders at each data collection location at the current moment. Based on the prediction results, areas with impending congestion risks can be identified, allowing for early intervention, such as diverting handling tasks to other non-congested paths or canceling lower-priority handling tasks, thereby ensuring smooth operation of the entire factory and maximizing the capacity of processing equipment. Attached Figure Description

[0010] The accompanying drawings illustrate various embodiments of various aspects of this disclosure, and they, together with the specification, serve to explain the principles of this disclosure. Those skilled in the art will understand that the specific embodiments shown in the drawings are merely exemplary and are not intended to limit the scope of this disclosure. In the drawings:

[0011] Figure 1 shows a layout diagram of one floor of a display panel factory as an example of a manufacturing plant according to an embodiment of the present disclosure.

[0012] Figure 2 shows a schematic flowchart of a control method for product transport in a factory according to an embodiment of the present disclosure.

[0013] Figure 3 illustrates a schematic diagram of constructing a training dataset for training a machine learning model according to an embodiment of the present disclosure.

[0014] Figure 4 illustrates a scenario where, at the current time tc, three transport tasks are about to be transported to the data acquisition location Pm.

[0015] Figure 5 illustrates an example flowchart of obtaining training and / or test datasets by sample splitting according to an embodiment of the present disclosure.

[0016] Figure 6 illustrates an example process of obtaining training and / or test datasets by sample splitting according to an embodiment of the present disclosure.

[0017] Figure 7 shows a structural block diagram of a control device for product transport in a factory according to various embodiments of the present disclosure.

[0018] Figure 8 shows a schematic block diagram of a computing device according to an embodiment of the present disclosure. Detailed Implementation

[0019] The technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0020] Before providing a detailed description of this disclosure, the relevant technical terms to be used will be explained.

[0021] STK: Short for Stocker, which can be viewed as the smallest enclosed cleanroom unit in a manufacturing plant (such as a display panel factory). It includes several processing machines or shelves on both sides for different processes and a slide rail in the middle, on which two trolleys (e.g., robots) run to move products.

[0022] OHCV: A conveyor belt connecting two storage spaces (STK) for bidirectional product transport.

[0023] OHS: A large loop track connecting several storage spaces STK, with several trolleys running counterclockwise on it for transporting products.

[0024] Job: A transport task is a task that starts from a certain processing equipment or shelf in a storage space STK and moves to a certain processing equipment or shelf in the same storage space STK or different storage space STKs. The completion of each job is equivalent to completing one transport schedule.

[0025] Figure 1 shows a layout diagram of one floor of a display panel factory as an example of a manufacturing plant according to an embodiment of the present disclosure.

[0026] As shown in Figure 1, each dashed box represents a storage space STK (STK1 to STK6 in the figure). Each storage space contains several processing devices (horizontal boxes within each STK in Figure 1) to process products using different techniques. Additionally, each processing device has dozens of shelves (not shown) above it. Each storage space STK also includes a slide rail (vertical box in Figure 1) on which two trolleys (e.g., robots) operate to move products back and forth between the processing devices / shelves within the STK, transporting only one unit of product at a time.

[0027] The passageway between the two storage spaces (STKs) in Figure 1 is the OHCV conveyor belt, which transports products for processing on the processing equipment within the two adjacent STKs. The solid box in the middle of Figure 1 is the large loop track (OHS), which can connect multiple storage spaces (6 STKs shown in the figure), transporting products for processing on the equipment within these 6 STKs. The small solid boxes adjacent to some storage spaces (STKs) are elevators, which transport products up and down floors for processing on the equipment within the storage spaces (STKs) on different floors.

[0028] Because the manufacturing processes for products such as display panels are complex and involve numerous steps, requiring multiple processing steps on different processing equipment in different storage spaces (STKs), and potentially involving moving between floors, it is necessary to traverse OHCV, OHS, and elevators multiple times to enter and exit each STK. A transport task is defined as a job, starting from a specific processing equipment or shelf in one STK and proceeding to a specific processing equipment or shelf in the same or different STKs. The completion of each job constitutes one transport scheduling operation, and a corresponding transport route is selected. As shown in Figure 1, the transport from start position A to end position D and from start position E to end position H constitute two jobs. The optional transport route for the job from start position A to end position D can be ABCD, and the optional transport route from start position E to end position H can be EFGH.

[0029] Manufacturing plants handle tens of thousands of transport tasks daily, creating a complex and interconnected traffic network. Furthermore, as mentioned earlier, the number of transport vehicles (also known as carts, such as robots) is limited, and each can only transport one unit of product at a time, constituting a transport job (corresponding to one transport command). Additionally, each processing unit has limited space, typically holding only 4-6 units. Therefore, new products can only arrive after finished products have been moved to the next processing unit to free up space. When the number of transport tasks is large, excessively long waiting times can occur at processing units, leading to congestion and idle equipment, thus severely impacting factory capacity.

[0030] For a large number of transport tasks, if the scheduling process is not optimized, it will lead to congestion, which will affect the smooth operation of the entire factory and the capacity of processing equipment.

[0031] Therefore, various embodiments of this disclosure propose a scheduling scheme for product handling in a factory. This scheme constructs a dataset (training dataset and / or test dataset) for training a machine learning model based on the product occupancy (corresponding to completed handling tasks), the number of scheduled handling orders (i.e., handling tasks scheduled to begin), and the estimated completion time (i.e., the time cost of completing the scheduled tasks) at each preset location within each storage space (STK), on the OHCV conveyor belt, and on the large loop track (OHS) over a historical period. The trained machine learning model then predicts the product occupancy, number of scheduled handling orders, and estimated completion time at each preset location within each storage space (STK), on the OHCV conveyor belt, and on the OHS large loop track over a future period. Based on the prediction results, areas with impending congestion risks can be identified, allowing for early intervention, such as diverting handling tasks to other non-congested paths or canceling lower-priority handling tasks, thereby ensuring smooth operation of the entire factory and maximizing the capacity of processing equipment.

[0032] The following description, in conjunction with Figures 2 to 8, provides further details of the scheduling schemes for product transport in a factory according to various embodiments of the present disclosure.

[0033] Figure 2 shows a schematic flowchart of a control method for product transport in a factory according to an embodiment of the present disclosure.

[0034] As shown in Figure 2, in step S210, the product occupancy at each data collection location in the factory at the current time, the reservation quantity representing the number of transport tasks scheduled to be transported to each data collection location at the current time, and the corresponding time cost required to complete the reservation quantity (referred to as the reservation completion time cost) are obtained.

[0035] For example, multiple data acquisition points are set up within the factory's storage space STK, on ​​the OHCV conveyor belt, and on the large loop track OHS. The transport route corresponding to each product transport task can be defined by one or more data acquisition points. As an example, in Figure 1, the data acquisition points may include positions A, B, C, D, E, F, G, and H, or positions S, P, Q, and R (not shown), etc., thereby forming multiple transport routes between these data acquisition points to transport products within the factory.

[0036] For example, sampling points can be set on each processing machine and its shelf, as well as on the OHCV and OHS. This can be understood as all inventory spaces (STK, OHCV, and OHS) being under monitoring, allowing for real-time detection of whether each sampling point is occupied by products. Furthermore, each data collection location can be associated with multiple sampling points; that is, each data collection location can correspond to an area containing multiple sampling points. The number of occupied sampling points within the corresponding area is calculated by summing these points to determine the product occupancy at that data collection location. For instance, multiple data collection locations can be set within an STK, each corresponding to an area within the STK, including the corresponding processing machine and multiple sampling points at its shelf. Additionally, each OHCV and each OHS can also have one or more data collection locations. Moreover, in the context of this disclosure, the transport route mentioned is defined relative to the data collection location.

[0037] Therefore, for each data collection location, the product occupancy, reservation quantity, and time cost required to complete each reservation (referred to as reservation completion time cost) can be obtained in real time at each sampling moment based on the sampling time interval, and the collection results obtained each time can be recorded.

[0038] Additionally, at the current moment, the reservation quantity at data collection location A represents the number of transport tasks (jobs) that will be moved to data collection location A at the current moment. Accordingly, the time cost for completing the reservation corresponding to this reservation quantity can be obtained. Unlike space occupancy and reservation quantity, which can be directly obtained from the status records, the time cost for completing the reservation can be estimated by fitting a data model, i.e., by constructing a data model. This will be described in detail later.

[0039] In step S220, a machine learning model is used to predict the transport status of each data collection location based on the product occupancy, reservation quantity, and the time cost of completing the reservation corresponding to the reservation quantity at each data collection location at the current moment.

[0040] Optionally, the machine learning model is trained based on the historical product availability, historical reservation volume, and historical reservation completion time cost corresponding to each historical reservation volume at various data collection locations for multiple historical sampling times. For example, the machine learning model can be a recurrent neural network (RNN) or a long short-term memory network (LSTM).

[0041] For example, as mentioned earlier, each STK, OHCV, and OHS has status records. Therefore, it's possible to record the historical product occupancy and reservation volume at different historical sampling times for each data collection location. Furthermore, the corresponding historical reservation completion time cost can be estimated using a data model. Thus, the data collected or estimated at different historical sampling times can form training and / or testing datasets for training machine learning models. The specific training process will be described later.

[0042] The trained machine learning model has the ability to predict the transport status of each data collection location over a future period based on the current product occupancy, reservation volume, and reservation completion time cost at each data collection location. For example, it can predict at least one of the product occupancy, reservation volume, and reservation completion time cost at future sampling times to determine the congestion status.

[0043] Optionally, when predicting the transport status corresponding to each data collection location, since each sample feature of the input machine learning model in the training dataset and test dataset of the machine learning model has a specific format, when using the trained machine learning model for prediction, it is also necessary to make the format of the input features of the input machine learning model the same as that of each sample feature. As will be described later, each sample feature of the input machine learning model is obtained by combining the product occupancy, reservation quantity and reservation completion time cost of l consecutive historical sampling times (time series length l). Therefore, the following operations can also be performed.

[0044] First, obtain the historical product occupancy, historical reservation volume, and historical reservation completion time cost corresponding to each historical reservation volume at each data collection location for the l-1 historical sampling times prior to the current time, where l is an integer greater than 1 and less than the total number of historical sampling times. Then, construct input features based on the product occupancy, reservation volume, and reservation completion time cost at each data collection location for the current time and the l-1 historical sampling times (therefore, a total of l times). Finally, use the input features as input to the machine learning model to predict at least one of the product occupancy, reservation volume, and reservation completion time cost at each data collection location at the next sampling time, as the transport status corresponding to each data collection location.

[0045] Therefore, by referring to the control method for product handling in the factory described in Figure 2, a trained machine learning model is used to predict one or more of the following in the future: product occupancy, reservation quantity, and time cost to complete a reservation at each data collection location within each inventory space STK, on ​​the OHCV conveyor belt, and on the large loop track OHS: i.e., the handling status. This predicted handling status can indicate areas where congestion risk is imminent; that is, if the congestion level in a certain area is high, the value of one or more of the following—product occupancy, reservation quantity, and time cost to complete a reservation—at the data collection location in that area will be higher.

[0046] Optionally, areas at risk of congestion can be identified based on the forecast results, allowing for early intervention to ensure smooth operation of the entire factory and maximize the capacity of processing equipment.

[0047] Therefore, method 200 may further include step S230, which schedules the product transport in the factory based on the transport status corresponding to each data acquisition location.

[0048] For example, the transport status corresponding to each data collection location can include the product occupancy at each location. In this case, a heatmap of each data collection location in the factory can be drawn based on the predicted product occupancy at each location, and different levels of product occupancy can be displayed using different markers (e.g., different colors), with higher product occupancy corresponding to a higher level of congestion at the data collection location. Then, transport tasks related to data collection locations with high congestion levels (e.g., product occupancy greater than or equal to a corresponding threshold) can be reassigned or lower-priority transport tasks related to that data collection location can be canceled. Therefore, intervention can be made in advance to prevent potential congestion, ensuring smooth operation of the entire factory and maximizing the capacity of processing equipment.

[0049] For example, the transport status corresponding to each data collection location can also include the time cost for completing the reservation corresponding to the reservation volume at each data collection location. In this case, a congestion risk index can be defined for each data collection location based on the time cost for completing the reservation corresponding to the reservation volume at each data collection location, where a higher product occupancy corresponds to a higher degree of congestion at the data collection location. Then, similarly, transport tasks related to data collection locations with high congestion risk indices (i.e., high congestion levels, such as a congestion risk index greater than or equal to the corresponding threshold) can be reassigned or lower-priority transport tasks related to that data collection location can be canceled.

[0050] In other words, the congestion risk index for each data collection location at a given time can be defined by predicting the time cost to complete an appointment, thereby quantifying the congestion status. Optionally, the congestion risk index for each data collection location at a given time is defined as the ratio of the difference between the time cost to complete an appointment corresponding to the number of appointments at that data collection location and the ideal time cost to the ideal time cost.

[0051] For example, the congestion risk index of each data collection location (1, ..., m) at time t can be defined by the following formula (1):

[0052] This represents the predicted transport time (time cost required to complete the reservation) for each data collection location (1, ..., m) at time t. This represents the ideal transport time (ideal time cost) for each data collection point (1, ..., m). Therefore, this process can monitor the future congestion risk index of each data collection location and provide early warnings for areas with high congestion indices, facilitating the relief of transport tasks and avoiding waste of equipment capacity due to increased waiting time caused by congestion.

[0053] Figure 3 illustrates a schematic diagram of constructing a training dataset for training a machine learning model according to an embodiment of the present disclosure.

[0054] As mentioned earlier, the historical product availability and reservation volume at each data collection location can be recorded at different historical sampling times. Furthermore, a data model can be used to estimate the historical reservation completion time cost corresponding to each reservation volume. Therefore, the data collected or estimated at different historical sampling times can form training and / or testing datasets.

[0055] As shown in Figure 3, the first matrix represents the historical product occupancy of each data collection location at different historical sampling times.

[0056] In the first matrix ( In the matrix, the number of rows represents the number of data collection locations (m in total), and the number of columns represents the time span of data collection (historical sampling times tn, t-n+1, ​​..., t-1, t, a total of n+1 historical sampling times). The most recently recorded historical sampling time is time t. Therefore, the first matrix has a dimension of m rows and (n+1) columns. Each row in the matrix represents the product occupancy at a data collection location from time tn to time t (i.e., times tn, t-n+1, ​​..., t-1, t), and each column represents the occupancy at m data collection locations at the same time. For example, f 1,tThis represents the product occupancy at time t at the first data collection location, and f m,t-n This represents the product occupancy at the m-th data collection location at time tn.

[0057] Second matrix and the third matrix The meaning of each matrix element in the first matrix is ​​the same as that in the second matrix. The same. For example, r 1,t Let r represent the number of reservations at the first data collection location at time t, and r m,t-n c represents the number of reservations for the m-th data collection location at time tn. 1,t This indicates that the first data acquisition location completed the scheduled r in the second matrix at time t. 1,t The estimated completion time and cost of each delivery task, and c m,t-n This indicates that the m-th data acquisition position completes the scheduled r in the second matrix at time tn. m,t The estimated time and cost for completing a delivery task.

[0058] As mentioned earlier, since multiple transport tasks may be scheduled, and each transport task's route may pass through multiple data collection locations, at a given sampling time, there may be multiple transport tasks about to deliver products to that data collection location. Therefore, the number of scheduled tasks at that data collection location at that time represents the number of these multiple transport tasks. Thus, we can determine the currently scheduled transport tasks about to be delivered to each data collection location for each historical sampling time to obtain the second matrix.

[0059] Furthermore, the third matrix represents the time cost corresponding to the reservation volume at each data collection location, i.e., the estimated time required for each data collection location to complete the reservation volume at different historical sampling times. For example, c m,t-n This indicates that the reserved r for the m-th data acquisition position in the second matrix at time tn has been completed. m,t The estimated time cost of each transport task.

[0060] The following describes the process of determining the time cost corresponding to the number of reservations for each data collection location at each historical sampling time.

[0061] At a historical sampling time td, there may be one or more transport tasks that are expected to transport the product to a certain data collection location (which serves as the destination of these transport tasks), and the location of the product at the historical sampling time td can also be obtained based on the historical transport scheduling records.

[0062] Each transport task corresponds to a transport route, which includes one or more sub-paths. For a specific data collection location, if one or more transport tasks are scheduled to transport the product to that data collection location at a historical sampling time td, that specific data collection location serves as the endpoint of one or more sub-paths. Since an endpoint can only receive one transport task at a time, if multiple transport tasks have scheduled to arrive at that endpoint, they must be completed one by one. The number of historical transport tasks with that specific data collection location as the endpoint, as mentioned above, can be recorded in real-time and thus obtained. Because it changes in real-time, it decreases by 1 for each completed transport task and increases by 1 for each scheduled transport task, hence it is defined as a real-time value.

[0063] Optionally, the transport routes associated with each transport task corresponding to the historical reservation volume can be obtained. For each transport route associated with the historical reservation volume, the current location of the product at the current historical sampling time td is determined, and the set of sub-paths between the current location and the data collection location is determined—that is, the remaining sub-paths that still require product transport—serves as the set of sub-paths for each transport task. For example, if there are no other data collection locations between the current location and the data collection location, the corresponding set of sub-paths includes one sub-path between the current location and the data collection location. If there are np data collection locations between the current location and the data collection location, the corresponding set of sub-paths includes np+1 sub-paths between the current location and the data collection location, where np is an integer greater than or equal to 1. Each transport route associated with the historical reservation volume has its own transported products; therefore, a set of sub-paths can be determined for each transport route / each transport task.

[0064] Then, when determining the historical completion appointment time cost corresponding to the historical appointment volume at each data acquisition location for a certain historical sampling time td, the remaining total path time cost from the transfer task to the data acquisition location can be determined based on the sub-path time cost of each sub-path in the sub-path set corresponding to each transfer task for that historical appointment volume. Next, the completion appointment time cost corresponding to that historical appointment volume can be determined based on the remaining total path time cost from each transfer task to the data acquisition location. For example, the remaining total path time costs corresponding to each transfer task can be added together to obtain the completion appointment time cost corresponding to that historical appointment volume. It should be noted that if at least two transfer tasks have different sub-path start points and different transfer tools, these transfer tasks can be parallel, and if the same transfer tool or reaching the same sub-path endpoint is required, they can be performed sequentially according to priority. In the scheme disclosed herein, considering the worst-case scenario (requiring sequential execution), the maximum possible completion appointment time cost is obtained by adding the remaining total path time costs corresponding to each transfer task.

[0065] Optionally, as mentioned above with reference to Figure 1, the product can be transported in different ways during the transport process. Therefore, each sub-path may correspond to a different transport type, and different transport types correspond to different transport characteristics that affect the time cost of the sub-path. Therefore, the sub-path time cost corresponding to each sub-path can be determined according to the transport type of each sub-path, thereby obtaining the remaining total path time cost of each transport task to the data collection location.

[0066] For example, when the transport type on the sub-path is elevator and OHCV transport (Type 1), it is a bidirectional transport task and there is no transport waiting time. Therefore, the sub-path time cost = the preset transport time cost of the sub-path (fixed value) * the number of transport tasks scheduled to the two endpoints of the sub-path (real-time value).

[0067] For example, if the transport type on the sub-path is transport in the storage space STK and the large loop (OHS) (Type 2), that is, the transport task is a one-way transport task that requires a trolley and generally requires transport waiting time, then the sub-path time cost = the preset transport time cost of the sub-path (fixed value) + the maximum transport waiting time of the sub-path in a single transport task * the number of transport tasks scheduled to the end of the sub-path (real-time value).

[0068] Optionally, the preset transport time cost for each sub-path refers to the time normally required to transport products on that sub-path. It is a fixed value and can be determined by calculating the average or median of the recorded preset transport time costs. Furthermore, the maximum transport waiting time for each sub-path in a single transport task is a variable and can be determined based on the average transport waiting time, the probability of completing a single transport task on that sub-path within the expected maximum transport waiting time, and the principle of exponential distribution.

[0069] For example, the calculation method is as follows:

[0070] The probability density function of the exponential distribution is shown in equation (2):

[0071] The expected value and variance are respectively The probability that a random variable X exceeds a specified value a is given by formula (3):

[0072] Based on the above probability density function, assuming the average delivery waiting time within the inventory space STK and on the sub-paths of the large loop is 25 seconds, then λ = 1 / 25. The delivery waiting time is a random variable X, and the probability of exceeding a specified value a (i.e., the maximum delivery waiting time) is set to 0.05, i.e., e. -λa =0.05, so a = 75 can be calculated. This means the probability of a delivery wait time exceeding 75 seconds is 0.05, and the probability of a delivery wait time less than 75 seconds is 0.95. Therefore, 75 seconds can be used as the maximum delivery wait time for this sub-path in a single delivery task. It should be understood that this is only an example. Users can set the probability of wanting to complete a single delivery task on a sub-path to the maximum delivery wait time, and then calculate the time (i.e., the maximum delivery wait time) to complete the delivery based on the principle of the exponential distribution mentioned above.

[0073] Therefore, the remaining total path time cost for each transport task corresponding to the historical reservation volume at this specific data collection location can be calculated. Since this specific data collection location can only receive one transport task at a time, and subsequent transport tasks must wait for the previous transport tasks to complete, i.e., the transport tasks are executed sequentially, the remaining total path time cost of each of the transport tasks can be added together as the time cost required to complete the historical reservation volume at this specific data collection location for the historical time tc (this gives the maximum time cost, because some transport tasks can be executed in parallel).

[0074] Similarly, when using a trained machine learning model for prediction, for each data collection location, the transport route associated with each transport task corresponding to the reservation volume can be obtained; and for each transport task associated with the reservation volume, the current location of the product at the current moment can be determined, and the set of sub-paths between the current location and the data collection location can be determined, thus obtaining the set of sub-paths corresponding to each transport task. Furthermore, determining the completion reservation time cost corresponding to the current reservation volume for each data collection location at the current moment is the same process as described above for calculating historical completion reservation time costs, i.e., based on the sub-path time cost of each sub-path in the set of sub-paths corresponding to each transport task corresponding to the current reservation volume, the remaining total path time cost from the transport task to the data collection location is determined; and based on the remaining total path time cost from each transport task to the data collection location, the completion reservation time cost corresponding to the reservation volume is determined.

[0075] As an example, Figure 4 shows that at the current time tc (referring to a historical sampling time or the current time actually used for prediction), three transport tasks are about to be moved to the data acquisition location Pm. Typically, the data acquisition location closest to the product at the current time is taken as the product's current location at that time.

[0076] As shown in Figure 4, the first transport task (job1) is set to travel from data acquisition position A1 (the starting position) through data acquisition positions B1 and C1 to data acquisition position Pm. Specifically, sub-path P11 from data acquisition position A1 to data acquisition position B1 uses an intra-STK transport method, sub-path P12 from data acquisition position B1 to data acquisition position C1 uses an elevator transport method, and sub-path P13 from data acquisition position C1 to data acquisition position Pm also uses an intra-STK transport method. At the current time tc, the product in the first transport task is located at data acquisition position C1.

[0077] The second transport task (job2) is configured to travel from data acquisition position A2 (the transport start position) to data acquisition position Pm via data acquisition position B2. Specifically, the sub-path P21 from data acquisition position A2 to data acquisition position B2 uses the STK (Site-Time Kinematic) transport method, and the sub-path P22 from data acquisition position B2 to data acquisition position Pm uses the OHCV (Ohm-On-Hand Vehicle) transport method. At the current time tc, the product in the second transport task is located at data acquisition position A2.

[0078] The third transport task (job3) is configured to travel from data acquisition position A3 (the transport start position) through data acquisition positions B3 and C3 to data acquisition position Pm. Specifically, the sub-path P31 from data acquisition position A3 to data acquisition position B3 uses an intra-STK transport method, and the sub-path P32 from data acquisition position B3 to data acquisition position Pm uses a large-loop OHS transport method. At the current time tc, the product in the third transport task is located at data acquisition position A3.

[0079] The fourth transport task (job4) is set to travel from data acquisition position O1 (the transport start position) to data acquisition position O2 via data acquisition position C1. At the current time tc, the product in the fourth transport task is located at data acquisition position O1.

[0080] For the first transport task, since the product in the first transport task is located at data acquisition position C1 at the current time tc, the sub-path set corresponding to the first transport task only includes sub-path P13 from data acquisition position C1 to data acquisition position Pm. Since the transport type of sub-path P13 is STK intra-transport, the time cost of sub-path P13 can be calculated as the preset transport time cost of the sub-path (a fixed value) + the maximum transport waiting time of the sub-path in a single transport task (a fixed value obtained through the probability density function of the exponential distribution) * the number of transport tasks currently scheduled to the end of the sub-path (here, the real-time value is 3). Regarding sub-paths P12 and P11, since the product has already been transported through these sub-paths, it is not necessary to calculate the time cost required on these sub-paths at this time.

[0081] Similarly, for the second transport task, since the product in the second transport task is located at data acquisition location A2 at the current time tc, the sub-path set corresponding to this second transport task includes sub-path P21 from data acquisition location A2 to data acquisition location B2 and sub-path P22 from data acquisition location B2 to data acquisition location Pm. Since the transport type of sub-path P22 is OHCV, with bidirectional endpoints B2 and Pm, and since there is also a sub-path in the fourth transport task ending at B2, the time cost of sub-path P22 can be the preset transport time cost of this sub-path (a fixed value) multiplied by the number of transport tasks currently scheduled to reach the bidirectional endpoints of this sub-path (here, the real-time value is 5, where endpoint B2 corresponds to two transport tasks and Pm corresponds to three transport tasks). Regarding sub-path P21, since the transport type of this sub-path is STK intra-transport and there is a fourth transport task with B2 as the destination, the time cost of sub-path P21 can be the preset transport time cost of this sub-path (fixed value) + the maximum transport waiting time of this sub-path in a single transport task (fixed value obtained through the probability density function of the exponential distribution) * the number of transport tasks currently scheduled to the destination of this sub-path (here the real-time value is 2, where the destination B2 corresponds to two transport tasks).

[0082] Similarly, for the third transport task, since the product in the third transport task is located at data acquisition location A3 at the current time tc, the sub-path set corresponding to this third transport task includes sub-path P31 from data acquisition location A3 to data acquisition location B3 and sub-path P32 from data acquisition location B3 to data acquisition location Pm. Since the transport type of sub-path P32 is OHS transport, the time cost of sub-path P32 can be calculated as: the preset transport time cost of the sub-path (a fixed value) + the maximum transport waiting time of the sub-path in a single transport task (a fixed value obtained through the probability density function of the exponential distribution) * the number of transport tasks currently scheduled to reach the endpoint of the sub-path (here, the real-time value is 3). Regarding sub-path P32, since the transport type of this sub-path is STK intra-transport and there are no other transport tasks ending at B3, the time cost of sub-path P32 can be the preset transport time cost of this sub-path (fixed value) + the maximum transport waiting time of this sub-path in a single transport task (fixed value obtained through the probability density function of the exponential distribution) * the number of transport tasks currently scheduled to the end of this sub-path (here the real-time value is 1).

[0083] Therefore, by using the sub-path time cost corresponding to each sub-path of each transport route, the remaining total path time cost corresponding to each transport task can be obtained, and finally, the completion reservation time cost corresponding to the reservation quantity (3 here) of the corresponding tc time and data collection location Pm can be obtained.

[0084] In this way, by calculating the time cost required to complete the reservation volume of each data collection location at each historical sampling time using the above method, the third matrix is ​​obtained.

[0085] Based on the three datasets (first matrix, second matrix, and third matrix) represented in matrix form above, training and / or test datasets can be obtained by splitting the samples for training machine learning models.

[0086] Figure 5 illustrates an example flowchart of obtaining a training dataset and / or a test dataset by sample splitting according to an embodiment of the present disclosure, and Figure 6 illustrates an example process of obtaining a training dataset and / or a test dataset by sample splitting according to an embodiment of the present disclosure.

[0087] First, as shown in Figure 5, in step S510, based on the historical product occupancy, historical reservation volume, and historical reservation completion time cost corresponding to each historical reservation volume at each data collection location for n+1 historical sampling times, a first matrix, a second matrix, and a third matrix with a dimension of m*(n+1) are generated respectively.

[0088] For example, as shown in Figure 6, the first matrix f corresponds to the historical product occupancy at each data collection location at n+1 historical sampling times, the second matrix r corresponds to the historical reservation quantity at each data collection location at n+1 historical sampling times, and the third matrix c corresponds to the historical reservation completion time cost corresponding to each historical reservation quantity in the second matrix.

[0089] In step S520, the first matrix, the second matrix, and the third matrix are split into samples based on a preset time series length l and a preset step size s to obtain a sample feature set.

[0090] For example, after determining the first sample feature corresponding to the previous three historical sampling times, the time step s can be slid sequentially to obtain multiple sample features.

[0091] Optionally, when splitting the three matrices into samples according to a preset time series length l, the sub-sample features obtained from the splitting of each matrix can be combined to obtain a corresponding sample feature.

[0092] As an example, as shown in Figure 6, with a time series length l of 3, the data of product occupancy, reservation volume, and reservation completion time cost (first matrix, second matrix, and third matrix) are taken at the three earliest historical sampling times to obtain sample features, as shown below:

[0093] in, These are the sub-sample features obtained from the first sample split of the first matrix, the second matrix, and the third matrix, respectively. These three sub-sample features are stacked to obtain the sample features obtained after the first sample split, with a dimension of 3m*3.

[0094] In step S530, based on the historical product occupancy, historical reservation volume, and historical reservation completion time cost corresponding to each historical reservation volume at each data collection location for each of the (l+1)th to (n+1)th historical sampling times in the n+1 historical sampling times, a prediction target value corresponding to each sample feature in the sample feature set is determined.

[0095] As shown in Figure 6, the sample feature Ds1 is used to predict the value at the next historical sampling time (the 4th historical sampling time), and the predicted target value is:

[0096] Similarly, the predicted target value can also be viewed as a combination of multiple sub-features, with a dimension of m*3.

[0097] Then, since the time step is s, that is, sliding s historical sampling moments (assuming s=1 here), the three matrices are slid sequentially. Similarly, the data of product occupancy, reservation volume, and time cost required to complete a reservation (first matrix, second matrix, and third matrix) are taken from the 2nd to 4th earliest historical sampling moments to obtain the sample features, as shown below:

[0098] Furthermore, similarly, this sample Ds2 is used to predict the value at the next historical sampling time (the 5th historical sampling time), and the predicted target value is:

[0099] Continue sliding sequentially until the last sample feature is obtained.

[0100] Therefore, by synchronously sliding these three matrices according to the sequence length to perform sample splitting, the sample features X = [Ds1, Ds2, ..., Ds] are obtained. n-l+1 ], and their respective prediction targets Y = [Y l+1 , ..., Y n+1 That is, the number of sample features is the same as the number of target values ​​to be predicted, which is n-1+1.

[0101] Therefore, after obtaining the dataset [X,Y], this dataset can be used to train a machine learning model, such as an LSTM or RNN model. For example, each sample feature in X can be used as input to the machine learning model, and the model can output a prediction result for each sample feature. Based on the prediction result corresponding to each sample feature and the corresponding prediction target value in Y, the parameters of the machine learning model can be adjusted.

[0102] Optionally, in some embodiments, the sample features obtained after sample splitting according to the above method, along with their corresponding prediction target values, can be divided into training and testing datasets. The training and testing datasets can be used for cross-validation of the machine learning model (e.g., 5-fold cross-validation). The ratio of the number of sample features in the training dataset to the number of sample features in the testing dataset is an integer greater than 1. For example, in the case of 5-fold cross-validation, all sample features are divided into 5 subsets, with the training dataset containing 4 subsets and the testing dataset containing 1 subset. Each time the model is trained and its performance is tested, one subset is sequentially selected from the 5 subsets as the current testing dataset, and the remaining 4 subsets are used as the training dataset for this iteration, thus allowing for 5 training and validation iterations.

[0103] In this way, a trained machine learning model can be obtained, which can be used to predict target values ​​based on the current product occupancy, reservation volume, and the time cost required to complete the reservation corresponding to the reservation volume. That is, the occupancy, reservation volume, and time cost to complete the reservation in the future (e.g., the next sampling time).

[0104] Therefore, according to the product handling scheduling scheme disclosed herein, a dataset (training dataset and / or test dataset) is constructed for training a machine learning model based on the product occupancy (corresponding to completed handling tasks), the number of scheduled handling orders (i.e., handling tasks scheduled to begin at that time), and the estimated completion time (i.e., the time cost of completing the scheduled handling orders) at each preset location within each inventory space STK, on ​​the conveyor belt OHCV, and on the large loop track OHS over a historical period (multiple historical sampling times). This trained machine learning model is used to predict the product occupancy, number of scheduled handling orders, and estimated completion time at each preset location within each inventory space STK, on ​​the conveyor belt OHCV, and on the large loop track OHS over a future period. Based on the prediction results, areas with impending congestion risks can be identified, allowing for early intervention, such as diverting handling tasks to other non-congested paths or canceling lower-priority handling tasks, thereby ensuring smooth operation of the entire factory and maximizing the capacity of processing equipment.

[0105] According to another aspect of this disclosure, a control device for product handling in a factory is also provided.

[0106] Figure 7 shows a structural block diagram of a control device for product transport in a factory according to various embodiments of the present disclosure.

[0107] As shown in Figure 7, the control device 700 may include an acquisition module 710, a prediction module 720, and an optional scheduling module 730.

[0108] The acquisition module 710 can be used to acquire the product occupancy at each data acquisition location in the factory at the current time, the reservation quantity representing the number of transport tasks scheduled to be transported to each data acquisition location at the current time, and the time cost for completing the reservation corresponding to the reservation quantity.

[0109] The prediction module 720 can be used to predict the delivery status of each data collection location by utilizing a trained machine learning model, based on the product occupancy, reservation volume, and the corresponding reservation completion time cost at each data collection location at the current time. The machine learning model is trained based on the historical product occupancy, historical reservation volume, and historical reservation completion time cost at each data collection location at multiple historical sampling times.

[0110] Optionally, the scheduling module 730 can be used to schedule the transport of products in the factory based on the predicted transport status corresponding to each data acquisition location.

[0111] Optionally, when predicting the transport status corresponding to each data collection location, the prediction module 720 is further configured to: obtain the historical product occupancy, historical reservation volume, and historical reservation completion time cost corresponding to each historical reservation volume for each data collection location at each of the l-1 historical sampling times prior to the current time, where l is an integer greater than 1 and less than the total number of historical sampling times; construct input features based on the product occupancy, reservation volume, and reservation completion time cost at each data collection location for the current time and the l-1 historical sampling times; and use the input features as input to the machine learning model to predict at least one of the product occupancy, reservation volume, and reservation completion time cost at each data collection location at the next sampling time as the transport status corresponding to each data collection location.

[0112] Optionally, the acquisition module 710 can also be configured to, for each data acquisition location, acquire the transport route associated with each transport task corresponding to the reservation volume at the data acquisition location; and, for each transport route associated with the reservation volume, determine the current location of the product at the current moment, and determine the set of sub-paths between the current location and the data acquisition location, thereby obtaining the set of sub-paths corresponding to each transport task. Furthermore, when acquiring the completion reservation time cost corresponding to the reservation volume at each data acquisition location at the current moment, the acquisition module 710 can be configured to: determine the remaining total path time cost of the transport task to the data acquisition location based on the sub-path time cost of each sub-path in the set of sub-paths corresponding to each transport task; and determine the completion reservation time cost corresponding to the reservation volume based on the remaining total path time cost of each transport task to the data acquisition location.

[0113] More details on the operation performed by module 710 to determine the time cost of completing the appointment can be found in the descriptions shown in Figures 3 and 4 above, and will not be repeated here.

[0114] Optionally, the scheduling device 700 may further include a training module for training a machine learning model (e.g., an RNN model or an LSTM model) based on the historical product occupancy, historical reservation volume, and historical reservation completion time cost corresponding to each historical reservation volume at each data collection location for n+1 historical sampling times. For example, it generates training and testing datasets based on these historical data and trains the machine learning model by cross-validating the machine learning model using the training and testing datasets (e.g., 5-fold cross-validation).

[0115] The training module, when constructing the training dataset and the test dataset, can be configured to: generate a first matrix, a second matrix, and a third matrix, each with a dimension of m*(n+1), based on the historical product occupancy, historical reservation volume, and historical reservation completion time cost corresponding to each historical reservation volume at each data collection location for n+1 historical sampling times; perform sample splitting on the first matrix, the second matrix, and the third matrix based on a preset time series length l and a preset step size s to obtain a sample feature set; determine the prediction target value corresponding to each sample feature in the sample feature set based on the historical product occupancy, historical reservation volume, and historical reservation completion time cost corresponding to each historical reservation volume at each data collection location for each of the (l+1)th to (n+1)th historical sampling times; and construct the training dataset and the test dataset based on the sample feature set and the prediction target value corresponding to each sample feature in the sample feature set.

[0116] More details about the operations performed by the training module can be found in the descriptions shown in Figures 5 and 6 above, and will not be repeated here.

[0117] Furthermore, although the modules described above are illustrated in Figure 7 by way of example, it should be understood that the device 700 may be divided into more or fewer modules depending on different functions, or each module may be divided into further sub-modules. In some example embodiments, the individual modules or further subdivided sub-modules may be implemented using electronic hardware (e.g., general-purpose processors, DSPs, ASICs, FPGAs or other programmable logic devices, discrete gate or transistor logic, discrete hardware components, etc.), computer software, programs or instruction sets (e.g., which may be stored in random access memory (RAM), flash memory, read-only memory (ROM), erasable programmable ROM (EPROM), etc.) or a combination of both.

[0118] According to another aspect of this disclosure, a computing device is also provided.

[0119] Figure 8 shows a schematic block diagram of a computing device according to an embodiment of the present disclosure.

[0120] As shown in Figure 8, the computing device 800 includes one or more processors, one or more memories connected via a system bus, and optional network interfaces, input devices, and displays. The memories include non-volatile storage media and internal memory. The non-volatile storage media of the terminal stores an operating system and may also store computer-executable programs or computer-readable code. When executed by the processor, the computer-executable program or computer-readable code enables the processor to perform various operations as described above with reference to Figures 2 to 6. The internal memory may also store computer-executable programs or computer-readable code. When executed by the processor, the computer-executable program or computer-readable code enables the processor to perform various operations as described above with reference to Figures 2 to 6.

[0121] The processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this disclosure. The general-purpose processor can be a microprocessor or any conventional processor, and can be based on an x84 architecture or an ARM architecture.

[0122] Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. It should be noted that the memory used in the methods described in this disclosure is intended to include, but is not limited to, these and any other suitable categories of memory.

[0123] The display screen of a computing device can be an LCD screen or an e-ink screen. The input device of a computing device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the terminal casing, or external keyboards, touchpads, or mice, etc.

[0124] According to another aspect of this disclosure, a computer-readable storage medium is also provided, which stores a computer program that, when executed by a processor, causes the processor to perform various operations as described above with reference to Figures 2 to 6.

[0125] According to another aspect of this disclosure, a computer program product is also provided, including a computer program that, when executed by a processor, implements various operations as described above with reference to Figures 2 to 6.

[0126] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of the methods and apparatus according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing at least one executable instruction for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0127] The exemplary embodiments of this disclosure described in detail above are merely illustrative and not restrictive. Those skilled in the art will understand that various modifications and combinations can be made to these embodiments or their features without departing from the principles and spirit of this disclosure, and such modifications should fall within the scope of this disclosure.

Claims

1. A control method for product conveying in a factory, comprising: Obtain the product occupancy, reservation quantity, and corresponding reservation completion time cost at each data collection location in the factory at the current moment, wherein the reservation quantity at each data collection location is used to represent the number of transport tasks that will be moved to the corresponding data collection location at the current moment. as well as Using machine learning models, and based on the product occupancy, reservation volume, and the time cost of completing the reservation corresponding to each data collection location at the current moment, the transport status corresponding to each data collection location is predicted.

2. The control method according to claim 1 further includes: Based on the predicted transport status corresponding to each data collection location, the transport of products in the factory is scheduled.

3. The control method according to claim 1, wherein Predict the transport status corresponding to each data collection location, including: Get the historical product occupancy, historical reservation volume, and historical reservation completion time cost corresponding to each historical reservation volume at each data collection location for l-1 historical sampling times before the current time, where l is an integer greater than 1 and less than the total number of historical sampling times. Based on the product occupancy, reservation volume, and reservation completion time cost at each data collection location for the current time and the l-1 historical sampling times, input features are constructed; and The input features are used as input to the machine learning model to predict at least one of the following at each data collection location in the next sampling time: the future product occupancy, the future reservation quantity, and the time cost for completing the reservation corresponding to the future reservation quantity, which is used as the transport status corresponding to each data collection location.

4. The control method according to claim 1, wherein The machine learning model is trained in the following ways: Based on the historical product occupancy, historical reservation volume, and historical reservation completion time cost corresponding to each historical reservation volume at each data collection location for n+1 historical sampling times, a training dataset and a test dataset are generated, where n is an integer greater than 1. as well as The machine learning model is trained by cross-validating it using the training dataset and the test dataset.

5. The control method according to claim 4, wherein The training dataset and the test dataset are constructed in the following way: Based on the historical product occupancy, historical reservation volume, and historical reservation completion time cost corresponding to each historical reservation volume at each data collection location for the n+1 historical sampling times, a first matrix, a second matrix, and a third matrix with a dimension of m*(n+1) are generated respectively. Based on a preset time series length l and a preset step size s, the first matrix, the second matrix, and the third matrix are split into samples to obtain a sample feature set; Based on the historical product occupancy, historical reservation volume, and historical reservation completion time cost corresponding to each historical reservation volume at each data collection location for each of the (l+1)th to (n+1)th historical sampling times in the n+1 historical sampling times, the prediction target value corresponding to each sample feature in the sample feature set is determined. as well as The training dataset and the test dataset are constructed based on the sample feature set and the prediction target value corresponding to each sample feature in the sample feature set.

6. The control method of claim 1, further comprising: For each data collection location, Obtain the transport route associated with each transport task corresponding to the reservation volume; as well as For each delivery task associated with the reservation volume, determine the current location of the product at the current moment, and determine the set of sub-paths between the current location and the data collection location to obtain the set of sub-paths corresponding to each delivery task.

7. The control method according to claim 6, wherein, Obtaining the time cost for completing a reservation corresponding to the reservation quantity at each data collection location at the current time includes: Based on the sub-path time cost of each sub-path in the sub-path set corresponding to each transport task, determine the remaining total path time cost from the transport task to the data collection location; and Based on the remaining total path time cost to the data collection location for each transport task, determine the completion appointment time cost corresponding to the appointment quantity.

8. The control method according to claim 6, wherein, Determining the set of sub-paths between the current location and the data acquisition location includes: When there are no other data acquisition locations between the current location and the data acquisition location, the sub-path set includes the sub-paths between the current location and the data acquisition location; and When there are np data collection locations between the current location and the data collection location, the sub-path set includes np+1 sub-paths between the current location and the data collection location, where np is an integer greater than or equal to 1.

9. The control method according to claim 6, wherein, For each subpath: When the transport type of the sub-path is the first type of transport, the time cost of the sub-path does not include the transport waiting time; as well as When the transport type of the sub-path is the second type of transport, the time cost of the sub-path includes the transport waiting time.

10. The control method according to claim 9, further comprising: The sub-path time cost of each sub-path in the sub-path set corresponding to each transport task is determined in the following way: When the transport type of the sub-path is type 1 transport, the sub-path time cost is determined as: the product of the preset transport time cost of the sub-path and the number of transport tasks scheduled to the two endpoints of the sub-path; and / or When the transport type of the sub-path is the second type of transport, the time cost of the sub-path is determined as: the preset transport time cost of the sub-path plus the sum of the products of the maximum transport waiting time of the sub-path in a single transport task and the number of transport tasks scheduled to the end of the sub-path.

11. The control method according to claim 10, wherein, The preset transport time cost of the sub-path is determined as the average or median of the historical transport time costs on the sub-path.

12. The control method according to claim 10, wherein, The maximum transport waiting time for a single transport task on a sub-path is determined based on the average transport waiting time, the probability of completing a single transport task on the sub-path within the maximum transport waiting time, and the principle of exponential distribution.

13. The control method according to claim 2, wherein the conveying status corresponding to each data acquisition position includes the product occupancy at each data acquisition position; The method further includes: For the predicted product occupancy at each data collection location, a heat map of each data collection location in the factory is drawn, and different levels of product occupancy are displayed with different display markers, where higher product occupancy corresponds to a higher degree of congestion at the data collection location.

14. The control method according to claim 2, wherein, The transfer status corresponding to each data collection location includes the time cost for completing the reservation corresponding to the reservation volume at each data collection location. The method further includes: Based on the time cost of completing a reservation corresponding to the reservation volume at each data collection location, a congestion risk index is defined for each data collection location. The higher the congestion risk index, the higher the degree of congestion at the data collection location.

15. The control method according to claim 13 or 14, wherein scheduling the product transport in the factory based on the predicted transport status corresponding to each data acquisition location includes: Tasks related to data collection locations with high congestion levels can be de-escalated, or lower-priority data collection tasks related to those locations can be cancelled.

16. The control method according to claim 14, wherein, The congestion risk index at each data collection location is defined as the ratio of the difference between the time cost of completing a reservation and the ideal time cost corresponding to the number of reservations at that data collection location to the ideal time cost.

17. A control device for product conveying in a factory, comprising: The acquisition module is used to acquire the product occupancy, reservation quantity, and the time cost for completing the reservation corresponding to each data acquisition location in the factory at the current time, wherein the reservation quantity at each data acquisition location is used to represent the number of transport tasks that are about to be moved to the corresponding data acquisition location at the current time. as well as The prediction module is used to predict the delivery status of each data collection location by using a trained machine learning model and based on the product occupancy, reservation quantity, and the time cost of completing the reservation corresponding to the reservation quantity at each data collection location at the current moment.

18. The control device according to claim 17, further comprising: The scheduling module is used to schedule the transport of products in the factory based on the predicted transport status corresponding to each data collection location.

19. A computing device, comprising: processor; as well as A memory having a computer-executable program stored thereon, which, when executed by the processor, causes the processor to perform the method as described in any one of claims 1-17.

20. A computer-readable storage medium having a computer-executable program stored thereon, which, when executed by the processor, causes the processor to perform the method as described in any one of claims 1-17.

Images