Multi-channel based dynamic scheduling method, device, equipment and production line

Abstract

The embodiment of the application provides a kind of based on the dynamic scheduling method, device and equipment of multi-channel and production line.It relates to vehicle production technical field.The method comprises: obtaining the event stream of multiple channels in the direct operation terminal of this adjustment period, the type of event stream includes formal order, intention order and browsing behavior;Standardized mapping and deduplication processing are carried out to event stream, and target event stream is obtained;Based on intention order and browsing behavior, conversion rate prediction is carried out according to the preset secondary confidence progressive rule, and the predicted order of each vehicle configuration in this adjustment period is obtained;The data fusion of predicted order and the unproduced order of unproduced order in formal order is carried out, and the order demand of each vehicle configuration in this adjustment period is obtained;Based on the order demand of each vehicle configuration, dynamic scheduling processing is executed through rolling time domain strategy, and the production scheduling plan of each vehicle configuration in this adjustment period is obtained.The application is used to achieve the effect of accurate and efficient capacity allocation.

Description

Technical Field

[0001] This application relates to the field of vehicle production technology, and in particular to a dynamic scheduling method, apparatus, equipment and production line based on multiple channels. Background Technology

[0002] As competition intensifies in the automotive industry, car manufacturers are increasingly adopting a direct-to-consumer (D2C) model, where consumers place orders directly through channels operated by the car manufacturer's brand. Under this model, order data shifts from a traditional multi-tiered dealership aggregation model to a real-time, direct-to-consumer model. The rapid changes in consumer preferences for vehicle configurations place higher demands on the responsiveness of production systems.

[0003] In related technologies, vehicle production scheduling systems receive sales data and sales forecasts on a monthly or weekly basis. Based on this data, they formulate production capacity and model plans and schedule vehicle production. However, this approach suffers from the inability to allocate production capacity in a timely and accurate manner, leading to shortages of popular models and stockpiles of slow-moving models. This, in turn, results in a disconnect between production and sales, inventory buildup, or wasted production capacity. Summary of the Invention

[0004] This application provides a multi-channel dynamic scheduling method, apparatus, equipment, and production line to achieve accurate and efficient capacity allocation.

[0005] In a first aspect, embodiments of this application provide a dynamic scheduling method based on multiple channels, including:

[0006] Acquire event streams from multiple channels within the directly operated terminals during this adjustment period. The types of event streams include formal orders, intended orders, and browsing behavior.

[0007] The event stream is standardized and deduplicated to obtain the target event stream;

[0008] Based on intended orders and browsing behavior, conversion rate is predicted according to a preset two-level confidence progression rule to obtain the predicted orders for each model configuration in this adjustment period. Among them, the probability weight of intended orders is higher than that of browsing behavior.

[0009] By fusing the forecasted orders with the unproduced orders that have not yet been scheduled for production in the formal orders, the order demand for each model configuration in this adjustment cycle can be obtained.

[0010] Based on the order demand for each vehicle configuration, dynamic scheduling is performed using a rolling time-domain strategy to obtain the production schedule plan for each vehicle configuration in this adjustment period.

[0011] In one possible implementation, the event stream is standardized, mapped, and deduplicated to obtain the target event stream, including:

[0012] Based on the signature table mapping mechanism, event streams from different channels are mapped in a unified manner to obtain a standard event stream with consistent representation. The signature table mapping mechanism includes vehicle configuration mapping and user identifier mapping.

[0013] The standard event stream is timestamped and then deduplicated within the same deduplication window for event streams with identical timestamps, configuration signatures, and user identifiers to obtain the target event stream.

[0014] In one possible implementation, based on intended orders and browsing behavior, conversion rate prediction is performed according to a preset two-level confidence progression rule to obtain predicted orders for each vehicle configuration during this adjustment period, including:

[0015] Based on the timestamps of intended orders and browsing behavior in the event stream, the first delay of intended orders and the second delay of browsing behavior in this adjustment period are determined;

[0016] If the first delay and the second delay do not meet the delay conditions, the predicted orders for each model configuration in the previous adjustment period will be used as the predicted orders for this adjustment period. The delay conditions include that the first delay is less than the first delay threshold and the second delay is less than the second delay threshold.

[0017] If the first and second time delays meet the time delay conditions, then based on the intended orders and browsing behavior, the conversion rate prediction model is used to predict the conversion rate according to the preset second-level confidence progression rule, and the predicted orders for each model configuration in this adjustment period are obtained.

[0018] In one possible implementation, the predicted orders include: intended demand and demand adjustment factors. Based on intended orders and browsing behavior, a conversion rate prediction model is used to predict the conversion rate according to a preset two-level confidence progression rule, resulting in predicted orders for each vehicle configuration during this adjustment period, including:

[0019] Based on the order conversion rate prediction model, the conversion rate of intended orders is predicted to obtain the first probability weight, and the conversion rate of browsing behavior is predicted to obtain the second probability weight, which is lower than the first probability weight.

[0020] The intended demand for each vehicle configuration is obtained by weighting the quantity of each vehicle configuration within the intended order based on the first probability weight.

[0021] The number of vehicle configurations within the browsing behavior is weighted based on the second probability weight to obtain the demand correction factor for each vehicle configuration.

[0022] The sum of the intended demand and the demand correction factor for each vehicle configuration will be used as the predicted order for each vehicle configuration in this adjustment period.

[0023] In one possible implementation, after obtaining the event streams from multiple channels within the directly operated terminals during the current adjustment period, the dynamic scheduling method based on multiple channels further includes:

[0024] Obtain the historical event stream corresponding to the anomaly detection window;

[0025] The mean of the corresponding vehicle configuration within the historical event stream is summed with the standard deviation of the preset weight to obtain the abnormal threshold of each vehicle configuration.

[0026] If the number of vehicle configurations represented by the event stream is greater than or equal to the abnormal threshold, the event stream data is determined to be abnormal, the event stream data is marked as abnormal, and manual review is triggered.

[0027] In one possible implementation, based on the order demand for each vehicle configuration, dynamic scheduling is performed using a rolling time-domain strategy to obtain the production schedule plan for each vehicle configuration in the current adjustment period, including:

[0028] Based on the order requirements for each vehicle model configuration, dynamic scheduling is performed through a fixed rolling cycle and scheduling time domain to obtain the scheduling results;

[0029] If the scheduling result meets the production line constraints, the scheduling result will be used as the production scheduling plan. The production line constraints include capacity constraints, material availability constraints, process route constraints, delivery date constraints, and personnel scheduling constraints.

[0030] If the scheduling result does not meet the production line constraints, the dynamic scheduling will be re-executed with a fixed rolling cycle and scheduling time domain until the scheduling result meets the production line constraints, or the number of recalculations of the dynamic scheduling reaches a preset threshold.

[0031] In one possible implementation, based on the order demand for each vehicle configuration, dynamic scheduling is performed using a rolling time-domain strategy to obtain the production schedule plan for each vehicle configuration in the current adjustment period. The multi-channel dynamic scheduling method further includes:

[0032] Implement production scheduling plans on the production line;

[0033] Collect data on the execution status of the production line's production schedule and calculate the deviation rate between the schedule and the actual execution.

[0034] When the deviation rate exceeds the dynamic threshold, dynamic scheduling recalculation is triggered, and the order conversion rate prediction model is updated.

[0035] Secondly, embodiments of this application provide a multi-channel-based dynamic scheduling device, comprising:

[0036] The acquisition module is used to acquire event streams from multiple channels within the directly operated terminals during this adjustment period. The types of event streams include formal orders, intention orders, and browsing behavior.

[0037] The processing module is used to standardize and deduplicate the event stream to obtain the target event stream; based on intended orders and browsing behavior, it predicts the conversion rate according to a preset two-level confidence progression rule to obtain the predicted orders for each vehicle configuration in this adjustment period, where the probability weight of intended orders is higher than that of browsing behavior; it merges the predicted orders with the unproduced orders in the formal orders to obtain the order demand for each vehicle configuration in this adjustment period; based on the order demand for each vehicle configuration, it performs dynamic scheduling processing through a rolling time domain strategy to obtain the production scheduling plan for each vehicle configuration in this adjustment period.

[0038] In one possible implementation, the processing module is specifically used to: map event streams from different channels in a unified manner based on a signature table mapping mechanism to obtain a standard event stream with consistent representation. The signature table mapping mechanism includes vehicle configuration mapping and user identifier mapping; align the standard event stream with timestamps, and perform deduplication processing on event streams with the same timestamp, configuration signature, and user identifier within the same deduplication window to obtain the target event stream.

[0039] In one possible implementation, the processing module is specifically configured to: determine the first delay of intended orders and the second delay of browsing behavior in the current adjustment period based on the timestamps of intended orders and browsing behavior in the event stream; if the first delay and the second delay do not meet the delay conditions, then the predicted orders of each vehicle configuration in the previous adjustment period are used as the predicted orders for the current adjustment period, wherein the delay conditions include that the first delay is less than the first delay threshold and the second delay is less than the second delay threshold; if the first delay and the second delay meet the delay conditions, then based on the intended orders and browsing behavior, the conversion rate is predicted using an order conversion rate prediction model according to a preset two-level confidence progression rule to obtain the predicted orders of each vehicle configuration in the current adjustment period.

[0040] In one possible implementation, the predicted order includes: intended demand and demand correction factor. The processing module is further configured to: predict the conversion rate of intended orders based on an order conversion rate prediction model to obtain a first probability weight, and predict the conversion rate of browsing behavior to obtain a second probability weight lower than the first probability weight; weight the quantity of each vehicle configuration within the intended order based on the first probability weight to obtain the intended demand for each vehicle configuration; weight the quantity of each vehicle configuration within the browsing behavior based on the second probability weight to obtain the demand correction factor for each vehicle configuration; and use the sum of the intended demand and demand correction factor for each vehicle configuration as the predicted order for each vehicle configuration in the current adjustment period.

[0041] In one possible implementation, after obtaining the event streams from multiple channels within the directly operated terminal during the current adjustment period, the processing module is further configured to: obtain the corresponding historical event streams within the anomaly detection window; sum the mean of the corresponding vehicle configurations within the historical event streams with the standard deviation of the preset weights to obtain the anomaly threshold for each vehicle configuration; if the number of vehicle configurations represented by the event stream is greater than or equal to the anomaly threshold, then the event stream data is determined to be abnormal, the event stream data is marked as abnormal, and manual review is triggered.

[0042] In one possible implementation, the processing module is specifically used to: perform dynamic scheduling based on the order requirements of each vehicle model configuration through a fixed rolling cycle and scheduling time domain to obtain the scheduling result; if the scheduling result meets the production line constraints, the scheduling result is used as a production scheduling plan, wherein the production line constraints include capacity constraints, material availability constraints, process route constraints, delivery date constraints, and personnel scheduling constraints; if the scheduling result does not meet the production line constraints, the dynamic scheduling is re-executed through a fixed rolling cycle and scheduling time domain until the scheduling result meets the production line constraints, or the number of recalculations of the dynamic scheduling reaches a preset threshold.

[0043] In one possible implementation, based on the order demand for each vehicle model configuration, dynamic scheduling is performed through a rolling time-domain strategy to obtain the production schedule plan for each vehicle model configuration in the current adjustment period. The processing module is further used to: execute the production schedule plan on the production line; collect the execution status of the production schedule plan on the production line and calculate the deviation rate between the schedule plan and the actual execution; when the deviation rate exceeds the dynamic threshold, trigger dynamic scheduling recalculation and update the order conversion rate prediction model.

[0044] Thirdly, embodiments of this application provide a dynamic scheduling device based on multiple channels, including: a memory and a processor;

[0045] The memory stores the instructions that the computer executes;

[0046] The processor executes computer execution instructions stored in memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0047] Fourthly, embodiments of this application provide a dynamic scheduling production line, on which multi-channel dynamic scheduling equipment as provided in the third aspect and / or various possible implementations of the third aspect is deployed, wherein:

[0048] The dynamic scheduling production line is used to generate a production schedule plan after the multi-channel dynamic scheduling method described in the first aspect and / or various possible implementations of the first aspect is executed by the multi-channel dynamic scheduling equipment, and to allocate and control the cycle time of the assembly tasks of each vehicle model on the production line.

[0049] Fifthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0050] Sixthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0051] The multi-channel dynamic scheduling method, apparatus, equipment, and production line provided in this application acquire event streams from multiple channels within the directly operated terminals during the current adjustment cycle. By standardizing and deduplicating these event streams, it can aggregate and purify multi-source demand information from formal orders, intent orders, and browsing behavior. Subsequently, by obtaining predicted orders for each vehicle configuration based on intent orders and browsing behavior, and by fusing these predicted orders with unproduced orders from the formal order pool, a more comprehensive and realistically changing order demand can be formed. Furthermore, based on the order demand for each vehicle configuration, a rolling time-domain strategy is used to execute dynamic scheduling, improving the real-time performance, accuracy, and ability to match fluctuations in terminal demand. Attached Figure Description

[0052] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0053] Figure 1 A schematic diagram illustrating a scenario of the multi-channel-based dynamic scheduling method provided in this application embodiment;

[0054] Figure 2 A flowchart illustrating a multi-channel-based dynamic scheduling method provided in an embodiment of this application;

[0055] Figure 3 A flowchart illustrating a multi-channel-based dynamic scheduling method provided in an embodiment of this application;

[0056] Figure 4 A schematic diagram of the structure of a multi-channel-based dynamic scheduling device provided in an embodiment of this application;

[0057] Figure 5 This is a schematic diagram of the structure of a multi-channel dynamic scheduling device provided in an embodiment of this application.

[0058] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0059] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application.

[0060] Production planning in automobile manufacturing involves data collaboration among sales terminals, order processing platforms, production scheduling systems, and manufacturing execution systems. Under a brand-direct sales model, orders originate from official apps, mini-programs, websites, directly operated stores, and online flagship stores. The configuration choices made by end-users continuously influence the demand structure for vehicle configurations.

[0061] In related technologies, production plans are often formulated based on sales forecasts or aggregated orders, and scheduling adjustments are performed according to fixed cycles. This data is typically first stored in the sales-side system before being transferred to the planning system. The production side then allocates capacity, materials, and process resources accordingly, forming production tasks for specific vehicle configurations. Under the direct sales model, terminal orders and behavioral data change frequently. Since data passes through multiple layers before entering the scheduling stage, time lags are likely to occur. There is a disconnect between the demand information received by the scheduling system and actual changes at the terminal level; the increase in demand for popular configurations and the decrease in demand for less popular configurations cannot be reflected in the production plan in a timely manner.

[0062] Meanwhile, fixed-cycle scheduling mainly revolves around formal orders, underutilizing demand signals from potential orders and browsing behavior. Scheduling decisions tend to be reactive, lacking proactive identification of short-term fluctuations. When promotional activities, new configurations, or regional hotspots emerge, production plans tend to continue using previous judgments, resulting in insufficient supply of some configurations and over-production of others, affecting inventory turnover, capacity utilization, and order delivery matching.

[0063] The multi-channel dynamic scheduling method provided in this application uses multi-channel direct-sales terminal event streams as the starting point for scheduling input. It places formal orders, intended orders, and browsing behavior into the same data processing link, first completing standardized mapping and deduplication, and then extracting predicted demand for each vehicle configuration from intended orders and browsing behavior. The fusion of predicted and formal orders forms order demand that more closely reflects changes at the terminal level. Dynamic scheduling in the rolling time domain can continuously refresh the production plan with the adjustment cycle. A direct connection is established between the scheduling process and direct-sales terminal data collection, data standardization processing, and production execution feedback. The production plan no longer relies solely on lagging summary results but continuously corrects demand judgments based on real-time events from multiple channels, improving the real-time performance, accuracy, and demand matching capability of scheduling.

[0064] Figure 1 The following is a schematic diagram illustrating a scenario of the multi-channel-based dynamic scheduling method provided in the embodiments of this application, such as... Figure 1 As shown, the specific application scenarios of this application embodiment include multiple channels 11, processing centers 12, and production lines 13, wherein:

[0065] Channel 11 is used to provide users with vehicle configuration selection, intention registration and order placement services, and generates an event flow that includes formal orders, intention orders and browsing behavior.

[0066] The processing center 12 communicates with multiple channels 11 to acquire event streams within each channel 11 and performs standardized mapping and deduplication on the event streams. Simultaneously, the processing center 12 predicts orders for each vehicle model configuration based on intended orders and browsing behavior. After merging the predicted orders with formal orders, it performs dynamic scheduling processing through a rolling time-domain strategy to generate production schedule plans for each vehicle model configuration.

[0067] Production line 13 is communicatively connected to processing center 12 to receive production scheduling plans issued by processing center 12. Production line 13 drives each workstation on the production line to perform vehicle assembly operations according to the production output, production sequence and time window of each vehicle model in the production scheduling plan.

[0068] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0069] Figure 2 This is a flowchart illustrating a multi-channel-based dynamic scheduling method provided in an embodiment of this application. Figure 2 As shown, the method includes:

[0070] S201. Obtain event streams from multiple channels within the directly operated terminals during this adjustment period. The types of event streams include formal orders, intended orders, and browsing behavior.

[0071] Multi-channel is used to limit the source range of the event stream. The event stream serves as the raw input data for dynamic scheduling. Formal orders represent vehicle orders for which users have already paid a deposit. Intended orders represent orders that have not yet been converted into formal orders but for which users have already expressed a purchase intention. Browsing behavior represents users' viewing and searching behavior regarding vehicle configurations across different channels, reflecting the level and changing trends of user interest in vehicle configurations.

[0072] The execution entity in this application embodiment can be a scheduling service deployed in an enterprise data platform or production planning platform. Optionally, the scheduling service establishes data communication connections with sales terminals, order processing platforms, and production scheduling systems, and aggregates data from multiple channels within the directly operated terminals according to a preset adjustment cycle.

[0073] Multiple channels can include online and offline direct sales channels. To ensure that data from different channels enters the same processing loop, formal orders, intention orders, and browsing behavior collected from each channel are transformed and processed to form a unified event stream, thereby providing input for subsequent standardized mapping, deduplication, prediction, and scheduling.

[0074] By integrating demand signals that were originally scattered across different direct-sales terminals into the scheduling chain, the time lag caused by the sales side aggregating the data before transmitting it to the planning side is reduced, allowing subsequent processing to revolve around the changes in actual order demand during this adjustment period.

[0075] In one possible implementation, during the acquisition of event streams from multiple channels within the directly operated terminals during the current adjustment period, if network fluctuations or peak terminal traffic occur, a multi-level priority queue is used to ensure the priority transmission of formal order data. Specifically, formal orders are placed in a high-priority queue, intended orders in a medium-priority queue, and browsing behavior in a low-priority queue. First, data is retrieved from the high-priority queue; data from the medium-priority queue is retrieved only when the high-priority queue is empty; and data from the low-priority queue is retrieved only when both the high-priority and medium-priority queues are empty.

[0076] Optionally, multiple channels may include at least one of the following: official application, mini-program, official website, directly operated store POS, and third-party e-commerce platform.

[0077] For example, obtaining event streams from multiple channels within the directly operated terminals during this adjustment period may include: collecting event streams in JavaScript Object Notation (JSON) format from the official application; obtaining event streams through the user data interface of the open platform, converting the user's OpenID on the mini-program into a unified user identifier using a mapping table; parsing the webpage clickstream logs of the official website, extracting configuration information through page path rules, converting the CookieID into a unified user identifier using a mapping table, and obtaining the event stream; connecting to the order interface of the point-of-sale (POS) system of directly operated stores, extracting store identifiers, salesperson identifiers, and order data, converting the store membership card number into a unified user identifier using a mapping table, and obtaining the event stream of the directly operated store POS system; and periodically batch-fetching order data through the open application programming interface (API) of a third-party e-commerce platform to obtain the event stream.

[0078] Optionally, when acquiring event streams from multiple channels within the directly operated terminals during this adjustment period, a batch request strategy can be adopted, triggering a batch retrieval every preset collection period. If the amount of data returned in a single batch retrieval reaches the upper limit of a single batch retrieval, a supplementary single batch retrieval is immediately triggered until the retrieval is complete. A mapping relationship is established between platform order numbers and brand internal order numbers, and platform user accounts are converted into unified user identifiers through a mapping table to ensure reduced data retrieval latency. Optionally, the collection period is 5 minutes. Optionally, the upper limit for a single batch retrieval is 1000 order records.

[0079] S202. Perform standardization mapping and deduplication on the event stream to obtain the target event stream.

[0080] Standardization mapping is used to unify proprietary expressions from different channels into a single data standard. Deduplication is used to remove duplicate events. The target event stream serves as the normalized input for subsequent order prediction and data fusion.

[0081] After acquiring event stream data, standardized mapping and deduplication are performed on the event streams within the current adjustment cycle. Standardized mapping can be achieved through mapping rules, mapping tables, or other unified encoding mechanisms to convert different expressions regarding users, vehicle models, and configurations from different channels into a unified expression format.

[0082] Optionally, configuration items that are not yet identified in the standardized mapping can be supplemented with additional mappings later.

[0083] Deduplication is used to identify and remove duplicate events, retaining valid event records that can be used for subsequent calculations.

[0084] By standardizing and deduplicating the event stream, the resulting target event stream has a unified expression and has eliminated duplicate counts, and can be directly used for requirement calculations at the configuration level of each vehicle model.

[0085] By standardizing mapping and deduplication, data from different channels, different expressions, and different reporting rhythms are grouped into a computable set of homogeneous events, reducing the bias caused by duplicate reporting and inconsistent expression in demand judgment.

[0086] S203. Based on intended orders and browsing behavior, conversion rate is predicted according to a preset two-level confidence progression rule to obtain the predicted orders for each model configuration in this adjustment period. Among them, the probability weight of intended orders is higher than the probability weight of browsing behavior.

[0087] Predicted orders are used to represent the potential order volume estimated based on weak signaling events such as intended orders and browsing behavior.

[0088] Specifically, in the second-order confidence progression rule, the probability weight of intended orders is used as the medium probability weight, and the probability weight of browsing behavior is used as the low probability weight. Then, by statistically analyzing intended orders and browsing behavior separately according to vehicle configuration, and combining this with historical demand changes, intended order and browsing predictions for each vehicle configuration within the current adjustment period are generated. Finally, the product of the probability weights of the intended order and the intended order is summed with the product of the probability weights of the browsing prediction and the browsing behavior to obtain the predicted orders for each vehicle configuration within the current adjustment period.

[0089] It should be understood that the probability weight of intended orders is used as a medium probability weight, and the probability weight of browsing behavior is used as a low probability weight, in order to ensure that the probability weight of intended orders is higher than the probability weight of browsing behavior.

[0090] Optionally, the probability weight of intended orders can be set to a range of 0.8 to 0.3 to treat intended orders as medium probability weights.

[0091] Optionally, the probability weight of browsing behavior can be set to a range of 0.05 to 0.2 to treat browsing behavior as a low-probability weight.

[0092] Browsing behavior represents a user's initial interest in a particular car model's configuration, representing the highest level of interest signal. An intention to order indicates that the user has completed a more in-depth comparison and consideration, entering the substantive decision-making stage. By incorporating intention to order and browsing behavior into medium and low probability weighting levels respectively, a signal gradient from weak to strong is essentially constructed. This allows order prediction to consider the accumulated user behavior at different stages, dynamically and progressively assessing the likelihood of order conversion. Furthermore, this approach maximizes the use of existing intention to order and browsing behavior without over-relying on formal orders, ensuring that order prediction maintains both high accuracy and sufficient timeliness.

[0093] Optionally, the predicted order should include at least the vehicle configuration identifier, the predicted quantity, and the calculation time.

[0094] By transforming intended orders and browsing behavior that were not originally formal orders into quantifiable order forecasts, scheduling decisions no longer rely solely on completed transaction data. Instead, they can form a forward-looking description of future demand as soon as configuration popularity changes, thus directly addressing the problems of delayed formal orders and difficulty in timely scheduling due to short-term demand fluctuations under the direct operation model.

[0095] S204. The predicted orders are merged with the unproduced orders in the formal orders that have not yet been scheduled for production to obtain the order demand for each model configuration in this adjustment cycle.

[0096] Data fusion is used to unify highly deterministic formal orders and probabilistic forecast orders under the same demand caliber. Order demand serves as the direct input for rolling time-domain dynamic scheduling.

[0097] First, the predicted orders and formal orders are linked according to the vehicle configuration, and the order demand for the corresponding vehicle configuration is obtained based on the combination of the two.

[0098] Optionally, the quantity of the same vehicle model configuration in the formal orders and the forecast orders in the same time period can be summed to obtain the order demand for the corresponding vehicle model configuration.

[0099] Optionally, based on the forecast orders from the previous adjustment period, vehicle configurations that have already been produced or are scheduled for production are removed from the formal orders to obtain the vehicle configurations for the unproduced orders that have not yet been scheduled for production in the formal orders. Then, the quantities of the vehicle configurations for the unproduced orders that have not yet been scheduled for production in the formal orders within the same time period are summed with the quantities of the same vehicle configurations in the forecast orders to obtain the order demand for the corresponding vehicle configurations.

[0100] Optionally, the order demand table should include at least the configuration code, the actual order quantity, the predicted order quantity, the total demand value, and the generation time.

[0101] By merging forecast orders with unproduced orders that have not yet been scheduled for production in the formal order book, a comprehensive order demand can be formed that includes both actual transaction demand and potential near-end demand, making the scheduling input more closely aligned with changes in the end market.

[0102] S205. Based on the order requirements for each vehicle configuration, dynamic scheduling is performed using a rolling time-domain strategy to obtain the production schedule plan for each vehicle configuration in this adjustment period.

[0103] The rolling time-domain strategy is used to continuously recalculate the plan within a finite scheduling time domain at a fixed refresh cycle. Dynamic scheduling processing is used to output production arrangement results for each vehicle model configuration under the combined influence of order demand, capacity, and manufacturing constraints. The production scheduling plan is used to guide subsequent production line execution.

[0104] Starting from the current moment, the system updates cyclically within the set scheduling time domain according to the rolling time domain strategy, and dynamically schedules the configuration of each vehicle model in combination with capacity, materials, processes and resource constraints.

[0105] After dynamic scheduling is completed, the production schedule plan corresponding to the order requirements of the vehicle configuration is output. Optionally, the production schedule plan should include at least the configuration code of the vehicle configuration, the planned output, the production line, the start time and the completion time.

[0106] By repeatedly executing the above process in a rolling time domain manner, the production schedule in the future time domain is refreshed based on the latest order demand in each adjustment cycle, so that both the deterministic changes in formal orders and the potential changes reflected by forecast orders can be continuously included in the scheduling calculation.

[0107] By directly coupling multi-channel terminal demand and forecast results with manufacturing-side resource constraints, and continuously recalculating the scheduling plan at a fixed refresh rate, the production plan no longer relies on lagging periodic summaries, but is adjusted in near real-time around the latest demand, thus obtaining a production scheduling plan that is more consistent with the actual needs of each vehicle configuration.

[0108] This application provides a multi-channel dynamic scheduling method, including acquiring event streams from multiple channels within the directly operated terminals during the current adjustment period; performing standardized mapping and deduplication on the event streams; obtaining predicted orders for each vehicle configuration during the current adjustment period based on intention orders and browsing behavior; fusing the predicted orders with unproduced orders from the formal orders to obtain the order demand for each vehicle configuration during the current adjustment period; and performing dynamic scheduling based on the order demand for each vehicle configuration using a rolling time-domain strategy to obtain the production scheduling plan for each vehicle configuration during the current adjustment period. By incorporating formal orders, intention orders, and browsing behavior from multiple directly operated channels into the same event processing link, and generating predicted orders based on the standardized and deduplicated target event streams, the predicted orders are then merged with formal orders to obtain the order demand for each vehicle configuration. Finally, based on the order demand for each vehicle configuration, dynamic scheduling is performed in the rolling time domain, expanding a single lagging order into a real-time terminal demand set. The production plan can be continuously refreshed with the adjustment period to allocate capacity effectively and promptly in the face of fluctuating market demand.

[0109] In one possible implementation, the adjustment period and the length of the rolling time domain can also be adjusted equivalently according to the complexity of the enterprise's vehicle models, the activity level of the channels, and the production organization method, as long as it can realize the update of order requirements and the generation of production scheduling plans based on multi-channel event flow.

[0110] Figure 3 This is a flowchart illustrating a multi-channel-based dynamic scheduling method provided in an embodiment of this application. Figure 3 As shown, in this embodiment... Figure 2 Based on the examples, a multi-channel-based dynamic scheduling method is described in detail, which includes:

[0111] In one possible implementation, step S202 may further include:

[0112] S2021. Based on the signature table mapping mechanism, event streams from different channels are mapped in a unified manner to obtain a standard event stream with consistent representation. The signature table mapping mechanism includes vehicle configuration mapping and user identifier mapping.

[0113] The original data is read based on the event type field, configuration field, and identity field of the event stream from different channels, and the signature table mapping mechanism is invoked to complete the field normalization.

[0114] Optionally, for vehicle configuration mapping, the original configuration name is matched with the standard configuration code. If multiple configuration versions exist, the current effective version is written into the corresponding mapping relationship.

[0115] Optionally, for user identifier mapping, a unified user primary key is generated based on the channel source and identity binding relationship, and the user primary key is written into the standard event stream.

[0116] Optionally, the standard event stream includes six core fields: event_type, user_id, timestamp, config_signature, channel_source, and confidence_level.

[0117] Optionally, a "configuration signature mapping table" can be established to map vehicle configurations. The core function of the configuration signature mapping table is to force the different field names, value formats, and encoding rules of different business terminals (such as App, POS, and official website) for the same vehicle configuration to converge under the unified standard configuration encoding system within the system. For example, "exterior color = starry blue" recorded on the App and "color code = BL-07" recorded on the POS, although expressed in different forms, will both point to the same standard code "EXT_COLOR_BL07" in the configuration signature mapping table, thereby achieving complete uniformity of configuration semantics across cross-channel event flows.

[0118] Optionally, to ensure the accuracy and timeliness of the "Configuration Signature Mapping Table" for vehicle configuration mapping, the product operations department is specifically responsible for maintaining the "Configuration Signature Mapping Table." Simultaneously, all newly added vehicle configurations will have their mapping entries added or updated within 48 hours. Optionally, each change to the "Configuration Signature Mapping Table" will generate a new version number, while the old version will be retained for 30 days for rapid rollback if necessary, thereby reducing data risks caused by mapping errors or business adjustments.

[0119] Optionally, for newly added configurations that have not yet been mapped, the data flow will not be simply discarded or blocked. Instead, they will be temporarily stored in their original channel codes and classified as "unknown configurations" for separate statistics. They will not be included in the calculation of the conversion rate prediction model for the time being. Within 72 hours after the mapping is completed, the historical temporary data will be automatically backfilled and then included in the prediction analysis of the subsequent rolling cycle. This design ensures the uninterrupted flow of data and avoids the interference of unmapped data on the accuracy of prediction.

[0120] Optionally, a set of "User ID Mapping Tables" corresponding to user identifiers can be established. The "User ID Mapping Table" is used to unify and associate scattered identity identifiers from different channels, such as WeChat OpenID, App User ID, official website CookieID, store membership card number, and e-commerce platform account, under a globally unique user identifier, so that the behavior of the same natural person in multiple channels can be linked together for comprehensive analysis.

[0121] Furthermore, it should be noted that the success rate of the "user ID mapping table" is not 100%. Its effectiveness largely depends on whether the user is logged in and whether there are identifiable cross-device behavioral clues. For cross-device browsing behaviors that are completely anonymous and cannot be associated with any logged-in identity, normalization cannot be achieved. These events will still be recorded separately in the form of independent anonymous IDs and cannot be included in user-level cross-channel behavior analysis.

[0122] In summary, by employing a dual mechanism of vehicle configuration mapping and user identifier mapping, the original multi-channel raw event streams, which were characterized by varying formats, chaotic coding, and dispersed identities, are transformed into a standard event stream with consistent expression, unified coding, and unified user identifiers. This provides a high-quality and reliable data foundation for subsequent advanced analysis tasks such as conversion rate prediction, cross-channel attribution, and user profile construction.

[0123] S2022. The standard event stream is timestamped and deduplicated within the same deduplication window for event streams with identical timestamps, configuration signatures, and user identifiers to obtain the target event stream.

[0124] Next, the original timestamps of the standard event stream are converted according to a unified time source, and the millisecond or second-level timestamps are standardized to a preset granularity so that events from different channels can be compared on the same time scale. Then, using a preset deduplication window as the judgment boundary, the standard event stream is compared by sliding comparison within the same deduplication window. When the timestamp, configuration signature, and user identifier of an event are all the same, it is judged as a duplicate event, and only one valid record is retained. The deduplication window can be set to a fixed duration according to the business sampling frequency.

[0125] Optionally, when the standard event stream enters the deduplication window, it is first sorted by timestamp, and then a joint deduplication key is constructed based on the configuration signature and user identifier. Duplicate records with the same deduplication key are deleted, and the target event stream is output after the window ends.

[0126] In this way, similar events from different channels are standardized and duplicates are removed before entering the predicted order generation stage. The configuration requirements expressed in the target event stream remain consistent, and repeated triggers by the same user across channels are not counted repeatedly. This provides a stable input for subsequent demand extraction based on intended orders and browsing behavior. Because timestamp alignment and joint deduplication work simultaneously, the target event stream can more accurately reflect real terminal changes and provide a consistent data foundation for subsequent order demand fusion and rolling time-domain dynamic scheduling.

[0127] In one possible implementation, step S203 may further include:

[0128] S2031. Based on the timestamps of intended orders and browsing behavior in the event stream, determine the first delay of intended orders and the second delay of browsing behavior in this adjustment period.

[0129] Extract the timestamps of intended orders and browsing behaviors from the event stream, and use the start time, cutoff time, or rolling update time of this adjustment cycle as reference values ​​to calculate the first delay corresponding to each intended order and the second delay corresponding to each browsing behavior.

[0130] S2032. If the first delay and the second delay do not meet the delay conditions, the predicted orders for each model configuration in the previous adjustment period shall be used as the predicted orders for this adjustment period. The delay conditions include that the first delay is less than the first delay threshold and the second delay is less than the second delay threshold.

[0131] If any type of delay exceeds the threshold, it is determined that the behavior data for this period is significantly lagging behind. The predicted orders for each vehicle configuration generated in the previous adjustment period are directly used, and the predicted orders for each vehicle configuration generated in the previous adjustment period are used as the predicted orders for this adjustment period.

[0132] Optionally, the first latency threshold is 10 minutes. Optionally, the second latency threshold is 12 minutes.

[0133] S2033. If the first delay and the second delay meet the delay conditions, then based on the intended orders and browsing behavior, the conversion rate prediction model is used to predict the conversion rate according to the preset second-level confidence progression rule, so as to obtain the predicted orders for each model configuration in this adjustment period.

[0134] When both types of latency do not exceed the preset threshold, it indicates that the currently collected terminal behavior is timely enough. Intended orders, browsing behavior, and historical conversion samples are then input into the order conversion rate prediction model. The order conversion rate prediction model can output the conversion probability for each vehicle configuration based on historical transaction records, configuration selection paths, and activity tags.

[0135] Then, based on the output of the order conversion rate prediction model according to the preset two-level confidence level progressive rule, the predicted orders are generated.

[0136] In this way, the generation of predicted orders is simultaneously constrained by behavioral timeliness and the output of the conversion model. When both types of delays do not exceed preset thresholds, the prediction result of the current cycle is used; when either type of delay exceeds the threshold, prediction continuity is maintained, thus synchronizing the demand input and adjustment cycle for each vehicle model configuration. The resulting predicted orders can stably participate in dynamic scheduling as part of the order demand, and enable the production plan to have a more continuous responsiveness to changes in end-user intentions.

[0137] In one possible implementation, the predicted order includes: intended demand and demand correction factor. Step S2033 may further include: predicting the conversion rate of intended orders based on the order conversion rate prediction model to obtain a first probability weight, and predicting the conversion rate of browsing behavior to obtain a second probability weight lower than the first probability weight; weighting the quantity of each vehicle configuration within the intended order based on the first probability weight to obtain the intended demand for each vehicle configuration; weighting the quantity of each vehicle configuration within the browsing behavior based on the second probability weight to obtain the demand correction factor for each vehicle configuration; and using the sum of the intended demand and demand correction factor for each vehicle configuration as the predicted order for each vehicle configuration in this adjustment period.

[0138] The order conversion rate prediction model performs a probability assessment on each intended order entering the current adjustment cycle, and multiplies the number of vehicle configurations corresponding to each intended order by the first probability weight to obtain the intended demand for the corresponding configuration.

[0139] Meanwhile, the order conversion rate prediction model performs probability assessments on the number of views, page dwells, or favorites that match the vehicle configuration during browsing behavior. The quantity corresponding to each configuration is multiplied by a second probability weight to obtain the demand correction factor for each vehicle configuration.

[0140] Next, the intended demand and demand adjustment factors for each vehicle configuration are summed on the same configuration dimension to obtain the predicted orders for each vehicle configuration in this adjustment cycle. The predicted orders can serve as demand inputs for subsequent integration with formal orders and participate in production scheduling updates in the rolling time domain.

[0141] By assigning different probability weights to intended orders and browsing behavior and then weighting and summarizing them, the predicted orders can simultaneously reflect both clear purchase intentions and potential demand changes. This ensures that the demand assessment for each vehicle configuration is consistent with changes in terminal popularity, improving the matching of order demand in this adjustment cycle and the timeliness of scheduling input.

[0142] In one possible implementation, after obtaining the event streams from multiple channels within the directly operated terminals during the current adjustment period, the dynamic scheduling method based on multiple channels further includes: obtaining the corresponding historical event streams within the anomaly detection window; summing the mean of the corresponding vehicle configurations within the historical event streams with the standard deviation of the preset weights to obtain the anomaly threshold for each vehicle configuration; if the number of vehicle configurations represented by the event stream is greater than or equal to the anomaly threshold, then the event stream data is determined to be abnormal, the event stream data is marked as abnormal, and manual review is triggered.

[0143] First, after receiving event streams from multiple channels within the directly operated terminals during this adjustment period, the current event streams are counted according to vehicle model configuration. Then, the quantity distribution of the corresponding vehicle model configuration within the historical time period is obtained. The mean is calculated based on the frequency of configuration occurrences in the historical event streams, and the standard deviation is weighted using pre-written weight parameters to obtain the anomaly threshold used for sensitivity adjustment. The anomaly detection window is a custom time period.

[0144] When the number of configurations for a certain vehicle model in the event stream reaches or exceeds the threshold, the event stream is written to the anomaly flag field, and the anomaly record is pushed to the manual review queue, where reviewers will verify the source channel, configuration signature, and data integrity.

[0145] If the review confirms that there are duplicate data collections, mismatched mappings, or abnormal data injections, the corresponding event streams will be removed or corrected before entering the subsequent scheduling process.

[0146] By comparing the number of vehicle configurations in the historical event stream with the number of vehicle configurations in the current event stream, the system automatically identifies abnormal fluctuations in vehicle configurations. Abnormal data is marked and transferred to manual review before entering scheduling, ensuring consistency and reliability between the data used for subsequent order demand integration and dynamic scheduling. Since the anomaly threshold is composed of the mean and weighted standard deviation, the threshold can be adjusted according to historical fluctuation levels, providing a more stable ability to identify high-frequency anomalies and sudden deviations.

[0147] Optionally, the anomaly detection window is a time window 15 minutes before and after the current minute within each day. Using the current minute as the detection base point, anomaly detection windows for the same time period within the past 30 days are traced back, and historical event streams for the same vehicle model configuration within the anomaly detection window are extracted. The mean and standard deviation of the corresponding vehicle model configuration in the historical event stream are then calculated. Furthermore, instead of using a one-size-fits-all fixed preset weight for different vehicle model configurations, a differentiated adaptive weighting strategy is implemented based on the market popularity of each vehicle model configuration, setting different preset weights for different vehicle model configurations.

[0148] For example, for popular car model configurations with an average daily order volume of more than 50, a preset weight of 2.5 is set, resulting in an anomaly threshold of the mean plus 2.5 times the standard deviation. For regular configurations with an average daily order volume between 5 and 50, a preset weight of 3 is set, tightening the anomaly threshold to the mean plus 3.0 times the standard deviation, balancing sensitivity and false alarm rate. For less popular configurations with an average daily order volume of less than 5, a preset weight of 4 is set, resulting in an anomaly threshold of the mean plus 4.0 times the standard deviation. This is because less popular configurations have a very low daily base, and small absolute value fluctuations can generate a large relative change rate. A more conservative threshold must be used to effectively filter random noise and avoid frequent false alarms triggered by isolated sporadic events.

[0149] When the number of events for a certain model configuration during the current adjustment period reaches or exceeds the aforementioned adaptively calculated anomaly threshold, it is determined that there is an abnormal fluctuation in the event stream. The event stream data for this adjustment period is automatically marked as "suspected anomaly," triggering a manual review process. The manual review involves operations or data analysis personnel investigating whether the anomaly is driven by a genuine market shift (such as promotional activities or public opinion events) or by systemic issues such as data collection failures, duplicate channel reports, or malicious data manipulation.

[0150] Through the above mechanism, real-time monitoring of massive vehicle configuration event streams is achieved. It can intelligently adjust the detection sensitivity according to the business characteristics of different configurations, and ensure the final accuracy of anomaly judgment through manual review, thus providing effective protection for subsequent data cleaning and the reliability of prediction models.

[0151] In one possible implementation, step S205 may further include: based on the order requirements of each vehicle configuration, performing dynamic scheduling through a fixed rolling cycle and scheduling time domain to obtain scheduling results; if the scheduling results meet the production line constraints, using the scheduling results as a production scheduling plan, wherein the production line constraints include capacity constraints, material availability constraints, process route constraints, delivery date constraints, and personnel scheduling constraints; if the scheduling results do not meet the production line constraints, re-executing dynamic scheduling through a fixed rolling cycle and scheduling time domain until the scheduling results meet the production line constraints, or the number of recalculations of dynamic scheduling reaches a preset threshold.

[0152] The order requirements for each vehicle model configuration are input into the rolling time-domain optimization model, and combined with the production line resource ledger, material inventory information, process route library, order delivery date information, and personnel shift schedule for linkage calculation to generate scheduling results.

[0153] Next, the production line constraints are checked item by item in the scheduling results. If all the scheduling results meet the constraints, they are fixed as the production schedule plan for this adjustment cycle. If any constraint is not met, dynamic scheduling is executed again until a scheduling result that meets the constraints is obtained, or the number of times dynamic scheduling is recalculated reaches a preset threshold.

[0154] This approach enables order demand and production line resources to be matched within the same rolling time domain, and the scheduling results can be directly converted into an executable plan after constraint verification. Because it is refreshed every time according to a fixed rolling cycle and recalculated within a limited scheduling time domain, the production scheduling plan can synchronously reflect changes in orders, material availability, and shift changes, thereby improving the consistency between the scheduling and the actual execution status of the production line.

[0155] In one possible implementation, based on the order demand for each vehicle configuration, dynamic scheduling is performed through a rolling time-domain strategy to obtain the production schedule plan for each vehicle configuration in the current adjustment period. The dynamic scheduling method based on multiple channels further includes: executing the production schedule plan on the production line; collecting the execution status of the production schedule plan on the production line and calculating the deviation rate between the schedule plan and the actual execution; when the deviation rate exceeds a dynamic threshold, triggering dynamic scheduling recalculation and updating the order conversion rate prediction model.

[0156] First, the production schedule obtained from the rolling time domain is distributed to the production line. The production line organizes the execution of processes according to the planned sequence and synchronously transmits the execution status of each workstation. The collected execution status is time-aligned and mapped to the plan. Combining the planned quantity and actual completed quantity for the corresponding vehicle model configuration, the deviation rate between the scheduled plan and the actual execution is calculated.

[0157] If the deviation rate does not exceed the dynamic threshold, the current scheduling plan will continue to be executed.

[0158] When the deviation rate exceeds the dynamic threshold, the dynamic scheduling is recalculated based on the latest execution status, current order demand, and remaining scheduling time domain to obtain an updated production arrangement. Simultaneously, deviation samples, execution lag characteristics, and demand change characteristics are input into the order conversion rate prediction model to correct subsequent conversion predictions for intended orders and browsing behavior. Furthermore, after the order conversion rate prediction model is updated, it participates again in subsequent rolling time domain scheduling, ensuring a more consistent match between predicted orders and actual production line capacity.

[0159] By linking execution status feedback with deviation thresholds, production scheduling is no longer limited to the initial plan. Instead, it can trigger recalculation based on the actual execution status of the production line and synchronously correct the order conversion rate prediction model. This ensures that subsequent scheduling inputs are more closely aligned with changes in end-user demand and production line execution capabilities. Since the recalculation conditions and model update conditions are uniformly bound to the deviation rate exceeding the limit scenario, the system can promptly correct scheduling results before the execution deviation expands, and continuously improve the synchronization between multi-channel order demand forecasting and production scheduling.

[0160] Figure 4 This is a schematic diagram of the structure of a multi-channel-based dynamic scheduling device provided in an embodiment of this application, as shown below. Figure 4 As shown, the multi-channel-based dynamic scheduling device 40 provided in this embodiment includes:

[0161] The acquisition module 401 is used to acquire event streams from multiple channels within the directly operated terminals during this adjustment period. The types of event streams include formal orders, intention orders, and browsing behavior.

[0162] Processing module 402 is used to standardize and deduplicate the event stream to obtain the target event stream; based on intended orders and browsing behavior, it predicts the conversion rate according to a preset two-level confidence progression rule to obtain the predicted orders for each vehicle configuration in this adjustment period, wherein the probability weight of intended orders is higher than the probability weight of browsing behavior; it merges the predicted orders with the unproduced orders in the formal orders that have not yet been scheduled for production to obtain the order demand for each vehicle configuration in this adjustment period; based on the order demand for each vehicle configuration, it performs dynamic scheduling processing through a rolling time domain strategy to obtain the production scheduling plan for each vehicle configuration in this adjustment period.

[0163] In one possible implementation, the processing module 402 is specifically used to: map event streams from different channels in a unified manner based on a signature table mapping mechanism to obtain a standard event stream with consistent representation. The signature table mapping mechanism includes vehicle configuration mapping and user identifier mapping; align the standard event stream with timestamps, and perform deduplication processing on event streams with the same timestamp, configuration signature, and user identifier within the same deduplication window to obtain the target event stream.

[0164] In one possible implementation, the processing module 402 is specifically configured to: determine the first delay of the intended orders and the second delay of the browsing behavior in the current adjustment period based on the timestamps of the intended orders and browsing behavior in the event stream; if the first delay and the second delay do not meet the delay conditions, then the predicted orders of each vehicle configuration in the previous adjustment period are used as the predicted orders for the current adjustment period, wherein the delay conditions include that the first delay is less than the first delay threshold and the second delay is less than the second delay threshold; if the first delay and the second delay meet the delay conditions, then based on the intended orders and browsing behavior, the conversion rate is predicted according to the preset two-level confidence progression rule through the order conversion rate prediction model to obtain the predicted orders of each vehicle configuration in the current adjustment period.

[0165] In one possible implementation, the predicted order includes: intended demand and demand correction factor. The processing module 402 is further configured to: predict the conversion rate of intended orders based on the order conversion rate prediction model to obtain a first probability weight, and predict the conversion rate of browsing behavior to obtain a second probability weight lower than the first probability weight; weight the quantity of each vehicle configuration in the intended order based on the first probability weight to obtain the intended demand for each vehicle configuration; weight the quantity of each vehicle configuration in the browsing behavior based on the second probability weight to obtain the demand correction factor for each vehicle configuration; and use the sum of the intended demand and demand correction factor for each vehicle configuration as the predicted order for each vehicle configuration in this adjustment period.

[0166] In one possible implementation, after obtaining the event streams from multiple channels within the direct-operated terminal during the current adjustment period, the processing module 402 is further configured to: obtain the corresponding historical event streams within the anomaly detection window; sum the mean of the corresponding vehicle configurations within the historical event streams with the standard deviation of the preset weights to obtain the anomaly threshold for each vehicle configuration; if the number of vehicle configurations represented by the event stream is greater than or equal to the anomaly threshold, then the event stream data is determined to be abnormal, the event stream data is marked as abnormal, and manual review is triggered.

[0167] In one possible implementation, the processing module 402 is specifically used to: perform dynamic scheduling based on the order requirements of each vehicle model configuration through a fixed rolling cycle and scheduling time domain to obtain the scheduling result; if the scheduling result meets the production line constraints, the scheduling result is used as a production scheduling plan, wherein the production line constraints include capacity constraints, material availability constraints, process route constraints, delivery date constraints, and personnel scheduling constraints; if the scheduling result does not meet the production line constraints, the dynamic scheduling is re-executed through a fixed rolling cycle and scheduling time domain until the scheduling result meets the production line constraints, or the number of recalculations of the dynamic scheduling reaches a preset threshold.

[0168] In one possible implementation, based on the order demand for each vehicle configuration, dynamic scheduling is performed through a rolling time-domain strategy to obtain the production schedule plan for each vehicle configuration in the current adjustment period. Then, the processing module 402 is further used to: execute the production schedule plan on the production line; collect the execution status of the production schedule plan on the production line and calculate the deviation rate between the schedule plan and the actual execution; when the deviation rate exceeds the dynamic threshold, trigger dynamic scheduling recalculation and update the order conversion rate prediction model.

[0169] The multi-channel dynamic scheduling device provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0170] Figure 5 This is a schematic diagram of the structure of a multi-channel-based dynamic scheduling device provided in an embodiment of this application. Figure 5 As shown, the multi-channel-based dynamic scheduling device 50 provided in this embodiment includes at least one processor 501 and a memory 502. Optionally, the multi-channel-based dynamic scheduling device 50 further includes a communication component 503. The processor 501, memory 502, and communication component 503 are connected via a bus 504.

[0171] In a specific implementation, at least one processor 501 executes computer execution instructions stored in memory 502, causing at least one processor 501 to perform the above-described method.

[0172] The specific implementation process of processor 501 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0173] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0174] This application provides a dynamic scheduling production line, which is equipped with a multi-channel dynamic scheduling device as described in the above embodiments and / or possible implementations of the embodiments. The dynamic scheduling production line is used to generate a production schedule plan after the multi-channel dynamic scheduling device executes the multi-channel dynamic scheduling method as described in the above embodiments and / or various possible implementations of the embodiments, and to perform job allocation and cycle control for the assembly tasks of each vehicle model on the production line.

[0175] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0176] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0177] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0178] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0179] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random-Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0180] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an application-specific integrated circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0181] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0182] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0183] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0184] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0185] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0186] Finally, it should be noted that other embodiments of this application will readily conceive of by those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes may be made without departing from its scope.

Claims

1. A dynamic scheduling method based on multiple channels, characterized in that, include: The event streams from multiple channels within the directly operated terminals during this adjustment period are obtained. The types of event streams include formal orders, intended orders, and browsing behavior. The event stream is standardized and deduplicated to obtain the target event stream; Based on the intended orders and the browsing behavior, the conversion rate is predicted according to the preset two-level confidence progression rule to obtain the predicted orders for each model configuration in this adjustment period, wherein the probability weight of the intended orders is higher than the probability weight of the browsing behavior. The predicted orders are merged with the unproduced orders that have not yet been scheduled for production in the formal orders to obtain the order demand for each model configuration in this adjustment cycle; Based on the order requirements for each vehicle configuration, dynamic scheduling is performed using a rolling time-domain strategy to obtain the production schedule plan for each vehicle configuration in this adjustment period.

2. The dynamic scheduling method based on multiple channels according to claim 1, characterized in that, The process of standardizing and deduplicating the event stream to obtain the target event stream includes: Based on the signature table mapping mechanism, event streams from different channels are uniformly mapped to obtain a standard event stream with consistent representation. The signature table mapping mechanism includes vehicle configuration mapping and user identifier mapping. The standard event stream is timestamped, and event streams with identical timestamps, configuration signatures, and user identifiers are deduplicated within the same deduplication window to obtain the target event stream.

3. The dynamic scheduling method based on multiple channels according to claim 1, characterized in that, Based on the intended orders and browsing behavior, conversion rate prediction is performed according to a preset two-level confidence progression rule to obtain predicted orders for each vehicle configuration in this adjustment period, including: Based on the timestamps of intended orders and browsing behavior in the event stream, the first delay of the intended orders and the second delay of the browsing behavior in this adjustment period are determined; If the first delay and the second delay do not meet the delay conditions, then the predicted orders for each vehicle configuration in the previous adjustment period will be used as the predicted orders for this adjustment period. The delay conditions include that the first delay is less than the first delay threshold and the second delay is less than the second delay threshold. If the first delay and the second delay meet the delay conditions, then based on the intended order and the browsing behavior, the conversion rate is predicted by the order conversion rate prediction model according to the preset two-level confidence progression rule, so as to obtain the predicted orders for each model configuration in this adjustment period.

4. The dynamic scheduling method based on multiple channels according to claim 3, characterized in that, The predicted orders include: intended demand and demand correction factors. Based on the intended orders and browsing behavior, the conversion rate is predicted using an order conversion rate prediction model according to a preset two-level confidence progression rule to obtain the predicted orders for each vehicle configuration in this adjustment period, including: Based on the order conversion rate prediction model, the conversion rate of the intended order is predicted to obtain a first probability weight, and the conversion rate of the browsing behavior is predicted to obtain a second probability weight that is lower than the first probability weight. The quantity of each vehicle configuration in the intended order is weighted based on the first probability weight to obtain the intended demand for each vehicle configuration; The number of vehicle configurations within the browsing behavior is weighted based on the second probability weight to obtain the demand correction factor for each vehicle configuration. The sum of the intended demand and the demand correction factor for each vehicle configuration will be used as the predicted order for each vehicle configuration in this adjustment period.

5. The dynamic scheduling method based on multiple channels according to any one of claims 1 to 4, characterized in that, After obtaining the event streams from multiple channels within the directly operated terminals during this adjustment period, the dynamic scheduling method based on multiple channels further includes: Obtain the historical event stream corresponding to the anomaly detection window; The mean of the corresponding vehicle configuration within the historical event stream is summed with the standard deviation of the preset weight to obtain the abnormal threshold of each vehicle configuration. If the number of vehicle configurations represented by the event stream is greater than or equal to the abnormal threshold, the event stream data is determined to be abnormal, the event stream data is marked as abnormal, and manual review is triggered.

6. The dynamic scheduling method based on multiple channels according to any one of claims 1 to 4, characterized in that, The order demands based on each vehicle model configuration are processed dynamically using a rolling time-domain strategy to obtain the production schedule plan for each vehicle model configuration in this adjustment period, including: Based on the order requirements for each vehicle model configuration, dynamic scheduling is performed through a fixed rolling cycle and scheduling time domain to obtain the scheduling results; If the scheduling result meets the production line constraints, the scheduling result shall be used as the production scheduling plan, wherein the production line constraints include capacity constraints, material availability constraints, process route constraints, delivery date constraints and personnel scheduling constraints; If the scheduling result does not meet the production line constraints, the dynamic scheduling is re-executed with a fixed rolling cycle and scheduling time domain until the scheduling result meets the production line constraints, or the number of recalculations of the dynamic scheduling reaches a preset threshold.

7. The dynamic scheduling method based on multiple channels according to any one of claims 1 to 4, characterized in that, After the order demand based on each vehicle configuration is processed through a rolling time-domain strategy to obtain the production schedule plan for each vehicle configuration in the current adjustment period, the multi-channel-based dynamic scheduling method further includes: Execute the aforementioned production scheduling plan on the production line; Collect the execution status of the production line in executing the production schedule plan, and calculate the deviation rate between the schedule plan and the actual execution. When the deviation rate exceeds the dynamic threshold, dynamic scheduling recalculation is triggered, and the order conversion rate prediction model is updated.

8. A dynamic scheduling device based on multiple channels, characterized in that, include: The acquisition module is used to acquire event streams from multiple channels within the directly operated terminals during this adjustment period. The types of event streams include formal orders, intention orders, and browsing behavior. The processing module is used to standardize and deduplicate the event stream to obtain the target event stream; based on the intended orders and the browsing behavior, it predicts the conversion rate according to a preset two-level confidence progression rule to obtain the predicted orders for each vehicle configuration in this adjustment period, wherein the probability weight of the intended orders is higher than the probability weight of the browsing behavior; it merges the predicted orders with the unproduced orders in the formal orders to obtain the order demand for each vehicle configuration in this adjustment period; based on the order demand for each vehicle configuration, it performs dynamic scheduling processing through a rolling time domain strategy to obtain the production scheduling plan for each vehicle configuration in this adjustment period.

9. A dynamic scheduling device based on multiple channels, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the multi-channel-based dynamic scheduling method as described in any one of claims 1-7.

10. A dynamically scheduled production line, characterized in that, The dynamic scheduling production line is equipped with the multi-channel based dynamic scheduling device as described in claim 9, wherein: The dynamic scheduling production line is used to perform work allocation and cycle control on the assembly tasks of each vehicle model on the production line based on the production schedule plan generated by the dynamic scheduling equipment after executing the multi-channel dynamic scheduling method.

Images