Multi-source sensing sorting system auxiliary line-stack removing device cooperative control method and system

Abstract

The invention discloses a multi-source sensing sorting system auxiliary line-stack removing device cooperative control method and system, and relates to the field of logistics intelligent sorting, and the method comprises the steps: connecting a target auxiliary line-stack removing device which comprises a plurality of parcel transmission lines and a stack removing unit, obtaining the current transmission line speed and a unit time parcel quantity sequence, and obtaining a parcel quantity sequence; the current unit time parcel processing amount of the stack removing unit is obtained; obtaining a current entrance parcel backlog amount of the destacking unit, configuring a parcel backlog predictor, inputting the plurality of transmission line speeds, the plurality of unit time parcel amount sequences and the unit time parcel processing amount into the parcel backlog predictor, and obtaining an entrance parcel backlog prediction sequence; and when the maximum backlog amount is greater than or equal to a preset backlog threshold value, obtaining an optimized unit time parcel processing amount according to the plurality of transmission line speeds, the plurality of unit time parcel amount sequences and the entrance parcel backlog prediction sequence, and configuring a destacking unit. The problems that supply and demand of logistics sorting are not matched, response is not timely, and the sorting efficiency is low are solved.

Description

Technical Field

[0001] This invention relates to the field of intelligent sorting in logistics, specifically to a collaborative control method and system for auxiliary lines and de-stacking devices in sorting systems based on multi-source sensing. Background Technology

[0002] In modern logistics sorting systems, logistics facilities that use automated assembly line operations to classify and process goods consist of control devices, sorting devices, conveying devices, and sorting lanes. Barcode identification enables rapid sorting of goods.

[0003] In existing technologies, the traditional passive response method is used for adjustment, which fails to utilize historical data for demand analysis. This leads to a mismatch between supply and demand between auxiliary line conveyors and parcel flow, resulting in untimely response to sorting backlogs, low sorting efficiency, and equipment wear and energy waste.

[0004] Therefore, there is an urgent need to provide a multi-source sensing sorting system auxiliary line-destacking device collaborative control method and system to achieve intelligent and precise matching of feeding and destacking rhythms. Summary of the Invention

[0005] This application provides a multi-source sensing sorting system auxiliary line-de-stacking device coordinated control method and system, aiming to solve the problems of supply and demand mismatch between auxiliary line conveying and package flow, untimely response to sorting backlog, and low sorting efficiency in the prior art.

[0006] In view of the above problems, this application provides a method and system for coordinated control of auxiliary line and destacking device in a sorting system with multi-source sensing.

[0007] In a first aspect, this application provides a multi-source sensing collaborative control method for auxiliary lines and destacking devices in a sorting system, including: Connect the target auxiliary line - de-overlap device, the target auxiliary line - de-overlap device includes multiple package transmission lines and de-overlap units, obtain the current transmission line speed and unit time package quantity sequence of the multiple package transmission lines, and the current unit time package quantity processed by the de-overlap unit; Obtain the current inbound package backlog of the de-overlap unit, configure a package backlog predictor based on the current inbound package backlog, and input multiple transmission line velocities, multiple unit time package quantity sequences, and unit time package quantity processed into the package backlog predictor to obtain the inbound package backlog prediction sequence; When the maximum backlog amount in the inlet package backlog prediction sequence is greater than or equal to the preset backlog threshold, the optimized package processing amount per unit time is obtained based on the multiple transmission line velocities, multiple unit time package quantity sequences, and the inlet package backlog prediction sequence, and the de-overlap unit is configured.

[0008] Secondly, this application provides a multi-source sensing sorting system auxiliary line-de-stacking device cooperative control system, including: The data acquisition module is used to connect to the target auxiliary line-de-overlap device, which includes multiple package transmission lines and de-overlap units. The module acquires the current transmission line speed and package quantity sequence per unit time of the multiple package transmission lines, as well as the current package quantity processed per unit time of the de-overlap unit. An inlet package backlog prediction module is used to obtain the current inlet package backlog of the de-overlap unit, configure a package backlog predictor based on the current inlet package backlog, and input multiple transmission line velocities, multiple unit time package quantity sequences, and unit time package processing quantity into the package backlog predictor to obtain an inlet package backlog prediction sequence. The de-overlap unit configuration module is used to configure the de-overlap unit when the maximum backlog amount in the inlet package backlog prediction sequence is greater than or equal to a preset backlog threshold, based on the multiple transmission line velocities, multiple unit time package quantity sequences and the inlet package backlog prediction sequence, to obtain the optimized unit time package quantity and configure the de-overlap unit.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application presents a multi-source sensing collaborative control method and system for the auxiliary line and de-stacking device in a sorting system. First, by connecting the target auxiliary line and de-stacking device, the current transmission line speed and parcel volume sequence per unit time of the parcel transport line, as well as the current parcel volume processed per unit time by the de-stacking unit, are obtained. Through multi-source sensing and prediction models, risks are identified and intervened in advance before congestion occurs, significantly improving system stability. Second, an inlet parcel backlog prediction sequence is obtained by constructing an inlet parcel backlog configuration parcel backlog predictor. Accurate structured prediction of future workload provides high-quality input for subsequent intelligent decision-making. This achieves high-precision, highly adaptive intelligent prediction, solving the online adaptability problem of the model and improving its practicality. When the maximum backlog volume in the inlet parcel backlog prediction sequence is greater than or equal to a preset backlog threshold, the parcel volume processed per unit time is optimized, and the de-stacking unit is configured. By temporarily reducing the transmission line speed, the global stability of the sorting node is ensured, achieving intelligent and refined adjustment of control parameters and forming targeted control of the scheduling strategy. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1A flowchart illustrating the collaborative control method for the auxiliary line and de-stacking device in a multi-source sensing sorting system; Figure 2 This is a schematic diagram of the structure of the collaborative control system for the auxiliary line and the de-stacking device of the multi-source sensing sorting system.

[0012] The labels in the attached diagram are explained as follows: Data acquisition module 11; inbound package backlog prediction module 12; de-overlapping unit configuration module 13. Detailed Implementation

[0013] This application provides a multi-source sensing collaborative control method and system for auxiliary line and de-stacking device in a sorting system, which is used to address the problems of supply and demand mismatch between auxiliary line conveying and package flow, untimely response to sorting backlog, and low sorting efficiency in the prior art.

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0015] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0016] Example 1, as Figure 1 As shown, this application provides a multi-source sensing-based auxiliary line-de-stacking device cooperative control method for a sorting system, the method comprising: S10: Connect the target auxiliary line-de-overlap device, which includes multiple package transmission lines and a de-overlap unit. Obtain the current transmission line speed and package quantity sequence per unit time of the multiple package transmission lines, as well as the current package quantity processed per unit time of the de-overlap unit. In this embodiment, the target auxiliary line - the de-stacking device is the specific controlled object. The auxiliary line refers to the preparatory conveyor line that transports packages from the storage or receiving area to the main sorting machine. The de-stacking device is a specialized device used to separate stacked packages to meet the main sorting machine's requirements for package spacing, thereby ensuring the accuracy of scanning and sorting.

[0017] The parcel conveyor line is the specific implementation of the auxiliary line, referring to multiple parallel conveyor belts or tracks used for parallel transport of parcels; the de-stacking unit is the core functional module of the de-stacking device, which usually includes adjustable speed belts, push rods, and baffles, and changes the spacing between parcels through actions such as starting, stopping, accelerating, and decelerating; the current conveyor line speed refers to the actual operating speed of each parcel conveyor line at the moment of data acquisition, and is a basic parameter that determines the speed of parcel transport.

[0018] The parcel volume sequence per unit time is a data sequence arranged in chronological order, reflecting the flow waveform of parcel arrivals.

[0019] Specifically, a communication connection is first established with the target auxiliary line—the de-collapse device—for data exchange and control. Next, key status data is collected in real time, acquiring the current transmission line speed and parcel volume sequence per unit time for multiple parcel transmission lines, as well as the current parcel volume processed per unit time by the de-collapse unit.

[0020] Step S10 in the method provided in this application embodiment includes: Obtain the maximum number of packages processed per unit time by the de-overlap unit; When the de-aliasing unit processes the parcel volume at the maximum unit time, if the maximum backlog volume in the inlet parcel backlog prediction sequence is still greater than or equal to the preset backlog threshold, the multiple transmission line speeds are adjusted and optimized to obtain multiple optimized transmission line speeds, and the multiple parcel transmission lines are configured.

[0021] In this embodiment of the application, firstly, the maximum number of packages processed per unit time by the de-overlapping unit is obtained.

[0022] The maximum number of packages processed per unit time is the highest processing rate that the de-folding unit can achieve stably within the allowable range. It is usually a fixed value or the upper limit of a range, determined by the equipment's technical specifications or through on-site calibration tests.

[0023] Specifically, the maximum number of packages that the de-aliasing unit can process per unit time is read in advance from the device database or configuration file. This maximum processing capacity of the de-aliasing unit per unit time indicates its maximum processing capability. The upper limit of the strategy's effectiveness can be increased later by enhancing the capabilities of the de-aliasing unit itself.

[0024] For example, according to the technical manual, a certain model of de-collapse unit has a maximum sustainable processing capacity of 36 packages per minute, which means that the maximum number of packages processed per unit time is 36 per minute, provided that the separation success rate and the mechanical life of the equipment are guaranteed.

[0025] Secondly, when the de-overlap unit is running at the maximum unit time to process the parcel volume, if the maximum backlog volume in the inlet parcel backlog prediction sequence is still greater than or equal to the preset backlog threshold, multiple transmission line speeds are adjusted and optimized to obtain multiple optimized transmission line speeds and configure multiple parcel transmission lines.

[0026] Among them, the preset backlog threshold is a preset value obtained from historical data. It is the benchmark value for judging whether there is a backlog of packages in the de-stack unit. It can be adaptively configured by technicians according to the actual scenario. The maximum backlog amount is the largest predicted value in the predicted value of the inlet package backlog prediction sequence. It represents the congestion peak that may occur in the future observation window.

[0027] Specifically, the de-tipping unit operates at maximum speed. Simultaneously, based on the current sequence, it re-runs the package backlog predictor and, under extreme conditions, determines whether to adjust or optimize the conveyor line speed. If, under extreme operating conditions, the maximum backlog still exceeds the preset backlog threshold, the operating speed of each conveyor line needs to be reduced, meaning the overall upstream material arrival speed needs to be reduced. By delaying the arrival time of packages at the de-tipping unit, the package flow is distributed over time, preventing a large influx of packages at the same time from causing backlog.

[0028] Multiple transmission line speeds are obtained through adjustment and optimization, the package flow is controlled from the source, and the de-aliasing unit is reconfigured according to the optimized transmission line speed to improve its processing capacity.

[0029] For example, if the preset backlog threshold is 20 packages, and the maximum package processing capacity is 36 packages / minute, and the predicted peak demand is 40 packages / minute, then processing at 36 packages / minute would result in a predicted maximum backlog of 25 packages, still exceeding the threshold of 20. Therefore, transmission line speed adjustment is triggered.

[0030] In step S10 of the method provided in this application embodiment, obtaining the current transmission line speed and unit time parcel quantity sequence of multiple parcel transmission lines includes: The current transmission line speed of the multiple package transmission lines is obtained by speed sensors deployed on the multiple package transmission lines; Industrial cameras deployed at the entrances of the multiple parcel conveyor lines are used to collect real-time data on the entry time and number of parcels into each parcel conveyor line. Obtain the length parameters of each package transport line, and calculate the estimated transport time of the package from the entrance of each package transport line to the de-stacking unit, based on the current transport line speed of each package transport line. A prediction time window is set, and the estimated arrival time of each package to the de-stacking unit is determined based on the entry time and the estimated transmission duration. Packages whose expected arrival time falls within the predicted time window are selected, and statistics are performed according to a preset unit time interval to generate a sequence of packages per unit time corresponding to each package transmission line.

[0031] In this embodiment of the application, firstly, the current transmission line speed of the multiple package transmission lines is obtained by speed sensors deployed on multiple package transmission lines.

[0032] Among them, the speed sensor is a physical sensor installed on the drive motor or transmission shaft of the package conveyor line to directly or indirectly measure the linear speed or rotational speed of the conveyor belt or transmission components, and can convert the output signal into the real-time linear speed of the conveyor line.

[0033] Deployment involves physically installing and electrically connecting speed sensors to the drive systems of each package transmission line to ensure that they can stably and accurately detect the physical quantity representing the speed of the transmission line.

[0034] Specifically, each package conveyor line is equipped with a dedicated speed sensor for independent monitoring of its operational status. A data acquisition network allows for real-time monitoring of the current conveyor speed of each line. Each package conveyor line is a dynamic system, and its operating speed determines the speed of package transport.

[0035] For example, in an auxiliary line consisting of three transmission lines L1, L2, and L3, a rotary encoder is installed on the drive motor of each transmission line. The control system reads the number of encoder pulses once per second. Using preset calibration coefficients, the speed values ​​of each line are calculated and updated in real time. The encoder pulse frequencies correspond to the following speeds: L1: 1.2 m / s, L2: 1.0 m / s, L3: 0.8 m / s. The resulting speed vector is [1.2, 1.0, 0.8] (unit: m / s), representing the operating conditions of the three packaged transmission lines.

[0036] Secondly, industrial cameras deployed at the entrances of multiple parcel conveyor lines are used to collect real-time data on the entry time and number of parcels into each conveyor line.

[0037] Industrial cameras are high-reliability, high-frame-rate image sensors used in industrial environments. They are typically integrated with light sources and lenses to form a vision sensing unit, used to continuously photograph the entry area of ​​transmission lines in order to capture visual information of packages entering.

[0038] Specifically, an industrial camera is fixedly installed above or to the side of the starting end of each package conveyor line to ensure that its field of view can completely cover the area where the package transitions from the upstream equipment to this section of the conveyor line. The industrial camera continuously captures images at a fixed frequency, counts the number of packages identified in the frame, obtains the entry time and number of packages for each package conveyor line, and transmits the image data to the image processing unit in real time.

[0039] For example, an industrial camera is configured above the entrance of the L2 transport line, and images are processed in 20-second cycles. At time 10, a new carton is detected to have fully entered the L2 line area, and the event is recorded as: {Transport Line: L2, Time: 10:00:00, Quantity: 1}.

[0040] Next, obtain the length parameters of each package conveyor line, and combine them with the current conveyor line speed of each package conveyor line to calculate the estimated transmission time of the package from the entrance of each package conveyor line to the de-stacking unit.

[0041] The length parameter is the physical path length of a packaged transmission line from its entrance to its end, which intersects with the entrance of the de-overlap unit. It is a fixed geometric parameter, which is usually measured and determined after the system is installed and debugged and then entered into the system database.

[0042] The estimated transmission time refers to the time it takes for a package to travel from the entry point to the de-stacking unit entry point, calculated based on an ideal model of uniform motion.

[0043] Specifically, the pre-stored length parameters of each transmission line are first retrieved from the configuration library. Then, when each package enters a transmission line, the length parameter of that transmission line is divided by the real-time speed of that line at the current moment to obtain the estimated transmission time for the package to travel from the entrance of each package transmission line to the de-stacking unit.

[0044] The estimated transmission time is calculated as: Length parameter / Real-time speed of the current transmission line. The more accurate the real-time speed measurement of the current transmission line, the more accurate the estimated transmission time.

[0045] For example, the length parameter of L2 is known to be 15 meters. At the moment the package enters, the current transmission line velocity of L2 is read as 1.0 m / s. Therefore, the estimated transmission time of the package is 15 meters / 1.0 m / s = 15 seconds.

[0046] In addition, a prediction time window is set to determine the expected arrival time of each package to the de-stacking unit based on the entry time and the expected transmission duration.

[0047] The prediction time window is a time interval extending into the future from the current moment. It is the scope of the system's prediction and control decisions, and all predictions and optimizations are performed on the state within this window. The expected arrival time is calculated by a formula and is the best time estimate for the package to arrive at the de-stacking unit entry point.

[0048] Specifically, a prediction time window is first set. Then, for each package, its estimated arrival time is calculated. The estimated arrival time = arrival time + estimated transmission time.

[0049] For example, the prediction time window is set to the next 30 seconds. It is expected that the package will arrive at the de-stacking unit entrance at 10:00:00 + 15 seconds = 10:00:15, with an estimated arrival time of 10:00:15. Similarly, the estimated arrival time of each package is calculated.

[0050] Finally, packages whose expected arrival time falls within the predicted time window are selected, and statistics are performed according to preset unit time intervals to generate a sequence of package quantities per unit time for each package transmission line.

[0051] The preset unit time interval is the basic time unit that further subdivides the prediction time window, used to segment and count consecutive arrival events; the unit time package quantity sequence is the final output data structure, where each element in the list corresponds to a unit time interval in chronological order, and the value of the element is the total number of packages expected to arrive within that interval.

[0052] Specifically, the process involves filtering to retain only packages expected to arrive within a future window, ignoring packages with arrival times too far in the future. This window is then divided into units at intervals, similar to creating a histogram. Figure 1 Similarly, the number of packages expected to arrive in each unit time interval is counted. Finally, a sequence of package counts per unit time is generated for each transmission line, representing the number of packages expected to arrive in the corresponding time.

[0053] For example, for transport line L2, the prediction time window is set to the next 30 seconds, and the unit time interval is 5 seconds. After filtering, the arrival times of 20 packages fall within the window. Statistically, 3 packages arrive within [T, T+5s), 8 arrive within [T+5s, T+10s), 5 arrive within [T+10s, T+15s), and 4 and 0 arrive in subsequent time periods, respectively. The resulting sequence of packages per unit time generated by the transport line is: [3,8,5,4,0,0].

[0054] In this embodiment, the current transmission line speed of each package conveyor is first obtained using a speed sensor; then, the arrival time and number of packages are captured using an industrial camera deployed at the entrance and a vision algorithm. Combining the known length parameters of each conveyor line, the estimated transmission time for each package to reach the de-stacking unit is calculated, thereby determining the estimated arrival time. Finally, by setting a prediction time window and a unit time interval, packages that will arrive within the window are selected, and a sequence of packages per unit time corresponding to each transport line is statistically generated.

[0055] Data collection through multi-source data sensing ensures the accuracy of real-time status injection. Backlog prediction then forecasts the risk of package backlog at the maximum parcel processing capacity per unit time on each parcel transport line. By comparing the maximum backlog volume with a preset backlog threshold, tiered collaborative optimization is performed, enabling intelligent and flexible decision-making. This avoids ineffective adjustments while controlling flow at the source, ensuring the normal operation of the transport lines and providing the most direct input for accurate prediction and collaborative control downstream.

[0056] S20: Obtain the current inbound package backlog of the de-aliasing unit, configure a package backlog predictor based on the current inbound package backlog, and input multiple transmission line velocities, multiple unit time package quantity sequences, and unit time processed package quantity into the package backlog predictor to obtain an inbound package backlog prediction sequence; In this embodiment, the current inlet package backlog is the number of packages already piled up at the entrance of the de-stacking device, reflecting the immediate level of congestion; the package backlog predictor is a trained mathematical model or algorithm that can predict changes in the backlog over a future period based on input conditions. The configuration involves setting the current inlet package backlog as the initial state of the prediction model.

[0057] The inlet package backlog prediction sequence is a time-ordered sequence of the predictor output, predicting the number of packages backlogged at the inlet of the de-stacking device at the end of each future unit time period.

[0058] Specifically, the current backlog amount is first obtained. Then, a pre-built package backlog predictor is invoked, inputting all the obtained dynamic data along with the current backlog amount into the predictor. The dynamic process of packages continuously arriving and the de-stacking device continuously processing is simulated to generate an inbound package backlog prediction sequence.

[0059] In step S20 of the method provided in this application embodiment, the construction step of the package backlog predictor includes: Collect historical package transmission records of the target auxiliary line-de-folding device, and establish sample multi-transmission line velocity set, sample multi-unit time package quantity sequence set, sample unit time package quantity processed set, and sample backlog change sequence set based on the historical package transmission records; Using the sample multi-transmission line velocity set, the sample multi-unit-time package quantity sequence set, and the sample unit-time package quantity processed set as inputs, and the sample backlog change sequence set as labels, an initial backlog predictor is trained and generated. A data integration layer is established, which is used to perform a superposition operation on the output of the initial backlog predictor and the inbound package backlog. The initial backlog predictor and the data integration layer are integrated to generate the package backlog predictor.

[0060] In this embodiment of the application, historical package transmission records of the target auxiliary line-de-overlap device are first collected, and sample multi-transmission line velocity set, sample multi-unit time package quantity sequence set, sample unit time package quantity set, and sample backlog change sequence set are established based on the historical package transmission records.

[0061] Among them, the historical package transmission record is a set of raw data with timestamps that are automatically recorded by various sensors and control logs during the operation of the target auxiliary line-de-overlap device over a period of time. It records the system's input, status and output under different operating conditions; the acquisition is the process of extracting and exporting these historical records from the historical database through the data interface.

[0062] The sample multi-transmission line velocity set is a vector set consisting of the velocity values ​​of each package transmission line at the time corresponding to each sample in the historical record; the sample multi-unit time package quantity sequence set is a sequence of future package arrivals per unit time predicted from the historical record based on the past arrival time and the transmission line velocity at that time.

[0063] The sample unit time package processing set is the actual average processing rate of the de-overlap unit at the corresponding time in the historical record; the sample backlog change sequence represents the difference between the backlog amount at each predicted time point and the backlog amount at its corresponding starting time, and the true sequence of the package backlog number at the entrance of the de-overlap unit changing with time.

[0064] Specifically, historical package transport records are collected, and data is gathered from the historical package transport records of the target sorting auxiliary line and the de-stacking device, including the transport line speed, the number of packages per unit time, and the number of packages that the equipment can process per unit time.

[0065] The historical backlog data refers not to the absolute value of the historical backlog, but to the change in the backlog.

[0066] Specifically, for each historical data segment, the difference between the actual backlog at the end of the prediction period and the backlog at the beginning of the period is calculated, and the resulting difference sequence is used as the sample backlog change sequence set.

[0067] For example, for transport lines L1, L2, and L3, 10,000 30-second segments were extracted from the historical database. For the 5000th segment, its sample characteristics are: multiple transport line velocities of [1.2, 1.0, 0.8] m / s; a sequence of parcel volumes per unit time in the next 5-second intervals of {[2, 5, 3, 1, 0, 0], [3, 8, 5, 4, 0, 0], [0, 1, 2, 4, 3, 2]}; and a sample parcel processing volume of 10 parcels per 30 seconds. The corresponding sample backlog change sequence is [5, 4, 7, 9, 12, 8, 5], indicating that the backlog starts at 5 parcels, then slightly decreases, then increases, and finally decreases again. These 10,000 sample pairs constitute the database for training the model.

[0068] Secondly, an initial backlog predictor is trained using a set of sample multi-transmission linear velocity, a set of sample multi-unit-time package quantity sequences, and a set of sample unit-time package processing quantities as inputs, and a set of sample backlog change sequences as labels.

[0069] For example, an LSTM time series network is used to train an initial backlog predictor. The Long Short-Term Memory (LSTM) network is a type of recurrent neural network designed to address the long-term dependency problem inherent in general recurrent neural networks. It is suitable for processing and predicting important events with very long intervals and delays in time series.

[0070] The specific steps for constructing the initial backlog predictor are as follows: Model data: The collected sets of sample multi-transmission linear velocities, sample multi-unit-time package volume sequences, and sample unit-time package processing volume sequences are used as input features of the model. The calculated sample backlog change sequence set is used as the training label of the model. The obtained sample data is normalized to eliminate the influence of units. Subsequently, feature fusion and sequence construction are performed: for a sample, the three input features are aligned and concatenated in the time dimension.

[0071] The input to each sample is constructed as a two-dimensional tensor including the total number of time slices within the prediction time window and the feature dimension. Each sequence in the corresponding sample backlog change sequence set is processed into a two-dimensional tensor including the end time and the change in backlog amount relative to the start of the sequence at the end, and all samples are divided into training, validation, and test sets in a 7:2:1 ratio.

[0072] Model Structure: An LSTM network model is designed, consisting of one or more stacked LSTM layers. The input shape of the first LSTM layer must match the shape of the preprocessed sample input tensor. The LSTM layer captures long-term dependencies in the input sequence through its internal gating mechanisms: input gate, forget gate, and output gate. After the LSTM layers, one or more fully connected layers are added to map the high-level temporal features output by the LSTM layers to the final prediction target dimension. The last fully connected layer has only one neuron and uses a linear activation function to directly output the predicted value of the stacking change for each time slice. The goal of the entire model is to learn a mapping function from the input feature sequence to the stacking change sequence.

[0073] Model Training: The LSTM model is trained in a supervised manner using the prepared training set. The mean squared error loss function is employed to calculate the difference between the model's predicted backlog change sequence and the actual sample backlog change sequence, enabling the model to accurately predict the magnitude and trend of backlog changes. The Adam adaptive learning rate optimizer is used to efficiently update the model weights. The training data is divided into mini-batches and input into the model for forward propagation, loss calculation, backpropagation, and parameter updates, all performed in multiple iterations throughout the process.

[0074] After each training epoch, the model performance is evaluated using a validation set. The change in validation set loss is monitored, and training is terminated early when the validation set loss no longer decreases to prevent the model from overfitting the training data. The final trained model yields an initial backlog predictor used to predict the trend and increment of the backlog.

[0075] Next, a data integration layer is established, which is used to perform superposition operations on the output of the initial backlog predictor and the inbound package backlog.

[0076] Among them, superposition operation refers to specific mathematical operations, usually addition; integration is the connection and encapsulation of modules in the software architecture to form a fully functional prediction function.

[0077] Specifically, the predicted backlog sequence can be obtained by adding the backlog change sequence output by the initial backlog predictor to the current inbound package backlog as input element by element. The predicted backlog amount = backlog change amount + current inbound package backlog amount.

[0078] Finally, the initial backlog predictor and the data integration layer are integrated to generate a package backlog predictor.

[0079] The data integration layer is a non-learning computational layer module that does not contain parameters that need to be trained and is defined by a fixed algorithm. The package backlog predictor is the final generated complete predictor, which is a composite model that includes the initial backlog predictor and the data integration layer.

[0080] Specifically, the initial backlog predictor module is connected and encapsulated with the data integration layer module through integration to obtain the final package backlog predictor.

[0081] Using the final generated parcel backlog predictor, the current backlog amount, conveyor line speed, future incoming material sequence, and current processing volume are input. The initial backlog predictor is invoked to predict the backlog change sequence based on the conveyor line speed, future incoming material sequence, and current processing volume. Then, the backlog change sequence is superimposed onto the current backlog amount through a data integration layer, outputting a predicted sequence of absolute backlog amounts for each future time point, starting from the current backlog amount. The final predicted value = current incoming backlog amount + predicted change amount; that is, the parcel backlog predictor receives all input information and directly outputs the predicted absolute backlog value for the future.

[0082] For example, the backlog of packages is 8. The integrated package backlog predictor is invoked. The initial backlog predictor inside the predictor is based on the current [transmission line speed, future incoming material sequence, current / test processing volume], and outputs the change sequence: [+2,-1,+4,+3,-2,1]. Then, combined with the data integration layer, the following is obtained: the predicted value at the end of the first time period = 8+2=10; the predicted value at the end of the second time period = 8+2-1=9; the predicted value at the end of the third time period = 8+2-1+4=13; and so on. The final output absolute backlog sequence is: [10,9,13,16,14,13].

[0083] In this embodiment, historical package transport records are collected, and data is gathered from the historical package transport records of the target sorting auxiliary line and the de-stacking device. Then, an LSTM time series network is used to train an initial backlog predictor, and through integration, the initial predictor is adapted to any real-time backlog state, resulting in extremely fast response speed and improved interpretability and reliability of the entire prediction system.

[0084] S30: When the maximum backlog amount in the inlet package backlog prediction sequence is greater than or equal to the preset backlog threshold, the optimized package processing amount per unit time is obtained based on the multiple transmission line velocities, multiple unit time package quantity sequences and the inlet package backlog prediction sequence, and the de-overlap unit is configured.

[0085] In this embodiment, the maximum backlog is the largest predicted value in the inlet package backlog prediction sequence, representing the potential congestion peak within the future observation window. The preset backlog threshold is a package backlog predictor generated by connecting and encapsulating the pre-set safety warning backlog predictor module and the data integration layer module. The hybrid architecture cleverly solves the problem of difficult state initialization in online applications of purely data-driven models, greatly reducing the complexity of model training and data requirements, ensuring the correctness of the causal logic of the prediction, and enabling the model to instantly determine the maximum backlog. When the predicted maximum backlog exceeds this threshold, unacceptable congestion risks will arise, necessitating intervention. Optimizing the package processing capacity per unit time reduces the predicted maximum backlog to below the safety threshold, requiring a reset of the processing capacity target value for the de-overlap unit.

[0086] Specifically, the maximum backlog is compared with a preset threshold to determine whether intervention is needed. Based on multiple transmission line velocities, multiple parcel volume sequences per unit time, and the inlet parcel backlog prediction sequence, the parcel volume processed per unit time is optimized to find a higher parcel volume processed per unit time, so that the maximum backlog of the new prediction sequence can be lower than the threshold. The de-aliasing unit is then reconfigured to improve its processing capacity.

[0087] In step S30 of the method provided in this application embodiment, the optimized parcel processing volume per unit time is obtained based on the plurality of transmission line velocities, the plurality of parcel volume sequences per unit time, and the inlet parcel backlog prediction sequence, including: Obtain the range of packages processed per unit time by the de-overlap unit; Using the range of parcels processed per unit time as a constraint, the parcels processed per unit time are optimized based on the multiple transmission line velocities and multiple parcels processed per unit time sequences to obtain the optimized parcels processed per unit time.

[0088] In this embodiment of the application, the range of packages processed per unit time of the de-overlap unit is first obtained.

[0089] The parcel processing capacity range per unit time is a dynamic range parameter characterizing the variable processing capacity of the de-collapse unit. It is a closed interval, including the minimum processing rate at which the de-collapse unit can operate stably while meeting basic functional requirements, and the maximum processing rate at which it can operate continuously while ensuring the safety of the equipment's mechanical structure, preventing motor overload, controlling noise and vibration, and achieving a separation success rate that meets process requirements. The parcel processing capacity range per unit time defines the legal range of the de-collapse unit's processing speed and is an important dynamic performance indicator of the equipment. It is usually not equal to the entire range from zero to the mechanical limit.

[0090] Acquisition is the process by which the control system reads the range of parcels processed per unit time from a preset equipment parameter database, configuration file, or through online calibration routines.

[0091] Specifically, the parcel processing volume range per unit time of the de-stacking unit is loaded from the knowledge base and read from the device controller [P]. min ,P max ].

[0092] Among them, P min This likely corresponds to the throughput P that the de-stabilizing conveyor belt maintains at a minimum creeping speed to prevent slippage. max This corresponds to the processing capacity calculated from the maximum safe linear velocity at the rated frequency of the frequency converter.

[0093] For example, a certain type of de-packing unit, after reading, shows that the number of packages processed per unit time is in the range of [20, 60] pieces / minute.

[0094] Secondly, with the range of parcels processed per unit time as a constraint, the parcels processed per unit time are optimized based on multiple transmission line velocities and multiple parcels processed per unit time sequences to obtain the optimized parcels processed per unit time.

[0095] Among them, constraints are the restrictions on the range of values ​​of solution variables in optimization theory, namely the range of parcels processed per unit time; the current parcels processed per unit time is the processing rate that the folding unit is executing at the optimization decision moment, which is usually the starting point of the optimization algorithm.

[0096] Specifically, after receiving the optimized control instructions, the pre-built package backlog predictor is used to make predictions based on multiple transmission line velocities and multiple unit time package volume sequences. By evaluating the future congestion situation of different de-overlap units within the constraint interval, the optimized unit time package volume that meets the target is finally selected.

[0097] For example, if the predicted maximum backlog of 26 exceeds the preset backlog threshold of 20, an optimized control command is triggered. The current parcel processing rate per unit time is 30 parcels / minute, and the device range is [20, 60]. Within the sub-range of [30, 60], a new parcel processing rate per unit time is sought. This might be tested using a certain search strategy: testing parcel processing rate per unit time = 45, the predicted maximum backlog is 15 < 20, which meets the standard; testing parcel processing rate per unit time = 40, the predicted maximum backlog is 18 < 20, which meets the standard; testing parcel processing rate per unit time P = 37, the predicted maximum backlog is 21 > 20, which does not meet the standard. Ultimately, it is determined that parcel processing rate per unit time = 40 is the minimum feasible value to satisfy the predicted maximum backlog < 20, and 40 is within the range [20, 60]. Therefore, the optimized parcel processing rate per unit time is determined to be 40 parcels / minute.

[0098] In step S30 of the method provided in this application embodiment, the current parcel processing volume per unit time is optimized based on the multiple transmission line velocities and multiple parcel processing volume sequences per unit time, constrained by the parcel processing volume range per unit time, to obtain an optimized parcel processing volume per unit time, including: Set a processing volume search range, where the lower limit of the processing volume search range is the number of packages processed per unit time, and the upper limit is the maximum number of packages processed per unit time within the range of package processing volumes per unit time. Calculate the average of the upper and lower limits of the processing volume search range, and use it as the test processing volume; The multiple transmission line velocities, the multiple unit time package quantity sequences, and the test processing quantity are input into the package backlog predictor to obtain the first update entry package backlog sequence. If the maximum backlog in the first updated entry package backlog sequence is greater than or equal to the preset backlog threshold, then the lower limit of the processing volume search range is updated to the test processing volume; If the maximum backlog in the first updated entry package backlog sequence is less than the preset backlog threshold, then the upper limit of the processing volume search range is updated to the test processing volume; The process is iteratively executed until a minimum test processing amount is found that makes the maximum backlog amount in the first update entry package backlog sequence less than the preset backlog threshold. This minimum test processing amount is then used as the optimized package processing amount per unit time.

[0099] In this embodiment of the application, firstly, a processing volume search range is set. The lower limit of the processing volume search range is the current number of packages processed per unit time, and the upper limit is the maximum number of packages processed per unit time in the range of package processing volume per unit time.

[0100] The processing capacity search range is the numerical range of the processing capacity of the de-overlapping unit evaluated during the optimization process. It is a dynamically shrinking range that initially covers all possible candidate solutions that are better than the current state. The lower limit is the minimum boundary of the search range, and the upper limit is the maximum boundary of the search range.

[0101] Specifically, an optimization algorithm is first used to initialize a processing capacity search range, where the current processing capacity is the lower limit of the search range, and the device's maximum safe processing capacity is the upper limit. The initial processing capacity search range represents the set of all possible solutions that could improve the current situation. If the current processing capacity is P1 pieces / minute and the device's maximum capacity is P2 pieces / minute, then the initial search range is [P1, P2]. Any solution less than P1 is invalid or inverse and is not considered; any solution greater than P2 is unsafe and is not considered.

[0102] For example, the initial processing volume search range is set to [35, 70] by optimizing the algorithm.

[0103] Secondly, the average of the upper and lower limits of the processing volume search range is calculated as the test processing volume.

[0104] Among them, the test processing volume refers to the processing rate value of the candidate de-overlap unit selected by the algorithm for simulation in the current iteration round, which will be input into the prediction model to check whether it can meet the requirements for eliminating the risk of future backlog.

[0105] Specifically, the average of the upper and lower limits of the processing capacity search range is calculated using the formula: Test processing capacity = (upper limit + lower limit) / 2. In binary search, the average value represents the most representative and balanced test point at both ends of the current search range. By using the essential binary search method, nearly half of the invalid search intervals can be eliminated.

[0106] For example, the initial search range is [35, 70]. The algorithm calculates an average value of (35+70) / 2 = 52.5. The test throughput is set at 52.5 pieces / minute. The algorithm will first evaluate whether increasing the stacking unit speed to 52.5 pieces / minute can resolve the congestion problem.

[0107] Next, multiple transmission line velocities, multiple unit time package quantity sequences, and test processing quantity are input into the package backlog predictor to obtain the first update entry package backlog sequence.

[0108] The first updated entry package backlog sequence is a predicted sequence of the number of packages backlogged at the entry point over a future period of time, calculated by the package backlog predictor based on the transmission line velocity and the package volume sequence per unit time, assuming the de-overlap unit is running at the test processing volume.

[0109] Specifically, the current real-time data, namely multiple transmission line speeds, multiple parcel volume sequences per unit time, and generated hypothetical parameters, along with the test processing volume, are taken as a complete input scenario and fed into the trained parcel backlog predictor. The parcel backlog predictor is used as the engine to quickly simulate the dynamic behavior of parcel backlog under the current specific parameters and output the predicted sequence of future backlog.

[0110] For example, for a test throughput of 52.5, the package backlog predictor is invoked, with the current transmission line speeds of the three lines [1.2, 1.0, 0.8] m / s, the future incoming material sequence, and the throughput of 52.5 pieces / minute as input. The package backlog predictor calculates the first updated inbound package backlog sequence as [10, 14, 19, 23, 21, 18]. The predicted maximum backlog is 23.

[0111] In addition, if the maximum backlog in the first update entry package backlog sequence is greater than or equal to the preset backlog threshold, the lower limit of the update processing volume search range is the test processing volume.

[0112] The update process dynamically modifies the boundary of the processing volume search range based on the judgment result, thereby reducing the search interval to half of its original size.

[0113] Specifically, the algorithm extracts the maximum value from the predicted sequence and compares it with a preset threshold. If the maximum backlog is greater than or equal to the threshold, it indicates that the current test processing capacity is insufficient to eliminate future congestion risks. Therefore, all values ​​less than or equal to the current test processing capacity can be excluded. At this point, the algorithm raises the lower limit of the search range to the current test processing capacity, indicating that the next round of search will be conducted within a larger processing capacity range.

[0114] For example, for a test processing volume of 52.5, if the maximum backlog is 23, which is greater than or equal to 20, then the search range is updated to [52.5, 70].

[0115] Meanwhile, if the maximum backlog in the first update entry package backlog sequence is less than the preset backlog threshold, the upper limit of the update processing volume search range is the test processing volume.

[0116] Specifically, if the maximum backlog is less than the threshold, it indicates that the current test processing capacity is effective and can control future backlog within a safe range. However, there may be smaller, more energy-efficient processing capacities. Therefore, all values ​​greater than the current test processing capacity, while valid, are not optimal. In this case, the algorithm lowers the upper limit of the search range to the current test processing capacity. This means that the next round of search will continue to look for potentially smaller efficient solutions within a smaller processing capacity range that includes the currently effective solutions.

[0117] For example, if the maximum backlog is 18 and less than 20, then the search range is updated to [35, 52.5].

[0118] Finally, the process is iterated until the minimum test processing amount is found that makes the maximum backlog amount in the first update entry package backlog sequence less than the preset backlog threshold. The minimum test processing amount is then used as the optimized package processing amount per unit time.

[0119] Among them, iterative execution refers to the process of repeatedly executing the calculation of test processing volume, obtaining the first update entry package backlog sequence, and updating the search range; minimum test processing volume refers to the running speed with the smallest value among all candidate processing volumes that can make the predicted maximum backlog less than the threshold.

[0120] Specifically, the loop iterates through the calculation of test throughput, obtains the first update entry package backlog sequence, and updates the search range. As the iteration progresses, the search interval continuously narrows. Finally, the loop stops when the difference between the upper and lower limits is less than a certain precision threshold.

[0121] For example, after several iterations, the search range is narrowed down to [45, 46], with a precision requirement of 0.5. At this point, 46 is a known valid processing volume, while 45 is invalid or unverified. The algorithm terminates, determining 46 packages / minute as the minimum test processing volume, and ultimately outputting it as the optimized package processing volume per unit time, instructing the de-folding unit to adjust the speed to 46 packages / minute.

[0122] In step S30 of the method provided in this application embodiment, the plurality of transmission line speeds are adjusted and optimized to obtain a plurality of optimized transmission line speeds, including: Obtain the transmission speed adjustment range for each package transmission line; Set a speed adjustment step size, and uniformly reduce the current transmission line speed of the multiple package transmission lines by the speed adjustment step size to obtain the adjusted transmission line speed of each package transmission line. Based on the adjustment of the transmission line speed of each package transmission line, a second updated inlet package backlog sequence is obtained; Determine whether the maximum backlog amount in the second update entry package backlog sequence is less than the preset backlog threshold; If the maximum backlog amount in the second update entry package backlog sequence is still greater than or equal to the preset backlog threshold, and the adjusted transmission line speed does not reach the lower limit of the transmission speed adjustment range, then the transmission line speed will continue to be reduced uniformly according to the speed adjustment step size. The process is iterated until the maximum backlog in the second update entry package backlog sequence is less than the preset backlog threshold, at which point the corresponding adjusted transmission line speed is used as the plurality of optimized transmission line speeds.

[0123] In this embodiment of the application, firstly, the transmission speed adjustment range of each package transmission line is obtained.

[0124] The transmission speed adjustment range refers to the safe and stable operating speed range allowed by the drive system of each auxiliary transmission line. It is usually represented as a closed range and is an inherent physical and electrical characteristic of the equipment. It is usually specified by the equipment manual or determined through on-site commissioning.

[0125] Specifically, when deciding to activate the transmission line speed control strategy, the first step is to define the transmission speed adjustment range for each controlled line. The predefined maximum and minimum transmission speeds for each transmission line are read in real time. This ensures that even under extreme control conditions, each transmission line can maintain basic transmission functionality, preventing secondary failures.

[0126] For example, for transport lines L1, L2, and L3, the control system obtains the speed adjustment range from the configuration file: L1: [0.5, 2.0] m / s, L2: [0.5, 1.8] m / s, L3: [0.6, 2.2] m / s. The current operating speeds are 1.5 m / s, 1.6 m / s, and 1.8 m / s, respectively. The obtained transmission speed adjustment ranges [0.5, 2.0], [0.5, 1.8], and [0.6, 2.2] serve as the standard for all subsequent speed reduction operations.

[0127] Secondly, a speed adjustment step size is set, and the current transmission line speed of multiple package transmission lines is uniformly reduced by the speed adjustment step size to obtain the adjusted transmission line speed of each package transmission line.

[0128] The speed adjustment step size is a pre-set positive constant representing the magnitude of a single speed adjustment, which determines the precision of the control. An excessively large step size may result in coarse adjustments and overshoot; an excessively small step size may result in a slow adjustment process, making it difficult to respond promptly to emergency congestion.

[0129] Uniform reduction refers to subtracting the calculated speed adjustment step size from the current speed of all package transmission lines involved in the control; adjusting transmission line speed refers to the new speed value reached by each transmission line after the uniform reduction operation, which is a temporary set of candidate speed settings for testing.

[0130] Specifically, based on the adaptive algorithm, a suitable speed adjustment step size is set. Then, for each transmission line, the following calculation is performed: Adjusted transmission line speed = Current transmission line speed - Speed ​​adjustment step size. All transmission lines requiring adjustment are adjusted synchronously, resulting in a uniform reduction and thus the adjusted transmission line speed for each packaged transmission line.

[0131] For example, the speed adjustment step is set to 0.15 m / s. The current speeds of the three transmission lines are uniformly subtracted: L1 new speed = 1.5 - 0.15 = 1.35 m / s; L2 new speed = 1.6 - 0.15 = 1.45 m / s; L3 new speed = 1.8 - 0.15 = 1.65 m / s. Thus, the set of adjustable transmission line speeds for this round of attempts is obtained: [1.35, 1.45, 1.65] m / s.

[0132] Next, based on adjusting the transmission line speed of each package transmission line, the second updated entry package backlog sequence is obtained.

[0133] The second updated inlet package backlog sequence is the predictor's output under the current transmission line speed adjustment. It is a predicted sequence of future inlet backlog changes, assuming the de-scrambling unit operates at maximum processing capacity and all package transmission lines operate at the adjusted transmission line speed calculated in this round.

[0134] Specifically, the calculated transmission line speeds for each line, the maximum number of packages processed per unit time by the de-overlapping unit, and the sequence of packages per unit time recalculated or adjusted based on the new speeds are used as complete input parameters to invoke the package backlog predictor. Based on the new parameters, the package backlog predictor simulates system operation and outputs the future second-updated inlet package backlog sequence.

[0135] For example, the adjusted transmission speed [1.35, 1.45, 1.65] m / s, the maximum processing capacity of the de-overlap unit, and the sequence of packages per unit time are input into the package backlog predictor. The package backlog predictor calculates and obtains the second updated entry package backlog sequence as [22, 25, 28, 26, 23, 20]. This sequence shows that even after the speed reduction, the predicted maximum backlog is still 28.

[0136] Additionally, it is determined whether the maximum backlog amount in the second update entry package backlog sequence is less than a preset backlog threshold.

[0137] Specifically, the maximum backlog in the second update entry package backlog sequence is compared with a preset backlog threshold.

[0138] If the maximum backlog in the second update entry package backlog sequence is still greater than or equal to the preset backlog threshold, and the adjusted transmission line speed does not reach the lower limit of the transmission speed adjustment range, then the transmission line speed will continue to be reduced uniformly according to the speed adjustment step size.

[0139] Specifically, if the maximum backlog is less than the preset backlog threshold, it indicates that the current adjustment of the transmission line speed is effective and can solve the problem. The loop can terminate; if the maximum backlog is greater than or equal to the preset backlog threshold and the adjusted transmission line speed is lower than the lower limit of the transmission speed adjustment range, it indicates that the current speed reduction is not enough and the loop needs to continue. In this case, the loop continues, and the transmission line speed is reduced uniformly according to the speed adjustment step size.

[0140] For example, the preset threshold is 20, and the maximum backlog predicted in the previous round was 28, which is greater than or equal to 20. The current adjustment speed is [1.35, 1.45, 1.65] m / s, all of which are higher than their respective lower limits [0.5, 0.5, 0.6] m / s. Therefore, the condition for further reduction is met. Based on [1.35, 1.45, 1.65], a further uniform reduction of 0.15 m / s is obtained, resulting in the speed to be tested in the next round: [1.20, 1.30, 1.50] m / s.

[0141] Finally, the process is iterated until the maximum backlog in the second update entry package backlog sequence is less than the preset backlog threshold, at which point the corresponding adjusted transmission line speed is used as multiple optimized transmission line speeds.

[0142] Specifically, the loop will continue iterating, with the speed decreasing by one step in each iteration, and then the effect will be evaluated. The loop will stop when the predicted maximum backlog is less than the threshold after a certain round of iterations, or when the speed has dropped to the lower limit but the success condition is not met before the start of a certain round of iterations.

[0143] For example, after several iterations, when the speed drops to [1.05, 1.15, 1.35] m / s, the predicted maximum value of the second update entry package backlog sequence is 18, which is less than the threshold of 20. The iteration stops. [1.05, 1.15, 1.35] m / s is used as the optimized transmission line speed, and instructions are immediately issued to set the speeds of the three transmission lines to 1.05, 1.15, and 1.35 m / s respectively, entering a stable operating state that matches the processing capacity.

[0144] In this embodiment, the range of parcels processed per unit time is first incorporated as a hard constraint into the optimization decision-making process to ensure the security of control commands and achieve refined energy-saving control. Secondly, the optimization search is combined with an intelligent prediction model to avoid trial and error on real equipment, ensuring the speed of solution evaluation and the scientific nature of optimization decisions, thus achieving refined energy-saving control. Finally, a fair and easily implementable multi-line collaborative strategy is provided by adopting a fixed small-step iterative speed reduction method.

[0145] The embodiments of this application, through the above specific implementation methods, achieve the following technical effects: In this embodiment, the current transmission line speed of each package conveyor is first obtained using a speed sensor; then, the arrival time and number of packages are captured using an industrial camera deployed at the entrance and a vision algorithm. Combining the known length parameters of each conveyor line, the estimated transmission time for each package to reach the de-stacking unit is calculated, thereby determining the estimated arrival time. Finally, by setting a prediction time window and a unit time interval, packages that will arrive within the window are selected, and a sequence of packages per unit time corresponding to each transport line is statistically generated.

[0146] Data collection through multi-source data sensing ensures the accuracy of real-time status injection. Backlog prediction then forecasts the risk of package backlog at the maximum parcel processing capacity per unit time on each parcel transport line. By comparing the maximum backlog volume with preset backlog thresholds, tiered collaborative optimization is performed, enabling intelligent and flexible decision-making. This avoids ineffective adjustments while controlling flow at the source, ensuring the normal operation of the transport lines and providing the most direct input for accurate prediction and collaborative control downstream.

[0147] Secondly, historical package transport records are collected, and data is gathered from the historical package transport records of the target sorting auxiliary line and the de-stacking device. Then, an LSTM time series network is used to train the initial backlog predictor, and through integration, the initial predictor is adapted to any real-time backlog state, with extremely fast response speed, which improves the interpretability and reliability of the entire prediction system.

[0148] Ultimately, the range of parcels processed per unit time is incorporated as a hard constraint into the optimization decision-making process to ensure the safety of control commands and achieve refined energy-saving regulation. Secondly, the optimization search is combined with intelligent prediction models to avoid trial and error on real equipment, ensuring the speed of solution evaluation and the scientific nature of optimization decisions, thus achieving refined energy-saving control. Finally, a fair and easily implementable multi-line collaborative strategy is provided by adopting a fixed small-step iterative speed reduction method.

[0149] Example 2, as Figure 2 As shown, this application provides a multi-source sensing sorting system auxiliary line-de-stacking device cooperative control system, the system comprising: The data acquisition module 11 is used to connect to the target auxiliary line-de-overlap device, which includes multiple package transmission lines and a de-overlap unit. The module acquires the current transmission line speed and the package quantity per unit time sequence of the multiple package transmission lines, as well as the current package quantity processed per unit time of the de-overlap unit. The inlet package backlog prediction module 12 is used to obtain the current inlet package backlog of the de-overlap unit, configure a package backlog predictor based on the current inlet package backlog, and input multiple transmission line velocities, multiple unit time package quantity sequences and unit time package quantity processed into the package backlog predictor to obtain the inlet package backlog prediction sequence. The de-overlap unit configuration module 13 is used to configure the de-overlap unit when the maximum backlog amount in the inlet package backlog prediction sequence is greater than or equal to a preset backlog threshold, based on the multiple transmission line velocities, multiple unit time package quantity sequences and the inlet package backlog prediction sequence, to obtain the optimized unit time package quantity and configure the de-overlap unit.

[0150] In one embodiment, the data acquisition module 11 is configured to: Obtain the maximum number of packages processed per unit time by the de-overlap unit; When the de-aliasing unit processes the parcel volume at the maximum unit time, if the maximum backlog volume in the inlet parcel backlog prediction sequence is still greater than or equal to the preset backlog threshold, the multiple transmission line speeds are adjusted and optimized to obtain multiple optimized transmission line speeds, and the multiple parcel transmission lines are configured.

[0151] This includes obtaining the current transmission line speed and the sequence of packages per unit time for multiple package transmission lines, including: The current transmission line speed of the multiple package transmission lines is obtained by speed sensors deployed on the multiple package transmission lines; Industrial cameras deployed at the entrances of the multiple parcel conveyor lines are used to collect real-time data on the entry time and number of parcels into each parcel conveyor line. Obtain the length parameters of each package transport line, and calculate the estimated transport time of the package from the entrance of each package transport line to the de-stacking unit, based on the current transport line speed of each package transport line. A prediction time window is set, and the estimated arrival time of each package to the de-stacking unit is determined based on the entry time and the estimated transmission duration. Packages whose expected arrival time falls within the predicted time window are selected, and statistics are performed according to a preset unit time interval to generate a sequence of packages per unit time corresponding to each package transmission line.

[0152] In one embodiment, the inlet package backlog prediction module 12 is used for: Collect historical package transmission records of the target auxiliary line-de-folding device, and establish sample multi-transmission line velocity set, sample multi-unit time package quantity sequence set, sample unit time package quantity processed set, and sample backlog change sequence set based on the historical package transmission records; Using the sample multi-transmission line velocity set, the sample multi-unit-time package quantity sequence set, and the sample unit-time package quantity processed set as inputs, and the sample backlog change sequence set as labels, an initial backlog predictor is trained and generated. A data integration layer is established, which is used to perform a superposition operation on the output of the initial backlog predictor and the inbound package backlog. The initial backlog predictor and the data integration layer are integrated to generate the package backlog predictor.

[0153] In one embodiment, the de-overlap unit configuration module 13 is configured to: Obtain the range of packages processed per unit time by the de-overlap unit; Using the range of parcels processed per unit time as a constraint, the parcels processed per unit time are optimized based on the multiple transmission line velocities and multiple parcels processed per unit time sequences to obtain the optimized parcels processed per unit time.

[0154] Specifically, taking the range of parcels processed per unit time as a constraint, and based on the multiple transmission line velocities and multiple parcel quantity sequences per unit time, the current parcel processing volume per unit time is optimized to obtain the optimized parcel processing volume per unit time, including: Set a processing volume search range, where the lower limit of the processing volume search range is the number of packages processed per unit time, and the upper limit is the maximum number of packages processed per unit time within the range of package processing volumes per unit time. Calculate the average of the upper and lower limits of the processing volume search range, and use it as the test processing volume; The multiple transmission line velocities, the multiple unit time package quantity sequences, and the test processing quantity are input into the package backlog predictor to obtain the first update entry package backlog sequence. If the maximum backlog in the first updated entry package backlog sequence is greater than or equal to the preset backlog threshold, then the lower limit of the processing volume search range is updated to the test processing volume; If the maximum backlog in the first updated entry package backlog sequence is less than the preset backlog threshold, then the upper limit of the processing volume search range is updated to the test processing volume; The process is iteratively executed until a minimum test processing amount is found that makes the maximum backlog amount in the first update entry package backlog sequence less than the preset backlog threshold. This minimum test processing amount is then used as the optimized package processing amount per unit time.

[0155] Among them, the multiple transmission line speeds are adjusted and optimized to obtain multiple optimized transmission line speeds, including: Obtain the transmission speed adjustment range for each package transmission line; Set a speed adjustment step size, and uniformly reduce the current transmission line speed of the multiple package transmission lines by the speed adjustment step size to obtain the adjusted transmission line speed of each package transmission line. Based on the adjustment of the transmission line speed of each package transmission line, a second updated inlet package backlog sequence is obtained; Determine whether the maximum backlog amount in the second update entry package backlog sequence is less than the preset backlog threshold; If the maximum backlog amount in the second update entry package backlog sequence is still greater than or equal to the preset backlog threshold, and the adjusted transmission line speed does not reach the lower limit of the transmission speed adjustment range, then the transmission line speed will continue to be reduced uniformly according to the speed adjustment step size. The process is iterated until the maximum backlog in the second update entry package backlog sequence is less than the preset backlog threshold, at which point the corresponding adjusted transmission line speed is used as the plurality of optimized transmission line speeds.

[0156] The embodiments of this application, through the above specific implementation methods, achieve the following technical effects: In this embodiment, firstly, the current transmission line speed of each package conveyor line is acquired using a speed sensor via the data acquisition module 11; secondly, the arrival time and quantity of packages are captured using an industrial camera deployed at the entrance and a vision algorithm. Combining the known length parameters of each conveyor line, the estimated transmission time for each package to reach the de-stacking unit is calculated, thereby determining the estimated arrival time. Finally, by setting a prediction time window and a unit time interval, packages arriving within the window are selected, and a sequence of packages per unit time corresponding to each transport line is statistically generated.

[0157] Data collection through multi-source data sensing ensures the accuracy of real-time status injection. Backlog prediction then forecasts the risk of package backlog at the maximum parcel processing capacity per unit time on each parcel transport line. By comparing the maximum backlog volume with a preset backlog threshold, tiered collaborative optimization is performed, enabling intelligent and flexible decision-making. This avoids ineffective adjustments while controlling flow at the source, ensuring the normal operation of the transport lines and providing the most direct input for accurate prediction and collaborative control downstream.

[0158] Secondly, the inlet package backlog prediction module 12 collects historical package transmission records and gathers data from the historical package transmission records of the target sorting auxiliary line and the de-stacking device. Then, an LSTM time series network is used to train the initial backlog predictor, and the initial predictor is adapted to any real-time backlog state through integration. The response speed is extremely fast, which improves the interpretability and reliability of the entire prediction system.

[0159] Finally, through the de-overlap unit configuration module 13, the range of parcels processed per unit time is incorporated as a hard constraint into the optimization decision-making process to ensure the safety of control commands and achieve refined energy-saving control. Secondly, the optimization search is combined with an intelligent prediction model to avoid trial and error on real equipment, ensuring the speed of solution evaluation and the scientific nature of optimization decisions, thus achieving refined energy-saving control. Finally, by adopting a fixed small-step iterative speed reduction method, a fair and easily implemented multi-line collaborative strategy is provided.

[0160] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0161] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A collaborative control method for auxiliary line and destacking device in a multi-source sensing sorting system, characterized in that, The method includes: Connect the target auxiliary line - de-overlap device, the target auxiliary line - de-overlap device includes multiple package transmission lines and de-overlap units, obtain the current transmission line speed and unit time package quantity sequence of the multiple package transmission lines, and the current unit time package quantity processed by the de-overlap unit; Obtain the current inbound package backlog of the de-overlap unit, configure a package backlog predictor based on the current inbound package backlog, and input multiple transmission line velocities, multiple unit time package quantity sequences, and unit time package quantity processed into the package backlog predictor to obtain the inbound package backlog prediction sequence; When the maximum backlog amount in the inlet package backlog prediction sequence is greater than or equal to the preset backlog threshold, the optimized package processing amount per unit time is obtained based on the multiple transmission line velocities, multiple unit time package quantity sequences, and the inlet package backlog prediction sequence, and the de-overlap unit is configured.

2. The method according to claim 1, characterized in that, The method further includes: Obtain the maximum number of packages processed per unit time by the de-overlap unit; When the de-aliasing unit processes the parcel volume at the maximum unit time, if the maximum backlog volume in the inlet parcel backlog prediction sequence is still greater than or equal to the preset backlog threshold, the multiple transmission line speeds are adjusted and optimized to obtain multiple optimized transmission line speeds, and the multiple parcel transmission lines are configured.

3. The method according to claim 1, characterized in that, Obtain the current transmission line speed and parcel quantity per unit time sequence for multiple parcel transmission lines, including: The current transmission line speed of the multiple package transmission lines is obtained by speed sensors deployed on the multiple package transmission lines; Industrial cameras deployed at the entrances of the multiple parcel conveyor lines are used to collect real-time data on the entry time and number of parcels into each parcel conveyor line. Obtain the length parameters of each package transport line, and calculate the estimated transport time of the package from the entrance of each package transport line to the de-stacking unit, based on the current transport line speed of each package transport line. A prediction time window is set, and the estimated arrival time of each package to the de-stacking unit is determined based on the entry time and the estimated transmission duration. Packages whose expected arrival time falls within the predicted time window are selected, and statistics are performed according to a preset unit time interval to generate a sequence of packages per unit time corresponding to each package transmission line.

4. The method according to claim 1, characterized in that, The steps for constructing the package backlog predictor include: Collect historical package transmission records of the target auxiliary line-de-folding device, and establish sample multi-transmission line velocity set, sample multi-unit time package quantity sequence set, sample unit time package quantity processed set, and sample backlog change sequence set based on the historical package transmission records; Using the sample multi-transmission line velocity set, the sample multi-unit-time package quantity sequence set, and the sample unit-time package quantity processed set as inputs, and the sample backlog change sequence set as labels, an initial backlog predictor is trained and generated. A data integration layer is established, which is used to perform a superposition operation on the output of the initial backlog predictor and the inbound package backlog. The initial backlog predictor and the data integration layer are integrated to generate the package backlog predictor.

5. The method according to claim 1, characterized in that, Based on the multiple transmission line velocities, multiple parcel volume sequences per unit time, and the inbound parcel backlog prediction sequence, an optimized parcel volume per unit time is obtained, including: Obtain the range of packages processed per unit time by the de-overlap unit; Using the range of parcels processed per unit time as a constraint, the parcels processed per unit time are optimized based on the multiple transmission line velocities and multiple parcels processed per unit time sequences to obtain the optimized parcels processed per unit time.

6. The method according to claim 5, characterized in that, Constrained by the range of parcels processed per unit time, the current parcel processing volume per unit time is optimized based on the multiple transmission line velocities and multiple parcel processing volume sequences per unit time to obtain an optimized parcel processing volume per unit time, including: Set a processing volume search range, where the lower limit of the processing volume search range is the number of packages processed per unit time, and the upper limit is the maximum number of packages processed per unit time within the range of package processing volumes per unit time. Calculate the average of the upper and lower limits of the processing volume search range, and use it as the test processing volume; The multiple transmission line velocities, the multiple unit time package quantity sequences, and the test processing quantity are input into the package backlog predictor to obtain the first update entry package backlog sequence. If the maximum backlog in the first updated entry package backlog sequence is greater than or equal to the preset backlog threshold, then the lower limit of the processing volume search range is updated to the test processing volume; If the maximum backlog in the first updated entry package backlog sequence is less than the preset backlog threshold, then the upper limit of the processing volume search range is updated to the test processing volume; The process is iteratively executed until a minimum test processing volume is found that makes the maximum backlog in the first update entry package backlog sequence less than the preset backlog threshold. This minimum test processing volume is then used as the optimized package processing volume per unit time.

7. The method according to claim 2, characterized in that, The multiple transmission line speeds are adjusted and optimized to obtain multiple optimized transmission line speeds, including: Obtain the transmission speed adjustment range for each package transmission line; Set a speed adjustment step size, and uniformly reduce the current transmission line speed of the multiple package transmission lines by the speed adjustment step size to obtain the adjusted transmission line speed of each package transmission line. Based on the adjustment of the transmission line speed of each package transmission line, a second updated inlet package backlog sequence is obtained; Determine whether the maximum backlog amount in the second update entry package backlog sequence is less than the preset backlog threshold; If the maximum backlog amount in the second update entry package backlog sequence is still greater than or equal to the preset backlog threshold, and the adjusted transmission line speed does not reach the lower limit of the transmission speed adjustment range, then the transmission line speed will continue to be reduced uniformly according to the speed adjustment step size. The process is iterated until the maximum backlog in the second update entry package backlog sequence is less than the preset backlog threshold, at which point the corresponding adjusted transmission line speed is used as the plurality of optimized transmission line speeds.

8. A multi-source sensing sorting system auxiliary line-de-stacking device collaborative control system, characterized in that, The system for implementing the method according to any one of claims 1-7 comprises: The data acquisition module is used to connect to the target auxiliary line-de-overlap device, which includes multiple package transmission lines and de-overlap units. The module acquires the current transmission line speed and package quantity sequence per unit time of the multiple package transmission lines, as well as the current package quantity processed per unit time of the de-overlap unit. An inlet package backlog prediction module is used to obtain the current inlet package backlog of the de-overlap unit, configure a package backlog predictor based on the current inlet package backlog, and input multiple transmission line velocities, multiple unit time package quantity sequences, and unit time package processing quantity into the package backlog predictor to obtain an inlet package backlog prediction sequence. The de-overlap unit configuration module is used to configure the de-overlap unit when the maximum backlog amount in the inlet package backlog prediction sequence is greater than or equal to a preset backlog threshold, based on the multiple transmission line velocities, multiple unit time package quantity sequences and the inlet package backlog prediction sequence, to obtain the optimized unit time package quantity and configure the de-overlap unit.

Images