A multi-hardware cooperative intelligent sorting production line control method and system

Abstract

The application relates to the technical field of intelligent logistics sorting, and discloses a multi-hardware cooperative intelligent sorting production line control method and system. The method comprises the following steps: automatically identifying and registering multiple hardware devices such as a code reading camera, a dynamic scale and a sorting machine; establishing an inter-device data exchange network based on asynchronous communication, realizing the standardized access of heterogeneous devices and event-driven communication; constructing an article flow dynamic model and embedding real-time state updating logic, continuously tracking the whole-process state of articles through periodic data acquisition; deploying a sorting rule engine, combining the model real-time state and rule output, generating control commands through a sorting instruction algorithm and distributing the control commands to corresponding devices to drive sorting. The application improves the device integration flexibility and system response speed through a unified asynchronous communication network, realizes intelligent sorting decision based on a real-time situation through global dynamic modeling, and improves the throughput efficiency and self-adaptive capacity of the production line.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent logistics sorting, in particular to a multi-hardware cooperative intelligent sorting production line control method and system. BACKGROUND

[0002] In a logistics sorting center, an automated production line is usually composed of multiple heterogeneous hardware devices such as code reading cameras, dynamic scales, sorting machines, etc. In the prior art, different manufacturers' devices often use exclusive communication protocols and interface standards, resulting in complex and rigid system integration process. The common control mode is to establish independent point-to-point connections between the central controller and each device. Under this architecture, data exchange between devices must be transferred and protocol-converted through the central controller, forming a tight coupling relationship. When a device type needs to be replaced or added, the communication logic of the entire system may need to be significantly modified or even redeveloped, lacking scalability.

[0003] The existing sorting control logic relies on pre-set, sequentially executed hard-coded processes. The control system triggers subsequent actions based on a single sensor event. This method builds control decisions on discrete, local information. As a continuous running dynamic system, the overall state of the production line, such as real-time item location, sorting port congestion, device health, etc. global information, is not systematically modeled and utilized. The decision-making process lacks awareness of the overall operation of the production line.

[0004] Due to the inability to obtain and integrate data from all chain devices in real time, the control system cannot predict and adaptively adjust to abnormal situations during the sorting process. When a sorting bin is blocked or a device experiences a temporary delay, the system can only passively handle the emergency after the fault occurs, and cannot plan alternative paths in advance. This limits the further improvement of production line efficiency and makes it difficult to cope with increasingly complex sorting rules and fluctuating cargo throughput requirements. There is a need for a control method that can achieve flexible communication between devices and make intelligent decisions based on real-time global state. SUMMARY

[0005] The purpose of the present application is to provide a multi-hardware cooperative intelligent sorting production line control method and system to solve the problems raised in the background.

[0006] To achieve the above purpose, the present application provides a multi-hardware cooperative intelligent sorting production line control method, which comprises:

[0007] performing an automatic identification and registration process of multiple hardware devices to generate a registered device set, wherein the multiple hardware devices encompass code reading cameras, volume cameras, dynamic scales, tumblers, bin buttons, bin block sensors, alarm lights, dual cameras, RFID reading devices, security scanners, inkjet printers, linear devices, and narrowband devices;

[0008] based on the registered device set, establishing an inter-device data exchange network, wherein an asynchronous communication mechanism is employed to configure interaction protocols among the multiple hardware devices;

[0009] based on the inter-device data exchange network, constructing an item flow dynamic model of the sorting production line, wherein the item flow dynamic model defines a movement trajectory of items from an entry point to an exit point, and embeds real-time state update logic;

[0010] deploying a sorting rule engine, wherein the sorting rule engine generates sorting path decisions based on a set of item attributes;

[0011] based on the inter-device data exchange network, initiating periodic data collection tasks to continuously monitor item flows on the sorting production line, and refreshing a state of the item flow dynamic model using collected data;

[0012] running a sorting instruction computation algorithm, wherein the sorting instruction computation algorithm generates control commands for each item based on a current state of the item flow dynamic model and an output of the sorting rule engine;

[0013] distributing the control commands to corresponding hardware devices through the inter-device data exchange network to drive sorting actions to be performed;

[0014] storing sorting process data, and performing sorting performance analysis routines based on the stored data.

[0015] Preferably, the performing an automatic identification and registration process of multiple hardware devices comprises:

[0016] scanning available devices in an industrial control network to detect a connection state and a response time of each hardware device;

[0017] performing parameter self-adaptive calibration on the detected devices to adjust a focal length of a code reading camera, a zero point of a dynamic scale, and an angle offset of a tumbler;

[0018] assigning a unique network identifier to each hardware device, and configuring a data frame format to ensure transmission consistency.

[0019] Preferably, the establishing an inter-device data exchange network comprises:

[0020] defining a message passing standard based on a publish-subscribe pattern to allow hardware devices to asynchronously exchange sensor data and control signals;

[0021] Setting a position trigger event, automatically activating the relevant equipment to collect data when the article reaches the volume camera or dynamic scale;

[0022] Implementing a failover protocol, rerouting data streams to backup equipment when detecting abnormality of the blockage sensor.

[0023] Preferably, the article flow dynamic model of the sorting production line comprises:

[0024] Drawing a physical path map of the article on the sorting production line, marking the code reading camera and the double camera as key data collection points;

[0025] Integrating the real-time blockage signal of the blockage sensor, dynamically updating the available slot state of the path map;

[0026] Processing the output image of the double camera, identifying double-piece articles and correcting the article quantity in the article flow dynamic model.

[0027] Preferably, the deployment of the sorting rule engine comprises:

[0028] Loading a set of article attributes, including size, weight, identification code, and security check results;

[0029] Configuring a priority rule table, calculating the priority level of the sorting path according to the set of article attributes;

[0030] Embedding the feedback data of the inkjet printer, dynamically adjusting the sorting path to match the inkjet content.

[0031] Preferably, the start of the periodic data collection task comprises:

[0032] Triggering the code reading camera to capture article identification information, and synchronously starting the double camera for double-piece verification;

[0033] Polling the status register of the blockage sensor, controlling the alarm light to flash and disabling the corresponding slot when detecting blockage;

[0034] Tracking the article position coordinates, updating the position data in the article flow dynamic model every fixed time interval.

[0035] Preferably, the running of the sorting instruction calculation algorithm comprises:

[0036] Matching the set of article attributes with the rule conditions of the sorting rule engine, determining the target slot number;

[0037] Evaluating the current state of the blockage sensor and the output of the security check machine, calculating the backup sorting path;

[0038] Generating the steering angle instruction of the pendulum sorting machine or the start-stop command of the linear equipment.

[0039] Preferably, the distributing the control command to the corresponding hardware device through the inter-device data exchange network comprises:

[0040] Sending a pulse signal to the balance wheel sorter to control it to guide the article to the target pocket, when the pocket blocking sensor reports a blockage, moving the article to the backflow channel through the narrowband device, coordinating the inkjet printer to trigger the inkjet action when the article passes to ensure that the inkjet content is consistent with the article identification.

[0041] Preferably, the storing the sorting process data and performing a sorting performance analysis routine based on the stored data comprises:

[0042] Recording the sorting time stamp, success sorting flag and error code of each article;

[0043] Counting the number of activations of the pocket blocking sensor and the duration of the alarm light, and calculating the pocket blockage rate;

[0044] Adjusting the rule weights of the sorting rule engine and the parameters of the article flow model according to the performance indicators.

[0045] Preferably, the application further comprises a multi-hardware collaborative intelligent sorting production line control system, the system comprises a memory, a processor and a computer program stored in the memory and running on the processor, and the processor implements the steps of the above-mentioned multi-hardware collaborative intelligent sorting production line control method when executing the computer program.

[0046] Compared with the prior art, the application has the beneficial effects that:

[0047] A unified inter-device data exchange network is established, and an asynchronous communication mechanism is used to configure an interactive protocol. This technical solution enables all connected heterogeneous hardware devices to interact with data on a standardized platform, breaking down the information barriers caused by the traditional point-to-point integration mode. The data generated by the devices, such as identification information and status signals, can be broadcast in a low-delay event-driven manner without waiting for the polling of the central controller. This decoupled architecture reduces the complexity of system integration, and when the production line needs to be upgraded or new devices are introduced, it only needs to be registered to the network and configure the corresponding communication protocol, greatly enhancing the flexibility and scalability of the system. Data processing has changed from serial to parallel, improving the overall throughput and response real-time performance of the system.

[0048] A dynamic model of the article flow is constructed for the sorting production line, and real-time state updating logic is embedded. The model fuses the dispersed and instantaneous device signals on the physical line into a continuous and global system view through continuous data collection, accurately reflecting the location, state of each article and the running situation of the whole line. This enables the sorting instruction calculation algorithm to make decisions based on rich context information far beyond the attributes of a single article. The generation of control commands not only considers the destination of the article, but also takes into account the real-time congestion of the target sorting port, the spacing between articles and the overall flow load on the line, thereby achieving dynamic path planning and load balancing. The sorting process changes from passive execution based on fixed rules to active optimization and predictive regulation based on real-time state, improving the sorting accuracy and overall throughput efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 The working principle diagram of the multi-hardware cooperative intelligent sorting production line control method described in the present application;

[0050] Figure 2 The flowchart for automatic identification and registration of multi-hardware devices;

[0051] Figure 3 The flowchart for constructing the dynamic model of article flow;

[0052] Figure 4 The data collection and article tracking diagram;

[0053] Figure 5 The control command distribution and execution diagram. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0055] Please refer to Figure 1The present application provides a kind of multi-hardware collaborative intelligent sorting production line control method, the method includes: automatic identification and registration process first scans the available equipment in industrial control network, detects the connection state and response time of reading code camera, volume camera, dynamic scale, balance wheel sorting machine, compartment button, block compartment sensor, alarm lamp, except double camera, RFID reading code equipment, security check machine, ink-jet printer, linear equipment and narrowband equipment, parameter self-adapting calibration is executed, such as adjusting the focal length of reading code camera and the zero point of dynamic scale, and each equipment is allocated unique network identifier to configure data frame format.Based on the registered device set, establish the data exchange network between equipment, adopt asynchronous communication mechanism to define the message passing standard of publish-subscribe mode, set position trigger event to activate data acquisition, and realize fault transfer protocol to handle exception.Article flow dynamic model is subsequently built, defines article movement track and integrates real-time state update logic, processes the output image of except double camera to identify double-piece article.Sorting rule engine generates sorting path decision based on article attribute set such as size, weight, identification code and security check result, configures priority rule table and embeds ink-jet printer feedback data.Periodic data acquisition task continuously monitors article flow, triggers reading code camera to capture identification information and polls block compartment sensor state, while tracking article position coordinate updates model.Sorting instruction calculation algorithm combines model state and rule output to generate control command, determines target compartment by matching article attribute and evaluates sensor state to calculate backup path.Control command is distributed to hardware equipment through network, such as sending pulse signal to balance wheel sorting machine or coordinating ink-jet printer action.Storage sorting process data and execute performance analysis routine, statistical indicators are used to adjust system parameters.

[0056] Example 1: refer to Figure 2In a specific implementation, the automatic identification and registration process starts with a comprehensive scan of the industrial control network, which detects available devices connected in the network by sending broadcast request packets, covering all hardware devices such as code reading cameras, volume cameras, dynamic scales, and swing wheel sorters. During the scanning process, the system records the Internet Protocol address and physical address of each device, and periodically polls the device port based on the heartbeat mechanism to detect the connection state. The response time is measured by calculating the transmission time difference between the request and confirmation packets to ensure network accessibility and communication stability. Parameter adaptive calibration is performed on the detected devices, which uses special routines for different device types, such as adjusting the focus of the code reading camera by dynamically adjusting the lens position based on the clarity index of the captured image until the key area of the image meets the identification requirements. When calibrating the zero point of the dynamic scale, the sensor output value is read multiple times under no load conditions, and the average offset is calculated and compensated. When correcting the angle offset of the swing wheel sorter, the actual position feedback from the encoder is compared with the preset standard position, and the deviation is corrected by adjusting the steering control signal. The calibration process is iteratively executed until the parameters are stable within the allowed error range. Each hardware device is assigned a unique network identifier, which is generated by combining the device manufacturer code and sequence number hash to ensure uniqueness. A uniform data frame format is configured, including preamble, target address, data payload, and cyclic redundancy check code fields, to ensure consistency and analyzability of data transmission between different devices.

[0057] When establishing a data exchange network between devices, a message passing standard based on the publish-subscribe model is defined. All hardware devices are registered as publishers or subscribers to the message broker server, which manages the topic routing logic. This supports asynchronous exchange of sensor data and control signals between devices, such as the code reading camera publishing the item identification code topic and the swing wheel sorter subscribing to the topic to obtain sorting information. Position trigger events are set, which bind trigger conditions to the detection areas of volume cameras, dynamic scales, and other devices. When an item reaches the trigger point, the relevant device is driven to start data collection through photoelectric sensor or encoder signal interruption, reducing invalid polling overhead. A failover protocol is implemented to continuously monitor the status heartbeat of critical devices such as blockage sensors. When communication timeout, signal anomalies, or other conditions are detected, the data stream is automatically rerouted to backup devices, and the session state and buffer data are synchronized to ensure data transmission continuity.

[0058] In specific implementation, the network scanning operation of automatic identification and registration process is deployed on a dedicated management node, which periodically sends user datagram protocol probe packets to a preset network segment and analyzes the source address and timestamp of the reply packet to construct a device list, the connection state assessment determines whether the device is online or offline based on a threshold of consecutive successful handshake times, and the response time calculation uses a sliding window average to filter transient fluctuations to improve accuracy. Parameter adaptive calibration is performed at the device driver layer, which calls an application programming interface to access hardware configuration registers, adjusts the focal length of the code reading camera through a stepper motor control loop, and adjusts the lens displacement according to image contrast feedback; dynamic zero-point calibration is performed by sampling weight values multiple times in a stable environment and calculating the arithmetic mean as the reference; the angle offset correction of the balance wheel sorting machine measures the deviation between the actual position and the command position by means of a laser range finder, and outputs a compensation pulse through a proportional-integral-derivative controller. The process of assigning a unique network identifier is integrated into the device registration service, which queries the unique identifier such as the media access control address in the device firmware and maps it to the internet protocol address allocated by the dynamic host configuration protocol, and the data frame format configuration follows industry standards to ensure field alignment and byte sequence uniformity. In specific implementation, the establishment of the inter-device data exchange network is based on the instantiation of a message broker at the middleware software layer, the definition of a topic tree structure organizes device data streams, and the asynchronous communication mechanism processes subscription messages through callback functions to avoid blocking the main control thread. The position trigger event configuration uses hardware interrupts or software timers, and generates a rising edge signal when an object blocks the photogate of the volume camera, triggering the data acquisition thread to start image capture and weight measurement. The fault transfer protocol implementation includes a health check module that periodically sends query commands to the blocking sensor, and marks the device as faulty and switches the data subscription path to a backup sensor instance if there is a timeout or no response, and network rerouting uses a load balancer to dynamically update the binding rules of the message broker.

[0059] In specific implementation, the parameter self-adaptive calibration process can be further refined, the focal length adjustment algorithm of the code reading camera is based on the principle of gradient maximization, the principle of gradient maximization drives the focal length fine adjustment by calculating the Laplacian operator response value of the image; the dynamic zero point calibration introduces a temperature compensation coefficient, which corrects the drift error according to the reading of the embedded temperature sensor; the angle offset correction of the balance wheel sorting machine fuses multi-sensor data, including inertial measurement unit and visual feedback to improve accuracy. The unique network identifier allocation scheme can extend the support of IPv6 address pool management, which automatically allocates globally unique addresses and configures the neighbor discovery protocol. The processing of the location trigger event can be optimized as an event-driven architecture, which uses an event bus to distribute trigger signals and reduces polling overhead. Therefore, it can be understood that the automatic identification and registration process reduces the need for manual intervention through standardized processes. The asynchronous communication mechanism of the inter-device data exchange network eliminates the timing problems caused by the performance differences of the devices through the buffer queue.

[0060] In specific implementation, the data frame format configuration can customize the header field, which contains the timestamp and sequence number for data reordering and packet loss detection. The message passing standard of the publish-subscribe mode can support multicast transmission, which efficiently distributes control commands to a group of similar devices. The setting of the location trigger event can integrate real-time operating system tasks, which schedule trigger processing routines with fixed priority. The implementation of the failover protocol can add a logging function, which tracks switching events for post-analysis. Optionally, the network identifier allocation can bind the device role, such as a sensor or an actuator, which affects the identifier encoding rules. Optionally, the message passing standard can include retry logic, which automatically resends messages when transmission fails.

[0061] Embodiment 2: refer to Figure 3In a specific implementation, the construction of the article flow dynamic model of the sorting production line starts from drawing a physical path diagram of the article on the sorting production line, the physical path diagram is defined in a Cartesian coordinate system, the origin is set at the center point of the sorting line entrance, the X axis extends along the running direction of the conveying belt, and the Y axis represents the transverse sorting displacement. The path diagram marks the code reading camera and the double camera as key data collection points. The code reading camera collection point is located at the coordinates (2.5, 0) for capturing article identification information, and the double camera collection point is located at the coordinates (3.8, 0) for double piece verification. The article movement trajectory is described by a differential equation to describe the change of position with time. The real-time jam signal of the integrated grid sensor is connected to the programmable logic controller through the digital input module. When the sensor detects that the grid is jammed, a high-level signal is generated, the available grid state of the path diagram is dynamically updated, the update logic marks the path segment corresponding to the jammed grid as a red forbidden passage state, and the article bypass path is automatically recalculated. The output image of the double camera is analyzed using a convolutional neural network model. The convolutional neural network model includes five convolutional layers and three fully connected layers. The double-piece article is identified based on the boundary overlap degree feature. The boundary overlap degree feature is obtained by calculating the ratio of the intersection area to the union area of the minimum bounding rectangle of the article contour. When the ratio exceeds the threshold value 0.3, it is determined as a double-piece article. The article quantity in the article flow dynamic model is corrected. The correction operation increases the single article count in the model to double the count.

[0062] When deploying the sorting rule engine, first load the article attribute set, which includes core attributes such as size data obtained from the volume camera, weight data obtained from the dynamic scale, identification code obtained from the code reading camera or RFID device, and security inspection results obtained from the security inspection machine. Configure the priority rule table, which stores the mapping relationship between rule conditions and target grid using a two-dimensional data table. Assign weight coefficients to different rules based on article attributes, and determine the priority level of the sorting path through weighted calculation.

[0063] In a specific implementation, the construction of the article flow dynamic model is exemplified. Assuming that the sorting production line has 12 grids, the physical path diagram models each grid as a graph node, the edges between the nodes represent the conveying belt path, the edge weight is calculated based on the actual distance, the code reading camera collection point is set at node A (coordinates (2.5, 0)), and the double camera collection point is set at node B (coordinates (3.8, 0)). The article movement trajectory is described by the following motion equation:

[0064]

[0065] wherein: represents the displacement of the article in the conveying belt direction at time t, is the base speed of the conveying belt, is the acceleration variable, represents the lateral swing displacement, is the i-th vibration source amplitude, is the angular frequency, is the phase angle, is the number of vibration sources.

[0066] In a specific implementation, the deployment example of the sorting rule engine involves specific rule conditions, and the priority rule table contains the fields: rule ID, attribute condition, weight coefficient, and target bin. For example, the entry of rule ID R001 defines the condition "weight > 5 kg and size volume > 0.01 m³", the weight coefficient is configured as [weight: 0.6, size: 0.4], the weighted score algorithm calculation formula is: priority score = weight normalized value × 0.6 + size normalized value × 0.4, and the items with a score higher than 0.8 are allocated to the priority bin G01. The feedback data of the inkjet printer is integrated by reading the inkjet content in real time through the OPCUA protocol. When the inkjet printer reports the inkjet "SH-202" but the rule calculation target is "BJ-101", the engine calls the path recalculation module to route the item to the backflow channel for reprocessing. The storage of the item attribute set can use a time series database, which optimizes the fast query of recent sensor data. Optionally, the update of the priority rule table can support hot deployment, which allows modifying the rule weight without stopping. The sorting rule engine realizes the quantitative processing of multi-attribute decision-making through the weighted scoring algorithm.

[0067] Embodiment 3: In a specific implementation, the periodic data acquisition task is started by triggering the system timer, which is configured to generate an interrupt signal at a fixed time interval. The interrupt service routine calls the data acquisition thread to perform the acquisition operation. First, trigger the code reading camera to capture the item identification information, and at the same time, start the double camera verification through the hardware synchronization signal line to ensure that the two sets of vision equipment collect the item data at the same time point, thereby improving the data consistency. Poll the state register of the jam sensor, read the jam state of each bin through the standard communication protocol, and when a jam is detected, immediately control the corresponding alarm lamp to flash at a fixed frequency, and disable the sorting allocation function of the bin in the system to avoid continuous accumulation of items. Track the item position coordinates, combine the encoder pulse count and the photoelectric sensor trigger timestamp to calculate the real-time position of the item on the path, and update the position data in the item flow model every fixed time interval to ensure that the model can accurately reflect the item flow trajectory.

[0068] When running the sorting instruction calculation algorithm, first format the item attribute set into a key-value pair format recognizable by the rule engine, traverse the rule conditions in the priority rule table for logical matching, filter out the rules that match successfully and determine the target bin number according to the weight coefficient. Evaluate the current state of the jam sensor and the output result of the security inspection machine. If the target bin is in a jammed state or the security inspection result is abnormal, use the path search algorithm to calculate the backup sorting path, and preferentially select the backup bin with the shortest physical distance and the lowest current load. According to the position information of the target bin or the backup bin, generate the steering angle instruction of the pendulum sorting machine or the start-stop and speed regulation command of the linear equipment. The instruction parameters are calculated based on real-time data such as the current position and moving speed of the item, to ensure accurate execution of the sorting action.

[0069] In specific implementation, taking the specific execution process of the data acquisition task as an example, it is assumed that the sorting line runs at a speed of 1.5 meters per second, and the timer interrupt service routine is triggered at time point , the read code camera trigger signal is set and a synchronization pulse is sent to the dual camera at the same time, and the two cameras complete image acquisition and output the item identification code and dual piece detection result within +3ms. The jam sensor state register polling adopts a multi-thread architecture, and the main thread sends Modbus query commands for 12 bins every 50ms. After the response data packet is parsed, the global state mapping table is updated. The item position tracking is based on the encoder resolution of 2000 pulses per meter and the photoelectric sensor trigger interval. The position coordinates are calculated through the following kinematics equation:

[0070]

[0071] Wherein: and represents the position of the item in the coordinate system at the kth sampling, is the cumulative pulse number of the encoder at time k, is the encoder resolution of 2000 pulses per meter, is the steering angle of the pendulum sorting machine.

[0072] In specific implementation, the synchronization mechanism of data acquisition task can be refined, the hardware synchronization signal adopts RS422 differential signal transmission, the rising edge time is less than 10 nanoseconds, and the reading camera and the double camera exposure time deviation are controlled within microseconds. The error retry mechanism is realized by polling the state register of the blocking grid sensor, and after three consecutive readings timeout, the sensor failure is marked and switched to the backup detection mode. The item position coordinate update introduces Kalman filter, the Kalman filter prediction model is based on the assumption of uniform speed, the observation value comes from the photoelectric sensor trigger position, and the coordinate jump is smoothed. The rule matching optimization of sorting instruction calculation algorithm is parallel processing, which uses multi-core CPU to evaluate multiple rule conditions at the same time, and reduces the calculation delay. The backup sorting path calculation caches the latest 10 available grid lists, and preferentially selects the backup grid with the shortest physical distance according to the nearest neighbor principle. The steering angle instruction generation of the wheel sorting machine increases the feedforward compensation, which dynamically adjusts the response gain of the steering engine according to the weight of the item.

[0073] In specific implementation, the reading camera trigger signal is configured as TTL level with pulse width of 100 microseconds, and the double camera synchronization pulse delay is controlled within 50 microseconds, so as to ensure that the image capture time difference is less than the frame exposure time. The state register of the blocking grid sensor adopts batch reading instruction, which obtains all grid states at a time to reduce communication overhead. The angle θ(k) in item position coordinate calculation is obtained by absolute value reading of the rotary encoder, which avoids cumulative error. The rule condition matching of sorting instruction calculation algorithm adopts a compiled rule engine, which precompiles rules into bytecode to improve execution efficiency. The backup sorting path calculation updates the path cost weight in real time, and the weight is dynamically adjusted according to the congestion degree of the grid. The steering angle instruction output of the wheel sorting machine increases the software limit protection to prevent mechanical overtravel. Optionally, data acquisition task can increase data validity check, including CRC check and numerical range check. Fuzzy logic can be introduced in sorting instruction calculation to handle rule matching uncertainty under boundary conditions.

[0074] Referring to Figure 4, which demonstrates the execution effect of data collection tasks and the real-time movement state of items on the sorting line. In the figure, multiple curves show the position change of the item in the coordinate system, the state monitoring of the blockage sensor, and the instantaneous speed calculated based on the encoder data. From the item position coordinate curve, we can see its movement trajectory on the sorting line: as time (horizontal axis, X axis) advances, the item position (vertical axis, Y axis) shows a stable linear growth trend, which reflects the uniform running characteristics of the conveyor belt; while the periodic fluctuations of the item position (vertical axis, Y axis) reflect the fine adjustment control of the pendulum sorting machine on the item position. The blockage sensor state is displayed in the form of green steps, clearly identifying the blockage events and their duration during system operation, which provides important environmental state input for sorting decisions. The orange speed curve reflects the stability of the system operation, and the small fluctuations in instantaneous speed are due to the accuracy of the encoder pulse counting and the dynamic adjustment during the movement process. Figure 4 The data in fully embodies the periodic data collection mechanism, including the timing trigger of the code reading camera, the state polling of the blockage sensor, and the real-time tracking of the item position. The coordinated work of these data provides complete input information for the sorting instruction calculation algorithm, ensuring the accuracy and real-time performance of the sorting decisions. By analyzing these data curves, the performance of the data collection system can be evaluated, and the sampling frequency and data processing algorithm can be optimized.

[0075] In a specific implementation, the control command distribution process first encapsulates the control commands generated by the sorting instruction calculation algorithm into device-specific protocol data units, which contain target device identifiers, command codes, parameter lists, and checksum fields. After encapsulation, the protocol data units are transmitted to the corresponding hardware devices through the message queue of the publish-subscribe mode. When sending pulse signals to the pendulum sorting machine, the pulse signals are generated using pulse width modulation technology, and the pulse width duty cycle has a linear correspondence with the steering angle. The required steering angle is calculated based on the target bin orientation, and then the high-level duration of the pulse signal is determined to control the pendulum sorting machine to accurately guide the item to the target bin. When the blockage sensor reports that the target bin is blocked, an exception handling process is triggered immediately, and the item is moved to the backflow channel through a narrowband device. The control parameters of the narrowband device (such as transmission speed, acceleration, and direction angle) are calculated based on the relative relationship between the current position of the item and the backflow inlet, ensuring smooth movement of the item. When coordinating the work of the code printer, the trigger time of the code printer is aligned with the time of the item arriving at the code printing head through a time synchronization protocol. Based on the item moving speed and the distance between the code printing head and the detection point, the trigger delay time is calculated to ensure that the code printer triggers the inkjet action at the best position of the item. At the same time, visual detection is used to confirm that the code content is consistent with the item identification, ensuring the accuracy of code printing. In a specific implementation, the control command distribution to the pendulum sorting machine is an example, assuming that the target bin is G05 and the steering angle is 45 degrees. The pulse width modulation parameter conversion formula is:

[0076]

[0077] wherein: represents the pulse signal high level duration, is the target steering angle 45 degrees, is the maximum pulse width 200 milliseconds. The pulse signal high level duration is calculated to be 25 milliseconds, the low level duration is 175 milliseconds, and the repetition period is 200 milliseconds. When the blockage sensor detects that the G05 slot is blocked, the blockage event immediately updates the sorting routing table, refer to Table 1, and the narrowband device receives the moving command parameters.

[0078] Table 1: Narrowband device control parameter table

[0079]

[0080] In the implementation of coordinated trigger of the inkjet printer, the inkjet head is installed at the position coordinate (5.2, 0), and when the article position sensor detects that the article reaches the coordinate (5.0, 0), the inkjet trigger delay time is calculated in advance. The delay time is calculated based on the current speed of the article 1.5 m / s and the distance difference 0.2 m, and it is concluded that the inkjet needs to be triggered after 133 milliseconds. The pulse signal generation can increase the soft start curve, and the soft start curve adopts a sine acceleration mode to reduce mechanical impact. Optionally, the narrowband device control can introduce an anti-collision algorithm to adjust the distance between adjacent devices in real time to prevent article accumulation. The control command distribution realizes the cooperative work of multiple devices through precise timing control.

[0081] In the specific implementation, the protocol data unit structure in the command encapsulation stage is specifically defined. The protocol data unit header includes a 4-byte synchronization word 0xAA55AA55, a 2-byte device identifier, a 2-byte command code, a variable-length data field, and a 2-byte CRC-16 check code at the tail. The impulse signal transmission of the balance wheel sorter adopts a differential signal transmission mode. The differential signal transmission uses the RS485 physical layer, the baud rate is set to 115200 bps, and the data frame includes a start bit, 8 data bits, no parity bit, and a stop bit. The processing priority of the blockage event reported by the blockage sensor is set to the highest level, and the system completes the narrowband device command distribution within 10 milliseconds after the event trigger. The execution of the narrowband device moving command includes three stages: the acceleration stage accelerates to 1.2 m / s with an acceleration of 0.5 m / s², the constant speed stage maintains the speed to move 2.8 meters, and the deceleration stage decelerates to stop with a deceleration of -0.5 m / s². The inkjet printer coordinated trigger adopts a hardware interrupt mode. After the rising edge interrupt of the position sensor trigger, a high-precision timer is started to count, and the inkjet printer trigger line level is pulled low for at least 10 milliseconds after the calculated delay time is reached.

[0082] In specific implementation, the transmission guarantee mechanism of the inter-device data exchange network, important control commands such as the rotation command of the tumbler sorter adopt a repeated sending strategy, and the same command packet is sent three times in succession, and the receiving end performs deduplication processing based on the sequence number. The jam sensor event report increases the confirmation mechanism, and the narrowband device must return the execution status word after receiving the transfer command, and if it is not confirmed within a timeout, the retransmission process is triggered. The inkjet printer trigger timing calibration uses the timestamp synchronization protocol, and the master controller and the built-in clock of the inkjet printer perform time synchronization every 30 seconds, and if the deviation exceeds 10 milliseconds, it is automatically corrected. The inkjet content verification link is coordinated, and after the inkjet printer completes the inkjet, the downstream vision system captures the inkjet image for optical character recognition comparison, and if they are inconsistent, an error log is recorded and the re-sorting process is triggered.

[0083] In specific implementation, the tumbler sorter pulse signal generation uses a dedicated timer hardware, the timer is configured in PWM mode, the output frequency is 5Hz adjustable, and the duty cycle resolution reaches 0.1%. The parameter optimization of the narrowband device transfer command is based on the kinematics model, and the kinematics model calculates the acceleration curve to make the article move smoothly and avoid overturning. The inkjet printer trigger delay calculation considers the speed fluctuation of the conveyor belt, and the speed fluctuation is dynamically corrected by real-time encoder feedback. The protocol data unit transmission increases the flow control mechanism, and when the receiving end buffer exceeds the threshold, it sends a pause frame to control the sending rate. The jam sensor event processing realizes priority preemption, and high-priority events can interrupt the execution of low-priority commands.

[0084] Referring to Figure 5 , the coordinated operation state of each execution device in the control command distribution process is shown, including the steering control of the tumbler sorter, the motion parameters of the narrowband device, and the trigger timing of the inkjet printer. In the figure, the complex timing relationship of multi-device collaborative control is clearly presented through curves and marker points of different colors. The tumbler steering angle curve shows the accurate control instructions generated by the system according to the sorting path decision, the step change of the angle corresponds to the sorting demand of different target pockets, and the smooth transition reflects the mechanical characteristics of the steering mechanism and the optimization effect of the control algorithm. The control parameters of the narrowband device include running speed and acceleration, and the coordinated change of these parameters ensures the stability of the article during the transfer process, especially the emergency response capability when handling jam events. The inkjet trigger event is displayed with triangular markers, and its time distribution reflects the system's accurate prediction and coordination control capability for the arrival time of the article. The data in the figure verifies the effectiveness of the command packaging mechanism and the inter-device coordination strategy. The time sequence distribution of the control command reflects the real-time response capability of the system, and the collaborative change of the parameters of each device demonstrates the complexity of multi-hardware collaborative work. By analyzing these control data, the command distribution algorithm can be optimized, and the overall operation efficiency and reliability of the system can be improved, especially in handling abnormal situations and device coordination.

[0085] In a specific implementation, the storage of sorting process data is implemented through a database management system, which uses a time series database architecture to store the sorting records of each item, each record containing a sorting timestamp, a successful sorting flag, and an error code field. The sorting timestamp records the complete time sequence from the entry of the item into the sorting line to the exit, with a precision of milliseconds. The successful sorting flag uses a Boolean value to indicate whether the item is correctly sorted into the target bin. The error code field encodes the types of abnormalities that occur during the sorting process, such as reading code failure or bin jamming. The activation times of the jam sensor and the duration of the alarm light are counted. The statistical process is achieved by querying the event log table in the database. The activation times of the jam sensor accumulate the number of times each bin changes from normal to jammed within the statistical period. The duration of the alarm light is the time difference from the start of the alarm to the shutdown, which is summed up to calculate the bin jamming rate using the following formula:

[0086]

[0087] where: represents the bin jamming rate, with a unit of percentage, is the total number of jam sensors, is the activation time of the i-th jam sensor within the statistical period, is the total item flow within the statistical period. Adjust the rule weight of the sorting rule engine and the parameters of the item flow dynamic model according to the performance indicators. The adjustment process is based on historical data analysis results. The rule weight modification is achieved by updating the weight coefficient matrix. The parameters of the item flow dynamic model are calibrated using the gradient descent method to optimize the coefficients in the motion equation.

[0088] In a specific implementation, when storing sorting process data, a time series database is used to record the complete sorting information of each item, including item ID, entry timestamp, exit timestamp, successful sorting flag, error code, and other core fields. The timestamp precision reaches milliseconds, and the error code is used to identify different abnormal types such as reading code failure, bin jamming, and inkjet abnormality. The activation times of the jam sensor and the duration of the alarm light are counted, and the database aggregation query function is used to group and count by bin number and time period. The jamming rate of each bin is calculated, and the jamming rate is calculated based on the total item flow within the statistical period, combined with the jam activation times, to intuitively reflect the bin running state.

[0089] The sorting performance analysis routine is executed based on the stored data, and the analysis indicators include sorting efficiency, bin jam rate, equipment failure rate, instruction execution delay, etc. The rule weight of the sorting rule engine is adjusted according to the analysis results, for example, when the jam rate of a certain bin is continuously high, the rule weight corresponding to the bin is reduced to reduce the distribution amount of the goods. The parameters of the goods flow dynamic model are adjusted, for example, the speed parameter in the model is corrected according to the actual moving speed of the goods to improve the position prediction accuracy of the model. The performance analysis routine supports generating statistical reports by day, week, and month to provide data support for production line operation and optimization, for example, by analyzing the equipment failure rules to determine the maintenance cycle, and by analyzing the sorting efficiency fluctuations to optimize the sorting rules.

[0090] The jam sensor activation frequency statistics realizes real-time stream processing, and the stream processing framework calculates the activation frequency every 5 seconds and triggers a warning. The alarm lamp duration statistics adds power consumption calculation, and the power consumption is evaluated based on the rated power of the alarm lamp and the product of the duration to evaluate energy efficiency. The bin jam rate formula is expanded for dimension-by-dimension calculation, and the dimension-by-dimension calculation respectively calculates the jam rate distribution according to the bin position and time period.

[0091] In specific implementation, the database management system configures a data retention policy, the data retention policy automatically archives detailed records that are more than 30 days old, and retains aggregated statistical results for long-term storage. The jam sensor activation frequency statistics realizes adaptive threshold, the adaptive threshold dynamically adjusts the activation judgment standard according to the historical baseline. The alarm lamp duration statistics adds anomaly detection, the anomaly detection identifies outliers with a duration that is more than 3 times the standard deviation. The bin jam rate calculation result is stored as a time series, and the time series is used for trend prediction and seasonal analysis. The rule weight adjustment process adds a manual review link, the manual review link allows operation and maintenance personnel to confirm the automatic adjustment suggestion. The goods flow dynamic model parameter calibration uses cross-validation, and the cross-validation uses multiple sets of data to verify the parameter generalization ability.

[0092] It should be noted that, in this text, relational terms such as first and second are used merely to distinguish one entity or action from another, without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device.

[0093] Although embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-hardware cooperative intelligent sorting production line control method, characterized in that, The method comprises: performing an automatic identification and registration process of multiple hardware devices to generate a registered device set, wherein the multiple hardware devices include a code reading camera, a volume camera, a dynamic scale, a tumbling sorter, a grid button, a blocked grid sensor, an alarm light, a dual camera, an RFID code reading device, a security inspection machine, a inkjet printer, a linear device, and a narrowband device; based on the registered device set, establishing an inter-device data exchange network, wherein an asynchronous communication mechanism is used to configure the interaction protocol between the multiple hardware devices; based on the inter-device data exchange network, constructing an item flow dynamic model of the sorting production line, wherein the item flow dynamic model defines the movement trajectory of the item from an entry point to an exit point, and embeds real-time state update logic; deploying a sorting rule engine, which generates a sorting path decision based on a set of item attributes; based on the inter-device data exchange network, starting a periodic data collection task to continuously monitor the item flow on the sorting production line, and using the collected data to refresh the state of the item flow dynamic model; running a sorting instruction calculation method, which generates a control command for each item based on the current state of the item flow dynamic model and the output of the sorting rule engine; distributing the control command to the corresponding hardware device through the inter-device data exchange network to drive the sorting action to be executed; storing sorting process data, and performing a sorting performance analysis routine based on the stored data; the establishment of the inter-device data exchange network comprises: defining a message passing standard based on a publish-subscribe mode, allowing hardware devices to exchange sensor data and control signals asynchronously; setting a location trigger event that automatically activates the relevant device to collect data when the item reaches the volume camera or the dynamic scale; implementing a failover protocol that reroutes data flow to a backup device when a blocked grid sensor anomaly is detected; the running of the sorting instruction calculation method comprises: matching the set of item attributes with the rule conditions of the sorting rule engine to determine the target grid number; evaluating the current state of the blocked grid sensor and the output of the security inspection machine to calculate the backup sorting path; generating the steering angle instruction of the tumbling sorter or the start-stop command of the linear device; the distribution of the control command to the corresponding hardware device through the inter-device data exchange network comprises: sending a pulse signal to the tumbling sorter to control it to guide the item to the target grid; when the blocked grid sensor reports a blockage, moving the item to a backflow channel through the narrowband device; coordinating the inkjet printer to trigger the inkjet action when the item passes through to ensure that the inkjet content is consistent with the item identification.

2. The multi-hardware cooperative intelligent sorting production line control method of claim 1, wherein, the automatic identification and registration process of the multiple hardware devices comprises: scanning the available devices in the industrial control network to detect the connection state and response time of each hardware device; performing parameter self-adaptive calibration on the detected devices to adjust the focal length of the code reading camera, the zero point of the dynamic scale, and the angle offset of the tumbling sorter; assigning a unique network identifier to each hardware device, and configuring the data frame format to ensure transmission consistency.

3. The multi-hardware cooperative intelligent sorting production line control method of claim 1, wherein, the construction of the item flow dynamic model of the sorting production line comprises: drawing a physical path map of the item on the sorting production line, and marking the code reading camera and the dual camera as key data collection points; Integrating the real-time blockage signal of the blockage sensor, dynamically updating the available slot state of the path map; Processing the output image of the dual-camera, identifying the dual-piece article and correcting the article quantity in the article flow dynamic model.

4. The multi-hardware cooperative intelligent sorting production line control method of claim 1, wherein, The deployment sorting rule engine includes: Loading the article attribute set, which contains size, weight, identification code and security check result; Configuring the priority rule table to calculate the priority level of the sorting path according to the article attribute set; Embedding the feedback data of the inkjet printer to dynamically adjust the sorting path to match the inkjet content.

5. The multi-hardware cooperative intelligent sorting production line control method of claim 1, wherein, The start periodic data collection task includes: Triggering the code reading camera to capture the article identification information and synchronously starting the dual-camera for dual-piece verification; Polling the status register of the blockage sensor, and when a blockage is detected, controlling the alarm light to flash and disabling the corresponding slot; Tracking the article position coordinates and updating the position data in the article flow dynamic model every fixed time interval.

6. The multi-hardware cooperative intelligent sorting production line control method of claim 1, wherein, The storage sorting process data and the execution of the sorting performance analysis routine based on the stored data include: Recording the sorting timestamp, success sorting flag and error code of each article; Counting the activation times of the blockage sensor and the duration of the alarm light, and calculating the slot blockage rate; Adjusting the rule weight of the sorting rule engine and the parameter of the article flow dynamic model according to the performance indicators. 7.A multi-hardware cooperative intelligent sorting production line control system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, The processor, when executing the computer program, realizes the steps of the multi-hardware cooperative intelligent sorting production line control method according to any one of claims 1 to 6.

Images