A multi-device collaborative control method for a press

Abstract

The application relates to the technical field of industrial automation control, and particularly provides a multi-device cooperative control method for a press, which comprises the following steps: decomposing a process task into multiple subtasks to be executed in sequence; establishing a unified time reference, dividing an action window for each subtask based on the action window of the press subtask, and defining a time-space forbidden area according to the action window; obtaining a communication delay parameter of a first robot and a second robot, and differentiating to calculate instruction issuing time of each device, wherein the advance amount of the second robot issuing time is greater than that of the first robot; issuing the instruction according to the instruction issuing time, and judging whether there is a behavior of violating the time-space forbidden area according to real-time feedback of the device and the current time reference, and if yes, triggering a safety response. Through the unified time reference, differentiating instruction scheduling and time-space forbidden area monitoring, high-precision cooperative control of heterogeneous devices is realized, and production safety and beat stability are improved.

Description

Technical Field

[0001] This invention belongs to the field of industrial automation control technology, specifically relating to a multi-device collaborative control method for presses. Background Technology

[0002] In automated stamping production lines, the collaborative operation of presses with various types of equipment, such as industrial robots and automata, is a key factor affecting production efficiency and product quality. Traditional collaborative control methods typically employ centralized timing control or simple interlocking logic based on programmable logic controllers, arranging the action sequences of each device according to a fixed timing sequence.

[0003] However, existing technologies have the following shortcomings: First, industrial robots and embodied intelligent robots differ significantly in terms of communication mechanisms, command response delays, and action execution characteristics. Traditional methods do not take this heterogeneity into account, resulting in insufficient timing planning accuracy. Second, traditional methods use static timing planning, which cannot cope with dynamic disturbances such as equipment performance fluctuations and workpiece batch differences, easily causing cycle instability or equipment interference. Furthermore, existing safety mechanisms are mostly spatial interlocks, lacking active protection in the time dimension, making it difficult to achieve the safety guarantees required for high-precision collaborative operations.

[0004] Therefore, how to achieve high-precision collaborative control of heterogeneous robots and presses, and possess dynamic adaptive capabilities, is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] In view of the above-mentioned shortcomings of the prior art, the present invention provides a multi-device collaborative control method for presses to solve the above-mentioned technical problems.

[0006] In a first aspect, the present invention provides a multi-device collaborative control method for a press, comprising:

[0007] The process task is broken down into multiple sub-tasks that are executed sequentially, including at least the first robot sub-task, the second robot sub-task, and the press sub-task;

[0008] On a globally unified timeline, based on the action window of the press subtask, action windows are divided for each subtask, and spatiotemporal no-go zones are defined based on the action windows of each subtask. The spatiotemporal no-go zones include: prohibiting the first robot and the second robot from entering the working area of ​​the press during the time period covered by the action window of the press subtask; and prohibiting the second robot from entering the movement path area of ​​the first robot during the time period covered by the action window of the first robot subtask.

[0009] Obtain the communication delay parameters between the first robot and the second robot, and calculate the instruction issuance time for the press, the first robot, and the second robot based on the communication delay parameters and each action window. The issuance time is earlier than the start time of the corresponding action window.

[0010] Send corresponding instructions to the press, the first robot, and the second robot according to the dispatch time;

[0011] Based on the real-time feedback status of the press, the first robot, and the second robot, and the current time reference, determine whether there is any behavior that violates the time and space restricted zone. If so, trigger a safety response.

[0012] In one optional implementation, the collaborative controller integrates a Modbus TCP communication interface and a TCP communication interface to construct a unified communication architecture. It establishes a Modbus TCP point-to-point communication connection with the first robot, a TCP local area network communication connection with the second robot, and a high-speed signal interaction connection with the press host, and adopts a unified clock synchronization mechanism. The first robot is an industrial robot, and the second robot is an embodied intelligent robot.

[0013] In one optional implementation, after the collaborative controller is powered on and initialized, it sends registration instructions to the industrial robot via Modbus TCP communication and to the embodied intelligent robot via TCP communication, and collects the operating status of the press host in real time.

[0014] In an optional implementation, the process task is decomposed into multiple sub-tasks to be executed sequentially, including:

[0015] Pre-store parameterized task diagram templates corresponding to various stamping process tasks;

[0016] When a stamping process task is received from the upper-level system, the corresponding parameterized task diagram template is selected from the template library according to the current process task identifier and equipment readiness status, and a task diagram structure containing the logical relationships between each subtask is generated.

[0017] The subtasks include at least the posture correction subtask of the embodied intelligent robot, the material handling subtask of the industrial robot, the stamping subtask of the press host, and the unloading and return subtask of the embodied intelligent robot. The collaborative controller uses the four subtasks as super nodes in the directed acyclic graph. Each super node encapsulates a fine-grained sequence of atomic operations. The atomic operations include at least the basic action units of moving to the alignment position, performing visual correction, grasping the workpiece, entering the mold cavity, and the press descending.

[0018] The edges between supernodes in the directed acyclic graph are used to define the temporal dependencies and spatial mutual exclusion constraints between different subtasks; among them, the posture correction subnode of the embodied intelligent robot and the material handling subnode of the industrial robot have a sequential dependency relationship, and the material handling subnode of the industrial robot and the stamping subnode of the press host have a spatial mutual exclusion relationship.

[0019] In one optional implementation, on a globally unified timeline, action windows are divided for each subtask based on the action window of the press subtask, including:

[0020] Fix the action window of the press subtask on the global timeline;

[0021] The first robot subtask is divided into action windows, with its start time later than the end time of the press subtask action window and its end time earlier than the start time of the press subtask action window in the next cycle, while reserving a safety margin.

[0022] Divide the action window for the second robot subtask so that it is staggered from the action window of the first robot subtask in time.

[0023] In an optional implementation, spatiotemporal restricted areas are defined based on the action windows of each subtask, including:

[0024] Define a first spatiotemporal restricted area for the press subtask. The spatial range of the first spatiotemporal restricted area is the working area of ​​the press, and the time range is the time period covered by the action window of the press subtask.

[0025] Define a second spatiotemporal no-go zone for the first robot subtask. The spatial range of the second spatiotemporal no-go zone is the area covered by the movement path of the first robot, and the temporal range is the time period covered by the action window of the first robot subtask.

[0026] The first time-space restricted area is used to prohibit any equipment other than the compressor from entering the space range within the time range;

[0027] The second spacetime restricted area is used to prohibit any device other than the first robot from entering the space area within the time range;

[0028] When the action windows of different subtasks overlap in time, the corresponding spatiotemporal restricted areas are configured to not overlap in space, or priority rules are set in the overlapping spatial areas to determine the devices allowed to enter.

[0029] In one optional implementation, the communication delay parameters between the first robot and the second robot are obtained, and the command issuance time for the press, the first robot, and the second robot is calculated based on the communication delay parameters and each action window, including:

[0030] Obtain a first communication delay parameter for the first robot and a second communication delay parameter for the second robot, wherein the value of the second communication delay parameter is greater than the value of the first communication delay parameter;

[0031] For the first robot, its instruction issuance time is calculated based on the difference between the start time of its corresponding first action window and the first communication delay parameter.

[0032] For the second robot, its instruction issuance time is calculated based on the difference between the start time of its corresponding second action window and the second communication delay parameter and the predicted adjustment amount, wherein the predicted adjustment amount is dynamically determined based on the current system load or historical communication fluctuation data.

[0033] For the press, the instruction issuance time is calculated based on the start time of its corresponding press subtask action window. The time difference between the instruction issuance time of the press and the start time of the press subtask action window is less than the instruction issuance time advance of the first robot and the second robot.

[0034] The calculated instruction issuance times are all earlier than the start time of the corresponding subtask action window, and the instruction issuance time advance of the second robot is greater than the instruction issuance time advance of the first robot.

[0035] In an optional implementation, before issuing corresponding instructions to the press, the first robot, and the second robot according to the issuance time, the method further includes:

[0036] Continuously monitor the latency jitter and packet loss rate of each communication link, and use filters to predict the communication latency for the next cycle;

[0037] The instruction issuance time for this round is calculated based on the predicted mean and standard deviation of communication delay. The calculation formula is T_issue=T_scheduled_start-D_cmd_estimated-α×σ_delay, where D_cmd_estimated is the predicted mean delay, σ_delay is the standard deviation of delay, α is the safety factor, and T_scheduled_start is the instruction issuance time.

[0038] For the second robot, an instruction preloading mechanism is adopted, which pre-delivers a set of parameterized instruction sequences to its local cache, and only sends a trigger signal at the time of execution.

[0039] In an optional implementation, it further includes:

[0040] After each production cycle, the timing deviation, position deviation, and load deviation are calculated; wherein, the timing deviation is the difference between the actual completion time of the subtask and the theoretically planned completion time, the position deviation is the difference between the actual position of the robot end effector and the theoretical target position, and the load deviation is the difference between the actual load of the robot end effector and the nominal load.

[0041] Based on the timing deviation and the position deviation, a preset deviation-compensation mapping table is queried to obtain the speed compensation coefficient of the first robot, and the movement speed of the first robot in the next production cycle is adjusted according to the speed compensation coefficient.

[0042] The actual cycle duration between the start times of two adjacent production cycles is measured, the cycle error between the actual cycle duration and the target cycle duration is calculated, the cycle error is input into the PID controller to obtain the dynamic adjustment amount of the press holding time, and the press holding time in the next production cycle is adjusted according to the dynamic adjustment amount to adjust the actual duration of the action window of the press subtask.

[0043] The deviation-compensation mapping table is pre-established through process testing and is used to map timing deviations and position deviations to corresponding speed compensation coefficients.

[0044] In an optional implementation, the deviation-compensation mapping table is a two-dimensional lookup table, with timing deviation and position deviation as input indices and the speed compensation coefficient of the first robot as the output value; the speed compensation coefficient is greater than or equal to 1.

[0045] The beneficial effects of this invention are that the multi-device collaborative control method for presses provided by this invention, by establishing a unified global time reference and defining spatiotemporal restricted areas based on the action windows of each sub-task, elevates safety protection from simple spatial interlocking to spatiotemporal joint constraints. The safety rule engine monitors the matching relationship between equipment status and the global time axis in real time, proactively identifying potential intrusion risks and triggering timely safety responses, effectively avoiding interference and collisions between devices. Furthermore, addressing the differences in communication latency and response characteristics between industrial robots and embodied intelligent robots, by acquiring the communication latency parameters of each device and calculating the command issuance time differently, precise synchronization of heterogeneous device actions is achieved, solving the problem of inaccurate timing coordination under traditional unified scheduling methods. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a schematic flowchart of a method according to an embodiment of the present invention.

[0048] Figure 2 This is a schematic diagram of the unified communication architecture of a method according to an embodiment of the present invention.

[0049] Figure 3 This is a schematic flowchart illustrating the process task decomposition of a method according to an embodiment of the present invention.

[0050] Figure 4 This is a schematic flowchart illustrating the division of action windows in a method according to an embodiment of the present invention. Detailed Implementation

[0051] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0052] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0053] The multi-device collaborative control method for presses provided in this embodiment of the invention is executed by a computer device, and correspondingly, the multi-device collaborative control system for presses runs in the computer device.

[0054] Figure 1 This is a schematic flowchart illustrating a method according to an embodiment of the present invention. Wherein, Figure 1 The executing entity can be a multi-device collaborative control system for the press. Depending on different requirements, the order of steps in this flowchart can be changed, and some steps can be omitted.

[0055] like Figure 1 As shown, the method includes:

[0056] S1. Decompose the process task into multiple sub-tasks to be executed sequentially, including at least the first robot sub-task, the second robot sub-task, and the press sub-task;

[0057] S2. On the global unified timeline, based on the action window of the press subtask, divide the action window for each subtask, and define a spatiotemporal restricted area based on the action window of each subtask. The spatiotemporal restricted area includes: prohibiting the first robot and the second robot from entering the working area of ​​the press during the time period covered by the action window of the press subtask; and prohibiting the second robot from entering the movement path area of ​​the first robot during the time period covered by the action window of the first robot subtask.

[0058] S3. Obtain the communication delay parameters between the first robot and the second robot, and calculate the instruction issuance time for the press, the first robot and the second robot based on the communication delay parameters and each action window. The issuance time is earlier than the start time of the corresponding action window.

[0059] S4. Send corresponding instructions to the press, the first robot, and the second robot according to the sending time;

[0060] S5. Based on the real-time feedback status of the press, the first robot, and the second robot, and the current time reference, determine whether there is any behavior that violates the time and space restricted area. If so, trigger a safety response.

[0061] Please refer to Figure 2 Build a unified communication architecture:

[0062] The collaborative controller integrates Modbus TCP and TCP communication interfaces to build a unified communication architecture. Specifically, the collaborative controller establishes a point-to-point communication connection with the industrial robot through the Modbus TCP communication interface. The Modbus TCP protocol has high reliability and determinism, making it suitable for industrial robot control with high real-time requirements. The collaborative controller establishes a local area network communication connection with the embodied intelligent robot through the TCP communication interface. The TCP protocol supports large data volume transmission and is suitable for embodied intelligent robots that need to transmit complex information such as visual images and posture data. The collaborative controller also establishes a high-speed signal interaction connection with the press host to transmit key status information such as press slider position, stroke status, and fault signals. This high-speed signal interaction can be implemented using industrial fieldbus or hardwiring.

[0063] It should be noted that in this embodiment, the first robot is an industrial robot and the second robot is an embodied intelligent robot. The two have significant differences in communication mechanisms and response characteristics. This embodiment lays the foundation for subsequent differentiated scheduling by constructing a heterogeneous communication architecture.

[0064] The collaborative controller employs a high-precision clock synchronization mechanism to ensure that the press, industrial robot, and embodied intelligent robot have a unified time base. In one preferred embodiment, the collaborative controller synchronizes the local clocks of each device using the IEEE 1588 precision time protocol, achieving microsecond-level time synchronization accuracy. In another preferred embodiment, the collaborative controller uses the slider stroke pulse signal of the press host as a unified hardware time base, taking the moment when the slider reaches the top dead center or bottom dead center as the start time of the cycle, synchronously triggering the action sequence of each device. By establishing a unified global time base, a prerequisite for subsequent accurate timing planning is provided.

[0065] After the collaborative controller is powered on and initialized, it executes the device registration and status acquisition process.

[0066] First, the collaborative controller sends a registration command to the industrial robot via Modbus TCP communication. This command includes the robot's device identifier, communication parameter configuration, coordinate system calibration information, and other details. Upon receiving the registration command, the industrial robot returns registration confirmation information, including its current device status, joint positions, and load parameters. After verifying the registration information, the collaborative controller marks the industrial robot as "registered."

[0067] Secondly, the collaborative controller sends a registration command to the embodied intelligent robot via TCP communication. This registration command includes the robot's device identifier, communication parameter configuration, vision system calibration parameters, and posture sensor calibration information. Upon receiving the registration command, the embodied intelligent robot returns registration confirmation information, including its current device status, body posture data, and vision system readiness status. After verifying the registration information, the collaborative controller marks the embodied intelligent robot as "registered."

[0068] Meanwhile, the collaborative controller collects real-time operating status data of the press host via high-speed signal interaction, including key information such as the press's current operating mode, slider position, stroke count, and fault codes. Once the collaborative controller confirms that the press host is in normal condition, it marks it as "ready".

[0069] Once the industrial robot, the embodied intelligent robot, and the press machine have all completed registration and are in normal working order, the collaborative controller determines that system initialization is complete and enters standby mode, awaiting process task instructions from the upper level. Throughout the operation, the collaborative controller continuously collects the operating status of each device at fixed intervals to ensure real-time monitoring and rapid response.

[0070] In one embodiment of the present invention, based on step S1, the following will provide a possible embodiment and its specific implementation will be described in a non-limiting manner, such as... Figure 3 As shown.

[0071] S101. Pre-storing and Building of the Task Graph Template Library

[0072] In the storage unit of the collaborative controller or its host computer system, parameterized task graph templates corresponding to various stamping process tasks are pre-built and stored. Each task graph template is organized in a directed acyclic graph data structure to describe the complete stamping process flow for a specific workpiece type.

[0073] Specifically, the task graph template contains the following core elements:

[0074] Template Identifier: A unique ID that identifies the task template and forms a one-to-one correspondence with the workpiece type;

[0075] A set of supernodes consists of multiple supernodes, each representing an independent subtask stage;

[0076] Edge set: Defines the temporal dependencies and spatial mutual exclusion constraints between supernodes;

[0077] Parameterized interface: Allows adjustment of execution parameters inside supernodes according to actual production conditions.

[0078] In this embodiment, the task diagram template has four core supernodes: the embodied intelligent robot posture correction sub-task node, the industrial robot material handling sub-task node, the press main unit stamping sub-task node, and the embodied intelligent robot unloading and return sub-task node. These four supernodes form a complete task chain according to the stamping process.

[0079] S102. Online generation and instantiation of task graphs

[0080] When the collaborative controller receives a stamping process task from the upper-level system (such as an MES system or human-machine interface), it executes the online parsing and instantiation process of the task diagram.

[0081] First, the collaborative controller parses the issued process task and extracts the current process task identifier, which is used to indicate the type of workpiece to be processed.

[0082] Secondly, the collaborative controller queries the current readiness status of each device, including the load capacity of the industrial robot, the vision system calibration status of the embodied intelligent robot, and the mold installation status of the press host, to obtain the device readiness status parameters.

[0083] Subsequently, the collaborative controller selects a matching parameterized task diagram template from the pre-stored task diagram template library based on the current process task identifier and equipment readiness status. If there are differences between the equipment readiness status and the template's default parameters, the collaborative controller adaptively adjusts the template parameters to generate an instantiated task diagram structure for the current batch of tasks.

[0084] S103. Internal structure encapsulation of supernodes

[0085] In the instantiated task graph structure, the cooperative controller treats four subtasks as supernodes in a directed acyclic graph, with each supernode encapsulating a fine-grained sequence of atomic operations. This hierarchical modeling approach decouples macroscopic process steps from microscopic equipment actions, providing a foundation for subsequent flexible planning.

[0086] Specifically, the atomic operation sequences encapsulated within each supernode include, but are not limited to, the following basic action units:

[0087] (1) Internal encapsulation of the super node of the posture correction subtask of the embodied intelligent robot: Atomic operation: move to the alignment position; Atomic operation: perform visual correction; Atomic operation: adjust the end pose to the target pose.

[0088] (2) Industrial robot material picking and placing sub-task super node internal encapsulation: Atomic operation: move to material picking standby position; Atomic operation: enter mold cavity material picking area; Atomic operation: perform workpiece gripping; Atomic operation: exit mold cavity area; Atomic operation: move to material placing standby position; Atomic operation: perform workpiece placement.

[0089] (3) Press main machine stamping sub-task super node internal encapsulation: Atomic operation: slide down to start; Atomic operation: contact workpiece; Atomic operation: pressure holding and forming; Atomic operation: slide return; Atomic operation: stop to top dead center.

[0090] (4) Internal encapsulation of the super node of the embodied intelligent robot unloading and return subtask: Atomic operation: move to the pick-up position; Atomic operation: perform workpiece grabbing; Atomic operation: move to the return conveyor position; Atomic operation: perform workpiece placement.

[0091] Through the above encapsulation method, each supernode becomes a functional module that can be independently planned, replaced, and parameterized.

[0092] S104. Definition of Constraint Relationships Between Supernodes

[0093] In the instantiated task graph structure, directed edges between supernodes are used to define logical constraints between different subtasks. This embodiment defines two types of core constraints:

[0094] (1) Temporal dependency:

[0095] A sequential dependency is established between the posture correction sub-node of the embodied intelligent robot and the material handling sub-node of the industrial robot. Specifically, a directed edge points from the posture correction sub-node to the material handling sub-node, indicating that the material handling sub-node can only begin execution after the posture correction sub-node has successfully completed. This constraint ensures that the workpiece has completed posture correction before being grasped by the industrial robot, avoiding grasping failure due to workpiece pose deviation.

[0096] (2) Spatial mutual exclusion relationship:

[0097] A spatial mutual exclusion relationship is established between the industrial robot's material handling sub-node and the press's stamping sub-node. Specifically, the directed edges corresponding to these two supernodes have a spatial mutual exclusion attribute label, indicating that neither can occupy the mold cavity area simultaneously. In subsequent timing planning, the collaborative controller will ensure that the action windows of these two subtasks are completely staggered in time, or set a spatiotemporal no-go zone for the mold cavity area, prohibiting either device from entering while the other is occupying it.

[0098] In addition, a spatial mutual exclusion relationship can be established between the industrial robot's material handling sub-node and the embodied intelligent robot's material unloading and return sub-node to ensure that the two will not interfere with each other in the intersection area.

[0099] S105. Application and Extension of Task Diagrams

[0100] The instantiated task graph structure serves as the basic input for subsequent timing planning and instruction scheduling. The spatiotemporal planner reads the sequence of supernodes and constraints in the task graph, combines them with a unified global time base, divides the action window for each supernode, and defines the corresponding spatiotemporal no-go zones based on spatial mutual exclusion relationships.

[0101] When production conditions change, the collaborative controller can quickly adapt to the production process of new workpieces without modifying the underlying program, by switching task diagram templates or adjusting the atomic operation sequences within supernodes. For example, when switching to sheet metal workpieces of different thicknesses, simply selecting the corresponding task diagram template will automatically adapt the holding time parameters of the press stamping sub-nodes and the gripping force parameters of the industrial robot.

[0102] In one embodiment of the present invention, based on step S2, the following will provide a possible embodiment and describe its specific implementation in a non-limiting manner.

[0103] S201. On the global unified timeline, based on the action window of the press subtask, divide the action windows for each subtask, such as... Figure 4 As shown:

[0104] (1) Establishment of a unified time standard

[0105] During the system initialization phase, the collaborative controller establishes a unified global time base. In this embodiment, the collaborative controller uses a high-precision clock synchronization protocol (such as the IEEE 1588 precision time protocol) to synchronize the local clocks of the press host, industrial robot, and embodied intelligent robot, achieving microsecond-level time synchronization accuracy. In another preferred embodiment, the collaborative controller uses the press host's slider stroke pulse signal as a unified hardware time base, defining the moment when the press slider reaches the top dead center as the start time T=0 of each production cycle. By establishing a unified global time base, a prerequisite is provided for subsequent accurate timing planning.

[0106] (2) Fixing the press subtask action window

[0107] The collaborative controller first obtains the process parameters of the press sub-task, including the slide descent time, holding time, and return time. Based on these parameters, the collaborative controller fixes the action window [T_press_start, T_press_end] of the press sub-task on the global time axis.

[0108] Taking a typical stamping process as an example, the action window configuration for the press subtask is as follows: the slide begins its descent at T=0, reaches the bottom dead center after 300ms, then enters a 200ms holding pressure phase, and finally returns to the top dead center after 300ms. Therefore, the action window for the press subtask is fixed at [T_press_start=0ms, T_press_end=800ms]. This window remains unchanged in each production cycle, serving as the benchmark for the entire timing plan.

[0109] (3) Division of sub-task action windows for industrial robots

[0110] The collaborative controller divides the action window for the industrial robot's sub-tasks, and this window is used to perform material handling operations. Considering that the industrial robot and the press share the same mold cavity area, and there is a spatial mutual exclusion relationship between the two, the action window of the industrial robot must avoid the time period when the press occupies the mold cavity.

[0111] The specific division rules are as follows: The start time of the action window for the industrial robot subtask must be later than the end time of the action window for the press subtask, ensuring that the press slide has fully returned to the top dead center and the mold cavity area is safely open; at the same time, the end time of this action window must be earlier than the start time of the action window for the press subtask in the next production cycle, ensuring that the industrial robot has completely exited the mold cavity area before the press descends again. In addition, a safety margin T_safe is reserved at both ends of the window to absorb time fluctuations during the action execution process.

[0112] Using the press window example above, the press window ends at T=800ms, and the next cycle press window starts at T=1000ms (assuming a production cycle time of 1000ms). Assuming a safety margin of T_safe=50ms, the action window for the industrial robot subtask can be divided into [T_robot_start=850ms, T_robot_end=950ms]. This window lasts 100ms and is used to complete the entire sequence of actions: picking up material from the mold cavity, transferring it to the unloading position, placing the workpiece, and returning to the standby position.

[0113] Before dividing the window, the collaborative controller presets the kinematic model of the industrial robot to estimate the shortest time required to complete the picking and placing actions. If the estimated time exceeds the available window duration, the collaborative controller will issue an alarm, requesting adjustments to process parameters or optimization of the robot path.

[0114] (4) Division of sub-task action windows of embodied intelligent robots

[0115] The collaborative controller divides action windows for the embodied intelligent robot's sub-tasks. These windows are used to perform attitude correction and material return operations. The working area of ​​the embodied intelligent robot overlaps with that of the industrial robot (such as the workpiece handover area), and the two are mutually exclusive in space. Therefore, the action windows of the embodied intelligent robot must be staggered with those of the industrial robot in time.

[0116] The specific division rules are as follows: The collaborative controller divides the action window of the embodied intelligent robot into two independent windows, corresponding to the attitude correction sub-task and the unloading and return sub-task, respectively. The attitude correction sub-task window is arranged before the industrial robot action window and is used to complete the attitude correction of the workpiece before the industrial robot picks up the material; the unloading and return sub-task window is arranged after the industrial robot action window and is used to transfer the finished part to the return line after the industrial robot finishes unloading the material.

[0117] Using the window example above, the embodied intelligent robot posture correction subtask window can be divided into [T_align_start=700ms, T_align_end=800ms]. This window overlaps with the latter half of the press window. However, since the embodied intelligent robot's working area is located at the correction station outside the mold cavity, there is no spatial conflict with the press mold cavity, thus allowing time overlap with the press window. The unloading and reflow subtask window can be divided into [T_transfer_start=950ms, T_transfer_end=1050ms]. This window starts at the end of the industrial robot window and partially overlaps with the press window of the next cycle. Similarly, due to the separation of spatial areas, there is no conflict.

[0118] It should be noted that if the estimated execution time of the attitude correction subtask or the unloading and reflow subtask exceeds the available window duration, the collaborative controller will perform conflict detection and window fine-tuning. For example, if attitude correction requires 150ms, but the current available window is only 100ms, the collaborative controller may consider starting the attitude correction window earlier, or adjusting the start and end times of the industrial robot window to free up more time resources.

[0119] (5) Validation and output of window division

[0120] After completing the action window division for all subtasks, the collaborative controller performs consistency verification on the division results to ensure that all windows meet the following conditions: the press window and the industrial robot window meet the safety margin requirements; the industrial robot window and the embodied intelligent robot window are completely staggered in time; and the duration of each window is not less than the minimum execution time of the corresponding subtask.

[0121] After successful verification, the collaborative controller outputs the action windows of each subtask as planning results for subsequent instruction scheduling steps. Simultaneously, the collaborative controller defines corresponding spatiotemporal restricted areas based on the time range of each window. For example, during the press window, the mold cavity area is marked as a restricted area, and during the industrial robot window, its motion path area is marked as a restricted area, providing a basis for safety monitoring.

[0122] S202. Define spatiotemporal restricted zones based on the action windows of each subtask:

[0123] (1) Definition of the spatiotemporal forbidden zone of the press subtask

[0124] The collaborative controller defines a first spatiotemporal restricted area for the press subtask. This restricted area covers the press's working area, specifically including the press cavity area, the slide motion envelope area, and the mold closing area between the upper and lower dies. The time frame of this restricted area completely overlaps with the press subtask's action window, i.e., from the moment the press slide begins its descent until the slide returns to its top dead center.

[0125] During the effective period of the first restricted space zone, the collaborative controller prohibits any equipment other than the press from entering the space. This means that neither industrial robots nor embodied intelligent robots may enter the mold cavity area during this period. When the real-time position feedback of any robot indicates that it has entered the restricted zone, the safety rule engine immediately triggers a safety response.

[0126] (2) Definition of the spatiotemporal forbidden zone of the first robot subtask

[0127] The collaborative controller defines a second spatiotemporal no-go zone for the first robot (industrial robot) subtask. This no-go zone covers the area traversed by the industrial robot's motion path, specifically including the material handling path, the material placement path, and the entire envelope space during workpiece transfer. The temporal scope of this no-go zone completely overlaps with the action window of the industrial robot subtask.

[0128] During the effective period of the second spatial-temporal restricted area, the collaborative controller prohibits any device other than industrial robots from entering this space. In particular, the working area of ​​the embodied intelligent robot overlaps with that of the industrial robot; therefore, during this time period, the embodied intelligent robot must avoid the movement path area of ​​the industrial robot.

[0129] (3) Handling spatial conflicts in time-overlapping scenarios

[0130] When the action windows of different subtasks overlap in time, spatial conflicts may occur in the corresponding spatiotemporal restricted areas. This embodiment handles this in the following two ways:

[0131] 1) Spatial Non-Overlapping Configuration: During the timing planning phase, the collaborative controller adjusts the start and end times of the action windows or optimizes the robot's motion path to ensure that the spatiotemporal exclusion zones of each device are completely separated in space within overlapping time windows. For example, the press sub-task window and the embodied intelligent robot posture correction window overlap in time, but since the press's working area is located inside the mold cavity, while the embodied intelligent robot's correction station is located in an independent area outside the mold cavity, their spatial areas do not overlap, thus avoiding conflict.

[0132] 2) Priority Rule Configuration: When the spatial areas corresponding to overlapping time windows cannot be completely separated, the collaborative controller sets priority rules for the overlapping areas to determine which devices are allowed to enter. For example, the industrial robot's material handling window and the embodied intelligent robot's material return window may partially overlap in time, and they may also overlap in the workpiece handover area. In this case, the collaborative controller sets the priority of the industrial robot to be higher than that of the embodied intelligent robot. During the overlapping time period, the industrial robot occupies the handover area first, and the embodied intelligent robot must wait for the industrial robot to exit before it can enter.

[0133] (4) Real-time monitoring and application of spatiotemporal restricted areas

[0134] The security rule engine of the collaborative controller continuously monitors the global timeline and the real-time feedback positions of each device. When the system time enters the time range of any spatiotemporal restricted area, the security rule engine checks whether the actual location of the corresponding device is within the spatial range of the restricted area. If unauthorized intrusion is detected, real-time interlocking is immediately triggered, and a pause or emergency stop command is sent to the relevant device.

[0135] In one embodiment of the present invention, based on step S3, the following will provide a possible embodiment and describe its specific implementation in a non-limiting manner.

[0136] S301. Acquisition and Calibration of Communication Delay Parameters

[0137] During the system initialization phase, the co-controller performs the process of acquiring and calibrating communication delay parameters.

[0138] The collaborative controller sends test commands to the first robot (industrial robot) via the Modbus TCP communication interface, records the time of command issuance and the robot's response and return time, calculates the round-trip delay, and takes half of it to obtain the one-way communication delay parameter. After multiple tests, the average value is taken to obtain the first communication delay parameter D1. In this embodiment, the Modbus TCP communication of the industrial robot has high determinism, and D1 is stable within the range of 5ms to 10ms.

[0139] The collaborative controller performs the same latency test procedure with the second robot (the embodied intelligent robot) via a TCP communication interface. Because the communication protocol stack of the embodied intelligent robot is relatively complex and may involve additional overhead such as visual data processing, its communication latency is typically higher than that of the industrial robot. After testing and calibration, the second communication latency parameter D2 stabilizes within the range of 20ms to 50ms, and its value is significantly greater than that of the first communication latency parameter D1.

[0140] The collaborative controller will continuously monitor fluctuations in communication latency and establish a historical fluctuation database for subsequent prediction and adjustment.

[0141] S302. Calculation of Industrial Robot Command Issuance Time

[0142] The collaborative controller calculates the command issuance time for the first robot. The calculation formula is as follows:

[0143] T_issue_1=T_start_1-D1

[0144] Where T_start_1 is the start time of the first robot subtask action window, and D1 is the first communication delay parameter.

[0145] Taking a certain production cycle as an example, the start time of the industrial robot's subtask action window is T_start_1 = 850ms, the first communication delay parameter D1 = 8ms, and the instruction issuance time T_issue_1 = 842ms. At this moment, the collaborative controller issues the action instruction to the industrial robot via the Modbus TCP protocol. After receiving the instruction, the industrial robot undergoes approximately 8ms of communication transmission and instruction parsing, and begins executing the predetermined action exactly at T = 850ms.

[0146] S303. Calculation of the time for issuing commands to the embodied intelligent robot

[0147] The collaborative controller calculates the command issuance time for the second robot. The calculation formula is as follows:

[0148] T_issue_2=T_start_2-D2-ΔT_pred

[0149] Where T_start_2 is the start time of the second robot subtask action window, D2 is the second communication delay parameter, and ΔT_pred is the predicted adjustment amount dynamically determined based on the current system load or historical communication fluctuation data.

[0150] The predicted adjustment amount ΔT_pred is calculated as follows: The coordinating controller continuously monitors the real-time fluctuations of communication latency, uses a sliding window to calculate the standard deviation of latency σ_delay over the most recent N periods, and sets a safety factor α (usually between 1.5 and 2.0). Then, ΔT_pred = α × σ_delay. When the system load is high or the network environment is unstable, σ_delay increases, and ΔT_pred increases accordingly, providing a larger time buffer for command issuance.

[0151] Taking a certain production cycle as an example, the start time of the action window for the posture correction subtask of the embodied intelligent robot is T_start_2 = 700ms, the second communication delay parameter D2 = 35ms, the historical fluctuation standard deviation σ_delay = 8ms, and the safety factor α = 1.5. Therefore, ΔT_pred = 12ms, and the instruction issuance time T_issue_2 = 700 - 35 - 12 = 653ms. At this moment, the collaborative controller issues instructions to the embodied intelligent robot via TCP communication, reserving a buffer time of 47ms to cope with communication fluctuations.

[0152] S304. Calculation of Press Command Issuance Time

[0153] The collaborative controller calculates the command issuance time for the press. The press and collaborative controller communicate via high-speed signals, resulting in extremely low communication latency (typically less than 1ms). Therefore, the difference between the press's command issuance time and its action window start time is much smaller than the robot's lead time.

[0154] The calculation formula is as follows:

[0155] T_issue_press = T_start_press -D_press

[0156] Where T_start_press is the start time of the press subtask action window, and D_press is the press communication delay parameter (usually within 1ms). Because the D_press value is very small, the advance time of the press command issuance is significantly smaller than that of industrial robots and embodied intelligent robots.

[0157] Taking a certain production cycle as an example, if the start time of the press subtask action window is T_start_press=0ms and D_press=0.5ms, then the instruction issuance time is T_issue_press=-0.5ms, that is, the instruction is issued 0.5ms before the end of the previous cycle to ensure that the press starts precisely at T=0.

[0158] S305. Unified Execution of Differentiated Scheduling

[0159] Based on the above calculations, the communication scheduler of the collaborative controller generates a timestamped instruction queue. During system operation, the communication scheduler monitors the unified global clock, and when the system time reaches the instruction issuance time for each device, it sends the action instruction to the corresponding device through the corresponding communication interface.

[0160] Through the aforementioned differentiated scheduling mechanism, despite the significant differences in communication latency between the industrial robot and the embodied intelligent robot, they can still execute precisely and synchronously according to the planned action window. The industrial robot receives instructions 8ms in advance, while the embodied intelligent robot receives instructions approximately 47ms in advance. Both initiate their actions precisely at the start of the window, achieving high-precision collaboration between heterogeneous devices.

[0161] The collaborative controller continuously monitors the actual execution status of each device. If it finds that the action start deviation caused by communication delay fluctuations exceeds the threshold, it dynamically adjusts the predicted adjustment amount ΔT_pred so that the system can adapt to changes in the network environment.

[0162] In one embodiment of the present invention, based on step S4, the following will provide a possible embodiment and describe its specific implementation in a non-limiting manner.

[0163] Based on the calculated command issuance time of each device, the collaborative controller sends action commands to the press, the first robot, and the second robot through the corresponding communication interfaces.

[0164] Based on the aforementioned differentiated instruction scheduling steps, this embodiment further provides a preferred instruction issuance implementation method. Through predictive fine-tuning and instruction preloading mechanisms, it further improves the accuracy of instruction issuance time and the reliability of heterogeneous device collaboration.

[0165] 1. Online monitoring and prediction of communication delays

[0166] During the execution of instruction scheduling, the collaborative controller continuously monitors the real-time status of each communication link, providing a data foundation for predictive fine-tuning.

[0167] For the Modbus TCP communication link with the first robot (industrial robot), the collaborative controller records the command issuance time and response return time in each communication cycle, calculates the instantaneous round-trip delay, and converts it into a one-way delay. For the TCP communication link with the second robot (embodied intelligent robot), the same method is used to continuously monitor latency jitter and packet loss rate. All monitoring data is stored in a sliding window buffer for statistical analysis.

[0168] The collaborative controller uses filters to predict the communication delay for the next cycle online. In a preferred embodiment, a Kalman filter is used to filter and predict the delay data. Based on historical delay observations and the system state transition model, the Kalman filter recursively calculates the optimal estimate of the delay for the next cycle, D_cmd_estimated, and its standard deviation σ_delay. This filter can effectively suppress measurement noise and dynamically adapt to the slow changing trend of the delay.

[0169] 2. Fine-tuning of instruction issuance time based on prediction results

[0170] The collaborative controller fine-tunes the command issuance time of each device based on the predicted communication delay parameters. The calculation formula is as follows:

[0171] T_issue=T_scheduled_start-D_cmd_estimated-α×σ_delay

[0172] Where: T_issue is the final determined instruction issuance time; T_scheduled_start is the planned start time of the corresponding subtask action window; D_cmd_estimated is the average communication delay of the next cycle predicted by the filter; σ_delay is the predicted delay standard deviation; α is a pre-set safety factor, which is usually in the range of 1.5 to 2.5, used to provide buffer margin to cope with delay fluctuations.

[0173] Taking a command issued by an embodied intelligent robot as an example, assuming the scheduled start time of its attitude correction subtask is T_scheduled_start = 700ms, the mean communication delay predicted by the Kalman filter for the next cycle is D_cmd_estimated = 35ms, the prediction standard deviation is σ_delay = 6ms, and a safety factor α = 2.0, then T_issue = 700 - 35 - 12 = 653ms is calculated. Compared to the static compensation method, this predictive fine-tuning adds a buffer margin of 12ms, which can effectively absorb sudden delays caused by network congestion or system load fluctuations.

[0174] For industrial robots, since the Modbus TCP communication delay is relatively stable and the σ_delay value is small, the buffer margin for predictive fine-tuning is correspondingly small, but it can still effectively compensate for periodic fluctuations.

[0175] 3. Implementation of the instruction preloading mechanism

[0176] In view of the complex command parsing and limited real-time processing capabilities of the second robot (embodied intelligent robot), this embodiment adopts an instruction preloading mechanism to further reduce the latency and uncertainty of real-time command issuance.

[0177] During system initialization or task switching phases, the collaborative controller pre-loads the complete sequence of parameterized instructions required by the embodied intelligent robot into its local cache. Taking the posture correction subtask as an example, the pre-loaded instruction sequence includes: target coordinates and velocity parameters for moving to the alignment position, trigger conditions and image processing parameters for visual correction, and target pose and motion planning parameters for end-effector posture adjustment. These instructions are encapsulated in the form of structured data packets and stored in the embodied intelligent robot's local memory.

[0178] During formal production operation, when the system time reaches the instruction issuance time T_issue, the collaborative controller only needs to send a brief trigger signal to the embodied intelligent robot. This signal contains the task identifier and timestamp information. Upon receiving the trigger signal, the embodied intelligent robot immediately retrieves the corresponding parameterized instruction sequence from its local cache and begins execution, without waiting for the complete instruction data transmission and parsing process.

[0179] Through the instruction preloading mechanism, the latency of real-time communication is reduced from the latency of transmitting the entire instruction to the latency of transmitting only the trigger signal. The latter typically involves only a few bytes of data, and the transmission time can be controlled within 1ms. This significantly reduces the latency uncertainty of embodied intelligent robots issuing instructions, enabling them to achieve more precise synchronization with industrial robots.

[0180] Based on the above mechanisms, the communication scheduler of the collaborative controller executes instruction issuance according to the following process:

[0181] The first step is to calculate the refined instruction issuance time for each device based on the filtered prediction results before the start of each production cycle.

[0182] The second step is to preload the complete parameterized instruction sequence into the local cache of the embodied intelligent robot during system idle periods or task switching.

[0183] The third step is that when the system clock reaches the calculated instruction issuance time, the communication scheduler issues instructions through the corresponding communication interface: for industrial robots and presses, complete action instructions are issued; for embodied intelligent robots, brief trigger signals are issued.

[0184] The fourth step involves the communication scheduler continuously monitoring the device response status after the command is issued, and feeding back the measured communication delay data to the filter for prediction and optimization in the next cycle.

[0185] Through the above preferred implementation method, this embodiment further introduces predictive fine-tuning and instruction preloading mechanisms on the basis of differentiated scheduling, which significantly improves the accuracy of instruction issuance time and the reliability of heterogeneous device collaboration. It is especially suitable for application scenarios where the communication environment fluctuates or the instruction processing of embodied intelligent robots is relatively complex.

[0186] In one embodiment of the present invention, based on step S5, a possible embodiment will be given below, and its specific implementation will be described in a non-limiting manner.

[0187] The collaborative controller continuously receives real-time feedback from the press, the first robot, and the second robot, using a unified time base. This feedback includes at least the device's current position, current action execution progress, and fault flags. Within each control cycle, the safety rule engine compares the current global time with the time range of each spatiotemporal restricted zone. If the current time falls within the time period of a certain spatiotemporal restricted zone, the real-time position of the corresponding device is obtained, and it is determined whether that position falls within the spatial range of that restricted zone. If any device is detected entering a prohibited area during a prohibited time period, or if a device that should have left remains in the restricted zone, a safety response is immediately triggered. A pause or emergency stop command is sent to all relevant devices, and the violation is recorded for subsequent analysis. The safety response has a higher priority than all regular scheduling commands.

[0188] Based on the above embodiments, in order to further improve the reliability of the multi-device collaborative control provided by the above embodiments, a deviation correction method is provided, including:

[0189] 1. Deviation monitoring and calculation

[0190] At the end of each production cycle, the collaborative controller executes the deviation monitoring and calculation process.

[0191] Calculation of timing deviation Δt: The collaborative controller records the actual completion time of each subtask and compares it with the theoretically planned completion time. Taking the material handling subtask of an industrial robot as an example, if the theoretically planned completion time is T=950ms and the actual completion time is T=972ms, then the timing deviation Δt=+22ms, indicating that the execution of this subtask was delayed by 22ms.

[0192] Calculation of position deviation ΔP: The collaborative controller obtains the actual position of the industrial robot end effector or the hand of the embodied intelligent robot and compares it with the theoretical target position. For example, if the actual position of the industrial robot end effector at the material unloading point deviates from the theoretical target position by 3mm in the X direction and 2mm in the Y direction, the position deviation ΔP≈3.6mm is obtained by calculating the spatial distance difference.

[0193] Calculation of load deviation ΔF: The collaborative controller obtains the actual load value through the force / torque sensors built into the robot body or end effector and compares it with the nominal load value. For example, if the nominal load is 5kg and the actual load is 5.3kg, then the load deviation ΔF = +0.3kg. This deviation is used to monitor load changes caused by fluctuations in workpiece weight or changes in gripping posture, and can serve as an input parameter for anomaly warning.

[0194] 2. Local rapid compensation

[0195] The collaborative controller queries a preset deviation-compensation mapping table based on the calculated timing deviation Δt and position deviation ΔP to obtain the speed compensation coefficient Kv of the industrial robot.

[0196] The deviation-compensation mapping table is a two-dimensional lookup table, using timing deviation Δt and position deviation ΔP as input indices and speed compensation coefficient Kv as the output value. This mapping table is pre-established during the production line debugging phase: various typical deviation scenarios are manually simulated (e.g., a 20ms increase in communication delay, a 5mm attitude deviation, etc.), and the industrial robot speed is manually fine-tuned until the optimal compensation parameters that restore stable cycle time are found. The corresponding relationships are then filled into the table. The speed compensation coefficient Kv is typically greater than or equal to 1, indicating speed-up compensation.

[0197] Assuming the query yields Kv=1.08, the collaborative controller calculates the industrial robot's movement speed for the next production cycle using the following formula: V_new = V_base×1.08, where V_base is the base speed. By increasing the robot's movement speed, local motion delays caused by timing and positional deviations are compensated for, allowing it to catch up on lost time.

[0198] 3. Global cycle time closed-loop control

[0199] The collaborative controller simultaneously executes global cycle time closed-loop control. It measures the actual cycle length T_actual between the start times of two adjacent production cycles and calculates the cycle error e = T_actual - T_target, where T_target is the target cycle length.

[0200] The periodic error e is input to the PID controller, which calculates the output regulation ΔT_hold according to the proportional-integral-derivative control law. The transfer function of the PID controller is expressed as:

[0201] ΔT_hold = K_p·e + K_i·∫e dt + K_d·(de / dt)

[0202] Wherein, K_p, Ki, and K_d are the proportional coefficient, integral coefficient, and derivative coefficient, respectively, which are pre-tuned according to the system response characteristics.

[0203] The collaborative controller calculates the press holding time for the next production cycle using the following formula:

[0204] T_hold_new = T_hold_base + ΔT_hold

[0205] When the actual cycle time is longer than the target cycle time (e>0), it indicates that the production line is too slow. The PID controller outputs a negative value ΔT_hold, shortening the press holding time to allow more time for the robots before and after, thus accelerating the overall cycle time. When the actual cycle time is shorter than the target cycle time (e<0), the PID controller outputs a positive value ΔT_hold, appropriately extending the holding time to avoid system instability due to excessive acceleration.

[0206] 4. Iterative Compensation Mechanism

[0207] The aforementioned local rapid compensation and global cycle time closed-loop control are executed cyclically in each production cycle. Local compensation adjusts the speed of the industrial robot to quickly respond to timing and positional deviations; global compensation adjusts the press holding time to stabilize the overall production cycle time. The two work together to dynamically converge and stabilize the actual cycle time near the target cycle time, achieving adaptive compensation for various disturbances.

[0208] This embodiment adopts a simple compensation mechanism that combines table lookup and PID control. It has low computational load, is intuitive to debug, and can effectively absorb dynamic disturbances such as equipment performance fluctuations and workpiece batch differences, thus maintaining the stability of production cycle.

[0209] Although the present invention has been described in detail with reference to the accompanying drawings and preferred embodiments, the present invention is not limited thereto. Various equivalent modifications or substitutions can be made to the embodiments of the present invention by those skilled in the art without departing from the spirit and essence of the invention, and such modifications or substitutions should all be within the scope of the present invention. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should also be covered within the protection scope of the present invention.

Claims

1. A multi-device collaborative control method for a press, characterized in that, The method, applied to a collaborative controller that is communicatively connected to a press, a first robot, and a second robot, includes: The process task is broken down into multiple sub-tasks that are executed sequentially, including at least the first robot sub-task, the second robot sub-task, and the press sub-task; On a globally unified timeline, based on the action window of the press subtask, action windows are divided for each subtask, and spatiotemporal no-go zones are defined based on the action windows of each subtask. The spatiotemporal no-go zones include: prohibiting the first robot and the second robot from entering the working area of ​​the press during the time period covered by the action window of the press subtask; and prohibiting the second robot from entering the movement path area of ​​the first robot during the time period covered by the action window of the first robot subtask. Obtain the communication delay parameters between the first robot and the second robot, and calculate the instruction issuance time for the press, the first robot, and the second robot based on the communication delay parameters and each action window. The issuance time is earlier than the start time of the corresponding action window. Send corresponding instructions to the press, the first robot, and the second robot according to the dispatch time; Based on the real-time feedback status of the press, the first robot, and the second robot, and the current time reference, determine whether there is any behavior that violates the time and space restricted zone; if so, trigger a safety response. On the global unified timeline, based on the action window of the press subtask, action windows are divided for each subtask, including: Fix the action window of the press subtask on the global timeline; The first robot subtask is divided into action windows, with its start time later than the end time of the press subtask action window and its end time earlier than the start time of the press subtask action window in the next cycle, while reserving a safety margin. Divide the action window for the second robot subtask so that it is staggered from the action window of the first robot subtask in time; Spatiotemporal restricted areas are defined based on the action windows of each subtask, including: Define a first spatiotemporal restricted area for the press subtask. The spatial range of the first spatiotemporal restricted area is the working area of ​​the press, and the time range is the time period covered by the action window of the press subtask. Define a second spatiotemporal no-go zone for the first robot subtask. The spatial range of the second spatiotemporal no-go zone is the area covered by the movement path of the first robot, and the temporal range is the time period covered by the action window of the first robot subtask. The first time-space restricted area is used to prohibit any equipment other than the compressor from entering the space range within the time range; The second spacetime restricted area is used to prohibit any device other than the first robot from entering the space area within the time range; When the action windows of different subtasks overlap in time, the corresponding spatiotemporal restricted areas are configured to not overlap in space, or priority rules are set in the overlapping spatial areas to determine the devices allowed to enter.

2. The method according to claim 1, characterized in that, The collaborative controller integrates Modbus TCP communication interface and TCP communication interface to build a unified communication architecture. It establishes a Modbus TCP point-to-point communication connection with the first robot, a TCP local area network communication connection with the second robot, and a high-speed signal interaction connection with the press host, and adopts a unified clock synchronization mechanism. The first robot is an industrial robot, and the second robot is an embodied intelligent robot.

3. The method according to claim 2, characterized in that, After the collaborative controller is powered on and initialized, it sends registration instructions to the industrial robot via Modbus TCP communication and to the embodied intelligent robot via TCP communication, and collects the operating status of the press host in real time.

4. The method according to claim 1, characterized in that, The process task is broken down into multiple sub-tasks that are executed sequentially, including: Pre-store parameterized task diagram templates corresponding to various stamping process tasks; When a stamping process task is received from the upper-level system, the corresponding parameterized task diagram template is selected from the template library according to the current process task identifier and equipment readiness status, and a task diagram structure containing the logical relationships between each subtask is generated. The subtasks include at least the posture correction subtask of the embodied intelligent robot, the material handling subtask of the industrial robot, the stamping subtask of the press host, and the unloading and return subtask of the embodied intelligent robot. The collaborative controller treats the four subtasks as supernodes in a directed acyclic graph. Each supernode encapsulates a fine-grained sequence of atomic operations. The atomic operations include at least the basic action units of moving to the alignment position, performing visual correction, grasping the workpiece, entering the mold cavity, and the press descending. The edges between supernodes in the directed acyclic graph are used to define the temporal dependencies and spatial mutual exclusion constraints between different subtasks; among them, the posture correction subnode of the embodied intelligent robot and the material handling subnode of the industrial robot have a sequential dependency relationship, and the material handling subnode of the industrial robot and the stamping subnode of the press host have a spatial mutual exclusion relationship.

5. The method according to claim 1, characterized in that, Obtain the communication delay parameters between the first robot and the second robot, and calculate the command issuance time for the press, the first robot, and the second robot based on the communication delay parameters and each action window, including: Obtain a first communication delay parameter for the first robot and a second communication delay parameter for the second robot, wherein the value of the second communication delay parameter is greater than the value of the first communication delay parameter; For the first robot, its instruction issuance time is calculated based on the difference between the start time of its corresponding first action window and the first communication delay parameter. For the second robot, its instruction issuance time is calculated based on the difference between the start time of its corresponding second action window and the second communication delay parameter and the predicted adjustment amount, wherein the predicted adjustment amount is dynamically determined based on the current system load or historical communication fluctuation data. For the press, the instruction issuance time is calculated based on the start time of its corresponding press subtask action window. The time difference between the instruction issuance time of the press and the start time of the press subtask action window is less than the instruction issuance time advance of the first robot and the second robot. The calculated instruction issuance times are all earlier than the start time of the corresponding subtask action window, and the instruction issuance time advance of the second robot is greater than the instruction issuance time advance of the first robot.

6. The method according to claim 1, characterized in that, Before issuing corresponding instructions to the press, the first robot, and the second robot according to the issuance time, the method further includes: Continuously monitor the latency jitter and packet loss rate of each communication link, and use filters to predict the communication latency for the next cycle; The instruction issuance time for this round is calculated based on the predicted mean and standard deviation of communication delay. The calculation formula is T_issue=T_scheduled_start-D_cmd_estimated-α×σ_delay, where D_cmd_estimated is the predicted mean delay, σ_delay is the standard deviation of delay, α is the safety factor, and T_scheduled_start is the instruction issuance time. For the second robot, an instruction preloading mechanism is adopted, which pre-delivers a set of parameterized instruction sequences to its local cache, and only sends a trigger signal at the time of execution.

7. The method according to claim 1, characterized in that, The method further includes: After each production cycle, the timing deviation, position deviation, and load deviation are calculated; wherein, the timing deviation is the difference between the actual completion time of the subtask and the theoretically planned completion time, the position deviation is the difference between the actual position of the robot end effector and the theoretical target position, and the load deviation is the difference between the actual load of the robot end effector and the nominal load. Based on the timing deviation and the position deviation, a preset deviation-compensation mapping table is queried to obtain the speed compensation coefficient of the first robot, and the movement speed of the first robot in the next production cycle is adjusted according to the speed compensation coefficient. The actual cycle duration between the start times of two adjacent production cycles is measured, the cycle error between the actual cycle duration and the target cycle duration is calculated, the cycle error is input into the PID controller to obtain the dynamic adjustment amount of the press holding time, and the press holding time in the next production cycle is adjusted according to the dynamic adjustment amount to adjust the actual duration of the action window of the press subtask. The deviation-compensation mapping table is pre-established through process testing and is used to map timing deviations and position deviations to corresponding speed compensation coefficients.

8. The method according to claim 7, characterized in that, The deviation-compensation mapping table is a two-dimensional lookup table, with timing deviation and position deviation as input indices and the speed compensation coefficient of the first robot as the output value; The speed compensation coefficient is greater than or equal to 1.

Images