Cooperative control method for industrial robot stacking

Abstract

The present application belongs to the technical field of industrial robot control, and particularly relates to a cooperative control method for industrial robots to stack, which comprises: writing a nominal occupation table for each robot according to a trajectory calculated swept voxel set; multiplexing a current sampling channel in a robot joint servo driver to extract a step component containing load information, and then inversely calculating an equivalent mass of a gear; calculating an actual stable duration, and combining a reference stable duration to calculate an inflation increment, and adding the inflation increment to an occupation time interval of a corresponding voxel in a space-time occupation grid to reconstruct a dynamic occupation envelope; and judging whether there is a time slice overlap of dynamic envelopes of adjacent robots on a shared voxel through a forward-looking scan of a multi-machine dynamic occupation total table, and if there is an overlap, issuing an entry delay signal to a robot that enters later to perform a rolling handshake arbitration scheduling. The present application adjusts a conservative margin from a static constant to an accurate inflation amount matching real-time load, and eliminates the restriction of excessive conservatism on a production line beat.

Description

Technical Field

[0001] This invention belongs to the field of industrial robot control technology, specifically relating to a collaborative control method for industrial robot palletizing. Background Technology

[0002] The palletizing stage of precision forged gears is the final process in gear manufacturing production lines. Its characteristics include relatively large individual product weight, regular shape, and multiple specifications mixed together. This requires multiple industrial robots to work collaboratively between the compact unloading table and rack to match the cycle time of the preceding forging line. In such multi-machine collaborative palletizing scenarios, achieving collision-free, high-cycle operation of multiple robotic arms within a limited physical space is a key technical challenge to ensuring the overall efficiency of the production line.

[0003] For this multi-robot collaborative palletizing scenario, existing technologies generally adopt a time-space scheduling scheme based on pre-planned trajectories. This involves generating fixed grasping, transporting, and placement trajectories for each robot through offline teaching or geometric path planning. Then, the timing of different robotic arms entering the shared space is controlled by time slice allocation in the shared work area or interlocking signals from the programmable logic controller. At the same time, a collision detection algorithm based on joint position feedback is combined to conduct pre-emptive checks and in-process monitoring of trajectory overlap at the geometric level, thereby achieving the basic safety baseline for multi-robot collaboration.

[0004] However, due to the objective working condition of mixed production of different specifications of precision forged gears on the same production line, the weight of the gear grasped by the robotic arm's end effector fluctuates significantly between 2kg and 18kg. This causes the decay time of the residual elastic oscillation of the end flange due to the unloading of the large inertial load after the robotic arm completes the placement action to change non-linearly. The actual time occupied in the shared space far exceeds the nominal cycle time given by the pre-planned trajectory. The existing solution based on fixed timing scheduling is completely unaware of the dynamic occupancy extension related to this load. Consequently, when adjacent robotic arms enter the shared space according to the nominal cycle time, the end effector of the previous robotic arm is still swinging within the visible amplitude, directly causing slight collisions or triggering an emergency stop of the safety light curtain. This forces the implementation of the project to leave a large static conservative margin in the time slice, slowing down the overall cycle time and significantly restricting the production line's fulfillment rate. Summary of the Invention

[0005] This invention provides a collaborative control method for industrial robot palletizing to solve the technical problem that in multi-robot collaborative palletizing, the residual oscillation at the end caused by gear variable load causes a nonlinear extension of the actual occupancy time of the shared space, while the fixed timing schedule is unaware of this extension, thus creating a dilemma between emergency stop due to collision and excessive conservatism.

[0006] This invention provides a collaborative control method for industrial robot palletizing, comprising the following steps: S1, discretize the overlapping areas within the reach of each robotic arm into voxels, construct a spatiotemporal occupancy grid for the shared workspace, and write the set of voxels swept by each robot based on its trajectory into a nominal occupancy table; S2, when the robotic arm completes the grasping action, the current sampling channel in the robot joint servo driver is reused to extract the step component containing load information, and the net load torque is separated by combining the torque contribution of the robotic arm itself, and then the equivalent mass of the gear is deduced. The deduction of the equivalent mass of the gear includes: obtaining the gravitational lever arm and gravitational acceleration of the end relative to the wrist joint under the current configuration, dividing the net load torque by the product of the gravitational lever arm and gravitational acceleration, and deducing the equivalent mass of the gear. S3. Based on the equivalent mass of the gear and the equivalent mass of the robotic arm's end effector wrist, calculate the actual stabilization time required for the residual displacement at the end effector to converge within the shelf positioning tolerance zone after the gear is placed. Combine this with the baseline stabilization time to calculate the expansion increment. Add the expansion increment to the occupancy time interval of the corresponding voxel in the spatiotemporal occupancy grid to reconstruct the dynamic occupancy envelope. The calculation method for the actual stabilization time is as follows: obtain the baseline stabilization time of the robot under no-load conditions; calculate the ratio of the equivalent mass of the gear to the equivalent mass of the robotic arm's end effector wrist, add 1 to the ratio, and then perform a square root operation; multiply the result of the square root operation by the baseline stabilization time to obtain the actual stabilization time required for the residual displacement at the end effector to converge within the shelf positioning tolerance zone after the gear is placed; subtract the baseline stabilization time from the actual stabilization time to obtain the expansion increment. S4 reports the dynamic occupancy envelope to the collaborative scheduling node. By looking ahead and scanning the multi-machine dynamic occupancy table, it determines whether the dynamic envelopes of adjacent robots overlap on the shared voxel. If there is an overlap, it sends an entry delay signal to the robot that entered later in order to perform rolling handshake arbitration scheduling.

[0007] Its effect is as follows: by inverting the actual mass of the gear through the multiplexed current sampling channel at the servo bottom layer, and deriving an improved relational formula for predicting the stable duration of the end residual oscillation in a simplified proportional form based on vibration dynamics, the expansion increment generated by this duration is added to the spatiotemporal occupancy grid of the shared workspace in real time. Through the rolling window-type handshake driving multi-machine collaborative scheduling, the dilemma caused by the nonlinear extension of the actual occupancy duration of the shared space due to the end residual oscillation caused by the gear variable load, and the fact that the fixed timing scheduling is completely unaware of this extension, is eliminated.

[0008] Furthermore, the overlapping areas accessible by each robotic arm are discretized using voxels to construct a spatiotemporal occupancy grid for the shared workspace, including: Based on the robot kinematics model and the layout of the work units, the overlapping areas accessible by each robotic arm are abstracted into a shared workspace. The shared workspace is discretized into voxels along three directions to form a three-dimensional voxel grid. The spatial coordinates of each voxel and the time interval in which it is occupied are recorded, and finally the spatiotemporal occupancy grid of the shared workspace is obtained.

[0009] Its effects are as follows: By defining the construction method of the spatiotemporal occupancy grid, first determining the shared workspace of multiple robotic arms overlapping, then performing three-dimensional voxel discretization and recording the coordinates and occupancy time interval of each voxel; it provides a standardized three-dimensional spatiotemporal data base for subsequent robot trajectory occupancy recording, dynamic expansion of time intervals, and conflict discrimination, realizing the structured expression of the space-time information of the shared workspace, and supporting the realization of subsequent dynamic correction, interference judgment and other functions.

[0010] Furthermore, the set of voxels scanned by each robot based on its trajectory is written into a nominal occupancy table, including: For each robot, based on the grasping trajectory, transport trajectory, and placement trajectory, calculate the set of voxels swept by the geometric envelope of the robotic arm body and end effector at each time slice according to the sampling step size; The set of voxels is written into the nominal occupancy table of the corresponding robot in the form of key-value pairs, and the baseline stabilization time of the robot under no-load conditions is included in the occupancy time interval of the last segment of the placement trajectory, thus forming the nominal occupancy table.

[0011] Its effect is that by discretizing the shared workspace into a writable and scalable three-dimensional voxel grid, and pre-recording the geometric occupancy trajectory of each robot over time in the form of a nominal occupancy table during the offline stage, it provides a low-level data foundation with a clear data structure for the subsequent dynamic expansion of residual oscillations, so that subsequent steps can complete the dynamic correction by simply adding time intervals without modifying the trajectory itself.

[0012] Furthermore, the current sampling channel within the robot joint servo driver is reused to extract the step component containing load information, and the net load torque is separated by combining it with the torque contribution of the robotic arm itself, including: At the instant the robotic arm completes the grasping action and the gear leaves the unloading table, the step component with the start-up edge is picked up from the current sampling channel of the wrist joint servo. Using the total electromagnetic torque given by the electromagnetic torque formula as the total amount, the contributions of the robot arm's own motion, self-damping, and unloaded self-weight terms are subtracted one by one to extract the net load torque introduced by the gears.

[0013] Furthermore, the expansion increment is added to the occupancy time interval of the corresponding voxel in the spatiotemporal occupancy raster to reconstruct the dynamic occupancy envelope, including: The occupancy time interval of the voxel corresponding to the end of the robot placement trajectory in the spatiotemporal occupancy grid is expanded by an increment based on the existing baseline stable duration in the nominal occupancy table, and finally the dynamic occupancy envelope is reconstructed.

[0014] Its effect is that by deriving an improved stable duration relationship containing only known parameters from the existing relationship of vibration dynamics, the occupancy duration of each voxel in the shared workspace is no longer a static constant bound to the trajectory, but a dynamic quantity that expands online with the mass of the gripping gear. This fills the gap between the nominal cycle time and the actual stable duration from a physical mechanism perspective, and provides accurate dynamic occupancy input for subsequent multi-machine timing arbitration.

[0015] Furthermore, the dynamic occupancy envelope is reported to the collaborative scheduling node, and the dynamic occupancy table of multiple robots is scanned in a look-ahead manner to determine whether there is time slice overlap between the dynamic envelopes of adjacent robots on shared voxels, including: After each grabbing action is completed, the equivalent mass and expansion increment of the gear are reported to the collaborative scheduling node; The collaborative scheduling node adds an expansion increment to the time interval at the end of the nominal occupancy table that has been constructed, forming a rolling updated multi-machine dynamic occupancy table; At the beginning of each control cycle, a look-ahead scan is performed on the dynamic occupancy table of multiple robots for the next few control cycles to determine whether there is time slice overlap on the shared voxels of the dynamic envelopes of adjacent robots.

[0016] Furthermore, if overlap exists, an entry delay signal is sent to the robot that enters later to perform a rolling handshake arbitration scheduling, including: If it is found that the dynamic envelopes of adjacent robots overlap in time slices on shared voxels, an entry delay signal is sent to the robot that enters later, and the delay amount is equal to the expansion increment of the preceding robot on the conflicting voxel relative to the nominal occupancy. If there is no overlap, a handshake signal is released, allowing subsequent robots to enter the shared workspace in the nominal time sequence, completing the rolling handshake arbitration scheduling.

[0017] Furthermore, after completing the rolling handshake arbitration scheduling, it also includes: After receiving the handshake signal, the robot controller performs the corresponding action according to its own working state; The actual entry and exit times are sent back to the collaborative scheduling node to correct the dynamic occupancy table of multiple machines in the next handshake window.

[0018] The beneficial effects are: it avoids the collision and sudden stop caused by the subsequent robotic arm entering the shared space before the end-effector oscillation of the previous robotic arm has subsided; and by compressing the conservative margin from a static fixed value to a precise expansion amount that strictly corresponds to the current load, it effectively eliminates the drag on the overall line cycle time caused by excessive conservatism, and ensures that the timing arbitration of multi-machine palletizing is strictly aligned with the actual physical occupancy. Attached Figure Description

[0019] Figure 1 This is a flowchart of the industrial robot palletizing collaborative control method in this invention.

[0020] Figure 2 This is a comparison curve of the end-point stabilization time as a function of gear mass in this invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] An embodiment of the industrial robot palletizing collaborative control method provided by the present invention: like Figure 1 As shown, the industrial robot palletizing collaborative control method includes the following steps: S1, construct the spatiotemporal occupancy raster and write it into the nominal occupancy table.

[0023] In an optional embodiment, for multiple palletizing robots in the precision forging gear production line area, firstly, based on the existing robot kinematics model and the CAD layout of the work unit, the overlapping areas accessible to each robotic arm are abstracted into a shared workspace. During the offline phase, the shared workspace is discretized into voxels along the X, Y, and Z directions to form a three-dimensional voxel grid, constructing a spatiotemporal occupancy grid for the shared workspace. Each voxel records its spatial coordinates and the time interval it occupies, serving as the basic data unit for subsequent multi-machine timing arbitration. Determining the voxel side length requires considering the geometric envelope dimensions of the robotic arm end effector and the magnitude of the material rack positioning tolerance zone, ensuring each voxel has sufficient spatial resolution to accurately capture potential geometric interference areas between adjacent robotic arms, while avoiding excessive subdivision that leads to grid expansion and increases the real-time computational pressure on the scheduling nodes.

[0024] Furthermore, for each robot, based on the grasping trajectory, transport trajectory, and placement trajectory, the set of voxels swept by the geometric envelope of the robotic arm body and end effector at each time slice is calculated point-by-point according to the sampling step size. This set is then written into the corresponding robot's nominal occupancy table in the form of key-value pairs. The nominal occupancy table uses the time slice as the primary key and the list of scanned voxel numbers as the values, forming the robot's dynamic occupancy trajectory in the shared workspace over time. It should be noted that the nominal occupancy table already includes the robot's baseline stable duration under no-load conditions. Include the time interval occupied at the end of the placement trajectory. These are the factory parameters provided by the manufacturer, based on actual measurements, for the robot under rated placement trajectory parameters and nominal palletizing configuration. The nominal palletizing configuration is represented by the middle layer of the end-height picking rack. When the actual placement height deviates from the nominal configuration, the maximum value of the measured values ​​for each configuration is taken as the standard. A conservative upper bound is used to ensure that the nominal occupancy table still has a basic safety margin under the most unfavorable geometry configuration.

[0025] Next, the nominal occupancy tables of each robot are uploaded to the collaborative scheduling node. The scheduling node determines whether there is a geometric conflict by traversing the intersection of the voxel sets occupied by each robot in the same time slice. If a conflict exists, a waiting time slice is inserted according to a predetermined priority to form the initial multi-robot collaborative nominal timing sequence; if there is no conflict, the current timing sequence is directly retained without inserting additional waiting. It should be noted that this step follows the mature spatiotemporal rasterization approach in the industry. Those skilled in the art can choose different voxel discretization precisions and conflict determination priority rules according to the actual production line layout, and are not limited to the specific implementation described above.

[0026] By discretizing the shared workspace into a writable and scalable three-dimensional voxel grid, and pre-recording the geometric occupancy trajectory of each robot over time in the form of a nominal occupancy table during the offline phase, a well-defined underlying data foundation is provided for the subsequent dynamic expansion of residual oscillations. This allows subsequent steps to achieve dynamic correction by simply adding time intervals without modifying the trajectory itself.

[0027] S2, extract the step component and inversely derive the equivalent mass of the gear.

[0028] In an optional embodiment, to perceive the actual mass of the gear without installing an end effector accelerometer and external vision equipment, when the robotic arm completes the grasping action, the current sampling channel within the robot joint servo driver is reused to extract the step component containing load information, picking up the electrical signal containing load information from the underlying hardware level. At the instant the robotic arm completes the grasping action and the gear leaves the unloading table, the end effector load jumps from zero to the gear's gravity term. This jump manifests as a step component with a clear start edge in the current sampling channel of the wrist joint servo, serving as a natural physical entry point for identifying the gear's mass. Since the mass of precision-forged gears of different specifications fluctuates significantly between 2 kg and 18 kg, the amplitude of this step component has a monotonic correspondence with the gear mass, possessing a stable and identifiable signal-to-noise ratio, meeting the basic requirements for engineering applications.

[0029] Furthermore, to separate the load torque contributed by gear gravity from the current sampling channel, the existing electromagnetic torque relationship of surface-mounted permanent magnet synchronous motors is introduced as a basis:

[0030] in, For electromagnetic torque, This represents the number of pole pairs of the motor, which can be directly read from the servo nameplate. This is the rotor permanent magnet flux linkage constant, which can be directly read from the servo's factory parameters. The above formula only provides the total electromagnetic torque at the motor end and cannot directly give the gear mass. Therefore, it is engineered for the palletizing scenario. The net load torque is separated by combining the robot arm's own torque contribution. The net load torque introduced by the gear is extracted by subtracting the robot arm's own torque contribution from each item of the total electromagnetic torque.

[0031] The meanings of the variables after the improvement are as follows: This is the net load torque equivalent to the gear's own weight at the wrist joint; The equivalent rotational inertia of the robotic arm in the current configuration is output in real time by the dynamic model provided by the robot manufacturer. and The angular acceleration and angular velocity of the wrist joint are obtained differentially from the existing encoder, respectively. The equivalent damping coefficient of the joint is given by the servo recognition factory parameters; The gravity term, representing the unloaded self-weight of the robotic arm in this configuration, is calculated in real-time by the dynamic model. The core improvement lies in: [the expression] given by the electromagnetic torque relationship... To calculate the total, subtract the robotic arm's own motion terms from each item. Self-damping term With unloaded tare weight The contribution is to retain only the net load torque introduced by the gear, thus achieving strict separation of load information and arm body information.

[0032] Then, the net load torque Divide by the gravitational lever arm of the distal end relative to the wrist joint in the current configuration. With gravitational acceleration This leads to the deduction of the equivalent mass of the gear. :

[0033] in, Direct output from forward kinematics Take the standard gravitational acceleration. The dynamic attribute fields of the robot in this handling task are written for subsequent steps. It should be noted that there are many other ways to achieve load quality perception in this field, including direct measurement by an end-effector torque sensor, and lookup table after visual identification of the model. Those skilled in the art can choose different perception methods according to the actual hardware configuration, and are not limited to the inversion method based on the current sampling channel mentioned above.

[0034] By reusing the existing current sampling channel within the servo driver, online identification of the actual gear mass during each gripping action is achieved without adding any external sensors. This provides accurate load input for the dynamic prediction of the subsequent residual oscillation stabilization duration, eliminating the installation, adaptation, and reliability risks associated with additional sensors at the hardware level.

[0035] S3 calculates the actual stable duration and reconstructs the dynamic occupancy envelope using the expansion increment.

[0036] In an optional embodiment, after load mass identification is completed, the constructed spatiotemporal occupancy grid is dynamically expanded to address the issue of residual end-effector oscillations caused by variable loads on the precision forged gear. Physically, after the robotic arm places the gear, the mechanical transmission chain connected to the end flange, due to its finite stiffness, will generate an underdamped free oscillation response at the moment of load unloading. The amplitude of this oscillation decays exponentially with time. The criterion for determining when the end-effector oscillation reaches stability is: the residual displacement at the end converges to within the material rack positioning tolerance zone after gear placement; that is, the amplitude of the residual displacement of the actual end position deviating from the placement target point converges to the material rack positioning tolerance zone. within, The structural design drawings are taken directly from the material rack and are objective parameters that can be directly read on the engineering site, requiring no additional calibration.

[0037] Furthermore, if the existing stability time relationship of a second-order underdamped system is directly adopted... As a basis for compensation for the time occupied, it is necessary to identify the damping ratio of each robotic arm under each gear specification on site. With natural frequency The engineering feasibility is extremely poor. Therefore, the existing fundamental relationship of the undamped natural frequency in vibration dynamics is used. Starting from this point, we derive an improved form for variable load conditions. Let be the equivalent stiffness of the drive train at the end effector of the robotic arm. After the placement action is completed, the equivalent mass of the oscillation system is changed from the equivalent mass of the wrist at the end effector of the robotic arm when unloaded. Transition to the grasping phase The corresponding natural frequencies are as follows:

[0038]

[0039] Since the settling time is inversely proportional to the natural frequency, that is... Assuming the damping ratio remains approximately constant, the ratio of the settling time in the two states is:

[0040] After simplification, an improved formula for variable load conditions is obtained. Based on the equivalent mass of the gear and the equivalent mass of the robotic arm's end effector wrist, the actual stabilization time required for the residual displacement at the end effector to converge to within the material rack positioning tolerance zone after the gear is placed is calculated.

[0041] in, The residual displacement at the end of the gear converges to the desired value after placement. The actual stabilization time required within this range, This is the baseline stabilization time for the robot under no-load conditions. The equivalent mass of the gear obtained from the inversion is... The equivalent mass of the wrist at the end of the robotic arm is given by the dynamic model. The entire derivation of this improved relationship relies solely on the existing relationship between natural frequency and mass, as well as the existing stable duration relationship, without introducing any empirical fitting parameters.

[0042] Next, a calculation and explanation will be performed using typical working conditions. For example, if a certain type of palletizing robot... , This identification ,but:

[0043] The settling time has expanded by 0.2 seconds compared to the baseline; For this expansion increment, the time interval occupied by the voxel where the robot's trajectory end is located in the already constructed spatiotemporal occupancy grid will be added to the nominal occupancy table. Extending further on the basis of The reconstructed dynamic occupancy envelope is obtained. The larger this value is, the greater the mass of the gear being grasped relative to the equivalent mass of the robotic arm wrist, the greater the decrease in the end-effector natural frequency, and the more significant the stability duration expansion. The smaller this value is, the smaller the ratio of the gear mass to the wrist equivalent mass, the closer the stability duration expansion is to zero, and the closer the dynamic occupancy envelope is to the static value of the nominal occupancy table.

[0044] By deriving an improved stable duration relationship containing only known parameters based on existing vibration dynamics relationships, the occupancy duration of each voxel in the shared workspace is no longer a static constant bound to the trajectory, but a dynamic quantity that expands online with the mass of the gripping gear. This physically bridges the gap between the nominal beat and the actual stable duration, providing accurate dynamic occupancy input for subsequent multi-machine timing arbitration.

[0045] S4, report the dynamic occupancy envelope and execute rolling handshake arbitration scheduling.

[0046] In an optional embodiment, each robot immediately retrieves the previously inverted data after each grasping action. Compared with the calculated expansion increment The information is reported to the collaborative scheduling node, which then appends the time interval for the last segment in the already constructed nominal occupancy table. This generates a dynamically updated multi-machine occupancy table. The scheduling node operates using a rolling window handshake arbitration method: at the beginning of each control cycle, it performs a look-ahead scan of the dynamically updated multi-machine occupancy table for the next few control cycles to determine whether there is time slice overlap between the dynamic occupancy envelopes of adjacent robots on shared voxels. If time slice overlap is found, an entry delay signal is sent to the robot that entered later. The delay amount is equal to the inflation increment of the preceding robot's relative nominal occupancy on the conflicting voxel. If there is no overlap, a handshake signal is immediately released, allowing subsequent robots to enter the shared workspace according to their nominal timing. The entire arbitration process is completed in milliseconds, and the arbitration result is sent to each robot controller via the real-time industrial bus.

[0047] Furthermore, upon receiving the handshake signal, each robot controller executes its corresponding action according to its own operational state and sends back the actual entry and exit times to the scheduling node for post-processing correction of the multi-machine dynamic occupancy table within the next handshake window. Since the arbitration result between different control cycles can be directly used as the initial state for the next look-ahead scan, the rolling window's progression is continuous. It eliminates the need for a full reconstruction of the occupancy table before each arbitration round; only incremental updates are required for the time intervals where changes occur, significantly reducing the real-time computational load on the scheduling node. It should be noted that there are many other ways to implement multi-machine timing arbitration in this field. Those skilled in the art can choose different look-ahead window lengths and priority rules based on actual production line cycle time requirements, and are not limited to the specific implementation of the rolling window-based handshake arbitration described above.

[0048] This forms a complete data flow path, including load identification during grasping, prediction of stabilization duration, expansion of occupancy envelope, rolling handshake arbitration, and execution result feedback. This transforms multi-machine collaborative scheduling from blindly following fixed timing sequences to accurately tracking the actual dynamic occupancy of each gear.

[0049] Thus, by using the dynamic occupancy envelope as input to drive the rolling window-style handshake arbitration, it not only avoids the collision and sudden stop caused by the subsequent robotic arm rashly entering the shared space before the end-effector oscillation of the previous robotic arm has subsided, but also effectively eliminates the drag on the overall line rhythm caused by excessive conservatism by compressing the conservative margin from a static fixed value to a precise expansion amount that strictly corresponds to the current load, so that the timing arbitration of multi-machine palletizing is strictly aligned with the actual physical occupancy.

[0050] like Figure 2 As shown, it can be observed that the stabilization time of the method of the present invention increases monotonically with the gear mass and is always lower than the fixed conservative upper limit of the existing method, indicating that the present invention effectively reduces the drag on the cycle time of the static conservative margin while ensuring safety.

[0051] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. An industrial robot palletizing collaborative control method, characterized in that, Includes the following steps: S1, discretize the overlapping areas within the reach of each robotic arm into voxels, construct a spatiotemporal occupancy grid for the shared workspace, and write the set of voxels swept by each robot based on its trajectory into a nominal occupancy table; S2, when the robotic arm completes the grasping action, the current sampling channel in the robot joint servo driver is reused to extract the step component containing load information, and the net load torque is separated by combining the torque contribution of the robotic arm itself, and then the equivalent mass of the gear is deduced. The deduction of the equivalent mass of the gear includes: obtaining the gravitational lever arm and gravitational acceleration of the end relative to the wrist joint under the current configuration, dividing the net load torque by the product of the gravitational lever arm and gravitational acceleration, and deducing the equivalent mass of the gear. S3. Based on the equivalent mass of the gear and the equivalent mass of the robotic arm's end effector wrist, calculate the actual stabilization time required for the residual displacement at the end effector to converge within the shelf positioning tolerance zone after the gear is placed. Combine this with the baseline stabilization time to calculate the expansion increment. Add the expansion increment to the occupancy time interval of the corresponding voxel in the spatiotemporal occupancy grid to reconstruct the dynamic occupancy envelope. The calculation method for the actual stabilization time is as follows: obtain the baseline stabilization time of the robot under no-load conditions; calculate the ratio of the equivalent mass of the gear to the equivalent mass of the robotic arm's end effector wrist, add 1 to the ratio, and then perform a square root operation; multiply the result of the square root operation by the baseline stabilization time to obtain the actual stabilization time required for the residual displacement at the end effector to converge within the shelf positioning tolerance zone after the gear is placed; subtract the baseline stabilization time from the actual stabilization time to obtain the expansion increment. S4 reports the dynamic occupancy envelope to the collaborative scheduling node. By looking ahead and scanning the multi-machine dynamic occupancy table, it determines whether the dynamic envelopes of adjacent robots overlap on the shared voxel. If there is an overlap, it sends an entry delay signal to the robot that entered later in order to perform rolling handshake arbitration scheduling.

2. The industrial robot palletizing collaborative control method according to claim 1, characterized in that, The overlapping areas accessible by each robotic arm are discretized using voxels to construct a spatiotemporal occupancy grid for the shared workspace, including: Based on the robot kinematics model and the layout of the work units, the overlapping areas accessible by each robotic arm are abstracted into a shared workspace. The shared workspace is discretized into voxels along three directions to form a three-dimensional voxel grid. The spatial coordinates of each voxel and the time interval in which it is occupied are recorded, and finally the spatiotemporal occupancy grid of the shared workspace is obtained.

3. The industrial robot palletizing collaborative control method according to claim 1, characterized in that, Write the set of voxels scanned by each robot based on its trajectory into a nominal occupancy table, including: For each robot, based on the grasping trajectory, transport trajectory, and placement trajectory, calculate the set of voxels swept by the geometric envelope of the robotic arm body and end effector at each time slice according to the sampling step size; The set of voxels is written into the nominal occupancy table of the corresponding robot in the form of key-value pairs, and the baseline stabilization time of the robot under no-load conditions is included in the occupancy time interval of the last segment of the placement trajectory, thus forming the nominal occupancy table.

4. The industrial robot palletizing collaborative control method according to claim 1, characterized in that, The current sampling channel within the robot joint servo driver is reused to extract the step component containing load information, and combined with the torque contribution of the robotic arm itself to separate the net load torque, including: At the instant the robotic arm completes the grasping action and the gear leaves the unloading table, the step component with the start-up edge is picked up from the current sampling channel of the wrist joint servo. Using the total electromagnetic torque given by the electromagnetic torque formula as the total amount, the contributions of the robot arm's own motion, self-damping, and unloaded self-weight terms are subtracted one by one to extract the net load torque introduced by the gears.

5. The industrial robot palletizing collaborative control method according to claim 1, characterized in that, The expansion increment is appended to the occupancy time interval of the corresponding voxel in the spatiotemporal occupancy raster to reconstruct the dynamic occupancy envelope, including: The occupancy time interval of the voxel corresponding to the end of the robot placement trajectory in the spatiotemporal occupancy grid is expanded by an increment based on the existing baseline stable duration in the nominal occupancy table, and finally the dynamic occupancy envelope is reconstructed.

6. The industrial robot palletizing collaborative control method according to claim 1, characterized in that, The dynamic occupancy envelope is reported to the collaborative scheduling node. A look-ahead scan of the multi-machine dynamic occupancy master table is used to determine whether there is time slice overlap between the dynamic envelopes of adjacent robots on shared voxels, including: After each grabbing action is completed, the equivalent mass and expansion increment of the gear are reported to the collaborative scheduling node; The collaborative scheduling node adds an expansion increment to the time interval at the end of the nominal occupancy table that has been constructed, forming a rolling updated multi-machine dynamic occupancy table; At the beginning of each control cycle, a look-ahead scan is performed on the dynamic occupancy table of multiple robots for the next few control cycles to determine whether there is time slice overlap on the shared voxels of the dynamic envelopes of adjacent robots.

7. The industrial robot palletizing collaborative control method according to claim 1, characterized in that, If overlap exists, an entry delay signal is sent to the robot that entered later to perform a rolling handshake arbitration scheduling, including: If it is found that the dynamic envelopes of adjacent robots overlap in time slices on shared voxels, an entry delay signal is sent to the robot that enters later, and the delay amount is equal to the expansion increment of the preceding robot on the conflicting voxel relative to the nominal occupancy. If there is no overlap, a handshake signal is released, allowing subsequent robots to enter the shared workspace in the nominal time sequence, completing the rolling handshake arbitration scheduling.

8. The industrial robot palletizing collaborative control method according to claim 7, characterized in that, After completing the rolling handshake arbitration scheduling, the following is also included: After receiving the handshake signal, the robot controller performs the corresponding action according to its own working state; The actual entry and exit times are sent back to the collaborative scheduling node to correct the dynamic occupancy table of multiple machines in the next handshake window.

Images