Multi-station collaborative operation robot group scheduling control method and system

Abstract

The invention discloses a multi-station collaborative operation robot group scheduling control method and system, and relates to the technical field of industrial robot control, and the method comprises the steps: obtaining and fusing multi-station vision and point cloud data, and determining the task pose information of each station of a robot group; according to the allocation task, the planning path and the registration divergence rollback re-collection data, obtaining an operation allocation result; then, correcting the pose by sensing the filtering binding force-displacement relation through near-field operation force; analyzing and identifying abnormity of the low-load test binding force data trend in operation; a scheduling control result is generated in combination with task feedback adjustment parameters; and finally, frequency optimization is constrained by an environment interference threshold value, and a multi-station cooperative scheduling scheme of the swarm robots is obtained. According to the method, the technical problems of insufficient pose precision, serious equipment interference and abnormal response lag of the robot group in multi-station operation are solved, and the technical effects of accurate pose control, timely abnormal identification and efficient cooperative scheduling of the multi-station robot group are achieved.

Description

Technical Field

[0001] This invention relates to the field of industrial robot control technology, and in particular to a method and system for scheduling and controlling robot groups in multi-workstation collaborative operations. Background Technology

[0002] In industrial production, multi-workstation operations are widespread. The scheduling and control of robot swarms are crucial to operational efficiency, accuracy, and equipment coordination, directly impacting product quality and production process stability. An efficient and precise scheduling and control system is key to ensuring smooth operation across multiple workstations. Current technologies often rely on single-vision localization or simple path planning for robot swarm scheduling, while collaborative control frequently ignores inter-device interference and dynamic operational errors. These methods may work in single-workstation or simple environments, but as industrial demands for multi-workstation collaboration increase, their limitations become apparent when applied to complex multi-workstation scenarios. This results in insufficient robot swarm operational accuracy, numerous collaborative interferences, and an inability to meet the demands for efficient and precise multi-workstation scheduling and control. Summary of the Invention

[0003] This application provides a method and system for scheduling and controlling robot groups in multi-station collaborative operations, which solves the technical problems of insufficient pose accuracy, severe equipment interference, and delayed abnormal response of robot groups in multi-station operations.

[0004] The first aspect of this application provides a method for scheduling and controlling a robot swarm in multi-station collaborative operation. The method includes: acquiring and fusing visual data and point cloud data from multiple workstations to determine the task pose information of the robot swarm at each workstation; based on the task pose information, performing task allocation and path planning on the robot swarm to obtain a task allocation result, and triggering backtracking and data re-acquisition when registration diverges; according to the task allocation result, the robot swarm performs compliant correction in near-field operation using force-sensing filtering and combining force-displacement relationships to obtain corrected pose information; during the corrected operation, the robot swarm performs low-load testing and combines force data and trend analysis to obtain anomaly identification results; based on the task allocation result, corrected pose information, and anomaly identification results, combined with task feedback and adaptive parameter adjustment, generating a scheduling control result; and based on the scheduling control result, performing frequency optimization with an environmental interference threshold as a constraint to obtain a multi-station collaborative scheduling scheme for the robot swarm.

[0005] A second aspect of this application provides a multi-station collaborative robot swarm scheduling and control system, the system comprising: a task pose information acquisition module, used to acquire and fuse visual data and point cloud data from multiple workstations to determine the task pose information of the robot swarm at each workstation; a task allocation result acquisition module, used to allocate tasks and plan paths for the robot swarm based on the task pose information to obtain a task allocation result, and trigger backtracking and data re-acquisition when registration diverges; a correction pose information acquisition module, used to perform compliant correction on the robot swarm during near-field operations based on the task allocation result, using force-sensing filtering and combining force-displacement relationships to obtain corrected pose information; an anomaly identification result acquisition module, used to perform low-load testing and combine force data and trend analysis during the corrected operation of the robot swarm to obtain anomaly identification results; a scheduling control result generation module, used to generate scheduling control results based on the task allocation result, corrected pose information, and anomaly identification results, combined with task feedback and parameter adaptive adjustment; and a collaborative scheduling scheme acquisition module, used to perform frequency optimization based on the scheduling control results and environmental interference threshold constraints to obtain a multi-station collaborative scheduling scheme for the swarm robots.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] This application acquires visual and point cloud data from multiple workstations and fuses them to determine the task pose information of a robot group. Through task allocation and path planning, near-field compliance correction is achieved by combining force-sensing filtering and force-displacement relationship. Then, anomalies are identified through low-load testing and force data trend analysis. Finally, parameters are adaptively adjusted based on task feedback, and frequency optimization is performed by constraining environmental interference thresholds to form a closed-loop scheduling control. This enables precise operation and efficient collaboration of a robot group across multiple workstations, meeting the high precision and high reliability requirements of multi-workstation scheduling control. The application achieves the technical effects of precise pose control, timely anomaly identification, and efficient collaborative scheduling of multi-workstation robot groups. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a flowchart illustrating the robot group scheduling and control method for multi-station collaborative operation provided in this application embodiment.

[0010] Figure 2This is a schematic diagram of the structure of a robot group scheduling and control system for multi-station collaborative operation provided in the embodiments of this application.

[0011] Figure labeling: Module 1 for task pose information acquisition, Module 2 for job allocation result acquisition, Module 3 for corrected pose information acquisition, Module 4 for anomaly identification result acquisition, Module 5 for scheduling control result generation, and Module 6 for collaborative scheduling scheme acquisition. Detailed Implementation

[0012] This application provides a method and system for scheduling and controlling robot groups in multi-station collaborative operations, which solves the technical problems of insufficient pose accuracy, severe equipment interference, and delayed abnormal response of robot groups in multi-station operations.

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0014] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices.

[0015] Example 1, as Figure 1 As shown, a robot group scheduling and control method for multi-station collaborative operation is provided, wherein the method includes:

[0016] Visual data and point cloud data from multiple workstations are acquired and fused to determine the task pose information of the robot group at each workstation.

[0017] Specifically, by collecting visual data in the form of two-dimensional images and point cloud data in the form of three-dimensional spatial depth information from each workstation, and through data registration and fusion processing, the task pose information of the robot group at each workstation is determined.

[0018] Based on the task pose information, task allocation and path planning are performed on the robot group to obtain the job allocation result, and backtracking and data re-acquisition are triggered when registration diverges.

[0019] Optionally, based on the task pose information, the operation requirements of each workstation are determined and assigned to the corresponding robot. Based on the requirements, the motion path between multiple workstations of the robot group is generated. When registration divergence occurs in the path planning, a backtracking is triggered and visual and point cloud data are re-acquired to obtain the operation assignment result.

[0020] Based on the task allocation results, the robot swarm performs compliance correction in near-field operations by using force-sensing filtering and combining force-displacement relationships to obtain corrected pose information.

[0021] In one embodiment of this application, when a group of robots is working in the near field, contact force data is collected, noise is removed by force-sensing filtering to obtain the effective force, and the robot group's pose is adjusted to achieve compliance correction in combination with the force-displacement relationship to obtain corrected pose information.

[0022] During the process of the robot group performing the corrected operation, a low-load test is conducted and force data and trend analysis are combined to obtain anomaly identification results.

[0023] Specifically, during the operation of the robot swarm after correction, force data is obtained through low-load testing, effective force data is extracted after preprocessing, and trend analysis is performed based on the effective force data to obtain anomaly identification results.

[0024] Based on the job allocation results, corrected pose information, and anomaly identification results, combined with task feedback and adaptive parameter adjustment, scheduling control results are generated.

[0025] Specifically, the task feedback information is obtained by comparing and analyzing the task allocation results, correcting the pose information and the anomaly identification results. Based on this, at least one control parameter such as pose control, path planning and force threshold is adaptively adjusted. Based on the adjusted parameters, the scheduling control result of the group of robots is generated to indicate the multi-station scheduling execution.

[0026] Based on the scheduling and control results, frequency optimization is performed with environmental interference threshold as a constraint to obtain a multi-station collaborative scheduling scheme for group robots.

[0027] Specifically, the scheduling requirements for multi-station operations of a group of robots are determined based on the scheduling and control results. Under these requirements, frequency optimization is performed with environmental interference thresholds as constraints, and a collaborative scheduling scheme for multi-station operations of the group of robots is generated based on the optimization results.

[0028] Furthermore, the method provided in this application embodiment includes:

[0029] The visual data includes two-dimensional image information collected from each workstation; the point cloud data includes three-dimensional spatial depth information collected from each workstation; based on the visual data and the point cloud data, the task pose information of the robot group at each workstation is determined through data registration and fusion processing.

[0030] Specifically, in industrial multi-station operation scenarios, basic data acquisition is first carried out to obtain the task pose information required by the robot swarm. For visual data, equipment such as industrial area scan cameras are used to acquire images of each workstation, obtaining two-dimensional image information of each workstation. These two-dimensional images can clearly present the planar layout, surface markings, and corner contours of the workpiece within the workstation, such as the position of the workpiece's mounting holes and the positioning marks printed on the surface. This allows for a preliminary determination of the approximate position range of the workpiece within the workstation, providing an initial planar reference for subsequent positioning.

[0031] Next, point cloud data from each workstation is collected to supplement the 3D spatial information. This is done using devices such as LiDAR or structured light scanners. These devices emit laser or structured light signals and then receive the reflected signals to calculate the 3D spatial coordinates of each point within the workstation, thereby generating point cloud data containing 3D spatial depth information of both the workstation and the workpiece. This data accurately reflects the 3D structure of the workpiece, such as its height, protrusion height, recess depth, and spatial distance between workpieces. It compensates for the lack of depth information in 2D images, providing 3D dimensional data support for precise positioning.

[0032] After completing the acquisition of both types of data, data registration and fusion processing are performed, which is crucial for enabling the two types of data to work synergistically. Data registration first involves feature extraction from the 2D image. The ORB algorithm is used to identify key feature points in the image, typically highly recognizable locations on the workpiece surface, such as corners, the center of circular holes, and special markings. Next, the corresponding 3D feature points are found in the point cloud data. The coordinate deviation between the two types of feature points in their respective coordinate systems is calculated, and the coordinate system parameters of the point cloud data or visual data are adjusted to align the 2D and 3D feature points in the same spatial coordinate system, eliminating coordinate system differences between data acquired from different devices, thus completing the registration. After registration, the fusion stage begins, mapping the planar details of the 2D image, such as workpiece surface texture and color markings, onto the 3D points in the point cloud data. This ensures that each 3D point contains both spatial coordinates and planar visual features, forming a complete 3D data model of the workstation that combines planar details and depth.

[0033] Finally, based on the fused complete 3D data model of the workstations, the task pose information of the robot group at each workstation is determined. The task pose information includes the robot's spatial position and working posture. By analyzing the workpiece's position to be worked on in the fused model, such as bolt fastening points and assembly interfaces, the precise 3D coordinates of this position in the robot's base coordinate system are calculated. At the same time, according to the orientation of the workpiece's working surface, such as the direction of the bolt hole axis and the direction of the normal to the assembly interface plane, the posture angle that the robot's end effector needs to adjust is determined to ensure that the end effector can dock with the workpiece's working part in the correct position and posture. Finally, the specific task pose information of each robot corresponding to each workstation is clarified.

[0034] By first collecting visual data in the form of two-dimensional images and point cloud data in the form of three-dimensional spatial depth from each workstation, and then performing registration processing such as feature extraction and coordinate alignment, as well as fusion processing that combines planar details and three-dimensional depth, the robot's spatial position and working posture are finally calculated based on the fused data. This achieves the effect of providing accurate spatial positioning basis for subsequent task allocation and path planning for the robot group.

[0035] Furthermore, the method provided in this application embodiment includes:

[0036] The task pose information is used to determine the work requirements of each workstation and assign them to the corresponding robots; the work requirements are used to generate the motion path of the robot group between multiple workstations; when registration divergence occurs during the path planning process, a rollback operation is triggered and the visual data and point cloud data are re-acquired.

[0037] Optionally, the task pose information includes the coordinates of the workpiece's position to be worked, its attitude angle, and the process requirements for each workstation, such as the precision grade of bolt tightening and the connection sequence of parts assembly. Based on this information, the specific operational requirements of each workstation can be clearly defined. For example, a workstation may require tightening three M8 bolts on a workpiece, with the positioning deviation of each bolt controlled within 0.3mm and the torque stabilized at 6-8 N·m. When allocating robots, the hardware parameters of each robot (such as the positioning accuracy, torque control range, and load capacity of the end effector) and its real-time working status (such as whether it is currently idle and the remaining energy consumption after completing the task) are combined to assign robots that match the workstation's operational requirements to the corresponding workstation. For example, a robot with an end-effector positioning accuracy of 0.1mm and closed-loop torque control function can be assigned to a workstation requiring high-precision bolt tightening, ensuring that the operational requirements of each workstation can be met by the robot's performance.

[0038] Next, after clarifying the task assignments for each robot, the starting and ending points of each robot's movement can be determined, which are the robot's current parking position or the end position of the previous workstation and the pending work position of the target workstation, respectively. When generating the movement paths between multiple workstations, a spatial environment model containing all workstations, fixed equipment, and passageway areas is first constructed, and the range of obstacles is marked. Obstacles include workstation supports, conveyor belts, and other robots in operation.

[0039] The A* algorithm is then used for path planning. The specific process is as follows: the spatial environment is divided into several uniform three-dimensional grids. Each grid is labeled as passable or impassable based on the presence of obstacles. Taking the grid corresponding to the robot's starting point as the initial node, the total cost of each node is calculated. The total cost = the actual distance traveled from the starting point to the node + the heuristic cost from the node to the destination, where the heuristic cost is calculated using the Euclidean distance between the node and the destination. Each time, the node with the minimum total cost is selected from the nodes to be expanded and expanded. This process is iterated until the node corresponding to the destination is expanded, forming a motion path that connects the starting point and the destination, is free from obstacle collisions, and has the optimal path length. At the same time, the paths of different robots are checked on a time axis to ensure that multiple robots do not enter the same spatial grid within the same time period, thus avoiding motion conflicts.

[0040] During path planning, the task pose information obtained by fusing visual data and point cloud data is used as the reference for the endpoint position. If, during data registration, the system detects a deviation between the fused pose data and the actual workstation environment that exceeds a set threshold, such as a 3D coordinate deviation exceeding 0.5mm or an attitude angle deviation exceeding 0.1°, it is determined to be a registration divergence. This situation is usually caused by interference from light reflection in the visual data or by the point cloud data lacking some depth information due to workpiece occlusion.

[0041] Then, a rollback operation is immediately triggered: First, the robot stops the current path planning process and returns to the stable position after the most recent pose calibration; then, the vision acquisition device is restarted to acquire the 2D image of the target workstation, and the laser scanning device is started to acquire the 3D point cloud data; after the new vision and point cloud data are acquired, the data registration process is re-executed to ensure that the newly generated task pose information is completely matched with the actual workpiece position and posture; finally, based on the calibrated task pose information, the operation requirements of each workstation are reconfirmed and the robot is assigned, and a new motion path is generated according to the above path planning process.

[0042] By matching robot capabilities with task pose information to determine task allocation, using the A* algorithm combined with environmental obstacles to generate conflict-free optimal motion paths, and backing up and re-collecting data to calibrate poses when registration diverges, the system achieves the effect of ensuring reasonable task allocation for the robot group, accurate motion paths, and the ability to cope with data deviations, thus providing a reliable execution foundation for subsequent near-field operations.

[0043] Furthermore, the method provided in this application embodiment includes:

[0044] The robot swarm collects contact force data during near-field operations; the acquired contact force data is processed by the force-sensing filter to remove noise and obtain effective force; based on the effective force and the force-displacement relationship, the pose of the robot swarm is adjusted for compliance, thereby completing the compliance correction and obtaining the corrected pose information.

[0045] Specifically, when robots are performing near-field operations, such as bolt pre-tightening and parts docking, their end effectors will make physical contact with the workpiece surface or the area to be worked on. At this time, the six-dimensional force sensor on the end effector will collect force data in real time during the contact process. This data includes linear contact forces along the X, Y, and Z axes as well as torque information around the three axes, which can completely reflect the force state when the robot contacts the workpiece. This provides the original force perception basis for subsequent posture correction, ensuring that the force changes at every moment of contact can be accurately captured, and avoiding subsequent adjustment deviations due to missing force data.

[0046] Next, after acquiring the raw contact force data, interference noise is removed through force-sensing filtering. The Kalman filter algorithm is used, and the specific process is as follows: First, a state equation for the force data is established based on the dynamic characteristics of the sensor, predicting the estimated force data value and the corresponding error covariance matrix at the current moment. This prediction references the optimal force data from the previous moment and the sensor's response delay characteristics. Then, the real-time acquired raw force data is used as the observation value and compared with the predicted estimate to calculate the Kalman gain. This gain is jointly determined by the prediction error covariance and the observation error covariance. If the observation error is small, the gain is large, emphasizing the correction of the predicted value with the observation value; otherwise, the predicted value is retained. The predicted force data estimate is then corrected based on the Kalman gain to obtain the optimal force data after fusing the observation information. Finally, the error covariance matrix is ​​updated to provide parameters for the next filtering iteration. Through this iterative process, interference components such as mechanical vibration and sensor circuit noise are effectively filtered out, ultimately outputting an effective force that truly reflects the contact state between the robot and the workpiece.

[0047] Then, the pose of the robot group is adjusted adaptively based on a preset force-displacement relationship. This force-displacement relationship is preset according to the specific work scenario (such as tightening bolts of different specifications or assembling parts of different materials), and specifies the pose adjustment amount corresponding to effective forces of different directions and magnitudes. For example, when the effective pressure along the Z-axis is greater than a preset threshold, the displacement adjustment amount in the negative Z-axis direction, i.e., away from the workpiece, is 0.02mm to avoid excessive pressure damaging the workpiece; when the torque around the X-axis is less than a preset lower limit, the corresponding adjustment amount is X-axis attitude angle +0.5° to ensure that the workpiece mating surfaces fit together.

[0048] Next, the system will first determine whether the direction and magnitude of the effective force are within the reasonable range allowed by the operation. If they are outside the range, the system will determine the required pose adjustment parameters based on the force-displacement relationship, including the position fine-tuning of the X, Y, and Z axes and the attitude angle adjustment around the three axes. Then, the drive unit controlling the robot will gradually correct the pose of the end effector according to the adjustment parameters. During this process, the system will continuously collect and filter the contact force data until the effective force stabilizes within the reasonable range. At this point, the robot's pose is the corrected pose information, and the compliance correction is completed.

[0049] By collecting near-field contact force data using the robot's end effector force sensor, removing noise using the Kalman filter algorithm to obtain effective force, and adjusting the pose by combining the preset force-displacement relationship, the robot group achieves compliant correction for near-field operations, improves pose accuracy, and avoids rigid contact damage to the workpiece.

[0050] Furthermore, the method provided in this application embodiment includes:

[0051] During the operation of the robot swarm after correction, a low-load test is performed to obtain force data; the force data is preprocessed to extract effective force data for analysis; and trend analysis is performed based on the effective force data to obtain anomaly identification results.

[0052] Specifically, during the process of the robot swarm performing tasks according to the corrected posture, a low-load test is initiated first. The load setting for the low-load test must be lower than the rated load for the formal operation. For example, if the formal bolt tightening requires an 8-10 N·m torque, the low-load test only applies 2-3 N·m of torque. This avoids damage to the workpiece or affecting the accuracy of subsequent formal operations due to excessive load during the test. Simultaneously, the force sensor on the robot's end effector is activated, collecting force data under low load in real time at a fixed sampling frequency, such as 100 Hz. This data includes the instantaneous value, peak value, and time-varying curve of the force, completely recording the dynamic changes of the force during the test, providing raw data support for subsequent judgment of the operational status.

[0053] Next, after acquiring the raw force data, preprocessing is performed to extract the effective force data. First, outlier removal is performed. Based on a preset reasonable force value range, for example, for the low load test of 2-3 N·m mentioned above, the reasonable range is set to 1.5-3.5 N·m. Extreme data that exceed this range, such as 0.5 N·m or 5 N·m due to instantaneous sensor interference, are directly removed to avoid extreme values ​​affecting the analysis results.

[0054] Subsequently, a moving average filter was used to smooth the data. The filter window size was set to 5 sampling points. The specific process was as follows: starting from the 5th sampling point, the filtered value of each sampling point was equal to the sum of the force data of that point and the previous 4 adjacent sampling points divided by 5. In this way, all data were gradually traversed to filter out high-frequency small-amplitude fluctuations caused by mechanical vibration and circuit noise, and finally, continuous, stable and effective force data that could truly reflect the stress state of the low-load test were obtained.

[0055] Then, based on the preprocessed effective force data, further trend analysis is conducted to obtain anomaly identification results. First, the normal force data variation trend under low-load testing for this type of operation is pre-defined. For example, under normal circumstances, the force under low load should show a slow linear increase with the operation time, such as an increase of 0.1-0.2 N·m per second, or remain stably within the range of 2-3 N·m. At this point, the least squares method is used to linearly fit the effective force data. The specific algorithm process is as follows: Let the effective force data be y1, y2, up to y... n It corresponds to n sampling times, with sampling times being x1, x2, up to x... n Assuming the fitted line equation is y = ax + b, where a is the slope and b is the intercept, the sum of squared deviations Σ(y) from all sampling points to the fitted line is calculated. i -(ax i +b)) 2 Find the minimum values ​​of a and b.

[0056] Finally, if the slope 'a' obtained from the fitting exceeds the normal range of 0.1-0.2 N·m / s, for example, a = 0.6 N·m / s, it indicates that the force is increasing too fast, or the effective force data is continuously exceeding the range of 2-3 N·m. If it is stable below 1 N·m, it indicates that the force is insufficient. In this case, it is determined that there is an abnormality in the operation. Anomaly identification results are generated based on the specific deviation, such as excessive force increase, suspected foreign object jamming on the workpiece surface, continuous low force, suspected stripped bolts, etc.

[0057] By collecting force data through low-load testing during operation, extracting effective force data through outlier removal and moving average filtering preprocessing, and using the least squares method to fit and analyze the force data trend, the system achieves the effect of timely identification of abnormalities in robot group operation without damaging the workpiece, providing a basis for subsequent operation adjustments.

[0058] Furthermore, the method provided in this application embodiment includes:

[0059] The task allocation results, corrected pose information, and anomaly identification results are compared and analyzed to obtain task feedback information; the control parameters are adaptively adjusted based on the task feedback information, and the control parameters include at least one of pose control parameters, path planning parameters, and force threshold parameters; a scheduling control result is generated based on the adaptively adjusted control parameters, and the scheduling control result is used to instruct the group of robots to schedule and execute at multiple workstations.

[0060] Specifically, before conducting comparative analysis, the core contents of the task allocation results, corrected pose information, and anomaly identification results should be clarified: the task allocation results include the workstations corresponding to each robot, the tasks to be performed (such as bolt tightening and parts assembly), and the task requirements (such as target pose accuracy and work load range); the corrected pose information is the actual three-dimensional spatial coordinates and posture angle data achieved by the robot after near-field compliance correction; the anomaly identification results record whether there are any anomalies in the operation (such as excessively rapid force growth or pose deviation exceeding the range) and the type and occurrence of the anomalies.

[0061] Next, during the comparative analysis, the deviation between the target pose in the job assignment results and the actual pose in the corrected pose information is first calculated. The Euclidean distance algorithm is used, with the target pose coordinates set as (x1, y1, z1) and the actual pose coordinates as (x2, y2, z2). The deviation value is... The system determines whether the deviation is within the required accuracy range, such as ≤0.3mm. It then correlates this with the anomaly identification results. If an anomaly of excessively rapid force increase exists, it needs to be checked whether the abnormal contact angle between the robot and the workpiece, uneven force distribution, or a mismatch between the assigned load parameters and the actual workpiece requirements are caused by pose deviation. Through this multi-dimensional comparison, task feedback information is ultimately generated. For example, if the actual pose of the first robot at the third station deviates from the target pose by 0.4mm, exceeding the 0.3mm accuracy requirement, and this deviation causes excessively rapid force increase during low-load testing, an anomaly warning is triggered.

[0062] After obtaining task feedback information, adaptive adjustments are made to the corresponding control parameters based on different types of feedback issues. On one hand, if the feedback indicates that the pose deviation is out of range, the pose control parameters are adjusted, for example, lowering the original positioning accuracy threshold from 0.3mm to 0.2mm, while optimizing the pose calibration sampling frequency, such as increasing it from 50Hz to 80Hz, to ensure the robot can correct pose deviations more frequently in subsequent operations. On the other hand, if the feedback shows that path planning causes excessive vibration when the robot reaches the workstation, affecting pose stability, the path planning parameters are adjusted, for example, reducing the robot's acceleration from 0.5m / s². 2 Reduced to 0.3 m / s2 At the same time, smooth transition nodes are added to the path, such as inserting two circular arc transition segments in a straight path to reduce motion impact.

[0063] On the other hand, if feedback indicates that the force threshold parameter is unreasonable, leading to abnormal misjudgments, the force threshold parameter will be adjusted. For example, the upper limit of the force for low-load testing will be adjusted from 3 N·m to 2.8 N·m, so that the force monitoring range better matches the actual force characteristics of the current workpiece. Each parameter adjustment is based on the specific problems in the task feedback information, ensuring that the adjusted parameters can specifically address the deficiencies in the operation.

[0064] After completing the adaptive adjustment of control parameters, the adjusted parameters are transformed into scheduling and control results executable by the robot swarm. First, based on the adjusted pose control parameters, the target pose accuracy requirements and calibration frequency for each robot at its corresponding workstation are redefined. For example, the second robot at the third workstation needs to achieve a positioning accuracy of 0.2mm and perform pose calibration every 10 seconds. Combined with the adjusted path planning parameters, the motion path details of the robots between multiple workstations are updated, including speed, acceleration, and transition node positions, to ensure a smoother path. Based on the adjusted force threshold parameters, the upper and lower limits of force monitoring and abnormal triggering conditions are set during operation. For example, during low-load testing, an early warning is triggered if the force value exceeds 2.8 N·m or falls below 1.5 N·m.

[0065] Finally, the scheduling and control results clearly define the order of task execution and coordination rules. For example, after the first robot completes its work at the first workstation, it must wait for the third robot to leave the passageway area of ​​the second workstation before initiating its movement to the second workstation, thus avoiding conflicts between multiple robot operations. The final scheduling and control results are presented in the form of an instruction list, clearly indicating the workstation, task requirements, motion parameters, and coordination rules for each robot, providing a clear basis for the multi-workstation scheduling and execution of a robot swarm.

[0066] By comparing the task allocation results, correcting the pose information, and identifying the anomaly, accurate task feedback is obtained. Parameters such as pose control, path planning, and force threshold are adjusted in a targeted manner. Then, based on the adjusted parameters, a scheduling control result containing specific execution instructions is generated. This achieves the effect of adapting the scheduling control result to actual task requirements and improving the accuracy and collaborative stability of multi-station operations of robot groups.

[0067] Furthermore, the method provided in this application embodiment includes:

[0068] Based on the scheduling and control results, the task scheduling requirements of the group of robots at multiple workstations are determined; under the task scheduling requirements, frequency optimization is performed with the environmental interference threshold as a constraint to obtain the frequency optimization result; and a multi-workstation collaborative scheduling scheme for the group of robots is generated based on the frequency optimization result.

[0069] In one embodiment, the scheduling control result includes the workstation allocation of the robot group, the specific task type (such as ultrasonic bolt inspection, visual scanning of parts), the task execution duration, and the equipment configuration requirements of each workstation. Based on this information, the core elements of the job scheduling requirements can be extracted. First, the target workstation and task attributes corresponding to each robot are extracted, and the equipment types required for different tasks are clarified (such as ultrasonic equipment for ultrasonic inspection, and visual equipment for visual scanning) and the basic frequency range of equipment operation are determined. Then, combined with the task execution duration, the working time window of each robot at each workstation is determined, and it is determined whether there are multiple robots or multiple types of equipment operating in parallel at adjacent workstations during the same time period.

[0070] Next, the spatial relationships of each workstation were analyzed to determine if there was a risk of frequency interference between devices due to the close proximity of workstations. For example, the scheduling control results instructed three robots to perform ultrasonic testing (requires ultrasonic equipment, base frequency range 2-5MHz) and visual scanning (requires visual equipment, base frequency range 10-20kHz) tasks at two adjacent workstations, and the operation time of all three robots was one hour consecutively. Therefore, the operation scheduling requirements were determined as follows: the two workstations must simultaneously support the operation of both types of equipment within their respective base frequency ranges, and frequency interference between adjacent workstations must be avoided within the one-hour operation window.

[0071] Next, given the clear job scheduling requirements, the environmental interference threshold is first determined. This threshold is the upper limit of the frequency difference pre-set based on the characteristics of the working environment (such as workstation spacing, air medium, and equipment anti-interference capability). When the operating frequency difference between two types of equipment is less than this threshold, signal superposition or attenuation will occur, affecting the job accuracy. Subsequently, frequency optimization is carried out. First, the basic frequency range of all participating equipment is listed, and then a greedy algorithm is used for frequency allocation: priority is given to allocating frequencies to equipment with a greater impact on job accuracy, such as ultrasonic testing equipment, whose frequency stability directly affects the accuracy of bolt defect detection. An initial frequency value is selected within its basic frequency range.

[0072] Next, frequencies are allocated to other devices. Each time a frequency is allocated, the difference between the new frequency and the already allocated device frequency is calculated, ensuring the difference is greater than the environmental interference threshold. If the currently selected frequency does not meet the constraints, a new frequency is selected within the device's base frequency range until all devices' frequencies meet the condition of being within their own base frequency range and having a frequency difference greater than the environmental interference threshold with other devices. This yields the final frequency optimization result. For example, if the environmental interference threshold is set to 1MHz, and the job scheduling requirements include two ultrasonic testing devices with base frequencies of 2-5MHz and one visual scanning device with a base frequency of 10-20kHz, first, 2.2MHz is allocated to the first ultrasonic testing device, and 3.5MHz is allocated to the second ultrasonic testing device. The frequency difference between the two is 1.3MHz > 1MHz. Then, 15kHz is allocated to the visual scanning device, with a frequency difference far exceeding 1MHz from both 2.2MHz and 3.5MHz, forming the final frequency optimization result.

[0073] Once the frequency optimization result is determined, it is combined with the operational details of the robot swarm to generate a multi-workstation collaborative scheduling scheme. First, the workstation, task type, and corresponding equipment operating frequency of each robot are associated to clarify the correspondence between robot-workstation-task-frequency. Then, combined with the task execution duration, specific operation time nodes are planned for each robot, including the time for equipment startup (initialization according to the optimized frequency), operation execution, and equipment shutdown, to avoid operation time conflicts between different robots at the same workstation.

[0074] Meanwhile, based on the spatial location of the workstation and the frequency characteristics of the equipment, the no-go zones for the robot's movement between workstations are marked, which are areas where the equipment signal may be abnormal due to frequency interference, to ensure that the robot does not enter these areas during its movement. Finally, this information is integrated into a structured scheduling scheme, which clarifies the complete work process of each robot, including time, location, frequency, actions, and coordination rules between multiple robots. For example, after a robot completes its work, it needs to send a signal before another robot can enter that workstation.

[0075] For example, based on the frequency optimization results, the first ultrasonic frequency is 2.2MHz, the second ultrasonic frequency is 3.5MHz, and the vision frequency is 15kHz. The generated collaborative scheduling scheme includes: the first robot equipped with the first ultrasonic frequency performs detection at 2.2MHz for 0-30 minutes at the first workstation; the second robot equipped with the second ultrasonic frequency performs detection at 3.5MHz for 30-60 minutes at the first workstation; and the third robot equipped with the vision device performs scanning at 15kHz for 0-60 minutes at the second workstation. The first and second robots need to be spaced 5 minutes apart when switching at the first workstation to complete the device frequency reset, and the robot movement path needs to avoid the area within 5 meters between the first and second workstations.

[0076] By decomposing the workstation, task type, and time window from the scheduling control results to determine the work scheduling requirements, a greedy algorithm is used to allocate frequencies to the equipment under the constraint of environmental interference threshold to obtain frequency optimization results. Then, the frequency, time, and path information are integrated to generate a structured scheme, which achieves the effect of ensuring that the robot group is free from frequency interference in multi-workstation operations, the work process is coordinated and orderly, and the overall work accuracy and efficiency are improved.

[0077] In summary, the robot group scheduling and control method for multi-station collaborative operation provided in this application has the following technical effects:

[0078] This application obtains and fuses visual and point cloud data from multiple workstations to determine the task pose information of a robot swarm. Through task allocation and path planning, and combining force-sensing filtering and force-displacement relationships, it completes near-field operation compliance correction for the robot swarm. Then, through low-load testing and force data trend analysis, it identifies operational anomalies. Subsequently, it adaptively adjusts parameters such as pose control, path planning, and force thresholds based on task feedback to generate scheduling control results. Finally, it optimizes the frequency using environmental interference thresholds as constraints, thus forming a multi-workstation collaborative scheduling scheme for the robot swarm. This achieves precise operation and efficient collaboration of the robot swarm across multiple workstations, meeting the high precision and high reliability requirements of multi-workstation scheduling control. It achieves the technical effects of precise pose control, timely anomaly identification, and efficient collaborative scheduling for multi-workstation robot swarms.

[0079] Example 2, as Figure 2 As shown, based on the same inventive concept as in Embodiment 1 above, this application provides a robot group scheduling and control system for multi-station collaborative operation, the system comprising:

[0080] Task pose information acquisition module 1 is used to acquire and fuse visual data and point cloud data from multiple workstations to determine the task pose information of the robot group at each workstation.

[0081] The task allocation result acquisition module 2 performs task allocation and path planning on the robot group based on the task pose information to obtain the task allocation result, and triggers backtracking and data re-acquisition when registration diverges.

[0082] The corrected pose information acquisition module 3 is used to obtain corrected pose information by performing compliant correction on the robot group in near-field operations based on the task allocation result, using force-sensing filtering and combining force-displacement relationship.

[0083] Anomaly identification result acquisition module 4 is used to perform low load testing and combine force data with trend analysis during the operation of the robot group after correction to obtain anomaly identification results.

[0084] The scheduling control result generation module 5 generates scheduling control results based on the job allocation results, corrected pose information, and anomaly identification results, combined with task feedback and adaptive parameter adjustment.

[0085] The collaborative scheduling scheme acquisition module 6 is used to perform frequency optimization based on the scheduling control results and with environmental interference threshold as a constraint, to obtain a multi-station collaborative scheduling scheme for the group robots.

[0086] Furthermore, the task pose information acquisition module 1 is used to perform the following steps:

[0087] The visual data includes two-dimensional image information collected from each workstation; the point cloud data includes three-dimensional spatial depth information collected from each workstation; based on the visual data and the point cloud data, the task pose information of the robot group at each workstation is determined through data registration and fusion processing.

[0088] Furthermore, the job assignment result acquisition module 2 is used to perform the following steps:

[0089] The task pose information is used to determine the work requirements of each workstation and assign them to the corresponding robots; the work requirements are used to generate the motion path of the robot group between multiple workstations; when registration divergence occurs during the path planning process, a rollback operation is triggered and the visual data and point cloud data are re-acquired.

[0090] Furthermore, the pose correction information acquisition module 3 is used to perform the following steps:

[0091] The robot swarm collects contact force data during near-field operations; the acquired contact force data is processed by the force-sensing filter to remove noise and obtain effective force; based on the effective force and the force-displacement relationship, the pose of the robot swarm is adjusted for compliance, thereby completing the compliance correction and obtaining the corrected pose information.

[0092] Furthermore, the anomaly identification result acquisition module 4 is used to perform the following steps:

[0093] During the operation of the robot swarm after correction, a low-load test is performed to obtain force data; the force data is preprocessed to extract effective force data for analysis; and trend analysis is performed based on the effective force data to obtain anomaly identification results.

[0094] Furthermore, the scheduling control result generation module 5 is used to perform the following steps:

[0095] The task allocation results, corrected pose information, and anomaly identification results are compared and analyzed to obtain task feedback information; the control parameters are adaptively adjusted based on the task feedback information, and the control parameters include at least one of pose control parameters, path planning parameters, and force threshold parameters; a scheduling control result is generated based on the adaptively adjusted control parameters, and the scheduling control result is used to instruct the group of robots to schedule and execute at multiple workstations.

[0096] Furthermore, the collaborative scheduling scheme acquisition module 6 is used to perform the following steps:

[0097] Based on the scheduling and control results, the task scheduling requirements of the group of robots at multiple workstations are determined; under the task scheduling requirements, frequency optimization is performed with the environmental interference threshold as a constraint to obtain the frequency optimization result; and a multi-workstation collaborative scheduling scheme for the group of robots is generated based on the frequency optimization result.

[0098] The robot group scheduling and control system for multi-station collaborative operation provided in the embodiments of the present invention can execute the robot group scheduling and control method for multi-station collaborative operation provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0099] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0100] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A multi-station collaborative work robot colony scheduling control method, characterized in that, The method comprises: Obtaining visual data and point cloud data of multiple work stations and fusing them to determine task pose information of the robot group at each work station; Based on the task pose information, task allocation and path planning are performed on the robot group to obtain work allocation results, and backtracking and data reacquisition are triggered when registration divergence occurs; According to the work allocation results, the robot group performs compliant correction in near-field work by using force perception filtering and combining force-displacement relationship to obtain corrected pose information; During the work process after the correction of the robot group, low-load testing is performed in combination with force data and trend analysis to obtain abnormal identification results; Based on the work allocation results, corrected pose information and abnormal identification results, task feedback and parameter adaptive adjustment are combined to generate scheduling control results; According to the scheduling control results, frequency optimization is performed with environmental interference threshold as the constraint to obtain a multi-station collaborative scheduling scheme of the robot group.

2. The method of claim 1, wherein, Obtaining visual data and point cloud data of multiple work stations and fusing them to determine task pose information of the robot group at each work station comprises: The visual data includes two-dimensional image information collected at each work station; The point cloud data includes three-dimensional spatial depth information collected at each work station; According to the visual data and the point cloud data, data registration and fusion processing are performed to determine the task pose information of the robot group at each work station.

3. The method of claim 1, wherein, Based on the task pose information, task allocation and path planning are performed on the robot group to obtain work allocation results, and backtracking and data reacquisition are triggered when registration divergence occurs, comprising: According to the task pose information, the work requirements of each work station are determined and allocated to the corresponding robot; Based on the work requirements, the motion path of the robot group between multiple work stations is generated; When registration divergence occurs during the path planning process, backtracking operation is triggered and the visual data and the point cloud data are reacquired.

4. The method of claim 1, wherein, According to the work allocation results, the robot group performs compliant correction in near-field work by using force perception filtering and combining force-displacement relationship to obtain corrected pose information, comprising: The robot group collects contact force data in near-field work; According to the obtained contact force data, the force perception filtering processing is performed to remove noise and obtain effective force; Based on the effective force and the force-displacement relationship, the pose of the robot group is adjusted for compliance, thereby completing the compliant correction and obtaining the corrected pose information.

5. The method of claim 1, wherein, During the work process after the correction of the robot group, low-load testing is performed in combination with force data and trend analysis to obtain abnormal identification results, comprising: During the work process after the correction of the robot group, low-load testing is performed to obtain force data; The force data is preprocessed to extract effective force data for analysis; Based on the effective force data, trend analysis is performed to obtain abnormal identification results.

6. The method of claim 1, wherein, Based on the work allocation results, corrected pose information and abnormal identification results, task feedback and parameter adaptive adjustment are combined to generate scheduling control results, comprising: The task assignment result, the corrected pose information and the abnormality identification result are compared and analyzed to obtain task feedback information; The control parameters are adaptively adjusted according to the task feedback information, and the control parameters include at least one of pose control parameters, path planning parameters and force threshold parameters; A scheduling control result is generated based on the adaptively adjusted control parameters, and the scheduling control result is used to instruct the group robot to perform scheduling in multiple stations.

7. The method of claim 1, wherein, According to the scheduling control result, frequency optimization is performed with the environmental interference threshold as a constraint to obtain a group robot multi-station collaborative scheduling scheme, including: Based on the scheduling control result, the work scheduling demand of the group robot in multiple stations is determined; With the environmental interference threshold as a constraint, frequency optimization is performed under the work scheduling demand to obtain a frequency optimization result; According to the frequency optimization result, the group robot multi-station collaborative scheduling scheme is generated.

8. A multi-station collaborative robot population scheduling control system, characterized by, Steps for implementing the robot group scheduling control method for multi-station collaborative work according to any one of claims 1 to 7, including: A task pose information acquisition module is configured to acquire visual data and point cloud data of multiple work stations and fuse the data to determine task pose information of a robot group in each station; A task assignment result acquisition module is configured to perform task assignment and path planning for the robot group based on the task pose information to obtain a task assignment result, and trigger backoff and data reacquisition when registration divergence occurs; A corrected pose information acquisition module is configured to perform soft correction by using force perception filtering and combining force-displacement relationship in near-field work of the robot group according to the task assignment result to obtain corrected pose information; An abnormality identification result acquisition module is configured to perform low-load testing and combine force data and trend analysis to obtain an abnormality identification result during the work performed by the robot group after correction; A scheduling control result generation module is configured to generate a scheduling control result based on the task assignment result, the corrected pose information and the abnormality identification result, and combine task feedback and parameter adaptive adjustment; A collaborative scheduling scheme acquisition module is configured to perform frequency optimization with the environmental interference threshold as a constraint according to the scheduling control result to obtain a group robot multi-station collaborative scheduling scheme.

Images