A method for online real-time dynamic calibration of a six-dimensional force sensor for robots

Abstract

This invention discloses an online real-time dynamic calibration method for a six-dimensional force sensor used in robots. This invention relates to the field of six-dimensional force sensor technology and solves the problem that existing methods often fail to perform in-depth analysis of deviation data from multiple calibrations, making it impossible to quickly locate the root cause of deviations. This invention uses a comparison of actual motion vector features with virtual vector features as the calibration trigger, rather than the fixed-cycle or manual judgment triggering mode of traditional methods. This allows for accurate identification of vector deviations caused by actual interference such as unevenness of the manipulated object and damping of components, initiating the calibration process only for robots with discrepancies. This design avoids the waste of time and computing power caused by "blind calibration when there is no deviation" and ensures "timely response when there is deviation," significantly improving the targeting and efficiency of calibration.

Description

Technical Field

[0001] This invention relates to the field of six-dimensional force sensor technology, specifically to an online real-time dynamic calibration method for a six-dimensional force sensor used in robots. Background Technology

[0002] As industrial automation rapidly develops towards higher precision and intelligence, robots are increasingly widely used in assembly, polishing, precision welding, and human-robot collaboration. As a core component for robots to achieve force perception and closed-loop force control, the measurement accuracy of six-dimensional force sensors directly determines the quality of robot operations.

[0003] However, in real industrial environments, the measurement results of six-dimensional force sensors are susceptible to interference from multiple factors, making it difficult to maintain stable accuracy. On the one hand, installation deviations during sensor assembly, the mass and center of gravity shift of the end effector, and changes in damping and friction during the operation of mechanical components directly introduce systematic errors. On the other hand, environmental interference such as uneven surfaces of the manipulated object and load fluctuations, as well as the sensor's own zero drift and temperature drift characteristics, further exacerbate measurement deviations. These problems cause significant differences between the robot's actual motion trajectory and the preset ideal trajectory (virtual running trajectory), severely affecting the accuracy of control tasks. For example, in precision assembly, it may lead to part jamming or damage, and in polishing operations, it may cause the workpiece surface roughness to exceed the standard.

[0004] Traditional six-dimensional force sensor calibration methods are mostly offline, requiring the robot to be disconnected from the production system and parameter calibration to be completed under fixed working conditions using dedicated calibration equipment (such as a six-degree-of-freedom platform and standard weights). This method not only disrupts the production process and is inefficient, but also cannot respond in real time to dynamic disturbances in actual operation (such as sudden loads or temperature changes), leading to a disconnect between calibration results and actual application scenarios. While some online calibration methods attempt to incorporate real-time data for correction, they are mostly limited to single-dimensional error compensation, lacking a systematic design that correlates "motion vector deviation - force / torque correction," making accurate dynamic calibration difficult. Furthermore, existing methods often fail to conduct in-depth analysis of deviation data from multiple calibrations, making it impossible to quickly locate the root cause of deviations (such as wear on specific components or defects in localized operating objects). This forces maintenance personnel to spend a significant amount of time troubleshooting, further increasing production and maintenance costs.

[0005] Therefore, developing an online real-time dynamic calibration technology for six-dimensional force sensors that can trigger calibration in real time, accurately correct parameters, and efficiently trace anomalies has become a key requirement for solving the current problems of insufficient force control accuracy, low calibration efficiency, and high maintenance difficulty in robots. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an online real-time dynamic calibration method for a six-dimensional force sensor used in robots. This method solves the problem that existing methods often fail to perform in-depth analysis of deviation data from multiple calibrations, thus failing to quickly pinpoint the root cause of the deviation.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for online real-time dynamic calibration of a six-dimensional force sensor for robots, comprising the following steps:

[0008] Step 1: Based on the operation instructions during the practical operation, confirm the motion vector characteristics associated with the robot during the practical operation. Then, based on the operation instructions, confirm the virtual vector characteristics of the robot in virtual space. Compare the confirmed motion vector characteristics with the virtual vector characteristics to determine whether the robot needs to perform a dynamic calibration process. The specific method is as follows:

[0009] The robot is controlled according to the operation instructions, and the running path associated with the robot monitoring node is monitored in real time. The monitored running path is combined with the set three-dimensional spatial coordinate system to confirm the displacement path associated with the monitoring node in different two-dimensional planes. The different displacement paths confirmed by multiple two-dimensional planes are used as the motion vector characteristics of the current robot.

[0010] Based on the relevant parameters associated with the operation instructions, within the set practical model, confirm the robot's associated running path within the practical model, and simultaneously use the same confirmation method as the motion vector characteristics to confirm the virtual vector characteristics associated with the robot within the practical model.

[0011] The confirmed motion vector features are compared with the virtual vector features to identify whether the two sets of vector features are completely consistent. If they are not completely consistent, the current robot is recorded as a robot to be calibrated.

[0012] If the two sets of vector features are completely identical, no processing is required;

[0013] Step 2: For robots requiring dynamic calibration, based on the motion vector characteristics and virtual vector characteristics of the actual operation, identify the motion point to be corrected, and based on the location of the motion point to be corrected, identify the surrounding reference points. Then, based on the difference characteristics between the motion point to be corrected and the reference points, identify the parameter correction characteristics of the motion point to be corrected. The specific method is as follows:

[0014] Based on the actual comparison process of motion vector features and virtual vector features, non-overlapping points are identified in different two-dimensional planes, and the identified non-overlapping points are recorded as motion points to be corrected.

[0015] Based on the two-dimensional plane where the point to be corrected is located, the displacement path associated with the point to be corrected is identified, and the point to be corrected is used as the center point. The reference point that is closest to the center point in terms of travel distance is identified on the displacement path. The reference point is the overlapping point that exists in the actual comparison process.

[0016] Based on the two-dimensional plane containing the reference point, identify the forces and moments associated with the reference point within the corresponding plane, and use them as standard forces and standard moments. The characteristic of the plane associated with the reference point is denoted as L. i Where i represents different axes in the two-dimensional plane, using: L i ÷ standard force = J1 i and L i ÷Standard torque = J2 i Confirm the measurement standard J1 associated with the standard force. i and the measurement standard J2 associated with the standard torque i Then, based on the associated motion point to be corrected, identify the associated point corresponding to the motion point to be corrected within the virtual vector feature, and confirm the straight-line distance between the motion point to be corrected and the associated point. The straight-line distance = feature distance of the motion point to be corrected - feature distance of the associated point.

[0017] Use: Straight-line distance ÷ J1 i =Correction force, linear distance ÷ J2 i =Correction torque, confirm the correction force and correction torque associated with the corresponding motion point to be corrected, and confirm the motion points to be corrected associated in different two-dimensional planes. Then, confirm the correction force and correction torque associated with different motion points to be corrected in turn, and generate the parameter correction features of the corresponding motion points to be corrected.

[0018] Step 3: Correct the features based on the different parameters associated with different motion points to be corrected, re-execute the operation instructions, and repeat Step 1 and Step 2 until the robot no longer needs to perform the dynamic calibration process.

[0019] Step 4: Identify the motion points to be corrected from different dynamic calibration processes, and confirm the clustering characteristics of different motion points to be corrected in three-dimensional space. Identify and display abnormal clusters. The specific method is as follows:

[0020] The motion points to be corrected in each dynamic calibration process are identified, and the location of each motion point to be corrected is confirmed in combination with the preset three-dimensional spatial coordinate system.

[0021] A set of points to be corrected is randomly selected and denoted as the midpoint of a sphere. The sphere is then treated as a ball with a radius R, which is a preset value. The space inside the sphere is denoted as the spherical space, and the points to be corrected within this space are denoted as embedded points. The number of embedded points is denoted as G.k , where k represents the ball at different positions;

[0022] And from the confirmed Gs k In the middle, select the maximum value G. k The sphere associated with max is selected, and the midpoint of the selected sphere's sphere is recorded as the outlier cluster point. The confirmed outlier cluster points are then displayed. If G... k If there are duplicate values ​​for max, then each G... k All spheres associated with max are selected spheres, and the midpoint of the sphere of each selected sphere is recorded as an anomaly clustering point.

[0023] Preferably, the selected ball also includes another method of determination:

[0024] Based on the preset spherical radius R, determine the volume of the sphere, and determine the different values ​​of G associated with each different position of the sphere. k , using: G k ÷ Volume = M k ;

[0025] Will satisfy: M k A ball with a value greater than Y1 is designated as the selected ball, where Y1 is the preset value; otherwise, no marking is performed.

[0026] This invention provides an online, real-time, dynamic calibration method for a six-dimensional force sensor used in robots. Compared with existing technologies, it has the following advantages:

[0027] This invention uses "comparison of actual motion vector features and virtual vector features" as the calibration trigger, rather than the fixed-cycle or manual judgment trigger mode in traditional methods. It can accurately identify vector deviations caused by actual interferences such as unevenness of the operating object and damping of components, and only initiate the calibration process for robots with discrepancies. This design avoids the waste of time and computing power caused by "blind calibration when there is no deviation" and ensures "timely response when there is deviation", which significantly improves the targeting and efficiency of calibration.

[0028] By associating the "point to be corrected with the reference point," J1i and J2i measurement standards are constructed based on the standard force and standard torque of the reference point. Then, the correction force and correction torque are calculated by combining the straight-line distance between the point to be corrected and the virtual associated point, realizing a closed loop of "deviation quantification - parameter correspondence - precise compensation." At the same time, by adjusting the increase or decrease of stress / torque through positive and negative distances, the correction direction is ensured to match the actual deviation. Furthermore, by repeatedly executing the "comparison-correction" process, the robot's motion vector is continuously driven to converge towards the virtual standard, effectively reducing the impact of assembly errors, environmental interference, and other factors on the six-dimensional force sensor measurement, and significantly improving the accuracy and stability of the robot's force control operation.

[0029] By mapping the motion points to be corrected in multiple calibrations to three-dimensional space, high-frequency deviation areas are filtered out using spherical clustering, and the midpoint of the sphere is displayed as anomaly cluster points. This design transforms scattered "point deviations" into concentrated "regional anomalies," enabling maintenance personnel to quickly locate the root cause of problems (such as component wear in specific areas, local defects in the operated object, etc.) without having to check massive amounts of calibration data one by one, significantly shortening the fault diagnosis time. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0031] 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 embodiments of the present invention, and not all embodiments. 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.

[0032] Please see Figure 1 This application provides an online real-time dynamic calibration method for a six-dimensional force sensor used in robots, comprising the following steps:

[0033] Step 1: Based on the operation instructions during the practical operation, confirm the motion vector characteristics associated with the robot during the practical operation. Then, based on the operation instructions, confirm the virtual vector characteristics of the robot in the virtual space. Compare the confirmed motion vector characteristics with the virtual vector characteristics to determine whether the robot needs to perform a dynamic calibration process. Specifically, there are specific monitoring points during the actual movement of the robot. These monitoring points are generally set at the center of the robot arm. Based on the specific operation process, the movement process of the monitoring points can be confirmed first to confirm the motion vector path in the three-dimensional space. At the same time, based on the corresponding operation instructions, there are also related virtual running paths in the corresponding virtual space. These paths are specific paths that exist in the corresponding virtual model.

[0034] Specifically, the method for identifying whether this robot needs to perform a dynamic calibration process is as follows:

[0035] The robot is controlled according to the operation instructions (which are preset instructions and are prepared in advance by relevant personnel), and the running path associated with the robot's monitoring nodes is monitored in real time. The monitored running path is combined with the set three-dimensional spatial coordinate system to confirm the displacement path associated with the monitoring node in different two-dimensional planes. The different displacement paths confirmed by multiple two-dimensional planes are used as the motion vector characteristics of the current robot.

[0036] Based on the relevant parameters associated with the operation instructions, within the set practical model (which is a preset model, pre-set by relevant personnel), confirm the robot's associated running route within the practical model, and simultaneously use the same confirmation method of motion vector characteristics to confirm that the virtual vector characteristics associated with the robot within the practical model have the same three-dimensional spatial coordinate system. The three-dimensional spatial coordinate point is generally a fixed point on the robot, which is convenient for comparing and verifying between vector characteristics.

[0037] The confirmed motion vector features are compared with the virtual vector features to identify whether the two sets of vector features are completely consistent. If they are completely consistent, no processing is required. If they are not completely consistent, the current robot is marked as a robot to be calibrated, which means that this robot needs to perform the subsequent dynamic calibration process. In actual operation, there are related errors, which cause the corresponding robot to not achieve a good accuracy. Specifically, by comparing and verifying the actual operation process with the virtual process, the errors existing in the robot in the actual operation process can be effectively identified. This is because there are related factors that interfere in the actual operation process, such as: the operation object is uneven, or the internal components of the robot have damping in the actual operation process, which leads to a large error between their vectors.

[0038] Step 2: For robots that need to perform dynamic calibration, identify the motion point to be corrected based on the motion vector characteristics and virtual vector characteristics of the actual operation process, identify the reference points in the vicinity based on the location of the motion point to be corrected, and then identify the parameter correction characteristics of the motion point to be corrected based on the difference characteristics between the motion point to be corrected and the reference points.

[0039] The specific method for confirming the parameter correction features of the motion point to be corrected is as follows:

[0040] Based on the actual comparison process of motion vector features and virtual vector features, non-overlapping points are identified in different two-dimensional planes, and the identified non-overlapping points are recorded as motion points to be corrected.

[0041] Based on the two-dimensional plane where the point to be corrected is located, the displacement path associated with the point to be corrected is identified, and the point to be corrected is used as the center point. The reference point that is closest to the center point in terms of travel distance is identified on the displacement path. The reference point is the overlapping point that exists in the actual comparison process.

[0042] Based on the two-dimensional plane containing the reference point, identify the forces and moments associated with the reference point within the corresponding plane, and use them as standard forces and standard moments. The characteristic of the plane associated with the reference point is denoted as L. iWhere i represents different axes in the two-dimensional plane (that is, the straight-line distance from the measurement axis in the two-dimensional plane; if it is the X and Y axis plane, then it is the distance associated with the X and Y axes), using: L i ÷ standard force = J1 i and L i ÷Standard torque = J2 i Confirm the measurement standard J1 associated with the standard force. i and the measurement standard J2 associated with the standard torque i Then, based on the associated motion point to be corrected, identify the associated point corresponding to the motion point to be corrected within the virtual vector feature (the associated point is consistent with the motion point to be corrected in terms of motion features; if the X-axis is used as the reference standard during the movement, then the X-coordinate of the associated point and the motion point to be corrected are the same coordinate; if the Y-axis is used as the reference standard, then the Y-coordinate is the same coordinate; all reference standards are preset standards, preset by the operator in advance). Confirm the straight-line distance between the motion point to be corrected and the associated point. The straight-line distance = the feature distance of the motion point to be corrected - the feature distance of the associated point. The feature distance can be directly obtained during actual operation and belongs to the straight-line height relative to the reference standard during operation.

[0043] Use: Straight-line distance ÷ J1 i =Correction force, linear distance ÷ J2 i =Correction torque, confirm the correction force and correction torque associated with the corresponding motion point to be corrected, and confirm the motion points to be corrected associated in different two-dimensional planes. Then, confirm the correction force and correction torque associated with different motion points to be corrected in turn, and generate the parameter correction features of the corresponding motion points to be corrected.

[0044] If the straight-line distance is positive, it means that the confirmed correction force and correction torque are also positive. In other words, in the corresponding two-dimensional plane, the corresponding force and torque are added in the direction of motion to ensure the standard of the motion vector process. If the straight-line distance is negative, it means that the force and torque are reduced in the corresponding direction of motion.

[0045] Step 3: Correct the features based on the different parameters associated with different motion points to be corrected, re-execute the operation instructions, and repeat Step 1 and Step 2 until the robot no longer needs to perform the dynamic calibration process.

[0046] Step 4: Identify the motion points to be corrected from different dynamic calibration processes, and confirm the clustering characteristics of different motion points to be corrected in three-dimensional space. Lock down abnormal clusters and display them. Subsequently, relevant personnel can quickly find the problem based on these abnormal clusters, identify whether there are any abnormalities in the associated components or operation objects, and make timely adjustments.

[0047] The specific method for locking down abnormal cluster points is as follows:

[0048] The motion points to be corrected in each dynamic calibration process are identified, and the location of each motion point to be corrected is confirmed in combination with the preset three-dimensional spatial coordinate system.

[0049] A set of points to be corrected is randomly selected and denoted as the midpoint of a sphere. The sphere is then treated as a ball with radius R, which is a preset value determined in advance by relevant personnel based on experience. The space inside the sphere is denoted as the spherical space, and the points to be corrected within this space are denoted as embedded points. The number of embedded points is denoted as G. k , where k represents the ball at different positions;

[0050] And from the confirmed Gs k In the middle, select the maximum value G. k The sphere associated with max is selected, and the midpoint of the selected sphere's sphere is recorded as the outlier cluster point. The confirmed outlier cluster points are then displayed. If G... k If there are duplicate values ​​for max, then each G... k All spheres associated with max are selected spheres, and the midpoint of the sphere of each selected sphere is recorded as an anomaly cluster point;

[0051] Specifically, it also includes another method for determining the selected ball:

[0052] Based on the preset spherical radius R, determine the volume of the sphere, and determine the different values ​​of G associated with each different position of the sphere. k G k ÷ Volume = M k ;

[0053] Will satisfy: M k Balls with a value greater than Y1 are designated as selected balls, where Y1 is a preset value. The specific value of Y1 is determined in advance by the operator based on experience. Otherwise, no marking is performed.

[0054] Specifically, in the actual operation process, there are spheres at different locations where the movement points to be corrected exist. Each sphere at a different location contains movement points at different locations. Based on the actual confirmation process, the clustering points can be effectively identified, and the points of abnormal clustering can be specifically displayed, which is convenient for subsequent maintenance personnel to carry out maintenance and repair.

[0055] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0056] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for online real-time dynamic calibration of a six-dimensional force sensor for robots, characterized in that, Includes the following steps: Step 1: Based on the operation instructions during the practical operation, confirm the motion vector characteristics associated with the robot during the practical operation. Then, based on the operation instructions, confirm the virtual vector characteristics of the robot in the virtual space. Compare the confirmed motion vector characteristics with the virtual vector characteristics to identify whether the robot needs to perform a dynamic calibration process. Step 2: For robots requiring dynamic calibration, based on the motion vector characteristics and virtual vector characteristics of the actual operation, identify the motion point to be corrected, and based on the location of the motion point to be corrected, identify the surrounding reference points. Then, based on the difference characteristics between the motion point to be corrected and the reference points, identify the parameter correction characteristics of the motion point to be corrected. The specific method is as follows: Based on the two-dimensional plane containing the reference point, identify the forces and moments associated with the reference point within the corresponding plane, and use them as standard forces and standard moments. The characteristic of the plane associated with the reference point is denoted as L. i Where i represents different axes in the two-dimensional plane, L i The straight-line distance from the axis of measurement in a two-dimensional plane is represented by: L i ÷ standard force = J1 i and L i ÷Standard torque = J2 i Confirm the measurement standard J1 associated with the standard force. i and the measurement standard J2 associated with the standard torque i Then, based on the associated motion point to be corrected, identify the associated point corresponding to the motion point to be corrected within the virtual vector feature, and confirm the straight-line distance between the motion point to be corrected and the associated point. The straight-line distance = feature distance of the motion point to be corrected - feature distance of the associated point. Use: Straight-line distance ÷ J1 i =Correction force, linear distance ÷ J2 i =Correction torque, confirm the correction force and correction torque associated with the corresponding motion point to be corrected, and confirm the motion points to be corrected associated in different two-dimensional planes. Then, confirm the correction force and correction torque associated with different motion points to be corrected in turn, and generate the parameter correction features of the corresponding motion points to be corrected. Step 3: Correct the features based on the different parameters associated with different motion points to be corrected, re-execute the operation instructions, and repeat Step 1 and Step 2 until the robot no longer needs to perform the dynamic calibration process.

2. The method for online real-time dynamic calibration of a six-dimensional force sensor for robots according to claim 1, characterized in that, In step one, the specific method for identifying whether the robot needs to perform a dynamic calibration process is as follows: The robot is controlled according to the operation instructions, and the running path associated with the robot monitoring node is monitored in real time. The monitored running path is combined with the set three-dimensional spatial coordinate system to confirm the displacement path associated with the monitoring node in different two-dimensional planes. The different displacement paths confirmed by multiple two-dimensional planes are used as the motion vector characteristics of the current robot. Based on the relevant parameters associated with the operation instructions, within the set practical model, confirm the robot's associated running path within the practical model, and simultaneously use the same confirmation method as the motion vector characteristics to confirm the virtual vector characteristics associated with the robot within the practical model. The confirmed motion vector features are compared with the virtual vector features to identify whether the two sets of vector features are completely consistent. If they are not completely consistent, the current robot is recorded as a robot to be calibrated.

3. The method for online real-time dynamic calibration of a six-dimensional force sensor for robots according to claim 2, characterized in that, In step one, if the two sets of vector features are completely identical, no processing is required.

4. The method for online real-time dynamic calibration of a six-dimensional force sensor for robots according to claim 1, characterized in that, In step two, the specific method for confirming the motion point to be corrected and the reference point is as follows: Based on the actual comparison process of motion vector features and virtual vector features, non-overlapping points are identified in different two-dimensional planes, and the identified non-overlapping points are recorded as motion points to be corrected. Based on the two-dimensional plane where the point to be corrected is located, the displacement path associated with the point to be corrected is identified. Using the point to be corrected as the center point, the reference point closest to the center point in terms of travel distance is identified on the displacement path. The reference point is the overlapping point that exists in the actual comparison process.

5. The method for online real-time dynamic calibration of a six-dimensional force sensor for robots according to claim 4, characterized in that, It also includes the following steps: Step 4: Identify the motion points to be corrected from different dynamic calibration processes, and identify the clustering characteristics of different motion points to be corrected in three-dimensional space, and lock down abnormal clusters for display.

6. The method for online real-time dynamic calibration of a six-dimensional force sensor for robots according to claim 5, characterized in that, In step four, the specific method for locking down abnormal cluster points is as follows: The motion points to be corrected in each dynamic calibration process are identified, and the location of each motion point to be corrected is confirmed in combination with the preset three-dimensional spatial coordinate system. A set of points to be corrected is randomly selected and denoted as the midpoint of a sphere. The sphere is then treated as a ball with a radius R, which is a preset value. The space inside the sphere is denoted as the spherical space, and the points to be corrected within this space are denoted as embedded points. The number of embedded points is denoted as G. k , where k represents the ball at different positions; And from the confirmed Gs k In the middle, select the maximum value G. k The sphere associated with max is selected, and the midpoint of the selected sphere's sphere is recorded as the outlier cluster point. The confirmed outlier cluster points are then displayed. If G... k If there are duplicate values ​​for max, then each G... k All spheres associated with max are selected spheres, and the midpoint of the sphere of each selected sphere is recorded as an anomaly clustering point.

7. The method for online real-time dynamic calibration of a six-dimensional force sensor for robots according to claim 6, characterized in that, The selected ball also includes another method of determination: Based on the preset spherical radius R, determine the volume of the sphere, and determine the different values ​​of G associated with each different position of the sphere. k , using: G k ÷ Volume = M k ; Will satisfy: M k A ball with a value greater than Y1 is designated as the selected ball, where Y1 is the preset value; otherwise, no marking is performed.

Images