Calibration method of touch sensor and object surface measurement method

Abstract

The invention discloses a calibration method of a tactile sensor and an object surface measurement method, and the method comprises the steps: determining a sensor contact point through a calibration needle; according to the calibration method for calibrating the initial pose of the mechanical arm and converting all subsequent poses into the initial pose of the mechanical arm, the influence of the initial absolute pose error of the mechanical arm and the absolute pose error of placement of a calibration base on the calibration result at the beginning of calibration is avoided, the normal vector detection precision of the surface of the sensor is improved, and when the tactile sensor with the irregular surface is calibrated, the calibration precision is improved. The normal vector positions of all positions of the surface of the tactile sensor are determined, sensor deformation caused in the manufacturing and using process can be corrected, or the tactile sensor with the irregular surface can be directly calibrated, finally, the surface of the sensor can be accurately modeled, and subsequent normal vector measurement is more accurate.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of three-dimensional reconstruction of objects operated by robots, in particular to a calibration method of a tactile sensor and an object surface measurement method. BACKGROUND

[0002] The tactile sensor calculates the size and direction of the contact force by the deformation of the contact surface when it contacts with the object. The surface of the tactile sensor of the type of tactile, magnetic force, etc. is usually a soft and deformable material such as silicone. When the soft surface contacts with the object, the deformation occurs, the image and magnetic field signal obtained inside the sensor change, and the size and direction of the contact force can be solved by the neural network according to the changed signal. In addition, the tactile sensor is also used to measure the normal vector of the contact surface, and the normal vector information of different positions of the measured object can be used to model the measured object. However, the existing technology can only use the sensor of the regular curved surface to detect the normal vector, and cannot use the non-regular curved surface to accurately detect the normal vector. At the same time, due to the manufacturing error of the sensor, the deformation of the material during use and other factors, the surface shape of the sensor designed as a regular surface shape deviates from the actual use, which affects the accuracy of subsequent measurement. SUMMARY The present application aims to provide a calibration method of a tactile sensor and an object surface measurement method to solve one or more technical problems in the prior art and at least provide a beneficial choice or create conditions.

[0003] The technical scheme adopted to solve the above technical problems is as follows: The present application provides a calibration method of a tactile sensor, which is suitable for a calibration device and a mechanical arm driving the movement of the tactile sensor, wherein the calibration device comprises a calibration base and a calibration needle, the calibration base comprises a calibration part and a fixed part arranged in sequence along the Z axis, the calibration part is a hemisphere with a diameter of R, a mounting position is arranged at the center position of the end of the calibration part away from the fixed part, the calibration needle is detachably mounted on the mounting position along the Z axis, the length of the calibration needle exposed from the calibration part is L, and the calibration method comprises: controlling the movement of the end of the mechanical arm to control the movement of the tactile sensor to just contact the end of the calibration needle away from the calibration part, recording the initial pose of the end of the mechanical arm at this moment , the pixel position of the contact point between the tactile sensor and the calibration needle on the tactile image of the tactile sensor ; subtracting L from the initial pose on the Z axis to obtain the corrected pose of the mechanical arm when the surface of the tactile sensor just contacts the calibration part ; by calculating the pixel position ; ; controlling to remove the calibration needle, controlling the end of the mechanical arm to move to other positions of the surface of the tactile sensor to contact the hemispherical surface of the calibration part; record the current contact pose of the end of the mechanical arm ; record the current physical position of the current contact point between the tactile sensor and the calibration part on the tactile image of the tactile sensor ; calculate the physical zero point and the current physical position The change matrix in the three-dimensional space , wherein the change matrix contains a rotation change R and a translation t. calculate the transformation matrix of the current contact pose relative to the corrected pose The transformation matrix , wherein the transformation matrix contains a rotation transformation matrix and a translation transformation ; calculate the normal vector of the calibration part at the current contact point; calculate the sensor normal vector at the current contact point of the tactile sensor; fit the collected contact point cloud and sensor normal vector data of the tactile sensor to obtain a sensor surface model.

[0004] The calibration method of the application has the following advantages: The calibration method of the application determines the sensor contact point by using the calibration needle, and converts all subsequent poses to the initial pose of the mechanical arm, which avoids the influence of the initial absolute pose error of the mechanical arm and the absolute pose error of the calibration base on the calibration result at the beginning of calibration, improves the detection accuracy of the normal vector of the sensor surface, and determines the position of the normal vector of the tactile sensor surface at each position when calibrating the tactile sensor on an irregular surface. Using this method, the deformation of the sensor caused by manufacturing and use can be corrected, or the tactile sensor on an irregular surface can be directly calibrated, and finally the sensor surface can be accurately modeled, so that the subsequent normal vector measurement is more accurate.

[0005] As a further improvement of the above technical solution, the calculation of the normal vector of the calibration part at the current contact point comprises: establish a coordinate system with the center of the calibration part as the coordinate origin, then the installation position coordinate on the calibration part is: ; The coordinates of the current contact point in the spatial coordinate system are: ; The normal vector of the current contact point on the calibration unit is:

[0006] in, for The modulus length; The normal vector containing the starting point position information of the normal vector is denoted as: ,in, The components of the current contact point on the three coordinate axes in space. The components of the normal vector of the current contact point on the three coordinate axes in space; The calculation of the sensor normal vector at the current contact point of the tactile sensor includes: Since the sensor normal vector at the current contact point of the tactile sensor is set to have opposite directions to the normal vector on the calibration unit, the sensor normal vector at the current contact point of the tactile sensor is: .

[0007] As a further improvement to the above technical solution, the initial pose... The pose information includes three positional degrees of freedom. and three degrees of freedom of posture ; The initial pose is calculated based on the length L of the calibration needle protruding from the calibration part. The correction is performed to determine the corrected pose of the robotic arm when the surface of the tactile sensor just contacts the calibration unit. ,include: The corrected pose The positional degree of freedom Z-axis component and the initial pose The positional degrees of freedom Z-axis components are related as follows The values ​​of other degrees of freedom are the same.

[0008] As a further improvement to the above technical solution, the step of calculating the pixel position... The physical zero point is calculated. ,include: Based on the preset pixel-to-physical distance conversion relationship, the pixel position is... The units are converted to physical lengths to obtain the physical zero point, denoted as . ).

[0009] As a further improvement to the above technical solution, the current physical location is denoted as... ; said physical zero point with said current physical position a change matrix in three-dimensional space , comprising: since and are both points on a two-dimensional image data, there is no rotation change relationship between them, and the rotation change R degenerates into a unit matrix:

[0010] Similarly, and There is no change relationship on the z-axis, only movement on the xy plane, so the translation transformation matrix t is:

[0011] said physical zero point with said current physical position a change matrix in three-dimensional space

[0012] As a further improvement of the above technical solutions, said calculating said current contact pose with respect to said modified pose a transformation matrix , comprising: during the process of said mechanical arm driving said tactile sensor from said modified pose to said current contact pose : Let the rotation angle of said mechanical arm around the Roll axis be , the rotation angle around the Pitch axis be , and the rotation angle around the Yaw axis be , then the rotation transformation matrix is: ; wherein:

[0013]

[0014]

[0015] And said translation transformation is:

[0016] wherein, is the position component of , is a position component of the current contact pose; the current contact pose is transformed to the modified contact pose by a transformation matrix:

[0017] As a further improvement of the above technical solution, the calculation of the sensor normal vector at the current contact point of the tactile sensor further comprises: According to the formula: , the normal vector at the current contact point of the tactile sensor is transformed into the coordinate system with the modified contact pose as the reference point; wherein is the inverse matrix of the transformation matrix , and is:

[0018] The application also provides an object surface measurement method, which is suitable for a calibration device and a mechanical arm driving a tactile sensor to move, wherein the calibration device comprises a calibration base and a calibration needle, the calibration base comprises a calibration part and a fixed part arranged in sequence along the Z axis, the calibration part is a hemisphere with a diameter of R, a mounting position is arranged at the center of the end of the calibration part away from the fixed part, the calibration needle is detachably mounted on the mounting position along the Z axis, and the length of the calibration needle exposed from the calibration part is L, and the object surface measurement method comprises: controlling the movement of the end of the mechanical arm to control the movement of the tactile sensor to contact different positions of an object; recording the current end pose of the mechanical arm , the contact position on the tactile sensor ; finding the corresponding sensor normal vector at the contact position according to the sensor surface model ; taking the current end pose as the normal vector starting point coordinate, transforming the sensor normal vector to the end pose , and transforming the surface normal vector position of the tactile sensor according to the different contact points to obtain the object contact point normal vector; fitting the collected object contact point cloud and object contact point normal vector data to obtain the surface model of the measured object.

[0019] As a further improvement of the above technical solution, the current end pose is taken as the normal vector starting point coordinate, and the sensor normal vector Transform to the end pose The surface normal vector position of the tactile sensor is changed according to different contact points to obtain the normal vector of the object contact point, including: According to the formula The normal vector of the object's contact point is calculated. ,in, The current end effector pose of the robotic arm Corrected pose relative to the robotic arm The transformation matrix, the transformation matrix Includes rotational changes Translation , set as .

[0020] As a further improvement to the above technical solution, the end-effector pose... The pose information includes three positional degrees of freedom. and three degrees of freedom of posture ; The object surface measurement method further includes: The end effector of the robotic arm is controlled to move so that the tactile sensor is brought into contact with the end of the calibration pin that is furthest from the calibration point, and the initial pose of the end effector of the robotic arm at this moment is recorded. The initial pose Subtract L on the Z-axis to obtain the corrected pose of the robotic arm when the surface of the tactile sensor just contacts the calibration unit. ; Calculate the current end effector pose of the robotic arm. Corrected pose relative to the robotic arm Transformation matrix : The robotic arm moves the tactile sensor from the corrected position. Move to end pose During the process: Let the rotation angle of the robotic arm around the Roll axis be denoted as The rotation angle around the Pitch axis is denoted as The rotation angle around the Yaw axis is denoted as Then the rotation transformation matrix is: ; in:

[0021]

[0022]

[0023] And the translation of the transformation is:

[0024] wherein, is the position component of the end pose with respect to the modified pose is:

[0025] Setting , the object surface normal vector containing point cloud coordinates is: .

[0026] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent from the description, or can be learned by practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0027] The present application will be further described below in conjunction with the accompanying drawings and examples; Figure 1 is a schematic diagram of an embodiment of the calibration device provided by the present application; Figure 2 is a flowchart of an embodiment of the calibration method provided by the present application; Figure 3 is a schematic diagram of the state of the tactile sensor contacting the calibration needle during the calibration process provided by the present application; Figure 4 is a schematic diagram of the state of the other position of the tactile sensor contacting the calibration part during the calibration process provided by the present application; Figure 5 is Figure 4 a partial enlarged view in Figure 6 is a flowchart of an embodiment of the object surface measurement method provided by the present application. DETAILED DESCRIPTION

[0028] Embodiments of the present application will be described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference signs represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by reference to the accompanying drawings are exemplary and are only used to explain the present application and cannot be understood as a limitation of the present application.

[0029] In the description of the present application, it should be understood that the orientation description, such as the orientation or position relationship indicated by up, down, etc., is based on the orientation or position relationship shown in the drawings, only for the purpose of facilitating the description of the present application and simplifying the description, and does not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation of the present application.

[0030] In the description of the present application, more refers to more than two. If there is a description of first, second, it is only for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features or implicitly indicating the sequence of indicated technical features.

[0031] In the description of the present application, unless otherwise explicitly limited, the words such as arrangement, installation, connection, etc. should be broadly understood, and the person skilled in the art can reasonably determine the specific meaning of the above words in the present application in combination with the specific content of the technical scheme.

[0032] The technical scheme of the present application will be described below in conjunction with the drawings. Obviously, the following described embodiments are part of the embodiments of the present application, not all embodiments.

[0033] For the widely existing irregular surface tactile sensor 205, due to the complex shape of the sensor surface itself, it is difficult to solve the contact surface normal vector through simple geometric calculation, and at the same time due to the existence of sensor manufacturing error, material deformation in use and other factors, the surface shape of the sensor designed as a regular surface shape deviates from the designed shape in actual use, affecting the accuracy of subsequent measurement, the present application provides a calibration method for tactile sensor 205, which can accurately obtain the distribution of the normal vector of the sensor surface in combination with the calibration device.

[0034] The calibration method for the tactile sensor 205 of the present application is suitable for the calibration device, the mechanical arm 204 driving the tactile sensor 205 to move.

[0035] As Figure 1 The calibration device of the present application comprises a calibration base 102 and a calibration needle 101, the calibration base 102 comprises a calibration part 1021 and a fixed part 1022 arranged in sequence along the Z axis, the calibration part 1021 is a hemisphere with a diameter of R, and a mounting position is provided at the center position of the end of the calibration part 1021 away from the fixed part 1022, the calibration needle 101 is detachably mounted on the mounting position along the Z axis, and the length of the calibration needle 101 exposed from the calibration part 1021 is L.

[0036] It is understood that the calibration base 102 is an integral part consisting of an upper hemisphere with a diameter of R and a lower cylinder. In this embodiment, the mounting position is a circular hole 103. The diameter of the calibration needle 101 is the same as the diameter of the circular hole 103. The length of the calibration needle 101 is longer than the depth of the circular hole 103. The calibration needle 101 can be inserted into the circular hole 103 and protrude a fixed length L.

[0037] In some other embodiments, the calibration needle 101 may be detachably mounted on the calibration base 102 in other ways.

[0038] like Figure 2 As shown, the calibration method of the present invention includes, but is not limited to, the following steps: Step S100: Control the end effector of the robotic arm 204 to move so that the tactile sensor 205 just contacts the end of the calibration pin 101 away from the calibration part 1021, and record the initial pose of the end effector of the robotic arm 204 at this moment. The pixel position of the contact point between the tactile sensor 205 and the calibration needle 101 on the tactile image of the tactile sensor 205. ; Step S200: Set the initial pose Subtract L on the Z-axis to obtain the corrected pose of the robotic arm 204 when the surface of the tactile sensor 205 just contacts the calibration unit 1021. ; Step S300: Calculate the pixel position The physical zero point is calculated. ; Step S400: Control the removal of calibration needle 101, and control the end of robotic arm 204 to move to another position on the surface of tactile sensor 205 to contact the hemispherical surface of calibration part 1021; Step S500: Record the current contact pose of the end effector of the robotic arm 204. ; Step S600: Record the current physical position of the current contact point between the tactile sensor 205 and the calibration unit 1021 on the tactile image of the tactile sensor 205. ; Step S700: Calculate the physical zero point With current physical location Transformation matrix in three-dimensional space ), where the change matrix Includes rotational transformation R and translational transformation t; Step S800: Calculate the current contact pose Relative to pose correction Transformation matrix , where the transformation matrix Includes rotation transformation matrix Translation ; Step S900: Calculate the normal vector of the calibration unit 1021 at the current contact point; Step S1000: Calculate the sensor normal vector at the current contact point of the tactile sensor 205; Step S1110: Fit the contact point cloud and sensor normal vector data of the collected tactile sensor 205 to obtain the sensor surface model.

[0039] In step S100, before calibration begins, the tactile sensor 205 is fixed to the end of the robotic arm 204. A calibration base 102 is fixed below the robotic arm 204 with the tactile sensor 205 mounted. A calibration pin 101 is inserted into the calibration base 102. The tactile sensor 205 is then installed at the end of the robotic arm 204. The robotic arm 204 is moved horizontally until it and the tactile sensor 205 are directly above the calibration base 102. The robotic arm 204 is then lowered, causing it to gradually approach the calibration pin 101. When the tactile sensor 205 just touches the calibration pin 101, the movement of the robotic arm 204 is stopped. Figure 3 As shown, at this moment, the tactile sensor 205 is in contact with the calibration needle 101, and a contact image 202 between the sensor and the calibration needle 101 appears on the signal of the tactile sensor 205. Simultaneously, the diameter of the calibration needle 101 is much smaller than the diameter of the calibration base 102. When it contacts the tactile sensor 205, the contact image left by the calibration needle 101 is smaller. Therefore, using the calibration needle 101 allows for more accurate acquisition of the initial contact position of the tactile sensor 205. Furthermore, due to the smaller diameter of the calibration needle 101, the contact image positions generated by the tactile sensor 205 contacting the calibration needle 101 at different angles are more similar, avoiding uncertainties caused by installation errors of the tactile sensor 205 and movement errors of the robotic arm 204.

[0040] like Figure 3 The initial pose 203 of the end effector of the robotic arm 204 shown is recorded at this moment. Initial pose The pose information includes three positional degrees of freedom. and three degrees of freedom of posture .

[0041] like Figure 3 The image 202 showing the contact between the sensor and the calibration needle 101 records the pixel position of the contact point between the tactile sensor 205 and the calibration needle 101 on the tactile image of the tactile sensor 205. Typically, the top left corner is used as the origin of the coordinate system to record the position of the contact point within the entire coordinate system. For example, [the following text is incomplete and likely refers to a different topic:]Figure 3 The contact image 202 recorded at the sensor and the calibration needle 101 is shown as , where a, b are in units of pixels.

[0042] In some embodiments, step S200 further comprises step S201: Step S201: Correcting the pose The Z-axis component of the position freedom of the corrected pose is related to the Z-axis component of the position freedom of the initial pose , and the values of other freedoms are the same.

[0043] The distance between the lower surface of the tactile sensor 205 and the calibration base 102 is the distance L at which the calibration needle 101 is exposed from the calibration base 102. The initial pose of the end of the robotic arm 204 is subtracted by L on the Z-axis, and the corrected pose of the robotic arm 204 when the surface of the tactile sensor 205 just contacts the calibration base 102 can be obtained. That is, it is assumed that the robotic arm 204 continues to move downward by L, and the highest point of the calibration base 102, i.e., the position of the circular hole 103 on the calibration base 102, can be just touched by the surface of the sensor. The Z-axis component of the position freedom of the corrected pose of the robotic arm 204 is related to the Z-axis component of the position freedom of the initial pose of the end of the robotic arm 204 as follows: , and the values of other freedoms are the same.

[0044] The distance between the lower surface of the tactile sensor 205 and the calibration base 102 is the distance L at which the calibration needle 101 is exposed from the calibration base 102. The initial pose of the end of the robotic arm 204 is subtracted by L on the Z-axis, and the corrected pose of the robotic arm 204 when the surface of the tactile sensor 205 just contacts the calibration base 102 can be obtained. That is, it is assumed that the robotic arm 204 continues to move downward by L, and the highest point of the calibration base 102, i.e., the position of the circular hole 103 on the calibration base 102, can be just touched by the surface of the sensor. The Z-axis component of the position freedom of the corrected pose of the robotic arm 204 is related to the Z-axis component of the position freedom of the initial pose of the end of the robotic arm 204 as follows: , and the values of other freedoms are the same.

[0045] The robotic arm 204 does not need to move to the corrected pose in the physical space, and the corrected pose is only a virtual pose calculated.

[0046] In some embodiments, step S300 further comprises step S301: Step S301: According to the preset pixel-physical distance conversion relationship, the pixel position The unit conversion of the physical length is obtained to obtain the physical zero point, denoted as ).

[0047] The preset pixel and physical distance conversion relationship is provided by the sensor manufacturer. The value in step S100 is converted into a physical unit. The sensor manufacturer will provide the physical size corresponding to one pixel in the sensor data. For example, one pixel point corresponds to 0.1 millimeter, and the physical zero point is: .

[0048] In step S400, the calibration needle 101 is removed, and the position of the calibration base 102 is kept unchanged. The end of the mechanical arm 204 is moved to contact the hemispherical surface of the calibration part 1021, as shown in Figure 4 , at this time, the tactile sensor 205 moves and rotates compared with the initial position, and leaves a contact image 206 on the sensor signal.

[0049] In step S500, the current contact pose of the end of the mechanical arm 204 is recorded , as shown in the mechanical arm 204 contact pose 212. Figure 4

[0050] In step S600, the current physical position of the current contact point between the tactile sensor 205 and the calibration part 1021 on the tactile image 206 of the tactile sensor 205 is recorded , denoted as , as shown in the contact image coordinates 207 of the sensor and the calibration part 1021, denoted as the current physical position Figure 5 , for the current physical position , the physical position and pixel point conversion method of the current physical position is the same as step S301.

[0051] As shown in Figure 5 , the contact part of the tactile sensor 205 and the calibration part 1021 is enlarged, the figure shows the virtual correction pose 209 of the mechanical arm 204, the contact pose 212 of the mechanical arm 204 when the sensor contacts the calibration part 1021, the contact point 210 of the sensor and the calibration needle 101, and the contact point 211 of the sensor and the calibration part 1021, and the image coordinates 207 of the sensor and the calibration part 1021 are placed together to show the image coordinates 208 of the sensor and the calibration needle 101.

[0052] In step S700, the change matrix of the physical zero point and the current physical position in the three-dimensional space is calculated​​​ Specifically, The transformation matrix is described in the form of homogeneous transformation matrix, and the transformation matrix contains a rotation transformation R and a translation transformation t, because and are both points on the two-dimensional image data, so there is no rotation relationship between them, and the rotation transformation R degenerates into a unit matrix:

[0053] Similarly, and There is no change relationship on the z-axis, only movement in the xy plane, so the translation transformation matrix t is:

[0054] In summary, the physical zero point and the current physical position The transformation matrix in three-dimensional space is:

[0055] Step S800: Calculate the current contact pose with respect to the modified pose The transformation matrix is specifically: The mechanical arm 204 drives the tactile sensor 205 to move from the modified pose to the current contact pose At the same time, it undergoes translation and rotation operations. For the rotation transformation, the rotation angle of the mechanical arm 204 around the Roll axis is denoted as , the rotation angle around the Pitch axis is denoted as , and the rotation angle around the Yaw axis is denoted as , then the rotation transformation matrix is: ; Where:

[0056]

[0057]

[0058] And the translation transformation is:

[0059] Wherein, is the position component of , and is the position component of . The current contact pose The transformation matrix relative to the corrected pose is:

[0060] In step S900, the normal vector of the calibration part 1021 at the current contact point is calculated, specifically: A coordinate system is established with the center of the calibration part 1021 as the coordinate origin, and the installation position coordinates on the calibration part 1021 are: ; The coordinates of the current contact point in the spatial coordinate system are

[0061] The normal vector of the current contact point on the calibration part 1021 is:

[0062] wherein, is the length of ; The normal vector containing the starting point position information of the normal vector is recorded as: , wherein, is the component of the current contact point on the three coordinate axes in space, is the component of the normal vector of the current contact point on the three coordinate axes in space.

[0063] In step S1000, the sensor normal vector at the current contact point of the tactile sensor 205 is calculated, specifically: According to the setting that the sensor normal vector at the current contact point of the tactile sensor 205 is opposite to the direction of the normal vector on the calibration part 1021, the sensor normal vector corresponding to the current contact point of the tactile sensor 205 is .

[0064] All normal vectors are transformed into the coordinate system with the corrected pose as the reference point to facilitate subsequent unified modeling, specifically: According to the formula: , the normal vector at the current contact point of the tactile sensor 205 is in the coordinate system with the corrected pose as the reference point; wherein is the inverse matrix of the transformation matrix , which is:

[0065] ​In this step, the surface of the tactile sensor 205 is tangent to the calibration portion 1021. Although the surface of the tactile sensor 205 is complex, the calibration portion 1021 is a standard sphere, so the normal vector of any point on the calibration portion 1021 is known. Therefore, the normal vector of the corresponding point on the surface of the tactile sensor 205 can be known.

[0066] Steps S400 to S1000 are repeated to collect the normal vectors of different points on the sensor surface.

[0067] In step S1110, an existing surface reconstruction algorithm such as BPA, LS, BF, etc. is used. The collected contact point cloud of the tactile sensor 205 and the sensor normal vector data are fitted to obtain a sensor surface model.

[0068] The calibration method of the present application uses the calibration needle 101 to determine the contact point of the sensor and converts all subsequent poses into the initial pose of the robot arm 204, which avoids the influence of the initial absolute pose error of the robot arm 204 and the absolute pose error of the calibration base 102 on the calibration result, improves the accuracy of the sensor surface normal vector detection, and determines the position of the normal vector of each point on the surface of the tactile sensor 205. The method can correct the deformation of the sensor caused during manufacturing and use, or directly calibrate the tactile sensor 205 on an irregular surface, and finally accurately model the sensor surface to make the subsequent normal vector measurement more accurate.

[0069] After the sensor calibration is completed, the accurate sensor surface model can be used to measure the surface of any object to be measured.

[0070] The present application also provides an object surface measurement method, which is suitable for a calibration device and a robot arm 204 for driving the tactile sensor 205 to move, as shown in FIG. 1, the object surface measurement method of the present application includes but is not limited to: Figure 6 Step M100: controlling the movement of the end of the robot arm 204 to control the movement of the tactile sensor 205 to contact the object at different positions; Step M200: recording the current end pose of the robot arm 204 , the contact point position on the tactile sensor 205 ; Step M300: finding the corresponding sensor normal vector at the contact point position according to the sensor surface model ; Step M400: taking the current end pose as the starting point coordinates of the normal vector, and taking the sensor normal vector ; ​Transform to end pose , and transform the surface normal vector position of the tactile sensor 205 according to the contact point to obtain the object contact point normal vector; Step M500: fitting the collected object contact point cloud and object contact point normal vector data to obtain the measured object surface model.

[0071] The step M400 of the embodiment of the present application further comprises the following steps: According to the formula The object contact point normal vector is calculated , wherein is the end pose of the current robot arm 204 is the transformation matrix of the modified pose of the robot arm 204 The transformation matrix contains a rotation change and a translation transformation , which is set to .

[0072] Wherein, the pose information of the end pose includes three position degrees of freedom and three attitude degrees of freedom , which are set to .

[0073] The step M400 of the embodiment of the present application further comprises the following steps: The transformation matrix of the end pose of the current robot arm 204 relative to the modified pose of the robot arm 204 is calculated as follows: During the movement of the robot arm 204 driving the tactile sensor 205 from the modified pose to the end pose : The rotation angle of the robot arm 204 around the Roll axis is denoted as , the rotation angle around the Pitch axis is denoted as , and the rotation angle around the Yaw axis is denoted as , then the rotation transformation matrix is: ; Wherein:

[0074]

[0075]

[0076] And the translation transformation is:

[0077] wherein, is the position component of In summary, the end pose is the transformation matrix relative to the modified pose

[0078] Further, the object surface normal vector containing the point cloud coordinates is: Repeat steps M100 to M400 until the required normal vector of the object surface is obtained.

[0079] In step M500, the contact point cloud and the object contact point normal vector data of the collected object are fitted by using existing surface reconstruction algorithms such as BPA, LS, BF, etc., to obtain the measured object surface model.

[0080] The present application combines the calibration device, first models the irregular surface of the tactile sensor 205, obtains the distribution of the surface normal vector of the tactile sensor 205, and then uses the calibrated surface normal vector of the tactile sensor 205 to solve the normal vector of the measured object surface in combination with the forward kinematics of the robot.

[0081] In the description of the present application, the description of the terms "one embodiment", "some embodiments", "exemplary embodiment", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present application, the exemplary description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0082] The embodiments of the present application are described in detail above in combination with the drawings, but the present application is not limited to the above embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the purpose of the present application.​

Claims

1. A method of calibrating a tactile sensor, characterized by, A mechanical arm suitable for a calibration device and moving a tactile sensor, wherein the calibration device comprises a calibration base and a calibration needle, the calibration base comprises a calibration part and a fixed part arranged in sequence along the Z axis, the calibration part is a hemisphere with a diameter of R, the center position of the end of the calibration part away from the fixed part is provided with a mounting position, the calibration needle is detachably mounted on the mounting position along the Z axis, the length of the calibration needle exposed from the calibration part is L, and the calibration method comprises: controlling the end of the mechanical arm to move to control the tactile sensor to be in contact with the end of the calibration needle away from the calibration part, and recording the initial pose of the end of the mechanical arm at this moment , the pixel position of the contact point between the tactile sensor and the calibration needle on the tactile image of the tactile sensor ; subtracting L on the Z axis to obtain a corrected pose of the robot arm when the surface of the tactile sensor just touches the calibration portion subtracting L on the Z axis to obtain a corrected pose of the robot arm when the surface of the tactile sensor just touches the calibration portion ; by calculating the pixel position converting the physical zero point ; controlling the removal of the calibration needle, and controlling the movement of the end of the mechanical arm to contact the hemispherical surface of the calibration part at other positions on the surface of the tactile sensor; recording a current contact pose of an end of the robot arm ; recording a current physical position of a current contact point between the tactile sensor and the calibration portion on a tactile image of the tactile sensor ; computing the physical zero point with the current physical position a change matrix in three-dimensional space wherein the change matrix comprises a rotational change R and a translational transformation t; computing the current contact pose a transformation matrix with respect to the modified pose where the transformation matrix comprises a rotation transformation matrix and a translation transformation ;​ calculating the normal vector of the calibration part at the current contact point; calculating the sensor normal vector at the current contact point of the tactile sensor; fitting the collected contact point cloud and sensor normal vector data of the tactile sensor to obtain a sensor surface model.

2. The calibration method of claim 1, wherein: the calculation of the normal vector of the calibration part at the current contact point comprises: A coordinate system is established with the spherical center of the calibration portion as a coordinate origin, and the mounting position coordinates on the calibration portion are: ; The coordinates of the current contact point in the spatial coordinate system are ; the normal vector of the current contact point on the calibration part (1021) is: wherein is the length of the module; The normal vector containing the normal vector starting point position information is denoted as: wherein, is the component of the current contact point on the three coordinate axes in space, is the component of the normal vector of the current contact point on the three coordinate axes in space; the calculation of the sensor normal vector at the current contact point of the tactile sensor comprises: According to the sensor normal vector at the current touch point of the tactile sensor is set opposite to the normal vector direction on the calibration portion, the sensor normal vector corresponding to the current touch point of the tactile sensor is .

3. The calibration method of claim 2, wherein: wherein The initial pose The pose information includes three position degrees of freedom And three attitude degrees of freedom ; The length L of the calibration portion exposed according to the calibration needle is calculated based on the initial pose of the robot arm is corrected to obtain a corrected pose of the robot arm when the surface of the tactile sensor just contacts the calibration portion , comprising: The modified pose The position freedom Z-axis component of the modified pose The position freedom Z-axis component of the initial pose The other freedom values are the same.

4. The calibration method of claim 3, wherein: said pixel position conversion of the physical zero point comprising: According to a preset pixel and physical distance conversion relationship, the pixel position is converted into a physical length unit to obtain the physical zero point, denoted as .

5. The calibration method of claim 4, wherein: The current physical location is noted as ; said calculating the physical zero point with the current physical position a change matrix in three-dimensional space ), comprising: Because With Both are points on the two-dimensional image data, so there is no rotational change relationship between them, and the rotational change R degenerates into a unit matrix: By analogy, With There is no relationship in the z-axis, only movement in the xy-plane, so the translation transformation matrix t is: The physical zero point With the current physical position The change matrix in three-dimensional space is:

6. The calibration method of claim 5, wherein: the current contact pose a transformation matrix relative to the modified pose , comprising: The mechanical arm moves the haptic sensor from the modified pose to the current contact pose in the process: The rotation angle of the mechanical arm around the Roll axis is denoted as , the rotation angle around the Pitch axis is denoted as , and the rotation angle around the Yaw axis is denoted as , and the rotation transformation matrix is: ; 7. The calibration method of claim 6, wherein: And the translation is: wherein is a position component of is a position component of the current contact pose a transformation matrix relative to the modified pose is: the calculation of the sensor normal vector at the current contact point of the tactile sensor further comprises: A mechanical arm suitable for a calibration device and moving a tactile sensor, wherein the calibration device comprises a calibration base and a calibration needle, the calibration base comprises a calibration part and a fixed part arranged in sequence along the Z axis, the calibration part is a hemisphere with a diameter of R, the center position of the end of the calibration part away from the fixed part is provided with a mounting position, the calibration needle is detachably mounted on the mounting position along the Z axis, the length of the calibration needle exposed from the calibration part is L, and the object surface measurement method comprises: According to the formula: The normal vector at the current contact point of the tactile sensor is referenced to the coordinate system of the corrected pose. wherein is the inverse matrix of the transformation matrix is the inverse matrix of the transformation matrix 8. An object surface measurement method characterized by, controlling the movement of the end of the mechanical arm to control the movement of the tactile sensor to just contact different positions of the object; fitting the collected contact point cloud and object contact point normal vector data to obtain a measured object surface model. record the current end position of the robot arm at this moment , touch point position on the tactile sensor ; finding the contact location according to a sensor surface model corresponding sensor normal vector ; at the current end pose as a normal vector origin coordinate, with the sensor normal vector transformed to the end pose and transform the surface normal vector position of the tactile sensor according to the different contact points to obtain the object contact point normal vector; 9. The object surface measurement method of claim 8, wherein: characterized in that: said current end pose as normal vector origin coordinates, with said sensor normal vector transformed to said end pose and transforming the surface normal vector position of said tactile sensor according to the different contact points to obtain the object contact point normal vector, comprising: According to the formula The object contact point normal vector is calculated wherein is the end pose of the current robot arm is the transformation matrix of the correction pose relative to the robot arm The transformation matrix contains a rotational change and a translational transformation is set to .

10. The object surface measurement method according to claim 9, the object surface measurement method further comprises: The end pose The pose information includes three position degrees of freedom And three attitude degrees of freedom ; wherein: controlling the end of the mechanical arm to move to control the tactile sensor to be in contact with the end of the calibration needle away from the calibration part, recording the initial pose of the end of the mechanical arm at this moment subtracting L from the initial pose in the Z axis to obtain the corrected pose of the mechanical arm when the surface of the tactile sensor is in contact with the calibration part subtracting L from the initial pose in the Z axis to obtain the corrected pose of the mechanical arm when the surface of the tactile sensor is in contact with the calibration part ; computing the end pose of the current robot arm a transformation matrix relative to the modified pose of the robot arm :​ The mechanical arm drives the tactile sensor from the correction pose to the end pose in the process: The rotation angle of the mechanical arm around the Roll axis is denoted as , the rotation angle around the Pitch axis is denoted as , and the rotation angle around the Yaw axis is denoted as , and the rotation transformation matrix is: ; ​ And the translation is: wherein is the position component the end pose a transformation matrix relative to the modified pose is: Set The object surface normal vector comprising the point cloud coordinates is: ).

Images