Shield segment visual servoing grasping and assembling method adaptive to motion of observation platform

Abstract

This invention discloses a visual servo grasping and assembly method for tunnel boring machine (TBM) segments adapted to the motion of an observation platform. The method includes the following steps: mounting a visual sensor on the observation platform to collect the three-dimensional coordinates of reference markers on the assembly machine, the assembly machine, the tunnel segments, and feature points at the target assembly position; unifying the local coordinate system of the observation platform with the base coordinate system of the assembly machine at the current moment; constructing an error vector field; based on the spatial distribution pattern of the error vector field, according to geometric decoupling mapping rules, allocating and mapping the overall error to the target motion increments of each independent joint of the assembly machine; iterative servo execution: executing the target motion increment commands in priority order, repeating the above steps until the magnitudes of all first spatial error vectors / second spatial error vectors are less than a set threshold. This invention relates to the field of automated equipment technology for tunnel construction and can solve the problem of poor engineering adaptability caused by traditional control methods relying on mechanism models and absolute pose calculations.

Description

Technical Field

[0001] This invention relates to the field of automated equipment technology for tunnel construction, and in particular to a visual servo grasping and assembly method for shield tunnel segments that adapts to the movement of an observation platform. Background Technology

[0002] Segment assembly is a crucial step in tunnel boring machine (TBM) construction, and its efficiency and precision directly affect the tunnel's forming quality, structural safety, and overall construction progress. Currently, most mainstream segment assemblies are operated manually via remote control or in a semi-automatic mode based on preset programs. This mode has the following significant drawbacks: the assembly machine is a complex series-parallel mechanism with multiple degrees of freedom and strong coupling; precise control of its end-effector position requires operators to possess extremely high spatial imagination and operational experience; operators find it difficult to accurately judge the relative distance and angle between the segment and the assembly machine's suction cups, which can easily lead to collisions and affect the safety of both the segment and the assembly machine itself; prolonged, high-stress remote control operations can easily cause visual and physical fatigue in operators, leading to operational errors.

[0003] To improve automation levels, existing automatic tunnel segment grabbing (or assembly) schemes are mainly divided into two categories: fixed vision + model-dependent control method and moving vision + pose calculation control method.

[0004] One method, fixed vision + model-dependent control, involves installing a camera at a fixed position on the tunnel boring machine (TBM). After visually measuring the segment pose, the machine relies on a precise kinematic model of the assembly machine for inverse kinematics control. This method places extremely high demands on the accuracy and calibration of the mechanism model. It suffers from poor reliability due to factors such as hydraulic system nonlinearity and mechanical backlash, and it cannot adapt to changes in the observation angle.

[0005] Mobile vision + pose calculation control method: Although this method uses a mobile observation platform (such as a robotic arm carrying a camera), it still requires first calculating the absolute six-degree-of-freedom pose of the suction cup and the tube segment, and then generating control commands through inverse kinematics. When the observation platform moves, the coordinate reference constantly changes, leading to accumulated pose calculation errors and poor control stability.

[0006] Therefore, there is a need to provide a visual servo grasping and assembly method for tunnel segments that adapts to the movement of the observation platform, which can solve the problem of poor engineering adaptability caused by the reliance on mechanism models and absolute pose calculations in traditional control methods. Summary of the Invention

[0007] The purpose of this invention is to provide a visual servo grasping and assembly method for tunnel lining segments that adapts to the movement of the observation platform, which can solve the problem of poor engineering adaptability caused by the reliance on mechanism models and absolute pose calculations in traditional control methods.

[0008] This invention is implemented as follows:

[0009] A visual servo grasping and assembly method for tunnel boring machine segments adapted to the movement of an observation platform includes the following steps:

[0010] Step 1: Mount a vision sensor on a mobile observation platform and use the vision sensor to collect the three-dimensional coordinates of the reference marker points on the assembly machine, as well as the three-dimensional coordinates of the feature points on the assembly machine, the segments, and the target assembly position.

[0011] Step 2: Unify the local coordinate system of the observation platform with the base coordinate system of the assembly machine at the current moment;

[0012] Step 3: Construct the error vector field;

[0013] Step 4: Based on the spatial distribution pattern of the error vector field, according to the preset geometric decoupling mapping rules, the overall error is allocated and mapped into the target motion increment of each independent joint of the assembly machine;

[0014] Step 5: Iterative servo execution: Execute the instructions for the target motion increment in a preset priority order, and repeat steps 1 to 5 until the magnitudes of all first spatial error vectors / second spatial error vectors are less than the set threshold.

[0015] In step 1, four reference markers are set up on the lifting beam of the assembly machine, denoted as... , i=1,2,3,4;

[0016] In step 1, during the segment gripping stage, the feature points include three first feature points set on the segment suction cup of the assembly machine and three second feature points set on the segment to be gripped. The first feature points are denoted as... Let i = 1, 2, 3, and the second feature point be denoted as... In the segment assembly stage, i=1,2,3; the feature points include three third feature points set on the segments to be assembled and three fourth feature points set on the target assembly position. The third feature points are denoted as... Let i = 1, 2, 3, and the fourth feature point be denoted as... , i=1,2,3.

[0017] Step 2 includes the following sub-steps:

[0018] Step 21: The assembly machine's base coordinate system {0} is located on the lifting beam of the assembly machine. Calculate the known coordinates of the four reference points in the assembly machine's base coordinate system {0}. , i = 1~4;

[0019] Step 22: Calculate the known coordinates of the four reference points in the local coordinate system {U} of the observation platform. , i = 1~4;

[0020] Step 23: Real-time calculation of the transformation relationship between the local coordinate system {U} of the observation platform and the base coordinate system {0} of the assembly machine at the current moment.

[0021] Step 23 includes the following sub-steps:

[0022] Step 231: The transformation matrix from the local coordinate system {U} of the observation platform to the base coordinate system {0} of the assembly machine is a 4×4 transformation matrix. 4×4 transformation matrix Including rotation matrix Translation vector It satisfies the following relation: Solve the 4×4 transformation matrix ;

[0023] Step 232: Solve for the rotation matrix using the solvePnP algorithm. Translation vector .

[0024] In step 3, during the segment grabbing stage, the coordinates of the three first feature points and three second feature points under the local coordinate system {U} of the observation platform are synchronously transformed to the base coordinate system {0} of the assembly machine using the transformation relationship from the local coordinate system {U} of the observation platform to the base coordinate system {0} of the assembly machine, and three sets of first spatial error vectors are calculated. The specific calculation method is as follows: i = 1, 2, 3;

[0025] Step 31: Calculate the three first feature points on the segment suction cup of the assembly machine respectively. and three second feature points Known coordinates in the local coordinate system {U} of the observation platform and , i = 1~3;

[0026] Step 32: Set the coordinates of the three first feature points on the segment suction cup of the assembly machine in the local coordinate system {U} of the observation platform. The coordinates of the three second feature points Transforming to the assembly machine's base coordinate system {0}, the transformation relationship is as follows:

[0027] ;

[0028] in, Let {0} be the coordinates of the first feature point in the assembly machine's base coordinate system. , Let {0} be the coordinates of the second feature point in the assembly machine's base coordinate system. ;

[0029] Step 33: Calculate the three sets of first spatial error vectors in the assembly machine's base coordinate system {0}. For i=1,2,3, the calculation formula is:

[0030] .

[0031] In step 3, during the segment assembly stage, the coordinates of the three third feature points and three fourth feature points under the local coordinate system {U} of the observation platform are synchronously transformed to the base coordinate system {0} of the assembly machine using the transformation relationship from the local coordinate system {U} of the observation platform to the base coordinate system {0} of the assembly machine, and three sets of second spatial error vectors are calculated. The specific calculation method is as follows: i = 1, 2, 3;

[0032] Step 31: Calculate the three third feature points on the segments to be assembled. and the three fourth feature points at the target assembly position Known coordinates in the local coordinate system {U} of the observation platform and , i = 1~3;

[0033] Step 32: Set the coordinates of the three third feature points on the segment to be assembled in the local coordinate system {U} of the observation platform. The coordinates of the fourth feature point at the three target assembly positions Transforming to the assembly machine's base coordinate system {0}, the transformation relationship is as follows:

[0034] ;

[0035] in, Let {0} be the coordinates of the third feature point in the assembly machine's base coordinate system. , Let {0} be the coordinates of the fourth feature point in the assembly machine's base coordinate system. ;

[0036] Step 33: Calculate the three sets of second spatial error vectors under the assembly machine's base coordinate system {0}. For i=1,2,3, the calculation formula is:

[0037] .

[0038] In step 4, the geometric decoupling mapping rules include:

[0039] When the three sets of first spatial error vectors are distributed in space in an approximately parallel manner, it indicates that the deviation is mainly manifested as an overall positional offset. The spatial average value of the three sets of first spatial error vectors is calculated, the first translation error component is extracted, and the first translation error component is mapped to the translation joint of the assembly machine.

[0040] When the three sets of second spatial error vectors are distributed in space in an approximately parallel manner, it indicates that the deviation is mainly manifested as an overall positional offset. The spatial average value of the three sets of second spatial error vectors is calculated, the second translation error component is extracted, and the second translation error component is mapped to the translation joint of the assembly machine.

[0041] When the three sets of first spatial error vectors change linearly, that is, when the three sets of first error vectors are parallel in direction in space but change systematically in magnitude, it indicates that the deviation is mainly manifested as the plane tilt angle of the assembly machine segment suction cup. By analyzing the distribution gradient characteristics of the three sets of first spatial error vectors, the first attitude rotation error component is extracted, and the first attitude rotation error component is mapped to the differential rotation control of the attitude rotation joint and lifting cylinder of the assembly machine.

[0042] When the three sets of second spatial error vectors change linearly, that is, when the three sets of second error vectors are parallel in direction in space but change systematically in magnitude, it indicates that the deviation is mainly manifested as the plane tilt angle of the segment to be assembled. By analyzing the distribution gradient characteristics of the three sets of second spatial error vectors, the second attitude rotation error component is extracted, and the second attitude rotation error component is mapped to the differential rotation control of the attitude rotation joint and lifting cylinder of the assembly machine.

[0043] When the first spatial error vector is distributed in a rotational manner in the plane, it indicates that the deviation is mainly manifested as torsion around the plane normal. By analyzing the directional change characteristics of the first spatial error vector in the plane, the rotational error component in the first plane is extracted, and the rotational error component in the first plane is mapped to the in-plane rotational joint of the assembly machine.

[0044] When the second spatial error vector exhibits a rotational distribution in the plane, it indicates that the deviation is mainly manifested as torsion around the plane normal. By analyzing the directional change characteristics of the second spatial error vector in the plane, the rotational error component in the second plane is extracted, and the rotational error component in the second plane is mapped to the in-plane rotational joint of the assembly machine.

[0045] In step 4, during the segment grasping stage, three sets of first spatial error vectors... For i=1,2,3, the mapping relationship to the target motion increment of each independent joint of the assembly machine is analyzed as follows:

[0046] ① The first target stroke of the large translational hydraulic cylinder :

[0047] Calculate the first average error vector The calculation formula is:

[0048] ;

[0049] in, The first average error vector The component in the X-axis direction, The first average error vector The component in the Y-axis direction, The first average error vector The component in the Z-axis direction; take the first average error vector. Components in the X-axis direction The first target stroke of the large translational hydraulic cylinder ;

[0050] ② The first target rotation angle of the slewing ring :

[0051] Calculate the plane normal vector of the segment suction cup of the assembly machine. and the plane normal vector of the segment to be captured The calculation formula is:

[0052] ;

[0053] get The first projection vector in the YOZ plane :

[0054] ;

[0055] get The second projection vector in the YOZ plane :

[0056] ;

[0057] Calculate the first projection vector With the second projection vector The first included angle The calculation formula is:

[0058] ;

[0059] First included angle This refers to the first target rotation angle of the slewing ring of the assembly machine. ;

[0060] ③ The first stroke differential of the lifting cylinder, including the first left stroke differential of the left lifting cylinder. Differential momentum of the first right stroke of the right lifting cylinder :

[0061] Obtain the first spatial error vector respectively and Components on the Z-axis and ;

[0062] Calculate components and The difference : ;

[0063] The differential momentum of the first left stroke of the left lifting cylinder is ,Right now ;

[0064] The differential momentum of the first right stroke of the right lifting cylinder is ,Right now ;

[0065] ④ The first target stroke of the pitch cylinder :

[0066] Obtain the plane normal vector of the segment suction cup of the assembly machine. The third projection vector in the XOZ plane : ;

[0067] Obtain the plane normal vector of the segment to be captured. The fourth projection vector in the XOZ plane : ;

[0068] Calculate the third projection vector With the fourth projection vector The second included angle The calculation formula is:

[0069] ;

[0070] According to the second included angle The first target stroke of the pitch cylinder was calculated. ;

[0071] ⑤ The first target stroke of the deflection cylinder :

[0072] Obtain the plane normal vector of the segment suction cup of the assembly machine. The fifth projection vector in the XOY plane :

[0073] ;

[0074] Obtain the plane normal vector of the segment to be captured. The sixth projection vector in the XOY plane :

[0075] ;

[0076] Calculate the fifth projection vector With the sixth projection vector The third included angle The calculation formula is:

[0077] ;

[0078] According to the third included angle The first target stroke of the deflection cylinder can be calculated. ;

[0079] ⑥ Lifting the first synchronous target stroke of the hydraulic cylinder The calculation formula is:

[0080] ;

[0081] Where N is a positive integer, N=2~4.

[0082] In step 4, during the segment assembly stage, three sets of second spatial error vectors... For i=1,2,3, the mapping relationship to the target motion increment of each independent joint of the assembly machine is analyzed as follows:

[0083] ① The second target stroke of the large translational hydraulic cylinder :

[0084] Calculate the second average error vector The calculation formula is:

[0085] ;

[0086] in, The second average error vector The component in the X-axis direction, The second average error vector The component in the Y-axis direction, The second average error vector The component in the Z-axis direction; take the second average error vector. Components in the X-axis direction The second target stroke of the large translational hydraulic cylinder ;

[0087] ② The second target rotation angle of the slewing ring :

[0088] Calculate the plane normal vector of the segment to be assembled. The plane normal vector of the target assembly position The calculation formula is:

[0089] ;

[0090] get The seventh projection vector in the YOZ plane :

[0091] ;

[0092] get The eighth projection vector in the YOZ plane :

[0093] ;

[0094] Calculate the seventh projection vector With the eighth projection vector The fourth included angle The calculation formula is:

[0095] ;

[0096] Fourth angle This refers to the second target rotation angle of the rotating ring of the assembly machine. ;

[0097] ③ The second stroke differential of the lifting cylinder, including the second left stroke differential of the left lifting cylinder. Differential momentum of the second right stroke of the right lifting cylinder :

[0098] Obtain the second spatial error vector respectively and Components on the Z-axis and ;

[0099] Calculate components and The difference : ;

[0100] The differential momentum of the second left stroke of the left lifting cylinder is ,Right now ;

[0101] The differential momentum of the second right stroke of the right lifting cylinder is ,Right now ;

[0102] ④ The second target stroke of the pitch cylinder :

[0103] Obtain the plane normal vector of the segment to be assembled. The ninth projection vector in the XOZ plane :

[0104] ;

[0105] Obtain the plane normal vector of the target assembly position. The tenth projection vector in the XOZ plane :

[0106] ;

[0107] Calculate the ninth projection vector With the tenth projection vector The fifth included angle The calculation formula is:

[0108] ;

[0109] According to the fifth angle The second target stroke of the pitch cylinder was calculated. 。;

[0110] ⑤ The second target stroke of the deflection cylinder :

[0111] Obtain the plane normal vector of the segment to be assembled. The eleventh projection vector in the XOY plane :

[0112] ;

[0113] Obtain the plane normal vector of the target assembly position. The twelfth projection vector in the XOY plane :

[0114] ;

[0115] Calculate the eleventh projection vector With the twelfth projection vector The sixth angle The calculation formula is:

[0116] ;

[0117] According to the sixth angle The second target stroke of the deflection cylinder can be calculated;

[0118] ⑥ Lifting the second synchronous target stroke of the hydraulic cylinder The calculation formula is:

[0119] ;

[0120] Where N is a positive integer, N=2~4.

[0121] In step 5, the priority order is set based on the geometric dependence of error elimination, and is as follows: ① Adjust the translation joint to eliminate positional deviation in the plane; ② Adjust the rotation joint in the plane to eliminate torsion about the plane normal; ③ Adjust the attitude rotation joint to make the plane of the assembly machine's segment suction cup parallel to the plane of the segment to be grasped during the segment grasping stage, and the segment to be assembled parallel to the target assembly position during the segment assembly stage; ④ Adjust the normal distance joint to achieve the final fit between the assembly machine's segment suction cup and the segment to be grasped during the segment grasping stage, and the final fit between the segment to be assembled and the target assembly position during the segment assembly stage.

[0122] In terms of the assembly machine control, the priority order of the segment grabbing stage is as follows: the first target stroke of the large translation cylinder, the first target rotation angle of the rotary ring, the first stroke differential of the lifting cylinder, the first target stroke of the pitch cylinder, the first target stroke of the yaw cylinder, and the first synchronous target stroke of the lifting cylinder.

[0123] Based on the priority order of the segment grasping stage, and in conjunction with the vision sensor, a closed-loop execution of visual perception and segment grasping is performed multiple times until the following conditions are met, at which point the segment grasping is considered successful:

[0124] threshold It is 2mm;

[0125] In terms of the control of the assembly machine, the priority order of the segment assembly stage is as follows: the second target stroke of the large translation cylinder, the second target rotation angle of the rotary ring, the second stroke differential of the lifting cylinder, the second target stroke of the pitch cylinder, the second target stroke of the yaw cylinder, and the second synchronous target stroke of the lifting cylinder.

[0126] Based on the priority order of the segment assembly stage, and in conjunction with visual sensors, a closed-loop execution of visual perception and segment assembly is performed multiple times until the following conditions are met, at which point the segment assembly is considered successful:

[0127] threshold It is 2mm.

[0128] Compared with the prior art, the present invention has the following advantages:

[0129] 1. The present invention has strong anti-interference ability: by synchronously solving the dynamic reference transformation of the same frame image, the visual measurement value is unified to the stable base coordinate system in real time, which completely eliminates the measurement drift caused by the movement of the observation platform itself and realizes "stability control in motion".

[0130] 2. This invention features model independence: It abandons the traditional "absolute pose calculation → inverse kinematics" path and directly performs decoupling mapping based on the geometric pattern of the error vector field, which has natural robustness to mechanism model errors and nonlinear factors.

[0131] 3. The closed-loop control of this invention is intuitive and reliable: "Three-point vector zeroing" is a clear physical objective, and the control process is transparent and observable, avoiding the unpredictable behavior that may be caused by traditional black box models.

[0132] 4. The present invention has good convergence: the priority adjustment order based on geometric dependence effectively avoids oscillations caused by multi-degree-of-freedom coupling, ensuring fast and stable convergence of each independent joint. Attached Figure Description

[0133] Figure 1 This is a schematic diagram of the operation of the segment grasping stage in the shield tunnel segment visual servo grasping and assembly method adapted to the movement of the observation platform of the present invention;

[0134] Figure 2 This is a schematic diagram of the operation of the tunnel segment assembly stage in the shield tunnel segment visual servo grasping and assembly method adapted to the movement of the observation platform of the present invention;

[0135] Figure 3 This is a diagram of the three-layer closed-loop control architecture of the shield tunnel segment visual servo grasping and assembly method adapted to the movement of the observation platform according to the present invention.

[0136] In the figure, the assembly machine's segment suction cup 1, the segment to be grabbed 2, the segment to be assembled 3, the target assembly position 4, the lifting beam 5, the observation platform 6, the slewing ring 7, the large translation cylinder 8, the lifting cylinder 9, the pitch cylinder 10, and the deflection cylinder 11. Detailed Implementation

[0137] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0138] Please see the appendix Figure 1 and attached Figure 2 A visual servo grasping and assembly method for tunnel boring machine segments adapted to the movement of an observation platform includes the following steps:

[0139] Step 1: Install a vision sensor on the mobile observation platform 6, and collect the three-dimensional coordinates of the reference marker points on the assembly machine through the vision sensor. At the same time, collect the three-dimensional coordinates of the feature points on the assembly machine, the segment and the target assembly position.

[0140] Preferably, the mobile observation platform 6 can adopt existing mobile platform structures such as drones and robotic arms, and can be selected according to the actual application conditions. The vision sensor can adopt existing industrial cameras, and can be selected according to the actual application conditions.

[0141] In step 1, four reference markers are set up on the lifting beam 5 of the assembly machine, and are denoted as follows: , i=1,2,3,4.

[0142] In step 1, during the segment gripping stage, the feature points include three first feature points set on the segment suction cup 1 of the assembly machine and three second feature points set on the segment 2 to be gripped. The first feature points are denoted as... Let i = 1, 2, 3, and the second feature point be denoted as... In the segment assembly stage, i=1,2,3; the feature points include three third feature points set on the segment 3 to be assembled and three fourth feature points set on the target assembly position 4. The third feature points are denoted as... Let i = 1, 2, 3, and the fourth feature point be denoted as... , i=1,2,3.

[0143] Step 2: Unify the local coordinate system of the observation platform with the base coordinate system of the assembly machine at the current moment.

[0144] Step 2 includes the following sub-steps:

[0145] Step 21: The assembly machine's base coordinate system {0} is located on the lifting beam 5 of the assembly machine. Calculate the known coordinates of the four reference points in the assembly machine's base coordinate system {0}. , i=1~4.

[0146] Calculating the coordinates of the four reference points on the lifting beam 5 sampled by the visual sensor in the coordinate system of the assembly machine is a conventional calculation method in this field, and will not be elaborated here.

[0147] Step 22: Calculate the known coordinates of the four reference points in the local coordinate system {U} of the observation platform. , i=1~4.

[0148] Since the relative position of the lifting beam 5 and the observation platform is determined, the calculation of the coordinates of the four reference markers on the lifting beam 5 sampled by the visual sensor in the local coordinate system of the observation platform is a conventional calculation method in this field, and will not be elaborated here.

[0149] Step 23: Real-time calculation of the transformation relationship between the local coordinate system {U} of the observation platform and the base coordinate system {0} of the assembly machine at the current moment.

[0150] Step 23 includes the following sub-steps:

[0151] Step 231: The transformation matrix from the local coordinate system {U} of the observation platform to the base coordinate system {0} of the assembly machine is a 4×4 transformation matrix. 4×4 transformation matrix Including rotation matrix Translation vector It satisfies the following relation: Solve the 4×4 transformation matrix .

[0152] The transformation matrix between the two coordinate systems is calculated using conventional methods in this field, and will not be elaborated here.

[0153] Step 232: Solve for the rotation matrix using the solvePnP algorithm. Translation vector .

[0154] The solvePnP algorithm is a conventional algorithm in existing technology, and will not be described in detail here.

[0155] Step 3: Construct the error vector field.

[0156] In step 3, during the segment grabbing stage, the coordinates of the three first feature points and three second feature points under the local coordinate system {U} of the observation platform are synchronously transformed to the base coordinate system {0} of the assembly machine using the transformation relationship from the local coordinate system {U} of the observation platform to the base coordinate system {0} of the assembly machine, and three sets of first spatial error vectors are calculated. Let i = 1, 2, 3. The specific calculation method is as follows:

[0157] Step 31: Calculate the three first feature points on the segment suction cup 1 of the assembly machine respectively. and three second feature points Known coordinates in the local coordinate system {U} of the observation platform and , i=1~3.

[0158] Calculating the coordinates of three first feature points and three second feature points sampled by the visual sensor in the local coordinate system of the observation platform is a conventional calculation method in this field, and will not be elaborated here.

[0159] Step 32: Set the coordinates of the three first feature points on the segment suction cup 1 of the assembly machine in the local coordinate system {U} of the observation platform. The coordinates of the three second feature points Transforming to the assembly machine's base coordinate system {0}, the transformation relationship is as follows:

[0160]

[0161] in, Let {0} be the coordinates of the first feature point in the assembly machine's base coordinate system. , Let {0} be the coordinates of the second feature point in the assembly machine's base coordinate system. .

[0162] Step 33: Calculate the three sets of first spatial error vectors in the assembly machine's base coordinate system {0}. (i=1, 2, 3), the calculation formula is:

[0163] .

[0164] In step 3, during the segment assembly stage, the coordinates of the three third feature points and three fourth feature points under the local coordinate system {U} of the observation platform are synchronously transformed to the base coordinate system {0} of the assembly machine using the transformation relationship from the local coordinate system {U} of the observation platform to the base coordinate system {0} of the assembly machine, and three sets of second spatial error vectors are calculated. (i=1, 2, 3). The specific calculation method is as follows:

[0165] Step 31: Calculate the three third feature points on the segment 3 to be assembled. and the three fourth feature points at target assembly position 4 Known coordinates in the local coordinate system {U} of the observation platform and (i=1~3).

[0166] Calculating the coordinates of three third and three fourth feature points sampled by the visual sensor in the local coordinate system of the observation platform is a conventional calculation method in this field, and will not be elaborated here.

[0167] Step 32: Set the coordinates of the three third feature points on the segment 3 to be assembled under the local coordinate system {U} of the observation platform. The coordinates of the three fourth feature points at target assembly position 4 Transforming to the assembly machine's base coordinate system {0}, the transformation relationship is as follows:

[0168] .

[0169] in, Let {0} be the coordinates of the third feature point in the assembly machine's base coordinate system. , Let {0} be the coordinates of the fourth feature point in the assembly machine's base coordinate system. .

[0170] Step 33: Calculate the three sets of second spatial error vectors under the assembly machine's base coordinate system {0}. (i=1, 2, 3), the calculation formula is:

[0171] .

[0172] Step 4: Based on the spatial distribution pattern of the error vector field (i.e., the three sets of first spatial error vectors in the segment grasping stage and the three sets of second spatial error vectors in the segment assembly stage), according to the preset geometric decoupling mapping rules, the overall error is allocated and mapped into the target motion increment of each independent joint of the assembly machine (including translational joints, attitude rotational joints and in-plane rotational joints).

[0173] The geometric decoupling mapping rule is based on the correspondence between the error vector field and the geometric effects of motion of each independent joint, rather than the inverse kinematics model of the mechanism.

[0174] In step 4, preferably, the geometric decoupling mapping rule includes:

[0175] When the three sets of first spatial error vectors are distributed in an approximately parallel manner in space, it indicates that the deviation is mainly manifested as an overall positional shift. By calculating the spatial average of the three sets of first spatial error vectors, the first translation error component can be directly extracted. , The first translation error components correspond to the first translation error components of the large translation cylinder 8 and the lifting cylinder 9 (including the left lifting cylinder and the right lifting cylinder), and the first translation error components are mapped to the translation joints of the large translation cylinder 8 and the lifting cylinder 9 (including the left lifting cylinder and the right lifting cylinder) of the assembly machine.

[0176] Preferably, when the angle between any two pairs of spatial error vectors is less than 0.1°, the first spatial error vector is determined to be approximately parallel in space.

[0177] Similarly, when the three sets of second spatial error vectors are distributed in space in an approximately parallel manner, it indicates that the deviation is mainly manifested as an overall positional shift. By calculating the spatial average of the three sets of second spatial error vectors, the second translation error component can be directly extracted. , The second translation error components correspond to the large translation cylinder 8 and the lifting cylinder 9 (including the left lifting cylinder and the right lifting cylinder), respectively, and the second translation error components are mapped to the translation joints of the large translation cylinder 8 and the lifting cylinder 9 (including the left lifting cylinder and the right lifting cylinder) of the assembly machine.

[0178] Preferably, when the angle between any two pairs of spatial error vectors is less than 0.1°, the second spatial error vector is determined to be approximately parallel in space.

[0179] When the three sets of first spatial error vectors exhibit a linear gradient change, meaning that the three sets of first error vectors are parallel in direction in space but their magnitudes change systematically, it indicates that the deviation is mainly manifested as the plane tilt angle of the assembly machine segment suction cup 1. By analyzing the distribution gradient characteristics of the three sets of first spatial error vectors, the pitch, yaw, and other first attitude rotation error components can be extracted. , , The first attitude rotation error components of the pitch cylinder 10 and the yaw cylinder 11, respectively, and the differential rotation control quantities of the left lifting cylinder and the right lifting cylinder are mapped to the attitude rotation joints of the pitch cylinder 10 and the yaw cylinder 11 of the assembly machine, as well as the differential rotation control of the lifting cylinder 9 (including the left lifting cylinder and the right lifting cylinder).

[0180] Similarly, when the three sets of second spatial error vectors exhibit linear gradient changes, meaning that the three sets of second error vectors are parallel in direction but exhibit systematic magnitude changes in magnitude, it indicates that the deviation is mainly manifested as the plane tilt angle of the segment 3 to be assembled. By analyzing the distribution gradient characteristics of the three sets of second spatial error vectors, the pitch, yaw, and other second attitude rotation error components can be extracted. , , The second attitude rotation error components corresponding to the pitch cylinder 10 and yaw cylinder 11, respectively, and the differential rotation control quantities of the left lifting cylinder and right lifting cylinder are mapped to the attitude rotation joints of the pitch cylinder 10 and yaw cylinder 11 of the assembly machine and the differential rotation control of the lifting cylinder 9 (including the left lifting cylinder and right lifting cylinder).

[0181] When the first spatial error vector exhibits a rotational distribution within the plane, it indicates that the deviation is primarily manifested as torsion around the plane normal. By analyzing the directional change characteristics of the first spatial error vector within the plane, the rotational error component within the first plane can be extracted. This is the first in-plane rotational error component), and the first in-plane rotational error component is mapped to the in-plane rotational joints such as the rotary ring 7 of the assembly machine.

[0182] Similarly, when the second spatial error vector exhibits a rotational distribution within the plane, it indicates that the deviation is primarily manifested as torsion around the plane normal. By analyzing the directional change characteristics of the second spatial error vector within the plane, the rotational error component within the second plane can be extracted. This is the rotational error component in the second plane), and the rotational error component in the second plane is mapped to the rotating joints in the plane, such as the slewing ring 7 of the assembly machine.

[0183] In step 4, during the segment grasping stage, three sets of first spatial error vectors... The mapping relationship between (i=1,2,3) and the target motion increments of each independent joint of the assembly machine is analyzed as follows:

[0184] ① The first target stroke of the large translation cylinder 8 :

[0185] Calculate the first average error vector The calculation formula is:

[0186] .

[0187] in, The first average error vector The component in the X-axis direction, The first average error vector The component in the Y-axis direction, The first average error vector The component along the Z-axis. Take the first average error vector. Components in the X-axis direction The first target stroke of the large translational hydraulic cylinder 8 .

[0188] ② The first target rotation angle of the slewing ring 7 :

[0189] Calculate the plane normal vector of the segment suction cup 1 of the assembly machine. and the plane normal vector of the segment 2 to be captured The calculation formula is:

[0190] .

[0191] get The first projection vector in the YOZ plane :

[0192] .

[0193] get The second projection vector in the YOZ plane :

[0194] .

[0195] Calculate the first projection vector With the second projection vector The first included angle The calculation formula is:

[0196] .

[0197] First included angle This refers to the first target rotation angle that the slewing ring 7 of the assembly machine needs to achieve. .

[0198] ③ The first stroke differential of lifting cylinder 9, including the first left stroke differential of the left lifting cylinder. Differential momentum of the first right stroke of the right lifting cylinder :

[0199] Obtain the first spatial error vector respectively and Components on the Z-axis and .

[0200] Calculate components and The difference : .

[0201] The differential momentum of the first left stroke of the left lifting cylinder is ,Right now .

[0202] The differential momentum of the first right stroke of the right lifting cylinder is ,Right now .

[0203] ④ The first target stroke of the pitch cylinder 10 :

[0204] Obtain the plane normal vector of the segment suction cup 1 of the assembly machine. The third projection vector in the XOZ plane : .

[0205] Obtain the plane normal vector of the segment 2 to be captured. The fourth projection vector in the XOZ plane : .

[0206] Calculate the third projection vector With the fourth projection vector The second included angle The calculation formula is:

[0207] .

[0208] According to the second included angle The first target stroke of the pitch cylinder 10 was calculated. .

[0209] Based on the hinge position between the free end of the piston rod of the pitch cylinder 10 and the base, the lifting beam 5, and the segment suction cup 1 of the assembly machine, the solution can be performed using the triangle cosine theorem. For the specific solution method, please refer to the invention patent "Analytical Method of Kinematics for Large Translation Lifting Clamp Type Segment Assembly Machine" (202311182397.6). This solution method is a conventional calculation method in this field and will not be elaborated here.

[0210] ⑤ The first target stroke of the deflection cylinder 11 :

[0211] Obtain the plane normal vector of the segment suction cup 1 of the assembly machine. The fifth projection vector in the XOY plane :

[0212] .

[0213] Obtain the plane normal vector of the segment 2 to be captured. The sixth projection vector in the XOY plane :

[0214] .

[0215] Calculate the fifth projection vector With the sixth projection vector The third included angle The calculation formula is:

[0216] .

[0217] According to the third included angle The first target stroke of the deflection cylinder 11 can be calculated. .

[0218] Based on the hinge position between the free end of the piston rod of the deflection cylinder 11 and the base, the lifting beam 5, and the segment suction cup 1 of the assembly machine, the solution can be performed using the triangle cosine theorem. For the specific solution process, please refer to the kinematic analysis method of the large translational lifting clamp-type segment assembly machine in the invention patent (202311182397.6). This solution method is a conventional calculation method in this field and will not be elaborated here.

[0219] ⑥ Lift the first synchronous target stroke of hydraulic cylinder 9 The calculation formula is:

[0220] .

[0221] Where N is a positive integer. Considering that the grabbing is to gradually approach the target rather than to complete it all at once, N=2~4 is preferred.

[0222] In step 4, during the segment assembly stage, three sets of second spatial error vectors... The mapping relationship between (i=1,2,3) and the target motion increments of each independent joint of the assembly machine is analyzed as follows:

[0223] ① The second target stroke of the large translation cylinder 8 :

[0224] Calculate the second average error vector The calculation formula is:

[0225] .

[0226] in, The second average error vector The component in the X-axis direction, The second average error vector The component in the Y-axis direction, The second average error vector The component along the Z-axis. Take the second average error vector. Components in the X-axis direction As the second target stroke of the large translational hydraulic cylinder 8 .

[0227] ② The second target rotation angle of the rotating ring 7 :

[0228] Calculate the plane normal vector of segment 3 to be assembled. The plane normal vector of the target assembly position 4 The calculation formula is:

[0229] .

[0230] get The seventh projection vector in the YOZ plane :

[0231] .

[0232] get The eighth projection vector in the YOZ plane :

[0233] .

[0234] Calculate the seventh projection vector With the eighth projection vector The fourth included angle The calculation formula is:

[0235] .

[0236] Fourth angle This refers to the second target rotation angle that the slewing ring 7 of the assembly machine needs to perform. .

[0237] ③ The second stroke differential of lifting cylinder 9, including the second left stroke differential of the left lifting cylinder. Differential momentum of the second right stroke of the right lifting cylinder :

[0238] Obtain the second spatial error vector respectively and Components on the Z-axis and .

[0239] Calculate components and The difference : .

[0240] The differential momentum of the second left stroke of the left lifting cylinder is ,Right now .

[0241] The differential momentum of the second right stroke of the right lifting cylinder is ,Right now .

[0242] ④ The second target stroke of the pitch cylinder 10 :

[0243] Obtain the plane normal vector of segment 3 to be assembled. The ninth projection vector in the XOZ plane :

[0244] .

[0245] Obtain the plane normal vector of the target assembly position 4. The tenth projection vector in the XOZ plane :

[0246] .

[0247] Calculate the ninth projection vector With the tenth projection vector The fifth included angle The calculation formula is:

[0248] .

[0249] According to the fifth included angle The second target stroke of the pitch cylinder 10 was calculated. .

[0250] Based on the hinge position between the free end of the piston rod of the pitch cylinder 10 and the base, the lifting beam 5, and the segment suction cup 1 of the assembly machine, the solution can be performed using the triangle cosine theorem. For the specific solution process, please refer to the kinematic analysis method of the large translational lifting clamp-type segment assembly machine in the invention patent (202311182397.6). This solution method is a conventional calculation method in this field and will not be elaborated here.

[0251] ⑤ The second target stroke of deflection cylinder 11 :

[0252] Obtain the plane normal vector of segment 3 to be assembled. The eleventh projection vector in the XOY plane :

[0253] .

[0254] Obtain the plane normal vector of the target assembly position 4. The twelfth projection vector in the XOY plane :

[0255] .

[0256] Calculate the eleventh projection vector With the twelfth projection vector The sixth angle The calculation formula is:

[0257] .

[0258] According to the sixth angle The second target stroke of the deflection cylinder 11 can be calculated. .

[0259] Based on the hinge position between the free end of the piston rod of the deflection cylinder 11 and the base, the lifting beam 5, and the segment suction cup 1 of the assembly machine, the solution can be performed using the triangle cosine theorem. For the specific solution process, please refer to the kinematic analysis method of the large translational lifting clamp-type segment assembly machine in the invention patent (202311182397.6). This solution method is a conventional calculation method in this field and will not be elaborated here.

[0260] ⑥ Lift the second synchronous target stroke of hydraulic cylinder 9 The calculation formula is:

[0261] .

[0262] Where N is a positive integer. Considering that the assembly and positioning is to gradually approach the target rather than to complete it all at once, N=2~4 is preferred.

[0263] Step 5: Iterative servo execution: Execute the instructions for the target motion increment in a preset priority order, and repeat steps 1 to 5 until the magnitudes of all first spatial error vectors / second spatial error vectors are less than the set threshold.

[0264] Preferably, the priority order is set based on the geometric dependence of error elimination, and is as follows: ① Adjust the translation joint to eliminate positional deviation in the plane; ② Adjust the rotation joint in the plane to eliminate torsion about the plane normal; ③ Adjust the attitude rotation joint so that the plane of the assembly machine's segment suction cup 1 is parallel to the plane of the segment 2 to be grasped during the segment grasping stage, and the segment 3 to be assembled is parallel to the target assembly position 4 during the segment assembly stage; ④ Adjust the normal distance joint to achieve the final fit between the assembly machine's segment suction cup 1 and the segment 2 to be grasped during the segment grasping stage, and the final fit between the segment 3 to be assembled and the target assembly position 4 during the segment assembly stage.

[0265] In terms of the assembly machine control, the priority order of the segment grabbing stage is as follows: the first target stroke of the large translation cylinder 8, the first target rotation angle of the rotary ring 7, the first stroke differential of the lifting cylinder 9, the first target stroke of the pitch cylinder 10, the first target stroke of the deflection cylinder 11, and the first synchronous target stroke of the lifting cylinder 9.

[0266] Based on the priority order of the segment grasping stage, and in conjunction with the vision sensor, a closed-loop execution of visual perception and segment grasping is performed multiple times until the following conditions are met, at which point the segment grasping is considered successful:

[0267] threshold The preferred size is 2mm.

[0268] In terms of the control of the assembly machine, the priority order of the segment assembly stage is as follows: the second target stroke of the large translation cylinder 8, the second target rotation angle of the rotary ring 7, the second stroke differential of the lifting cylinder 9, the second target stroke of the pitch cylinder 10, the second target stroke of the deflection cylinder 11, and the second synchronous target stroke of the lifting cylinder 9.

[0269] Based on the priority order of the segment assembly stage, and in conjunction with visual sensors, a closed-loop execution of visual perception and segment assembly is performed multiple times until the following conditions are met, at which point the segment assembly is considered successful:

[0270] threshold The preferred size is 2mm.

[0271] Preferably, a grasping control unit can be provided to execute steps 2 to 5 above. The grasping control unit specifically includes: a coordinate transformation module for real-time calculation of coordinate transformation relationships; an error decoupling mapping module with built-in geometric decoupling mapping rules for generating target motion increments for each independent joint of the assembly machine based on the error vector field; and an iterative control module for managing priority order and closed-loop iteration.

[0272] Please see the appendix Figure 3 This invention employs a visual servoing three-layer closed-loop control architecture based on a visual sensor mounted on the observation platform 6:

[0273] The first layer is the dynamic benchmark alignment layer, which solves the real-time alignment problem between the local coordinate system of the observation platform and the base coordinate system of the assembly machine. Its core technology adopts real-time PnP calculation based on common-view markers. By dynamically photographing the benchmark marker points through the movement of the observation platform 6, the transformation relationship from the local coordinate system of the observation platform to the base coordinate system of the assembly machine located on the lifting beam of the assembly machine is calculated through PnP calculation.

[0274] The second layer, the geometric error calculation layer, addresses the problem of extracting executable error information from visual sensor data. Its core technology employs three-point error vector field analysis and geometric decoupling mapping. Based on three first spatial error vectors / second spatial error vectors in the assembly machine's base coordinate system, geometric decoupling mapping rules are formulated to achieve the allocation and calculation of the target motion increment for each independent joint.

[0275] The third layer, the priority execution layer, is used to solve the problem of achieving convergence by driving each independent joint in an orderly manner. Its core technology is the sequential iteration of defining geometric dependencies. By defining the motion priority sequence of each independent joint, and through multiple iterations and action execution, the accurate alignment of the assembly machine's segment suction cup 1 with the segment 2 to be gripped, and the segment 3 to be assembled with the target assembly position 4 is finally achieved.

[0276] This invention provides a servo control method that directly drives each joint based on real-time visual feature errors through decoupling mapping. This fundamentally avoids the dependence on mechanism models and absolute pose calculations in existing technologies, and maintains the stability of the control reference when the observation platform moves. It avoids motion interference of the observation platform and the accumulation of pose calculation errors, and is especially suitable for dynamic working conditions where there is relative motion between the independent joints of the observation platform and the assembly machine.

[0277] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the invention. Therefore, any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A visual servo grasping and assembly method for tunnel boring machine segments adapted to the movement of an observation platform, characterized by: Includes the following steps: Step 1: Install a visual sensor on the mobile observation platform (6) and collect the three-dimensional coordinates of the reference mark points on the assembly machine through the visual sensor, and at the same time collect the three-dimensional coordinates of the feature points on the assembly machine, the segment and the target assembly position; Step 2: Unify the local coordinate system of the observation platform with the base coordinate system of the assembly machine at the current moment; Step 3: Construct the error vector field; Step 4: Based on the spatial distribution pattern of the error vector field, according to the preset geometric decoupling mapping rules, the overall error is allocated and mapped into the target motion increment of each independent joint of the assembly machine; Step 5: Iterative servo execution: Execute the instructions for the target motion increment in a preset priority order, and repeat steps 1 to 5 until the magnitudes of all first spatial error vectors / second spatial error vectors are less than the set threshold.

2. The shield tunnel segment visual servo grasping and assembly method adapted to the movement of the observation platform according to claim 1, characterized in that: In step 1, four reference markers are set up on the lifting beam (5) of the assembly machine, and are denoted as follows: , i=1,2,3,4; In step 1, during the segment gripping stage, the feature points include three first feature points set on the segment suction cup (1) of the assembly machine and three second feature points set on the segment (2) to be gripped. The first feature points are denoted as follows: Let i = 1, 2, 3, and the second feature point be denoted as... i=1,2,3; During the segment assembly stage, the feature points include three third feature points set on the segment to be assembled (3) and three fourth feature points set on the target assembly position (4). The third feature points are denoted as Let i = 1, 2, 3, and the fourth feature point be denoted as... , i=1,2,3.

3. The shield tunnel segment visual servo grasping and assembly method adapted to the movement of the observation platform according to claim 1, characterized in that: Step 2 includes the following sub-steps: Step 21: The base coordinate system {0} of the assembly machine is located on the lifting beam (5) of the assembly machine. Calculate the known coordinates of the four reference points in the base coordinate system {0} of the assembly machine. , i = 1~4; Step 22: Calculate the known coordinates of the four reference points in the local coordinate system {U} of the observation platform. , i = 1~4; Step 23: Real-time calculation of the transformation relationship between the local coordinate system {U} of the observation platform and the base coordinate system {0} of the assembly machine at the current moment.

4. The shield tunnel segment visual servo grasping and assembly method adapted to the movement of the observation platform according to claim 3, characterized in that: Step 23 includes the following sub-steps: Step 231: The transformation matrix from the local coordinate system {U} of the observation platform to the base coordinate system {0} of the assembly machine is a 4×4 transformation matrix. 4×4 transformation matrix Including rotation matrix Translation vector It satisfies the following relation: Solve the 4×4 transformation matrix ; Step 232: Solve for the rotation matrix using the solvePnP algorithm. Translation vector .

5. The shield tunnel segment visual servo grasping and assembly method adapted to the movement of the observation platform according to claim 1, characterized in that: In step 3, during the segment grabbing stage, the coordinates of the three first feature points and three second feature points under the local coordinate system {U} of the observation platform are synchronously transformed to the base coordinate system {0} of the assembly machine using the transformation relationship from the local coordinate system {U} of the observation platform to the base coordinate system {0} of the assembly machine, and three sets of first spatial error vectors are calculated. The specific calculation method is as follows: i = 1, 2, 3; Step 31: Calculate the three first feature points on the segment suction cup (1) of the assembly machine respectively. and three second feature points Known coordinates in the local coordinate system {U} of the observation platform and , i = 1~3; Step 32: Set the coordinates of the three first feature points on the segment suction cup (1) of the assembly machine under the local coordinate system {U} of the observation platform. The coordinates of the three second feature points Transforming to the assembly machine's base coordinate system {0}, the transformation relationship is as follows: ； in, Let {0} be the coordinates of the first feature point in the assembly machine's base coordinate system. , Let {0} be the coordinates of the second feature point in the assembly machine's base coordinate system. ; Step 33: Calculate the three sets of first spatial error vectors in the assembly machine's base coordinate system {0}. For i=1,2,3, the calculation formula is: 。 6. The shield tunnel segment visual servo grasping and assembly method adapted to the movement of the observation platform according to claim 1, characterized in that: In step 3, during the segment assembly stage, the coordinates of the three third feature points and three fourth feature points under the local coordinate system {U} of the observation platform are synchronously transformed to the base coordinate system {0} of the assembly machine using the transformation relationship from the local coordinate system {U} of the observation platform to the base coordinate system {0} of the assembly machine, and three sets of second spatial error vectors are calculated. The specific calculation method is as follows: i = 1, 2, 3; Step 31: Calculate the three third feature points on the segment (3) to be assembled. and the three fourth feature points at the target assembly position (4) Known coordinates in the local coordinate system {U} of the observation platform and , i = 1~3; Step 32: Set the coordinates of the three third feature points on the segment (3) to be assembled under the local coordinate system {U} of the observation platform. The coordinates of the three fourth feature points at the target assembly position (4) Transforming to the assembly machine's base coordinate system {0}, the transformation relationship is as follows: ； in, Let {0} be the coordinates of the third feature point in the assembly machine's base coordinate system. , Let {0} be the coordinates of the fourth feature point in the assembly machine's base coordinate system. ; Step 33: Calculate the three sets of second spatial error vectors under the assembly machine's base coordinate system {0}. For i=1,2,3, the calculation formula is: 。 7. The shield tunnel segment visual servo grasping and assembly method adapted to the movement of the observation platform according to claim 1, characterized in that: In step 4, the geometric decoupling mapping rules include: When the three sets of first spatial error vectors are distributed in space in an approximately parallel manner, it indicates that the deviation is mainly manifested as an overall positional offset. The spatial average value of the three sets of first spatial error vectors is calculated, the first translation error component is extracted, and the first translation error component is mapped to the translation joint of the assembly machine. When the three sets of second spatial error vectors are distributed in space in an approximately parallel manner, it indicates that the deviation is mainly manifested as an overall positional offset. The spatial average value of the three sets of second spatial error vectors is calculated, the second translation error component is extracted, and the second translation error component is mapped to the translation joint of the assembly machine. When the three sets of first spatial error vectors change linearly, that is, when the three sets of first error vectors are parallel in direction in space but change systematically in magnitude, it indicates that the deviation is mainly manifested as the plane tilt angle of the assembly machine segment suction cup (1). By analyzing the distribution gradient characteristics of the three sets of first spatial error vectors, the first attitude rotation error component is extracted, and the first attitude rotation error component is mapped to the differential rotation control of the attitude rotation joint and lifting cylinder (9) of the assembly machine. When the three sets of second spatial error vectors change linearly, that is, when the three sets of second error vectors are parallel in direction in space but change systematically in magnitude, it indicates that the deviation is mainly manifested as the plane tilt angle of the segment (3) to be assembled. By analyzing the distribution gradient characteristics of the three sets of second spatial error vectors, the second attitude rotation error component is extracted, and the second attitude rotation error component is mapped to the differential rotation control of the attitude rotation joint and lifting cylinder (9) of the assembly machine. When the first spatial error vector is distributed in a rotational manner in the plane, it indicates that the deviation is mainly manifested as torsion around the plane normal. By analyzing the directional change characteristics of the first spatial error vector in the plane, the rotational error component in the first plane is extracted, and the rotational error component in the first plane is mapped to the in-plane rotational joint of the assembly machine. When the second spatial error vector exhibits a rotational distribution in the plane, it indicates that the deviation is mainly manifested as torsion around the plane normal. By analyzing the directional change characteristics of the second spatial error vector in the plane, the rotational error component in the second plane is extracted, and the rotational error component in the second plane is mapped to the in-plane rotational joint of the assembly machine.

8. The shield tunnel segment visual servo grasping and assembly method adapted to the movement of the observation platform according to claim 1, characterized in that: In step 4, during the segment grasping stage, three sets of first spatial error vectors... For i=1,2,3, the mapping relationship to the target motion increment of each independent joint of the assembly machine is analyzed as follows: ① The first target stroke of the large translational hydraulic cylinder (8) : Calculate the first average error vector The calculation formula is: ； in, The first average error vector The component in the X-axis direction, The first average error vector The component in the Y-axis direction, The first average error vector The component in the Z-axis direction; take the first average error vector. Components in the X-axis direction The first target stroke of the large translational hydraulic cylinder (8) ; ② The first target rotation angle of the slewing ring (7) : Calculate the plane normal vector of the segment suction cup (1) of the assembly machine. The plane normal vector of the segment (2) to be captured The calculation formula is: ； get The first projection vector in the YOZ plane : ； get The second projection vector in the YOZ plane : ； Calculate the first projection vector With the second projection vector The first included angle The calculation formula is: ； First included angle This refers to the first target rotation angle of the rotating ring (7) of the assembly machine. ; ③ The first stroke differential of the lifting cylinder (9), including the first left stroke differential of the left lifting cylinder. Differential momentum of the first right stroke of the right lifting cylinder : Obtain the first spatial error vector respectively and Components on the Z-axis and ; Calculate components and The difference : ; The differential momentum of the first left stroke of the left lifting cylinder is ,Right now ; The differential momentum of the first right stroke of the right lifting cylinder is ,Right now ; ④ The first target stroke of the pitch cylinder (10) : Obtain the plane normal vector of the segment suction cup (1) of the assembly machine. The third projection vector in the XOZ plane : ; Obtain the plane normal vector of the segment (2) to be captured. The fourth projection vector in the XOZ plane : ; Calculate the third projection vector With the fourth projection vector The second included angle The calculation formula is: ； According to the second included angle The first target stroke of the pitch cylinder (10) was calculated. ; ⑤ The first target stroke of the deflection cylinder (11) : Obtain the plane normal vector of the segment suction cup (1) of the assembly machine. The fifth projection vector in the XOY plane : ； Obtain the plane normal vector of the segment (2) to be captured. The sixth projection vector in the XOY plane : ； Calculate the fifth projection vector With the sixth projection vector The third included angle The calculation formula is: ； According to the third included angle The first target stroke of the deflection cylinder (11) can be calculated. ; ⑥ The first synchronous target stroke of lifting cylinder (9) The calculation formula is: ； Where N is a positive integer, N=2~4.

9. The method for visual servo grasping and assembling of tunnel segments adapted to the movement of the observation platform according to claim 1, characterized in that: In step 4, during the segment assembly stage, three sets of second spatial error vectors... For i=1,2,3, the mapping relationship to the target motion increment of each independent joint of the assembly machine is analyzed as follows: ① The second target stroke of the large translational cylinder (8) : Calculate the second average error vector The calculation formula is: ； in, The second average error vector The component in the X-axis direction, The second average error vector The component in the Y-axis direction, The second average error vector The component in the Z-axis direction; take the second average error vector. Components in the X-axis direction As the second target stroke of the large translational hydraulic cylinder (8) ; ② The second target rotation angle of the slewing ring (7) : Calculate the plane normal vector of the segment (3) to be assembled. The plane normal vector of the target assembly position (4) The calculation formula is: ； get The seventh projection vector in the YOZ plane : ； get The eighth projection vector in the YOZ plane : ； Calculate the seventh projection vector With the eighth projection vector The fourth included angle The calculation formula is: ； Fourth angle This refers to the second target rotation angle of the rotating ring (7) of the assembly machine. ; ③ The second stroke differential of the lifting cylinder (9) includes the second left stroke differential of the left lifting cylinder. Differential momentum of the second right stroke of the right lifting cylinder : Obtain the second spatial error vector respectively and Components on the Z-axis and ; Calculate components and The difference : ; The differential momentum of the second left stroke of the left lifting cylinder is ,Right now ; The differential momentum of the second right stroke of the right lifting cylinder is ,Right now ; ④ The second target stroke of the pitch cylinder (10) : Obtain the plane normal vector of the segment (3) to be assembled. The ninth projection vector in the XOZ plane : ； Obtain the plane normal vector of the target assembly position (4). The tenth projection vector in the XOZ plane : ； Calculate the ninth projection vector With the tenth projection vector The fifth included angle The calculation formula is: ； According to the fifth angle The second target stroke of the pitch cylinder (10) was calculated. 。; ⑤ The second target stroke of the deflection cylinder (11) : Obtain the plane normal vector of the segment (3) to be assembled. The eleventh projection vector in the XOY plane : ； Obtain the plane normal vector of the target assembly position (4). The twelfth projection vector in the XOY plane : ； Calculate the eleventh projection vector With the twelfth projection vector The sixth angle The calculation formula is: ； According to the sixth angle The second target stroke of the deflection cylinder (11) can be calculated; ⑥ The second synchronous target stroke of lifting cylinder (9) The calculation formula is: ； Where N is a positive integer, N=2~4.

10. The shield tunnel segment visual servo grasping and assembly method adapted to the movement of the observation platform according to claim 1, characterized in that: In step 5, the priority order is set based on the geometric dependence of error elimination, and is as follows: ① Adjust the translation joint to eliminate the positional deviation in the plane; ② Adjust the rotation joint in the plane to eliminate the torsion around the plane normal; ③ Adjust the attitude rotation joint so that the plane of the assembly machine's segment suction cup (1) is parallel to the plane of the segment to be gripped (2) during the segment gripping stage, and the segment to be assembled (3) is parallel to the target assembly position (4) during the segment assembly stage; ④ Adjust the normal distance joint to achieve the final fit between the assembly machine's segment suction cup (1) and the segment to be gripped (2) during the segment gripping stage, and the final fit between the segment to be assembled (3) and the target assembly position (4) during the segment assembly stage. In terms of the assembly machine control, the priority order of the segment grabbing stage is as follows: the first target stroke of the large translation cylinder (8), the first target rotation angle of the rotary ring (7), the first stroke differential of the lifting cylinder (9), the first target stroke of the pitch cylinder (10), the first target stroke of the deflection cylinder (11), and the first synchronous target stroke of the lifting cylinder (9). Based on the priority order of the segment grasping stage, and in conjunction with the vision sensor, a closed-loop execution of visual perception and segment grasping is performed multiple times until the following conditions are met, at which point the segment grasping is considered successful: threshold It is 2mm; In terms of the control of the assembly machine, the priority order of the segment assembly stage is as follows: the second target stroke of the large translation cylinder (8), the second target rotation angle of the rotary ring (7), the second stroke differential of the lifting cylinder (9), the second target stroke of the pitch cylinder (10), the second target stroke of the deflection cylinder (11), and the second synchronous target stroke of the lifting cylinder (9). Based on the priority order of the segment assembly stage, and in conjunction with visual sensors, a closed-loop execution of visual perception and segment assembly is performed multiple times until the following conditions are met, at which point the segment assembly is considered successful: threshold It is 2mm.

Images