JP2009279700A - Work gripping method and work gripping device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and a device for recognizing the shape and position of a flexible work and gripping the work. <P>SOLUTION: This work gripping device in which the attitude of the fingers is changed in three degrees of freedom comprises a rotating shaft drive section 101 which has an installation surface 105 for mounting on a robot arm and a center axis in the direction normal to the installation surface 105; a rotating drive section 102 connected to the rotating shaft drive section 101 in such a manner as to have a center axis positioned at a predetermined angle α1 relative to the center axis of the rotating shaft drive section 101; a rotating drive section 103 connected to the rotating shaft drive section 102 in such a manner as to have a center axis positioned at a predetermined angle α2 relative to the center axis of the rotating shaft drive section 102; a finger mechanism for holding the flexible work connected to the rotating drive section 103, and a robot controller for controlling these operations. The robot controller grips the work in a desired gripping attitude by calculating the attitude of the finger mechanism. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a method for gripping a workpiece having an indeterminate form or flexibility arranged in a non-aligned state (for example, a variable sheet-like article such as a cloth on a square, hereinafter sometimes referred to as “flexible object”). The present invention relates to a device, for example, a technique for grasping a characteristic part of one laundry washed from a pile of stacked laundry.

The selection process in a conventional laundry factory is performed according to the following procedure. First, the soiled cloth pieces received are collected together and washed or washed with water or a cleaning agent, and then the cloth pieces are cleaned, and then introduced into the dehydration process. The cloth pieces are dehydrated with a dehydrator. Then, the cloth pieces in which the cloth pieces are entangled with each other are untangled and the cloth pieces are piled up.
The cloth piece is flexible and easily deformed, so that it is difficult to mechanize the cloth piece automatically and it is difficult to take out one piece of the cloth piece from the pile and unfold it. The cloth pieces were spread and folded by hand, or the cloth pieces were spread by hand, and then the edges (sides) of the cloth pieces were aligned and supplied to the folding device for folding.

However, it takes a lot of labor to pick up each piece of cloth piled up by hand under a hot and humid adverse environment and arrange it in a folding device. There were problems such as high.
Therefore, a device for folding cloth pieces was devised, but the cloth pieces after washing and dewatering were piled up, picked up one by one from the piled cloth pieces, and the diagonal corners of the cloth pieces were ridged There is a device (for example, Patent Document 1) in which the folding means is mechanized, but the conventional device has a problem that the operation is complicated and time-consuming and the folding accuracy is poor.

Therefore, an affiliated company of the applicant proposed a method and apparatus for grasping and hanging an arbitrary part of a rectangular cloth piece and grasping one of the corners of the suspended cloth piece (Patent Document 2). .
Further, an applicant's affiliated company has proposed a device for folding a rectangular cloth piece such as a hand towel into a triangular shape (Patent Document 3).

By the way, in order to perform the same work as a human being with a robot hand, it is necessary that the work exists in a workable space of the robot hand, that is, the hand effector of the robot hand is reachable and the hand effector is desired. It is necessary to exist in a space that can take the attitude of This workable space is a subset of the reachable space, and the space in which a desired posture can be expressed is further limited.
In the 6-axis manipulator robot hand, it is easy to control and control the joint angle for expressing the tool posture corresponding to the workpiece posture. However, since the reachable space contains a singular posture that cannot be represented, I may not be able to express. From another point of view, it is necessary to program the robot hand to perform work while the work is reachable. In such a conventional robot hand, it is extremely difficult to accurately grasp and spread an irregular shaped workpiece.

As a gripping method for recognizing a workpiece by a visual recognition sensor and positioning and determining a posture of a robot hand, there are methods described in Patent Documents 4 and 5, for example. However, when the tool posture for corresponding to the directionality of the workpiece posture that appears includes a singular point, the problem that the posture cannot be expressed has not been solved.
Japanese Patent Laid-Open No. 2004-228561 has a problem to solve the singularity problem in which two axes are coaxially positioned and an axis positioned in the middle performs an irregular operation. Paying attention to the fact that the robot hand described in this document often moves linearly with the hand effector in the same posture, a linear motion axis is provided at the tip of the articulated arm, and the linear motion axis It is disclosed that the position of the singular point is shifted by changing the relative position of the hand effector with respect to the arm.

Japanese Patent Publication No. 5-34262 Japanese Patent No. 3379857 Japanese Patent No. 3801758 JP 2003-94367 A JP 2006-130580 A JP 2001-54889 A

If a linear motion axis or a rotation axis is added to avoid a singular point, the combination of joint angles for realizing the posture (tool posture) of a hand effector such as a finger is doubled and the control becomes complicated, The problem of increased cost and weight arises.
On the other hand, in a robot hand having 5 axes or less, there is a case where there is no calculation solution for matching the workpiece posture and the tool posture, and there is a problem that the posture cannot be gripped or the posture after gripping is not clean.

  An object of the present invention is to provide a method and an apparatus for recognizing and gripping the shape and position of a workpiece that is a flexible object.

The outline of the means for solving the above problems is as follows.
(1) Expression of gripping posture (1-1) A posture conversion tool for gripping a workpiece is provided. This posture conversion tool can be attached to the arm unit, and is a hand effector (for example, a finger mechanism), a drive mechanism for posture conversion, and a means for calculating the amount of movement and rotation of the drive actuator for posture expression. Have
(1-2) A posture expression including virtual posture conversion is realized. Making use of the characteristics of the workpiece, which is a flexible object, the hand effector after gripping and the posture of the workpiece (pose) are accurately matched. The robot hand joint angle control can be calculated by including virtual fingers and virtual joints, or the deformation of the cloth within the allowable range can be used to allow the robot hand to work with a small degree of freedom. .
(2) Calculation of joint angle In order to express a desired gripping posture by the posture conversion tool, a robot controller having means for calculating the movement amount / rotation amount of all the drive actuators is required. Normally, if control values such as coordinates and tool parameters are input, the movement amounts and rotation amounts of all the drive actuators can be calculated by inverse kinematic calculation. However, when the control value includes a singular posture (singular point), a solution cannot be obtained by a normal calculation. Here, the peculiar posture refers to a posture in which the number of joints and motors is sufficient, but it cannot move in a specific direction. Where this unique posture is located is determined from the relationship between the length of each link of the manipulator and the orientation in which the joint is fixed with respect to each link. In order to avoid the singular posture, the present invention takes the following two methods.
(2-1) When the posture of the hand effector (tool posture) matches the workpiece posture is a singular point, the joint rotation angle of the posture conversion tool is changed, and calculation is performed for different tool postures.
(2-2) A list of tool postures not including singular points is stored in a table array, and an optimum tool posture for the set work is selected from the table array.

Usually, a manipulator has a combination of joint angles that makes the Jacobian matrix unique. Manipulators have singularities at the boundaries of their workspace, and most have singularity trajectories inside the workspace. Singular points can be classified into the following two types. One is a work space boundary singularity, which exists in a state where the manipulator is fully extended or folded and overlapped, and the hand effector is in the vicinity of the work space or in a boundary shape. The other is a singular point inside the work space, generally occurring when two or more joint axes are aligned. The posture conversion tool avoids these singular points by changing the tool posture with respect to both of these singular points.
For the determination of the singular posture, it is preferable to determine whether the matrix is a regular matrix by calculating a determinant of the Jacobian matrix. When the Jacobian matrix is singular, that is, when there is no inverse matrix, the matrix is not a regular matrix and can be determined by the determinant being 0. Here, the Jacobian matrix is a matrix that represents a relationship such as how much the hand moves in Cartesian coordinates in response to each joint performing a slight rotation. This is because inverse kinematics calculation is generally complicated, but once the Jacobian matrix is obtained, the movement of the robot hand can be easily calculated.

Means for solving the problems based on the above technical idea will be described below.
[1] A robot hand system including a gripping unit that grips a workpiece with fingers that open and close, a driving unit that includes multiple axes that move the gripping unit in space, a gripping position information acquisition unit, and a calculation unit. A first step of determining a gripping position of a flexible object on the basis of information on a characteristic part of the workpiece acquired by the gripping position information acquisition unit, and the calculation unit is in a finger in an open state. A second step of setting the path and posture of the gripping unit so that the gripping position of the workpiece is positioned at the second position, and a third step of gripping the gripping position of the workpiece by moving the gripping unit by the driving unit. The gripping method of the workpiece.
[2] The drive unit enables posture conversion with two degrees of freedom, and the calculation unit matches the origin of the coordinate system of the workpiece with the origin of the coordinate system of the finger and two axes in the second step. A step of setting an allowable rotation angle with respect to the first axis of the workpiece, setting a virtual axis for rotating the gripping portion within the allowable rotation angle, and setting a path and posture of the gripping portion. The work gripping method according to [1].
[3] The drive unit enables posture conversion with one degree of freedom, and the calculation unit causes the coordinate system of the workpiece and the origin of the coordinate system of the finger and two axes to coincide in the second step. A first virtual axis that sets an allowable rotation angle with respect to the first axis of the workpiece and rotates the gripping portion within the allowable rotation angle, and a second axis that is orthogonal to the first axis of the workpiece. [1], including a step of setting an allowable rotation angle, setting a second virtual axis for rotating the grip portion within the allowable rotation angle, and setting a path and posture of the grip portion. The workpiece gripping method described.
[4] The work gripping method according to any one of [1] to [3], wherein the work is a sheet-like cloth.
[5] A workpiece gripping device that performs finger posture conversion with three degrees of freedom, having a mounting surface (105) for the robot arm, and a rotation axis having a central axis in a normal direction to the mounting surface (105) A drive unit (101), and a rotary drive unit (102) connected to the rotary shaft drive unit (101) so as to have a central axis that forms a predetermined angle α1 with the central axis of the rotary shaft drive unit (101) A rotation drive unit (103) connected to the rotation axis drive unit (102) so as to have a central axis that forms a predetermined angle α2 with the central axis of the rotation axis drive unit (102), and the rotation drive unit (103 A finger mechanism for holding a workpiece that is a flexible object, and a robot controller for controlling the operation of the workpiece. The robot controller calculates a posture of the finger mechanism to obtain a desired workpiece. A workpiece gripping device that grips in a different gripping posture.
[6] A workpiece gripping device that performs finger posture conversion with two degrees of freedom, having a mounting surface (105) for the robot arm, and a rotation axis having a central axis in a normal direction to the mounting surface (105) A drive unit (101), and a rotary drive unit (102) connected to the rotary shaft drive unit (101) so as to have a central axis that forms a predetermined angle α1 with the central axis of the rotary shaft drive unit (101) A finger mechanism for holding a workpiece, which is a flexible object, connected to the rotary shaft drive unit (102) so as to have a central axis that forms a predetermined angle α2 with the central axis of the rotary shaft drive unit (102). A robot controller for controlling these operations, wherein the robot controller sets a virtual axis in the deformation direction of the workpiece and calculates the posture of the finger mechanism to hold the workpiece in a desired holding posture. Gripping device.
[7] The workpiece gripping device according to [5] or [6], further including a robot arm that controls positioning of the tip of the finger mechanism with Cartesian coordinates.
[8] A workpiece gripping device that changes the posture of a finger with one degree of freedom, and has a mounting surface (105) to the robot arm, and forms a predetermined angle α1 with the normal direction to the mounting surface (105). The rotation drive unit (102) connected so as to have a central axis and a rotation axis drive unit (102) having a central axis that forms a predetermined angle α2 with the central axis of the rotation axis drive unit (102). A finger mechanism for holding a workpiece, which is a flexible object, and a controller that controls the operation of the finger mechanism. When the gripping posture becomes a specific posture, the controller drives the rotation drive unit (102 ) Is rotated by a predetermined angle, and the posture of the finger mechanism is calculated by using the joint of the robot arm of 6 degrees of freedom connectable to the mounting surface (105) for posture conversion, thereby gripping the workpiece in a desired gripping posture. A workpiece gripping device that makes it possible to
[9] An attachment surface (105) to the robot arm, a finger mechanism for holding a workpiece that is a flexible object connected to the attachment surface (105), and a controller, wherein the controller has a gripping posture. In the case of a unique posture, a virtual axis is set in the deformation direction of the workpiece, and the posture of the finger mechanism is calculated by using the joint of the robot arm with 6 degrees of freedom connectable to the mounting surface (105) for posture transformation. Thus, a workpiece gripping device that enables a workpiece to be gripped in a desired gripping posture.
[10] The apparatus further includes gripping position information acquisition means for acquiring information on one to three feature portions that define the gripping position of the workpiece, and the robot controller calculates a workpiece posture based on the information on the feature portion, A calculation step of calculating a gripping posture of the finger mechanism based on the posture, and a determination step of determining whether the gripping posture is a singular point and calculating the gripping posture again when it is determined that the gripping posture is a singular point The workpiece gripping device according to [8] or [9].

According to the present invention, it is possible to grip a workpiece feature such as an indeterminate and flexible cloth with a tool posture within an allowable range for the workpiece posture.
Further, even with a robot hand that is considered to have insufficient degrees of freedom in the prior art, it can be gripped so that the workpiece posture after gripping matches the tool posture.
Furthermore, it is possible to always set a gripping posture that avoids singular points.

The best mode of the present invention will be described with an example of a handling system for a cloth such as a washed towel.
As shown in FIG. 1, the handling system of the present invention is for loading a work 5 conveyed by a conveyor device 9 into a folding machine 8 by a robot hand A10 and a robot hand B20.
The components of the handling system of the present invention are arranged in the frame 1. This handling system includes the pull-up hand 2, the entire imaging mechanism 3, the recognition base 7, the robot hand A10, and the robot hand B20 as substantial components.

In addition, the following "term" used in this specification has the following meaning unless there is particular notice.
Position: Position and orientation expressed in 3D Cartesian coordinates: Direction of another coordinate system expressed in 3D Cartesian coordinates Pose: Position and orientation, and another coordinate system in a certain coordinate system Position and orientation coordinates: coordinate values expressed in a certain coordinate system

<< Imaging mechanism >>
An overall imaging mechanism 3 that is a stereo visual sensor is provided above the frame 1.
The entire imaging mechanism 3 has a projector 31 for projecting the projection pattern 60. It is possible to identify the height of the workpiece and the position of the end by imaging with the whole imaging mechanism 3.

  The hand image pickup mechanism A13 mounted on the robot hand A10 and the hand image pickup mechanism B23 mounted on the robot hand B20 are also configured by the tilt optical system. The hand imaging mechanism 13 and the hand imaging mechanism 23 include a projector 41 similar to the projector 31 in order to project the projection pattern 60.

《Robot hand》
The robot hands 10 and 20 of the present invention include arm portions 11 and 21 formed of multi-axis joints, and a posture conversion tool 100 attached to the arm portions 11 and 21. The configuration of the robot hand will be described in detail in the embodiments.

<< 1. Outline of work process >>
FIG. 2 is a drawing for explaining the outline of the work process in the handling system of the present invention.
(a) Detection of highest point In this step, the highest point of the workpiece peak 6 is detected based on the imaging data of the entire imaging mechanism 3. Here, the pile 6 of the work is a pile of non-patterned monochromatic towels stacked after being washed and dehydrated. Grab the highest point of the workpiece peak 6 with the robot hand A or B.
(b) Grasping the workpiece edge
(i) The work 5 gripped by the robot hand is developed on the recognition table 7.
(ii) A location that is considered to be one of the end portions of the workpiece is determined by the hand imaging mechanism, and the location is grasped by the finger mechanism.

(c) Grasping the corner of the workpiece Transfer the workpiece gripped by the finger mechanism to the pull-up hand 2. The corner of the work suspended by the pull-up hand 2 is determined by the hand imaging mechanism, and the corner of the work is gripped by the finger mechanism.
(d) Holding the other corner of the workpiece While holding the corner of the workpiece, pull the workpiece upward by the robot hand. Then, the other corner is positioned below the workpiece. The other corner of the workpiece is determined by the hand imaging mechanism 4 and the other corner of the workpiece is gripped by the finger mechanism of the other robot hand.
(e) Inserting the work into the folding machine The work gripped at the two corners is placed on the conveyor of the folding machine in a state of being extended by two robot hands. The work is folded at a predetermined number of folds by the folding machine.

<< 2. Grip by finger mechanism >>
FIG. 3 is a flowchart of a procedure for gripping a workpiece by the finger mechanism.
First, grip position information is acquired by a measurement command. By moving the measurement sensor attached to the robot arm, the measurement origin is moved to the setting pose for measurement. The grip position information of the workpiece 5 is input by imaging the workpiece 5. The workpiece 5 is imaged by the hand imaging mechanism 4, image data including feature point information is transmitted to the controller, and image processing is performed on the imaged data to determine the gripping position. More specifically, a position where an end point (P1), a long side point (P2), and a short side point (P3) as feature points are detected is set as a gripping position (STEP 1). In the present embodiment, the XYZ coordinates of feature points are measured, and the posture of the workpiece 5 can be calculated.
Next, the pose information of the imaged measurement origin is input to the controller and used as basic information for converting the coordinate value measured in STEP 1 into robot coordinates (STEP 2).
Then, coordinate transformation is performed to obtain the pose of the workpiece end in the robot coordinates, and this pose becomes the finger grip pose (STEP 3).
When the gripping position of the workpiece 5 is determined, the path / posture of the robot hand up to the gripping position is calculated (STEP 4). Subsequently, it is checked whether or not the path / posture expression is impossible because the calculated path of the robot hand includes a singular point (STEP 5). When there is a defect in the route / posture, the route is recalculated until it is confirmed that the route does not include the defect. When the route is fixed, route information is transmitted to the robot hand (STEP 6).

  Details of the present invention will be described in the examples. However, the present invention is not limited to the examples.

[1] Structure The robot hand according to the first embodiment includes an arm unit 11 having three or more degrees of freedom and a posture changing tool 100 having three degrees of freedom attached to the tip thereof.
The arm unit 11 is composed of a plurality of manipulators, and may be, for example, a Cartesian (Cartesian coordinate system) robot arm having three axes of XYZ as shown in FIG. 17, or a vertical articulated joint as shown in FIG. An arm may be used. Each joint shaft of the arm unit 11 includes, for example, an actuator such as a servo motor, a planetary gear type speed reducer, and a position detector such as an encoder, and the actuator is driven and controlled by a robot controller (not shown). .
In the present embodiment, the position of the tip point of the finger does not change due to the posture change by the posture conversion tool 100. The position of the tip point is expressed by the robot arm 11, and the gripping posture is expressed by the posture conversion tool 100. A robot controller (not shown) controls the position of the finger tip and the posture of the finger.
The posture conversion tool 100 is configured as shown in FIGS. 7 and 8, and can change the posture of the hand effector 104 with respect to the attachment surface 105 with three degrees of freedom. The tool 100 is connected to the robot arm unit 11 by an attachment surface 105. 7 and 8, Σ _MB is a pose (coordinate system) of a tool mounting surface (mechanical interface) 105, and Σ ₃ is a coordinate system that defines the posture of the finger. In order from the mounting surface 105, a first joint 101, a second joint 102, and a third joint 103 are arranged to determine each joint pose. In FIG. 7, these joint axes are arranged so as to intersect at one point of the finger tips. The tip of the hand effector 104 is provided with a two-finger finger that opens and closes and constitutes a gripping portion. This gripping part is attached to the rotary actuator 103 and can define a desired posture. At the other end of the rotary actuator 103, a rotating shaft inclined by α ₂ degrees with respect to the finger length direction (W direction) is arranged, and the rotating shaft is connected to the rotary actuator 102. At the other end of the rotary actuator 102, a rotation axis inclined by α ₁ degree with respect to the rotation axis is arranged, and the rotary actuator 101 is connected to the rotation axis. Here, α ₁ and α ₂ are both acute angles. Further, α ₁ and α ₂ are preferably set to the same angle for the reason described later.
Note that the posture conversion tool 100 of this embodiment is intended only for posture change. Therefore, it is desirable to provide a moving mechanism for translating the tool 100 or the workpiece so that the translational positional relationship between the workpiece end and the finger end can be controlled.

In the above, as α ₁ is set larger, the range in which the posture can be expressed freely becomes larger. When α = 90 degrees, the W-direction posture of the tool posture can theoretically be expressed in degrees of freedom from the entire surface of the spherical surface toward the center, but it is realistic because the tool itself easily interferes with the workpiece when gripping the workpiece. Not. Therefore, non extensive posture conversion, it is advantageous to set small alpha ₁ Since the length of the link is shortened it becomes small and light. Α ₁ is preferably selected from, for example, 18 degrees, 22.5 degrees, and 45 degrees. Although α ₁ and α ₂ can be controlled even if they are different, the absolute values of the numerical values are preferably the same. This is because when the difference occurs, a work impossible range is generated near the θ1 axis.
In the configuration of the present invention, since the extension lines of the axes intersect as described above, the origin of the joint pose is arranged at a tolerance point. Accordingly, when parameters are determined so that they can be calculated most simply, di and a _i−1 are 0 for all joints. Further, since the intersection position of the finger tips becomes the origin position of the calculated joint pose, the pose of the joint 3 becomes the tool posture as it is.
By determining the link parameters in this manner, the tool posture (finger pose) when the rotation angles θ1 to 3 of each joint are rotated can be calculated.
In this embodiment, a necessary posture expression range is set while calculating a trade-off with the rigidity of the tool 100. For example, when α ₁ = α ₂ = 18 degrees, the fingers express a gripping posture within a cone having an apex angle of 72 degrees. For example, when α ₁ = α ₂ = 45 degrees, the apex angle is 180 degrees, that is, the fingers express the gripping posture in the hemisphere. Table 1 below shows the link parameters of the robot hand described above in DH notation (Denabit-Hartenberg notation). According to this link parameter, the position of the tip of the grip portion does not change. That is, the translation positions of the tool attachment surface pose Σ _MB and the finger origin pose Σ ₀ are unchanged even when θ rotates.
For conversion of ^MB T ₁ , d1 should be included in the calculation, but for the sake of simplicity, the format is ^MB T ₀ ⁰ T ₁ . For the actual position control of the tip, the translational movement amount is calculated. At this time, the position control of the finger tip can be performed by adding a correction corresponding to d0 included in ^MB T ₀ to the control value.
Further, in FIG. 8, for the sake of simplicity of calculation, the position of Σ _{0 is set at} a position d ₀ away from the origin of the mechanical interface pause Σ _MB . General, articulated by using a robot or Cartesian robot controller of can express desired sigma ₀ posture by inputting and setting the link parameters between the robot the free end shaft and sigma _0, gripping posture from the posture In order to take this, the rotation angle of θi (i = 1 to 3) may be calculated.

In the DH notation, the common normal that the joint i-1 axis and the i axis have only one is taken as the X axis, the axial direction of each axis is considered as the Z axis, and the common normal between the joints i-1 to i ( The twist angle around (X axis) is α _i-1 (see FIG. 8iii), and the angle of each common normal (X axis) around the Z axis of joint i is θi (see FIG. 8iv). The length of the common normal is a _i-1 . In addition, the shift of the common normal on the Z axis is denoted by di.

[2] Representation of gripping posture As a transformation representation method of the tool posture of the hand effector, a method using a homogeneous transformation matrix is known. The posture expression of the tool posture when the rotation angle is given to each joint of the tool 100 described above can be calculated by forward kinematics calculation, and is expressed by the following equations 1 and 2. Here, θ1 corresponds to the rotation axis denoted by reference numeral 101, θ2 corresponds to the rotation axis denoted by reference numeral 102, and θ3 corresponds to the rotation axis denoted by reference numeral 103. Further, α1 = 18 ° and α2 = −18 degrees. In the notation, s _i = sin θ _i and c _i = cos θ _i .

  If the world coordinates are set to match the control coordinates of the robot, for example, each base axis component of the world coordinates of each axis of the UVW of the tool posture can be expressed in a matrix. The tool posture that matches the workpiece posture is the gripping posture.

[3] Identification of workpiece posture The workpiece posture is acquired based on information on a workpiece feature obtained by an external sensor such as a stereo visual sensor. An example of a connection mode between the external sensor and the robot hand is shown in FIG. 13, and an example of a communication processing flow between the external sensor and the robot hand is shown in FIG. However, the external visual sensor of the present invention is not limited to this, and other known configurations can be applied.
The workpiece posture is the workpiece posture defined by the feature portion of the workpiece edge. For example, as shown in FIG. 4, the Cartesian of the end portion P1, the point P2 on the long side, and the point P3 on the short side. It is defined by base coordinate values (XYZ coordinate values) in coordinates (orthogonal coordinates). Here, when the work is a flexible object such as a cloth, since the shape changes, there is a problem that it is often difficult to estimate the pose by matching with a known shape model. For this reason, a process of identifying a vector in the long side direction and the short side direction from three points of the identified characteristic part of the workpiece end and performing pose estimation is performed.
The posture conversion tool 100 according to the present embodiment has the degree of freedom of rotation of the three joints so that the tool posture of the gripper is the same as the workpiece posture. A procedure for converting the above-described coordinate information of P1 to P3 into target gripping posture information when calculating the joint angle for this posing will be described below with reference to FIG.
In the following, the calculation method when the short side direction of the workpiece posture and the W direction of the tool posture are matched will be described, but the same procedure is performed when the long side direction of the workpiece posture is matched with the V direction of the tool posture. Of course you can do that.

(1) A short side vector P3 and a long side vector P2 are derived from Pn (Xn, Yn, Zn) (n = 1, 2, 3).
Pn (u, v, w) = Pn (u, v, w)-P1 (u, v, w)
(2) The inner product of the short side vector and the long side vector is determined, and the perpendicularity defect of the feature is determined. If it is defective, look for a different edge or abandon it to identify the next workpiece. This is because, in the case of a workpiece such as an irregular and flexible cloth, the recognized end shape clearly does not form a right angle. If the vectors are orthogonal, the inner product is 0, so if it is less than an appropriate threshold value, it is judged good and the process proceeds.
(3) Normalize the direction of the short side vector to obtain vector W.
(4) The cross product of the short side vector and the long side vector is calculated and normalized to obtain the vector U.
(5) Calculate the outer product W × U and normalize it to a vector V.
(6) and the rotation matrix R formed of vector UVW expressed in the base of the world coordinates, the end point position P1 as homogeneous transformation matrix ^G T _M whose components, which is referred to as work gripping target posture (see Equation 3).
Here, the work pose is obtained in the sensor coordinate system and converted into world coordinates, but the coordinate value in the three sensor coordinate system is converted into the world coordinate value and then the work pose in world coordinates is obtained. Good.

  When a relative posture difference is provided between the workpiece posture and the tool posture, a gripping posture may be obtained by applying a rotation matrix expressing the relative posture to the workpiece posture (see Expression 11).

[4] Calculation of Joint Rotation Angle In the posture conversion tool 100 of this embodiment, the rotation angle of each joint when expressing the grip posture is based on the Wx, Wy, and Wz components of the rotation matrix R given as the workpiece posture. The joint angles of J1 ′ to J3 ′ included in the rotary shaft driving units 103 to 101 can be calculated.
Specifically, in the first embodiment, when the rotation angle from J1 ′ to J3 ′ is given from θ1 to θ3, the rotation matrix indicating the gripping posture is expressed as Equation 4 when the rotation by θn is expressed as ⁿ⁻¹ T _n. become.

The right side of the above equation 3 is the work pause information obtained by measurement, and an equation with this as the right side of the above equation 4 is considered. Here, the inverse kinematic calculation formula for obtaining the rotation angle of θ1 to θ3 for matching the workpiece posture shown by the matrix ⁰ T ₃ and the finger posture represented by the above equation is given by the following equations 5-7. It is prescribed.
First, from the Wz component, θ2 can be calculated based on Equation 5 below.

Next, θ1 can be calculated by applying the following 6 which is a formula for obtaining θ when acos θ + bsin θ = C to the Wx component.

Finally, from the Vz component, θ3 can be calculated based on Equation 5 below.

The above calculation can be similarly obtained from components other than Wz, Wy, and Wz in the right-hand side of the matrix of Equation 4.
There are cases in which division is not possible due to the characteristics of the trigonometric function, etc., but θ is estimated from a prior check and its state. This makes it possible to calculate the joint angle rotation angle of the three-axis manipulator for taking a gripping posture when given three Cartesian coordinate values.
In addition, when this tool is connected to the mounting surface of a robot controlled by Cartesian coordinates and the finger tip position is controlled, the following translational movement operator is applied from the left side of the matrix in Equation 4 to move forward. Therefore, the kinematic calculation formula including the translational movement amount is as shown in the following formulas 8 and 9.
Therefore, the translational movement amount becomes the vectors Px, Py, and Pz by solving the inverse kinematics of the above equation. Since the finger tip position of the present embodiment has the feature that it does not change with the rotation of θ1 to 3, it is only necessary to simply calculate the translational movement vector.

[5] Calculation of joint rotation angle by matching finger posture and workpiece posture When calculation by inverse kinematics is difficult, or when the choice of workpiece posture is limited, in order to perform calculation easily What happens to the rotation matrix when the rotation angles are combined for θ1, θ2, and θ3 calculated in advance may be stored and a combination that best matches the rotation matrix of the workpiece posture may be employed. Table 2 exemplifies a conversion table when options of θ1 and θ2 are set every 45 degrees from −135 degrees to 180 degrees. In Table 2, θ3 = 0 is set, but actually a similar table is prepared for the options of θ3.
When the hand effector is a two-finger finger, since it can be gripped even in an inverted posture, a matrix of inverted postures (a matrix in which the signs of vectors U and V are exchanged) may be included in the options.
The vector W is an outer product represented by W = U × V. The degree of attitude matching determines the inner product of each UVW vector constituting the finger attitude and each base vector of the UVW coordinate axes constituting the workpiece attitude. Here, as to which axis is to be preferentially matched, a vector in a direction in which the workpiece can be more easily rotated by contact that occurs in the gripping process is preferentially matched.
In Table 2 below, the choices for each θi are set every 45 °. However, if more precise gripping is required, it is naturally possible to prepare and match the finer choices of θi.

Example 2 relates to a posture conversion tool having two degrees of freedom. The present embodiment is different from the posture conversion tool of the first embodiment in that it does not have the rotation axis driving unit 103. The present embodiment uses the deformation of a workpiece such as a cloth, and performs control to match the workpiece posture deformed by gripping with the tool posture.
[1] Freedom of work The work posture has 3 degrees of freedom other than 3 degrees of translation. In this regard, since the posture conversion tool 100 according to the present embodiment has only two degrees of freedom, a tool posture that matches the workpiece posture may not be expressed. However, when the workpiece is a flexible object such as a cloth, the workpiece can be gripped if the end portion of the workpiece is positioned between the two fingers even if there is an error between the workpiece posture and the tool posture. In the present embodiment, control is performed on the premise of an allowable error between the workpiece posture and the tool posture by utilizing the fact that the workpiece is deformed by gripping.

[2] Permissible posture error The posture error between the workpiece posture and the tool posture needs to be examined for each of the U, V, and W axes that define the workpiece posture.
The following law was confirmed by the experimental results (see FIG. 12).
(1) The posture error generated with respect to the U axis is not corrected even after the workpiece is gripped.
(2) A posture error generated with respect to the V-axis and the W-axis is corrected by deformation of the workpiece accompanying gripping if it is within an allowable range.

The parameters for determining the allowable range of posture error include, in the case of two finger fingers, the maximum workpiece thickness, finger release width, finger width, finger length (apparent), accuracy of workpiece posture identification, Examples include the accuracy and reproducibility of robot hand motion control. More specifically, if the clearance D obtained by FIG. 16 and the following equation 10 is larger than the accuracy and reproducibility error of the robot operation, the gripping can be performed.

[3] Virtual axis and virtual finger By specifying the allowable posture error, it is possible to use a robot hand with a low degree of freedom. Further, even when the tool posture that matches the workpiece posture becomes a unique posture, it is possible to perform gripping with a tool posture different from the unique posture.
In this embodiment, the deformation of the workpiece side with respect to the V-axis and the W-axis is regarded as being caused by one operation of the additional shaft. That is, by defining the deformation of the cloth as virtual joints and finger rotations, it is possible to express the gripping posture even with a posture changing tool with a small degree of freedom.
For example, when the posture change tool has one degree of freedom and two degrees of freedom, the gripping posture is calculated with a virtual finger having a virtual axis. At this time, the calculation is performed on the assumption that the virtual finger takes a posture that completely matches the workpiece posture. If the posture error between the virtual finger and the actual finger is within an allowable error range that can be gripped, gripping is possible.
For example, as shown in FIG. 11, when gripping a cloth piece that is suspended along the V-axis direction, the end of the cloth piece has a direction in which the long side direction (V direction) is downward. It becomes. As a result of the experiment, it was confirmed that most of the cloth pieces were present in a cone having a central axis in the V direction of 30 degrees. If the gripping posture is taken within this range, the posture of the cloth deformed by the gripping and the posture of the finger coincide.
In addition, the calculation can be performed simply and at high speed by adding a virtual finger connected to the virtual joint and performing the calculation. That is, on the one hand, the virtual finger is accurately adjusted to the work posture, and on the other hand, the actual joint and the rotation angle of the virtual joint are calculated, and the virtual joint is checked when checking the path from entry to grip for gripping. It is determined whether the rotation angle is within the above-described allowable error, and if it is within the allowable error, a calculation process of gripping is performed.

[4] Calculation Method The indirect rotation angle θ1 of the rotation shaft driving unit 103 and the indirect rotation angle θ2 of the rotation shaft driving unit 102 are calculated by the following procedure.
First, the rotation matrix R of the workpiece posture is calculated based on information from the stereo visual sensor. Next, attention is focused on the W-axis direction of the workpiece posture.
When the W axis is in a cone with an apex angle of θ1 + θ2 and outside the cone with an apex angle of θ1, there must be a posture in which the W axis can be matched but the U and V axes cannot be matched. become. Except when the W axis coincides with the Z direction, when the W axis coincides, the relative posture with the workpiece is uniquely determined, and two types of combinations are possible. Of these, the combination having the smallest relative posture difference is selected, and if the posture difference is within the limit, gripping can be performed without any change in the final gripping posture. Specifically, the calculation is performed according to the following procedure.
(1) Calculate rotation angles (θ1, θ2) for matching the W components.
(2) The rotation angle of the virtual finger around the W axis when the tool component is matched to the workpiece posture is calculated by matching the W component.
(3) A pass / fail judgment is made by comparison with an allowable value.

In the above (1), θ3 may be fixed to an actual value in the inverse kinematic calculation of the posture conversion tool having three degrees of freedom.
In the above (2) and (3), since the orthogonality around the W axis has only to be determined, the inner product of the V vector of the workpiece posture and the V vector of the tool posture is calculated. Since the inner product can determine the orthogonality of the vector, it is a numerical value representing the angle at which the work rotates around the W axis in accordance with the gripping operation of the fingers.
Using the gripping condition parameter of Equation 8 above, a threshold value of the grippable rotation error range θ is provided, the value of the inner product that applies to this threshold value is determined, and the one with the smaller orthogonality (the side where the vectors match) is adopted. .

[1] Six-degree-of-freedom arm unit The third embodiment has a configuration in which the driving unit 105 (θ1 rotation axis) of the posture conversion tool is simulated as the J6 rotation axis of the arm unit, and the driving unit 103 (θ3 rotation axis) is eliminated. . The posture conversion tool of the present embodiment has one degree of freedom in posture conversion by itself. In addition, the tool posture W axis is tilted at the same angle as the tilt angles α ₁ and α ₂ of the rotation axis.

A processing flow when gripping a workpiece, which is a flexible object, by a gripping device including an external visual sensor that is not limited to a stereo visual sensor, an arm unit, and a posture conversion tool 100 is as follows.
(1) Supply workpieces.
(2) Acquire information such as the shape of the workpiece using a visual sensor.
(3) The finger gripping posture is calculated from the workpiece posture information.
(4) It is determined whether or not the posture is unique based on the calculated Jacobian matrix of the gripping posture.
(5-1) If the posture is determined to be unique, the posture of the gripper is converted by the posture conversion tool.
(5-2) After changing the posture of the gripping part, the gripping posture is recalculated to determine whether or not it is a unique posture.
(5-3) If the singular posture cannot be avoided even if the above steps are repeated a predetermined number of times, the processing is stopped.
(6-1) When it is determined that the posture is not unique, parameters such as a gripping posture and posture information of the tool 100 are transmitted to the robot hand.
(6-2) The robot hand takes the final gripping posture.

[2] Details of singularity avoidance
a) As to the state of the workpiece, for example, (a) in the case of piles of towels piled up, (b) when one piece is taken out and placed on a flat plate with wrinkles etc. ) A state in which the outline end of one towel is held by a robot hand and is suspended with the same end height. The gripping by the posture conversion tool 100 can be performed by (i) posture conversion or (ii) a posture specified in advance according to the necessity of work.
In the cases (a) and (b) described above, since it is not necessary to match the tool posture with the workpiece posture, gripping is performed with a fixed downward posture as the gripping posture.
In the case of the above (c), for example, since it is necessary to grip with a tool posture that matches the posture of the end feature portion of the suspended cloth, the workpiece is brought to a specified position using another transport means as a previous step. Grasp after transporting.

b) The workpiece shape can be recognized by the visual sensor in the cases (a) and (b) above, even with only one grip position information. At this time, it is preferable to fix the visual sensor to the base frame 12 and to have the measurement origin of the sensor coordinate system and the sensor coordinate system in a homogeneous transformation matrix format with the world coordinate system.
In the case of (c) above, three points of position information (position information of P1 to P3) are necessary. In general, when working with a robot hand, three-point coordinate values are measured using robot coordinates (in this embodiment, integrated with world coordinates). However, as in the first embodiment, the origin position of the workpiece posture may be processed so as to match the origin position of the tool posture.
The visual sensor was attached to the free end of a 6-axis manipulator. In this configuration, the three-point information ^S T _M becomes coordinates in the coordinate system of the visual sensor, and is converted into coordinates in world coordinates. Here, the mechanical interface of the six-axis manipulator orientation, measurement origin posture ^G T _S of the vision sensor is known, setting this attitude difference as tool parameters (see FIG. 18), the control point measuring origin of the visual sensor As a robot hand.
Thus, world coordinates of each point can be converted into the world coordinate value from the internal coordinates of the visual sensor coordinate transformation equation _{^{(Tm = G T S S T}} M) ( see FIG. 15).

Next, the work posture is converted into the format of the rotation matrix R from the position information of the three points in the same procedure as in the first embodiment. Then, together with the position vector of the workpiece origin from the robot base coordinates, a homogeneous transformation matrix is formed. Here, since the pose of the work posture and the tool posture are made the same, the grip posture is made to coincide with the work posture. Subsequently, a determinant of the Jacobian matrix when the gripping posture is taken is calculated to determine whether the matrix is regular. If it is determined that the matrix is not a regular matrix, it is a singular posture, and since it is difficult to express the posture, it is avoided.
The posture conversion tool 100 of the present embodiment can cause the hand effector to take eight types of postures with respect to the base plate 12 and the attachment surface 105. That is, as shown in FIG. 10, the J7 rotation axis can be selected from 0, 90, 180, and 270 degrees of rotation angle from the origin (four types of rotation angle selection), forward grip and reverse hand grip Can be selected, and a total of eight types of gripping postures can be selected.
Based on the Jacobian matrix, it is determined whether or not the target position is a singular posture. If it is a singular posture, a different J7 rotation angle is reset, a regular determination is made, and a J7 rotation angle that does not become a singular posture is determined as a gripping posture. Select as. In addition, if it is sent to the robot simulator and becomes a singular posture before being sent to the robot controller, recalculation using a different J7 option can avoid the robot controller from being restarted due to an alarm. Therefore, it is preferable. If all the J7 rotation angles are in the unique posture, the gripping cannot be performed and the processing is stopped (for example, the work is released and the process returns to the beginning of the control program).
When a gripping posture that is not a unique posture is found, the gripping posture is expressed and gripped. The gripped work is transferred to the next process.

[3] Other aspects of singularity avoidance In the above [2], the singularity is determined using the Jacobian matrix. The unique posture can also be avoided by limiting the posture of the J5 joint to a few, keeping the posture as much as possible, and matching the workpiece posture and the tool posture by rotating the J6 axis and the J7 axis.
For example, when the workpiece is in the state (c), the distribution of the workpiece posture in the W direction is often horizontal with respect to the ground. In this case, the distribution of the tool posture in the V direction is generally vertical. In most cases, the rotation around the U axis falls within ± 30 degrees.

If the two robot hands are arranged opposite to each other as shown in FIG. 18, in the case of the relationship between the robot and the work space, the posture handling range of one robot hand can be limited to half of the work range. In the example, the posture charge range of the robot A is A01 to A14. Then, according to the W direction of the workpiece posture, the posture of the posture conversion tool is changed by applying to a classification table as shown in Table 3 that defines the tool parameters.
More specifically, two basic postures of the J5 joint are selected in a posture that does not extend in the work area, and a tool posture that can be expressed by the basic posture is determined by rotation of the J7 joint (see FIG. 19). Then, after the tool parameter is set, control for applying the workpiece posture and the tool posture is performed. By increasing the tool parameter options with J7 joints, it is possible to cover a 90-degree range when the W vector is viewed from the top with one J5 posture, and it is possible to cope with a 180-degree workpiece posture change by the reversed posture. is there. The posture of the J5 joint at this time can design a gripping operation in the work space without taking a specific posture in which the J4 and J6 joints are in a straight line (see FIG. 19). Therefore, in the gripping work in this range, it is not necessary to perform regular judgment of the Jacobian matrix, and the control becomes simple. Another advantage is that the assumed posture can be limited in advance, so that even if the wiring of the visual sensor or posture conversion tool attached to the robot arm is not damaged in all robot joint operations, damage due to unexpected motion is prevented. it can.

The above tool parameters define the relationship between the posture of the mechanical interface MI and the tool posture, and are part of the link parameters from the sixth axis joint pose to the hand effector pose of the robot hand. In a general 6-axis robot controller, the translation distance to the finger control point with respect to the origin pose of the MI is represented by x, y, and z. Further, the posture change expressed by rotating in the order of the X, Y, and Z axes with respect to the origin coordinate axis direction of the MI is represented by rotation angles A, B, and C around the x, y, and z axes. This relationship is expressed by the following formula 11 when α = A, β = B, and γ = C.
Here, in the description of Equation 4 above, it is possible to calculate tool parameters when the posture of the finger is changed by rotation from θ ₁ to θ ₃ . In the case of rotation only by θ ₂ as in this embodiment, if the actual fixed values are applied to θ ₁ and θ ₃ , the right side of the following equation 11 can be calculated, and therefore the rotation angle ABC can be calculated.

The 6-axis robot controller expresses the posture if the control value and the tool parameter in the robot coordinates are known, so that the singular point is avoided by determining the J7 rotation angle not to be a singular point. The options for the J7 rotation angle are every 90 degrees. In this embodiment, the tilt angle is 18 degrees, and the finger pose that is the control point changes as shown in the figure, so that the relative posture with respect to the work can be held within 9 degrees at any rotation angle. Therefore, the change in the posture of the list can be limited to within 9 degrees from the line connecting the control point and the center of the list, so that the unique posture does not occur within the work range set in the cube shape shown in FIG.
In addition, even when the J7 posture changing mechanism is not provided as described above, a workpiece deformed like a cloth has a posture within an allowable rotation angle around the V or W axis in the workpiece coordinate system as shown in FIG. By inputting the tool parameters including the error, it is possible to avoid the singular posture and accurately match the work pose and the finger pose. The tool parameters at this time can be set by calculating the tool parameters from the joint pose at the free end to the finger pose obtained by rotating the virtual axis set around the V or W axis of the actual finger pose by the above equation 11.

As the two-finger finger used in this embodiment, an appropriate one from the following options is adopted.
The fingers and grippers are adopted from a parallel gripping type, a fulcrum opening / closing type, and a parallel ring type. As shown in FIG. 9, the parallel gripping type is a type in which the nail attachment portion of the gripper slides. The fulcrum opening / closing type is provided with a fulcrum at the attachment portion of the claw or gripper, and the claw swings and grips as a part of the lever (not shown). In the parallel link type, as shown in FIG. 20, the claws are constituted by parallel links, and the claws are gripped along a track determined by the link shape and restraint conditions.
Since the parallel gripping type has a small gripping stroke and a high gripping force, the parallel gripping type is suitable for the intended operation as described in [2] (a) and (b) above. The reason is that a large number of workpieces are gripped when the gripping stroke is large, and the workpieces can be pulled up without slipping even when they are entangled.
Since the fulcrum opening / closing type has a large gripping stroke, the fulcrum opening / closing type is suitable for the purpose [2] (c). This is because the gripping stroke is large, so that gripping can be performed even if the position recognition accuracy and operation accuracy are poor.
The parallel link type shown in FIG. 20 holds the finger tips along the circular orbit with the same posture. In the relative relationship between the flexible workpiece and the fingers, an appropriate amount of clearance D when the fingers are released as shown in FIG. 16 is required. This is because if the value of the clearance D is small, it is difficult to grip, and if it is too large, the possibility of gripping something other than the gripping feature increases. Here, the clearance D can be controlled by correcting and changing the position of the finger in the approach axis direction. In the parallel link type, the tip position when the finger is normally closed is controlled as a control point, and the actual clearance D is small due to the posture difference θ when the workpiece thickness t is indefinite or calculated using a virtual axis. In this case, there is an advantage that it can be dealt with by correcting the grip control point to the back side. Therefore, it can be suitably used for any of the above [2] (a), (b), and (c).
Of course, the opening / closing width itself may be controlled by directly controlling the actuator of the gripper even if it is not a parallel link type.
Supplementing the gripping posture, in the case of [2] (a) and (b) above, the cloth is gripped so as to be pinched in a downward fixed posture. Especially in case of (A), when the workpiece is spread by taking the posture that the tip of the parallel hand claw is arranged parallel to the flat plate as shown in FIG. Is possible. If the shape of the nail is configured in this way, the same gripping is possible in the parallel gripping type and the fulcrum opening / closing type, but the parallel link type grips while approaching the back, so that it is easier to grip. Has an effect.

  According to the present invention, sheets, wraps, pillow covers, clothes, sick clothes, bath towels, face towels, etc. used in hospitals, hotels, etc. are rented and collected after being used by the user, washed and cleaned at the factory. After that, it will be possible to automate the entire linen supply business line. More specifically, for example, work that has been performed manually by a conventional worker, such as work for loading into a washing machine, work for picking an end point of linen after washing and loading into a finishing machine, etc. It becomes possible.

1 is an overall configuration diagram of a handling system of the present invention. It is drawing for demonstrating the outline | summary of the work process of the handling system of this invention. It is a flowchart of the procedure which hold | grips a workpiece | work by the finger mechanism of this invention. (i) It is explanatory drawing of the characteristic part imaging by the imaging mechanism of this invention, (ii) It is explanatory drawing of the attitude | position change of the finger mechanism of this invention. It is explanatory drawing of an imaging mechanism provided with the projection apparatus of this invention. It is a whole block diagram of the robot hand of this invention. It is a whole block diagram of the attitude | position conversion tool of this invention. It is an image figure of attitude | position control by the attitude | position conversion tool at the time of carrying out XYZ fixed angle. It is drawing for demonstrating attitude | position conversion of a hand effector. It is drawing for demonstrating a mode that the rotation angle from an origin is selected with the J7 rotating shaft of an attitude | position conversion tool. It is a cone model for explaining the existence range of a virtual axis and the long side of a work. (i) Drawing which shows the state which the attitude | position of the workpiece | work and virtual finger | toe matched, and the drawing which showed the actual finger mechanism and the attitude | position of a workpiece | work in the case of (ii) (i). It is drawing which shows an example of the connection aspect of an external sensor and a robot hand. It is drawing which shows an example of the communication processing flow of an external sensor and a robot hand. It is drawing for demonstrating the procedure which converts the coordinate information of the characteristic part of a workpiece | work into target holding | grip attitude | position information. It is a figure for demonstrating the element for calculating the clearance D for determining the tolerance | permissible_range of an attitude | position error. It is an external appearance perspective view which shows the structural example by a Cartesian robot arm. It is a top view for demonstrating the attitude charge range of the robot hand at the time of arrange | positioning two robot hands facing each other. It is a figure explaining that the attitude | position change of 90 degree | times can be performed by one list attitude | position. It is drawing of a parallel link type finger gripper.

Explanation of symbols

1 Frame 2 Pull-up hand (hanging hand)
3 Whole imaging mechanism (stereo vision sensor for whole)
4 Hand imaging mechanism (stereo vision sensor for hands)
5 Work (cloth)
6 Work pile 7 Recognition base 8 Folding machine 9 Conveyor device 10 Robot hand A
11 Arm 12 Base plate 20 Robot hand B
31, 41 Projector (Projector)
32, 33, 42, 43 Camera (imaging device)
51 Long side 52 Short side 60 Projection pattern 71 Robot base pose 72 Work base pose 73 XYZ translational movement vector 100 Posture conversion tool 103-101 Rotary actuator (rotary axis drive unit)
104 End effector (hand effector)
105 Mounting surface

Claims (10)

A workpiece that is a flexible object by a robot hand system that includes a gripping unit that grips a workpiece with fingers that open and close, a drive unit that includes multiple axes that move the gripping unit in space, a gripping position information acquisition unit, and a calculation unit. A method of gripping
A first step of determining the gripping position of the flexible object based on the information on the characteristic part of the workpiece acquired by the gripping position information acquiring unit;
A second step of setting the path and posture of the gripping part so that the gripping position of the workpiece is positioned in the finger in the open state,
A third step of moving the gripping part by the driving part to grip the gripping position of the workpiece;
A method of gripping a workpiece, comprising: The drive unit enables posture conversion with two degrees of freedom,
In the second step, the calculation unit matches the origin of the coordinate system of the workpiece and the coordinate system of the finger and two axes, sets an allowable rotation angle with respect to the first axis of the workpiece, and sets the allowable rotation The method for gripping a workpiece according to claim 1, further comprising a step of setting a virtual axis for rotating the gripper within an angle and setting a path and posture of the gripper. The drive unit enables posture conversion with one degree of freedom,
In the second step, the calculation unit matches the origin of the coordinate system of the workpiece and the coordinate system of the finger and two axes, sets an allowable rotation angle with respect to the first axis of the workpiece, and sets the allowable rotation An allowable rotation angle with respect to a first virtual axis for rotating the gripping portion within the angle and a second axis orthogonal to the first axis of the workpiece is set, and the gripping portion is rotated within the allowable rotation angle. 2. The workpiece gripping method according to claim 1, further comprising a step of setting a second virtual axis and setting a path and posture of the gripping portion.   The method for gripping a workpiece according to claim 1, wherein the workpiece is a sheet-like cloth. A workpiece gripping device that performs posture conversion of fingers with three degrees of freedom,
A rotating shaft drive unit (101) having a mounting surface (105) to the robot arm and having a central axis in a direction normal to the mounting surface (105);
A rotary drive unit (102) connected to the rotary shaft drive unit (101) so as to have a central axis that forms a predetermined angle α1 with the central axis of the rotary shaft drive unit (101);
A rotary drive unit (103) connected to the rotary shaft drive unit (102) so as to have a central axis that forms a predetermined angle α2 with the central axis of the rotary shaft drive unit (102);
A finger mechanism for holding a workpiece which is a flexible object connected to the rotation drive unit (103);
A robot controller for controlling these operations,
The robot controller is a workpiece gripping device that grips a workpiece in a desired gripping posture by calculating a posture of a finger mechanism. A workpiece gripping device for performing finger posture conversion with two degrees of freedom,
A rotating shaft drive unit (101) having a mounting surface (105) to the robot arm and having a central axis in a direction normal to the mounting surface (105);
A rotary drive unit (102) connected to the rotary shaft drive unit (101) so as to have a central axis that forms a predetermined angle α1 with the central axis of the rotary shaft drive unit (101);
A finger mechanism for holding a workpiece, which is a flexible object, connected to the rotation axis drive unit (102) so as to have a center axis that forms a predetermined angle α2 with the center axis of the rotation axis drive unit (102);
A robot controller for controlling these operations,
The robot controller is a workpiece gripping device that grips a workpiece in a desired gripping posture by setting a virtual axis in the deformation direction of the workpiece and calculating the posture of the finger mechanism.   The workpiece gripping apparatus according to claim 5 or 6, further comprising a robot arm that controls positioning of the tip of the finger mechanism with Cartesian coordinates. A workpiece gripping device that performs posture conversion of a finger with one degree of freedom,
A rotation drive unit (102) having a mounting surface (105) to the robot arm and connected so as to have a central axis that forms a predetermined angle α1 with a normal direction to the mounting surface (105);
A finger mechanism for holding a workpiece, which is a flexible object, connected to the rotation axis drive unit (102) so as to have a center axis that forms a predetermined angle α2 with the center axis of the rotation axis drive unit (102);
A controller for controlling these operations,
When the gripping posture becomes a unique posture, the controller rotates the rotation drive unit (102) by a predetermined angle and uses the joint of the 6-degree-of-freedom robot arm that can be connected to the mounting surface (105) for posture conversion. A workpiece gripping device that can grip a workpiece in a desired gripping posture by calculating the posture of the finger mechanism. A mounting surface (105) to the robot arm, a finger mechanism for holding a work that is a flexible object connected to the mounting surface (105), and a controller.
The controller sets a virtual axis in the deformation direction of the workpiece when the gripping posture becomes a unique posture, and uses a joint of a 6-degree-of-freedom robot arm that can be connected to the mounting surface (105) for posture conversion. A workpiece gripping device that can grip a workpiece in a desired gripping posture by calculating the posture of the finger mechanism. Gripping position information acquisition means for acquiring information on one to three feature parts defining the gripping position of the workpiece;
The robot controller calculates a workpiece posture based on the information on the feature, calculates a gripping posture of the finger mechanism based on the workpiece posture, determines whether the gripping posture is a singular point, and is a singular point. 10. The workpiece gripping apparatus according to claim 8, further comprising a determination step of calculating a gripping posture again when the determination is made.