JP2003062775A - Teaching system for human hand type robot - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve substantially the difficulty of smooth teaching caused by a workspace difference based on a geometrical dimension difference between a robot hand and a human hand. SOLUTION: A distant environment where a robot is put is taken by a camera 226 of a robot control system 210, and then it is sent to a robot teaching system 250 through a vision controller 218 and a supervisor 212. A hypothesis environment is established by a virtual simulation system 252 from the visual picture sent to the robot teaching system 250 and prior information on an operated object. A human operator attaches a force feed back to the hand and he performs the series of operation that he wants to make the robot operate. During the operation the three-dimensional position of the hand of human operator is measured with a three-dimensional measurement device 264. The observed value of the series of operation is sent to a hand operability analysis system 254 and it is analyzed to maximize the robot hand operability.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a robot having a humanoid hand with multiple fingers and multiple joints (hereinafter referred to as a humanoid hand robot), and more particularly to a teaching system for the humanoid hand.

[0002]

TECHNICAL BACKGROUND An example of a humanoid hand robot is, for example, Gifu Hand developed by the inventors.
II (see Figure 1) (H. Kawasaki, K. Tsuneo,
K. Uchiyama, & T. Kurimoto, Dexterous Anthropomorph
ic Robot Hand with Distributed Tactile Sensor: Gif
u Hand II, Proc. of IEEE Int. Conf. on System, Ma
n and Cybernetics pp. II-782 ・ See 787, 1999).
The Gifu Hand II is designed so that its shape and mechanism are almost the same as those of a human hand. That is,
This humanoid hand robot has a thumb 110 and four fingers 1.
22, 124, 126, 128 and the thumb 1
10 has four joints and four degrees of freedom, and each finger 122,
124, 126, and 128 also have four joints and three degrees of freedom. All servo motors are mounted in the frame of the hand. A 6-axis force sensor is attached to each fingertip, and 624 tactile sensors are distributed on the surface of the finger such as the palm 140 or the thumb. And this Gifu Ha
nd II is configured so that not only the shape of the human hand but also the movement space is the same. In the teaching method of such a humanoid hand robot, movements of the palm and fingers of a human being a master have been conventionally performed (hereinafter, when pointing both the palm and the finger, it is called a human hand, and in the case of a robot, a hand). Is generally taught by a direct master / slave method in which a target value is directly transmitted to a robot which is a slave. However, the sizes of the palm and fingers of the master and slave are different, the work areas of the fingers do not match, and the motor functions are also different, so even if the knuckle angle of the master is realized by the slave,
In some cases, the work cannot be performed because the reach area of the fingertip is different, and the human needs to move the hand of the human being the master in consideration of the difference, and it is often difficult to smoothly teach. As another method, the position / orientation of the palm of a human being, who is the master, can be directly realized by the robot / hand, and the position of the fingertip of the robot can be adjusted by adjusting the angle of the finger to minimize the position error. There is also a method of absorbing differences in pod motor function (RN Rohling
and JM Hollerbach, "Optimized FingertipMapping
for Teleoperation of Dexterous Robot Hand, Proc. o
f IEEE International Conference on Robotics and Au
tomation, pp.769-775, 1993). However, since the position and posture of the palm are fixed, the common work area is small unless the geometrical dimensions of the human hand and robot hand are similar, so the master's fingertip position should be realized by the slave. Is often difficult, and even in the movable range of the slave, it is difficult to generate a predetermined force or speed with the fingertip because of the vicinity of the kinematic singular point.

[0003]

SUMMARY OF THE INVENTION An object of the present invention is to teach a motion from a motion of a human being as a master to a humanoid hand robot as a slave, a difference in geometrical dimensions between them, and a finger joint. An object of the present invention is to provide a robot teaching method that significantly improves smooth teaching which is difficult due to a difference in work area and work ability due to a difference in movable range, a difference in the number of degrees of freedom of joints, and the like.

[0004]

In order to achieve the above object, the present invention provides a palm, a finger, and an object in human work in an environment in a teaching system for generating a target trajectory of a motion of a humanoid hand robot. A contact point between the finger of the robot hand and the target object is made to coincide with a contact point between the human finger and the target object by using a measurement unit that measures the three-dimensional movement of the object and the data from the measurement unit. It is characterized by comprising an operability analysis unit for analyzing the operability of the robot hand in the movable range of the finger mechanism for the position and orientation of the palm of the robot hand. The measurement unit may use, as measurement data, a contact point between a target object and a human finger when a human operates the target in the environment. In addition, this system
Virtual that creates a virtual environment for humans to work
It is also possible to provide a simulation unit and perform human work in the environment with a force feedback glove for feeding back the action on the target object to the human hand from the virtual simulation unit.

[0005]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. FIG. 2 is a configuration example of a teaching system for a humanoid hand robot according to the embodiment of the present invention. As shown in FIG. 2, the humanoid hand robot teaching system 200 includes a robot control system 210 and a robot teaching system 250. Robot control system 21
0 is composed of an arm 222 controlled by an arm controller 214, a hand 224 controlled by a hand controller 216, and a camera 226 controlled by a vision controller 218, under a supervisor 212. In addition, the robot teaching system 250 includes a data acquisition system 262, a three-dimensional position measuring device 264, a force feedback glove 30.
0, virtual simulation system 252,
It is composed of a hand operability analysis system 254 and a robot simulation system 256.

The operation of the humanoid hand robot teaching system 200 according to this embodiment will be described. The environment in which the remote robot is placed is captured by the camera 226 of the robot control system 210 and sent to the robot teaching system 250 via the vision controller 218 and the supervisor 212. A virtual environment is constructed by the virtual simulation system 252 from the visual image of the vision sent to the robot teaching system 250 and the prior information of the object to be operated. In the constructed virtual environment, at least a virtual object corresponding to an object and a virtual hand corresponding to a human hand are presented. Humans attach a force feedback to their hands and perform a series of operations in a virtual environment that they want a robot to operate on a virtual object. The three-dimensional position of the human hand at this time is measured by the three-dimensional measuring device 264 and sent to the virtual simulation system 252 via the data acquisition system 262. At this time, the virtual simulation system 2
When the virtual hand and the virtual object are spatially interfered with each other, 52 calculates a virtual drag force and performs force feedback.
Send to gloves. The virtual drag force is calculated for each part of the fingertip and the palm that come into contact with the virtual object. This virtual drag force is controlled in the force feedback glove as the target value of the force at the fingertip or palm. For this reason, when there is interference, a person can obtain a touch feeling or a grip feeling, and can operate a virtual object with a feeling similar to an actual object operation. The target robot hand may be, for example, the previously mentioned Gif.
Like u Hand II, it has five fingers like a human hand, and the thumb has four joints and four degrees of freedom, and the other fingers have four joints and three degrees of freedom.

An example of the structure of the force feedback glove used is shown in FIG. In a virtual environment, human peculiarity is necessary for humans to feel that they are manipulating real objects in the real world. Force information, especially regarding contact, is very important. Forces, along with visual effects, should be presented to humans. In this system, the force on the human hand grasping a virtual object is presented to the human hand by using a force feedback glove. As shown in FIG. 3, the force feedback glove 300 includes a force feedback mechanism and a data glove (for example, Virtual Technolobe) for measuring a joint angle of a human hand.
gies Cyber Glove). Force feedback mechanism is located on arm 316 1
It has zero servo motors 352. The torque generated by the servo motor 352 is transmitted to each finger by the wire 342. The force on the finger is detected by the force sensors 338, 33.
4 is measured. Humans feel force with two points on each finger. Force Feedback Globe 30 in this example
The force resolution at 0 is about 30 grams. By using this configuration, virtual force is transmitted to humans. In order to use the force feedback glove 300, the positional relationship between the hand and the virtual target object,
Accurate information on contact point and contact force is required. For example, the following five parameters are measured as motion data. ^ref v _object : Velocity of target object with respect to reference coordinate system ^object v _hand : Velocity of ^{hand with} respect to target coordinate system ^hand p _{i-th finger} : Position of _{i-th finger with} respect to hand coordinate system ^hand v _{i-th finger} : Hand coordinate Velocity of the _{i-th finger in the} system ^hand f _{i-th finger} : virtual force of the _{i-th finger in the} hand coordinate system

Now, the observed values of a series of operations are
Hand-operable via the acquisition system 262
Sent to the sex analysis system 254. Figure 4 is for teaching
FIG. Figure 4 (a) shows a virtual simulation
3 illustrates a human hand within the motion system 252.
FIG. 4B shows a humanoid robot hand as a control target.
Is shown. Teaching grasps the object with the fingertips of the human hand
And to perform that action with a humanoid hand robot
The purpose is to set a target value. Shown in Figure 4 (a)
In this way, the target object coordinate system is set for the target object, and
The hand coordinate system is set. Position and orientation of target object and palm
Is defined as the position / orientation of these coordinate systems. In the figure
The circle marks indicate contact points. In such a teaching system
In other words, the shape of the human hand and the robot hand and the exercise machine
Position and posture of the palm expressed in the reference coordinate system due to different functions
^refr_handAnd fingertip position^refP_fingersThe humanoid hand
Depending on the bot, it may not be possible at the same time,
However, it is difficult to operate the target object due to the limit of the movable range of the mechanism.
There are things. The position of the palm when considering the gripping action of the object
·posture^refr_handRather than the grip position--that is, the fingertip position
^refP_fingers--- are important, so prioritize them
hand^refr_handIt is desirable to correct At this time,
Fingertip position indicated by measurement data^refP_fingersCan be realized
Position / posture of palm ^refr_handExist innumerable in a certain area
However, it is not the only one. Therefore, of the observation data, the standard
Position / orientation of target object expressed in coordinate system^refr_objectAnd finger
Destination position^{r ef}P_fingersAnd power^refF_handIs teaching data
Position and posture of the palm^refr_{h and}About
Instead of using that value as the indicator data,
Maximum or maximum evaluation index that represents the operability of
Analyze the position and posture of the palm to maximize the
The fruit is used as teaching data. The position and shape of the palm above
Momentum^refr_hand， Fingertip position^refP_fingers， Target object position
·posture^refr_objectIs a vector value.

As a method for evaluating the mechanism of the robot arm, there is an analysis of characteristics of the Jacobian matrix J. For a general robot arm, the speed r of the hand position / orientation and the joint speed q
The relationship is expressed by the following equation.

[Equation 1] At this time, the area that the velocity r of the hand position / orientation can take depends on the Jacobian matrix J, and its size is det (J ^T J)
Is almost proportional to. Therefore, det (J ^T J) can be an index for evaluating the work performance of the robot arm mechanism.
However, this index does not include consideration of the range of motion of the joint angle of the robot. Therefore, when the teaching command is generated, the hand operability evaluation function PI for obtaining the optimum hand position / orientation is given by the following equation.

[Equation 2] And the position / posture of the palm is determined so that the value of PI becomes maximum. Here, i is a subscript indicating the finger of the hand, and is represented by 1 to 5 in order from the thumb to the little finger. The first term on the right side indicates the manipulability evaluated by the Jacobian matrix J _i of the mechanism of each finger, and w _i indicates its weight. The second term on the right side indicates the manipulability evaluated by the movable range of the mechanism of each finger, and ρ _i indicates its weight. The function PI for evaluating the movable range is shown by the following equation.

## EQU3 ## PI = −Σ _j {(q _ij −a _ij ) ⁻² + (b _ij −
q _ij ) ⁻² } (a _ij <b _ij ) where q _ij is the angle of the j-th joint of the finger i.
a _ij and b _ij are the limit angles of the j-th joint of the finger i.
In a _ij <q _ij <b _ij , when q _ij changes its value and approaches a _ij or b _ij , the value of PI changes rapidly and becomes a large negative value. Clearly, PI is a function of knuckle angle.

Now, the robot hand, which is currently being taught, has five fingers, and the thumb has four joints and four degrees of freedom.
Since the other fingers have a 4-joint 3-degree-of-freedom mechanism, the Jacobian matrix J _i of the mechanism of each finger has the following relationship.

[Equation 4] here,^handv_i-th fingerIs the finger in hand coordinates
It is the speed ahead. q_{i-th finger}Is the angular velocity of the controllable joint
Degree and q for fingers other than the thumb_{i-th fin ger}=
[Q_i-1, Q_i-2, Q_i-3]^T(I = 2-5) and thumb
Has 4 joints and 4 degrees of freedom q_1st finger= [Q_1-1, Q_1-2, Q
_1-3, Q_1-4]^TAnd^handv_1st finger3 degrees of freedom
And has one degree of freedom redundancy. Q for each finger_i-1Strange
Is a movement of adduction and abduction, q_ij(J> 1) is in front
Bend and backbend. Thumbs have redundant degrees of freedom
Therefore, the posture of the palm and fingertips is the root of the finger mechanism related to forward bending and backward bending.
Total value of joint angle from original to fingertip θ = q_1-2+ Q_1-3+
q_1-4And use this as an adjustable parameter
ing. Therefore, the evaluation function PI is the finger joint vector
q_i-th fingerExpressed as a function of (i = 1-5) and θ
It On the other hand, the position and posture of the palm^refr_hand= [X_hand, Y
_hand, Z_hand, Θ_hand, Φ_hand, Ψ_hand]^TAnd fingertip position
^refP_fingersAnd θ are given, the knuckle vector is
Decided. Where x _hand, Y_hand, Z_handIs the reference coordinate
The position of the palm is shown. θ_hand, Φ_{ha nd}, Ψ_hand
Is the rotation that represents the rotation from the reference coordinates to the hand coordinates.
-Z Euler angle. Fingertip position^refP_fingersTeach
Given as data. Therefore, the inverse kinematics of the finger
The evaluation function PI is^refr_handAnd θ function PI (^refr
_hand, Θ). Weight w_iAnd ρ_iWhich finger
Are 1.0 and 10.0 respectively^-TenSet to. PI up
Make big^refr_handAnd θ are calculated using the steepest descent method
It To perform the steepest descent calculation,^refr_handand
An initial value of θ is required, and this initial value is within the movable range of the finger.
Must fit in. Therefore, the initial value is human
The position and orientation of the palm of the robot and the value of θ, which is the robot hand.
When it is not possible to achieve with
Find the initial value that fits. Target robot
・ If the hand is similar to that of a human, the condition can be easily
It is possible to find an initial value to satisfy. Hand operability
Position / posture of the palm, fingertip position, finger position obtained by the analysis system
The previous force is described in the object coordinate system set for the object.
And these in the robot simulation system
Simulation of work execution based on teaching data
After confirming that the teaching data is good, enter the robot control system.
Teaching data is sent.

In this embodiment, the robot teaching based on the work of the human being the master in the VR space is described, but the robot hand is not limited to the work in the VR space, and the robot hand can be used in the human work in the real space. Needless to say, the robot teaching for adjusting the position / posture of the palm of the robot hand so as to maximize the operability of the robot or its vicinity is also a similar teaching method. Further, since the thumb of the present embodiment has four joints and four degrees of freedom, the parameter θ representing the posture of the finger is introduced. However, when the thumb has three degrees of freedom, it is not necessary to introduce such a parameter. Needless to say. Further, although the five-finger robot hand is used in the present embodiment, it is needless to say that the same teaching method can be executed by the human teaching according to the index of the robot even with four or less robot hands. Yes. Also,
The above-described teaching system is also applied to a robot teaching system described by one of the inventors (see Japanese Patent Laid-Open No. 2000-308985) when interpreting the intention of an operation and creating a teaching command for the robot. be able to.

[0012]

With the above configuration, even if the human hand and the robot hand have different work areas, the position / posture of the palm of the robot hand that maximizes or almost maximizes the manipulability of the hand is calculated. By using as the teaching data, the contact point where a human operates the object can be realized by the robot hand, and since the hand operability is maximized, the speed and force that can be generated at the fingertips of the robot hand are maximized. , It becomes possible to realize the operation of the object as taught.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of a hand of a humanoid hand robot.

FIG. 2 is a diagram showing a configuration example of a teaching system according to an embodiment of the present invention.

FIG. 3 is a diagram showing a configuration example of a force feedback glove of the present invention.

FIG. 4 is an illustration of the teaching of an embodiment of the present invention.

[Explanation of symbols]

110 thumb 120 each finger 122, 124, 126, 128 finger 140 palm 200 human type hand robot teaching system 210 robot control system 212 supervisor 214 arm controller 216 hand controller 218 vision controller 222 arm 224 hand 226 camera 250 robot teaching system 252 Virtual simulation system 254 Hand manipulability analysis system 256 Robot simulation system 262 Data acquisition system 264 Three-dimensional position measuring device 300 Force feedback glove 316 Arm 342 Wire 352 Servo motor 338, 334 Force sensor

Claims (3)

[Claims] 1. A teaching system for generating a target trajectory of a movement of a humanoid robot, a measuring section for measuring three-dimensional movements of a palm, a finger, and a target object in human work in an environment, and a measuring section from the measuring section. Using the data, the contact point between the finger of the robot hand and the target object is made to match the contact point between the human finger and the target object, and the position and orientation of the palm of the robot hand are measured in the movable range of the finger mechanism. A human-type hand robot teaching system comprising: an operability analysis unit for analyzing operability of a robot hand. 2. The teaching system for a humanoid hand robot according to claim 1, wherein the measuring unit measures a contact point between a target object and a human finger when the human operates the target object in the environment. A humanoid hand robot teaching system characterized by using data. 3. The teaching system for a humanoid hand robot according to claim 1, further comprising a virtual environment for generating a virtual environment for a human to work.
A human body working in the environment is performed by attaching a force feedback glove for feeding back the action on the target object to the human hand from the virtual simulation unit. Teaching system for humanoid hand robot.