KR100818059B1 - Method for controlling human arm motion reproduction of humanoid robot - Google Patents

Abstract

A method for controlling reproduction of human arm motion of a humanoid robot is provided to allow the humanoid robot to reproduce minute human arm motion regardless of size and arm freedom of the humanoid robot by considering displacement variation amount and displacement speed variation amount of an arm joint of the humanoid robot and obtaining optimized motion. A method for controlling reproduction of human arm motion of a humanoid robot comprises the steps of: obtaining a time track capturing the human arm motion(S10); calculating an object function for minimizing difference between the human arm motion and humanoid robot arm motion with respect to each time value of the captured time track and searching for optimal joint displacement amount(S30); and controlling the humanoid robot so as to perform changed motion by using the optimal joint displacement amount per each hour(S60).

Description

METHODO FOR CONTROLLING HUMAN ARM MOTION REPRODUCTION OF HUMANOID ROBOT}

1 is a flowchart of a method for controlling the reproduction of a human arm motion of a humanoid robot according to the present invention.

2 is an exemplary diagram for explaining the difference in arm length between a human and a humanoid robot. FIG. 2A is a human model and FIG. 2B is a model diagram of a humanoid robot.

3 is an exemplary view for explaining the phase difference between the upper arm of a human and a humanoid robot, FIG. 3A is a human model, and FIG. 3B is a model diagram of a humanoid robot.

FIG. 4 is an exemplary diagram for describing unnecessary motions occurring in a singular posture of a humanoid robot, and FIG. 4A is a screen for monitoring unnecessary motions occurring in a singular posture before adding a third term in the objective function of the present invention. 4b is a screen for monitoring the removal of unnecessary motion occurring in a particular posture after adding the third term.

The present invention relates to controlling the motion of a humanoid robot, and more particularly, to a method of controlling the motion so that the humanoid robot can reproduce the motion similar to the human arm motion.

Humanoid robots are robots that provide various services on behalf of humans or through collaboration with humans by imitating human intelligence, behavior, and interaction. With the development of humanoid robots, the scope of use of robots used to implement automation systems in specific industries can be extended to human life.

In order for humans and humanoid robots to coexist and interact smoothly, not only the appearance of humanoid robots but also their human behavior must be friendly and predictable. If humanoid robots are in a poor and unpredictable state, they cannot communicate with each other smoothly but rather act as a threat to humans.

Among these, the most basic movement that should be naturally imitated for communication between humans and humanoid robots will be arm movements. This is because humans use various arm movements in communication to express their content and feelings in abundance.

In the past, efforts to imitate human motion in a humanoid robot have been steadily continued by various scholars as follows.

Dasgupta and Nakamura (A. Dasgupta and Y. Nakamura.Making feasible walking of humanoid robots from human motion capture data.In International Conference on Robotics and Automation, pages 1044 ~ 1049, May 1999.Detroit, Michigan) Capture data for human walking was used to obtain. In this method, Fourier expansion was used to obtain the time trace of the target ZMP (Zero Moment Point) from the capture data. Based on this, the reaction force of the foot against the ground according to the trajectory of the target ZMP was obtained using the optimization technique. However, such a technique is not easy to be applied to the reproduction of the arm motion due to the difference in the degree of freedom of the foot and the arm and the difference of the behavior boundary.

Zhao and Huang et al. (X. Zhao, Q. Huang, Z. Peng, and K. Li. Kinematics mapping and similarity evaluation of humanoid motion based on human motion capture.In International Conference on Intelligent Robots and Systems, volume 1, pages 840 845, September 28 ~ October 2 2004. Sendai, Japan) defined similar functions and limited the method of applying captured motions to humanoid robots. The similar function is defined to minimize the difference in joint displacement between human and humanoid robots. However, here, there is a problem that the degree of reality is assuming that the degree of freedom of the humanoid robot is the same as the human.

Although these papers provide various ideas for converting human motions into humanoid robots, they only describe the techniques of applying imitations of human motions to humanoid robots without simply reflecting the differences between them. Therefore, it is not easy for a humanoid robot to reproduce the actual human arm motion naturally.

Accordingly, it is an object of the present invention to provide a method for controlling the reproduction of the human arm motion of a humanoid robot such that the humanoid robot can reproduce a smooth and flexible motion as a human motion.

It is another object of the present invention to provide a method for controlling the reproduction of the human arm motion of the humanoid robot, which enables to reproduce the delicate human arm motion regardless of the size of the humanoid robot and the degree of freedom of the arm (more than 4 degrees of freedom). .

The method for controlling the reproduction of the human arm motion of the humanoid robot of the present invention for solving the above problems comprises the steps of: obtaining a visual trajectory capturing the human arm motion; Calculating an objective function for minimizing the difference between the arm motion of the human arm and the arm motion of the humanoid robot for each time value of the captured time trajectory to obtain an optimal amount of joint displacement; And controlling the humanoid robot to perform the converted operation by using the obtained optimum amount of joint displacement per hour.

Here, the method for controlling the human arm motion representation of the humanoid robot comprises the step of equalizing the boundary between the arm work space of the human and the humanoid robot in consideration of the difference between the human arm length and the arm length of the humanoid robot; It includes more.

Here, the optimal joint displacement amount corresponds to a range between the minimum displacement change amount and the maximum displacement change amount of the arm joint of the humanoid robot and the minimum displacement speed change amount and the maximum displacement speed change amount of the arm joint of the humanoid robot. Minimize the function.

The controlling of the humanoid robot to perform the converted operation may include performing a curve on the captured time trajectory time range of the joint trajectory using the optimal joint displacement amount, and performing the curved joint. Determining the displacement and velocity trajectory of each joint using the trajectory, and controlling the humanoid robot to perform the switched motion using the determined displacement and velocity trajectory as the target trajectory of the motion controller. .

Here, the objective function is a first term for minimizing a positional difference between a plurality of markers attached to a human hand and virtual points of a humanoid robot corresponding to the plurality of markers when capturing the human arm motion; The sum of the second term for minimizing the phase difference between the upper arm and the upper arm of the humanoid robot, and the third term for minimizing the amount of change in the joint value of the human arm and the arm of the humanoid robot.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the drawings so that those skilled in the art may easily implement the present invention.

1 is a flowchart of a method for controlling the reproduction of a human arm motion of a humanoid robot according to the present invention.

First, a time trajectory for capturing a human arm motion to be reproduced in a humanoid robot is obtained (S10). Human arm motion can generally be obtained by using the motion capture system. The motion capture system is a device that attaches a plurality of markers to a human body to obtain a time trace and monitors the movement of each marker to obtain a time trace according to the motion.

Next, it adjusts to match the work space boundary of the arm between the human and the humanoid robot (S20). Since the working space that the arm can reach is different between the human and the humanoid robot due to the difference in arm length, if the time trajectory data capturing the motion of the human arm using the motion capture system is directly applied to the humanoid robot as described above. Similar behavior cannot be reproduced. This will be described in more detail with reference to FIG. 2.

2 is an exemplary diagram for explaining the difference in arm length between a human and a humanoid robot. FIG. 2A is a human model and FIG. 2B is a model diagram of a humanoid robot. As shown in FIG. 2, there is a difference in arm length between a human and a humanoid robot. Therefore, it is necessary to similarly adjust the outer boundary of the working space of the two objects by using the arm length ratio between the two objects.

를 이용하여 로봇의 팔 길이를 인간의 팔 길이와 동일하게 조절한다. 여기서, L_human과 L_robot은 각각 인간과 로봇의 팔 길이이다. 그러므로, 위 비율을 이용하여 인간형 로봇의 위팔과 아래팔의 길이는

과

로 조절된다. 이로 인해 인간형 로봇 팔의 작업 공간의 경계는 인간 팔의 작업 공간의 경계와 유사하게 조절될 수 있다. That is, the ratio of the arm length of the humanoid robot to the length of the human arm

Adjust the arm length of the robot to be the same as the length of the human arm. Where L _human and L _robot are the arm lengths of the human and the robot, respectively. Therefore, using the above ratio, the length of the upper and lower arms of the humanoid robot is

and

Is adjusted. As a result, the boundary of the working space of the humanoid robot arm can be adjusted similarly to the boundary of the working space of the human arm.

However, even after the arm length is adjusted, the humanoid robot has less freedom than the human and the difference in the ratio of the length of the lower arm to the upper arm does not exactly match the human workspace. In order to solve this problem, the general inverse kinematics can be used to change the motion of the robot using the direction and position of the human hand. However, imitation of human arm motion by this method can make a significant difference in the behavior of the two individuals. In particular, the position and posture that the elbow or hand can reach within the workspace may vary according to the difference in the length and freedom of the upper and lower arms.

In order to solve this problem that occurs even after matching the working space boundary of the arm, the present invention proposes a method of switching to an optimized operation form by introducing an objective function as follows.

Next, the objective function is calculated for each time value of the captured time trajectory to obtain an optimal joint displacement amount (S30).

The joint value at the i th time of the entire operation time may be summarized by Equation 1 below.

for j=1~N

for j = 1 to N

은 전체 동작 시간 중 i번째 시간에서의 j번째 관절 값을 의미한다.

는 j번째 관절의 i-1번째 시간과 i번째 시간 사이의 관절 값 변화량이다. N은 인간형 로봇 팔의 관절 수이다.

은 이전 시간, 즉, i-1번째 시간 에 푼 최적해로써 i번째 시간에서는 알고 있는 값이다. 따라서,

을 알기 위해서는

를 계산하여야 한다. 그러므로, 최적의 관절 변위량

를 최적화 문제의 변수로 사용한다. here,

Denotes the j-th joint value at the i-th time of the entire operation time.

Is the amount of change in the joint value between the i-1 th time and the i th time of the j th joint. N is the number of joints of the humanoid robot arm.

Is the optimal solution solved for the previous time, i-1th time, and is known at the i time. therefore,

To know

Should be calculated. Therefore, the optimal amount of joint displacement

Is used as a variable in the optimization problem.

The objective function to be minimized at the i th time, that is, t = t _i , is defined as in Equation 2 below.

로 표시된다. Here, the first term is a term for minimizing an error between the position of the human hand and the humanoid robot, the second term is a term for minimizing the phase difference between the human upper arm and the upper arm of the humanoid robot, and the third term is the term of the human Minimized amount of change in the joint value of the arm and the arm of the humanoid robot. In order to reconcile the differences between the operations defined in the respective terms, each term has a predetermined weight W, W ₁₀ appropriate to each term. And W ₁₁ is added. At the i time, the joint angle variation vector is

Is displayed.

The first term consists of terms that minimize the positional difference between the markers attached to the human hand and the virtual points of the humanoid robot in order to consider an error between the position of the human hand and the humanoid robot.

, 로

및

는 각각 인간의 팔목, 엄지, 약지에 위치한 마커에서 측정된 궤적이고,

,

및

는 인간형 로봇의 팔목, 엄지, 약지에 정의된 점의 궤적이다. 이때 정의된 각 점들은 관절값으로 표현되고, 이러한 관절값은 다시 관절 변위량

에 의해 얻어진다. here,

, By

And

Are traces measured from markers located on the human wrist, thumb, and ring finger, respectively.

,

And

Is the trajectory of the point defined in the wrist, thumb and ring finger of the humanoid robot. Each defined point is expressed as a joint value, and this joint value is again a joint displacement amount

Obtained by

When the first term is minimized, the difference in the position of the human hand and the position of the humanoid robot can be minimized to reproduce the hand motion similarly. However, even in this case, there is a possibility that the elbow movements are different from each other due to the phase difference between the upper arm of the human upper arm and the upper arm of the humanoid robot. Therefore, the second paragraph was added to compensate for this.

The second term is to minimize the phase difference between the human upper arm and the upper arm of the humanoid robot. 3 is an exemplary view for explaining the phase difference between the upper arm of a human and a humanoid robot, FIG. 3A is a human model, and FIG. 3B is a model diagram of a humanoid robot.

의 끝 점과 그 점 의

으로의 수직 투영점을 연결하는 벡터로써, 이 벡터 크기의 제곱은 다음과 같이 관절 변위의 변화량

에 수학식 3으로 표현될 수 있다. Referring to FIG. 3, there is a phase difference represented by the vector S _clbow between the human upper arm and the upper arm of the humanoid robot. S _clbow is shown as

End point of and that point of

A vector that connects the vertical projection points to, the square of the magnitude of the vector is the amount of change in joint displacement as

It can be represented by the equation (3).

,

이며,

과

는 시간 t_i 에서 캡쳐된 인간의 팔꿈치와 어깨의 위치벡터이고,

과

는 시간 t_i 에서 계산된 인간형 로봇의 팔꿈치와 어깨의 위치벡터이다. here,

,

Is,

and

Is the position vector of human elbow and shoulder captured at time t _i ,

and

Is the position vector of the elbow and shoulder of the humanoid robot calculated at time t _i .

By minimizing the value of the second term as described above, the phases of the upper arm of the human upper arm and the upper arm of the humanoid robot can be matched to match the elbow movements.

In the general motion, the motions were properly converted due to the above terms, but the motion still occurred in a few singular positions. For example, in a posture with the arm facing down, even if the upper arm of the humanoid robot rotates in the opposite direction to the lower arm, the position and phase of the humanoid robot's hand coincide with that of the human, so that it is necessary in a particular posture. Without this, the arm can rotate 360 degrees, and so on. Therefore, the third term was added to solve this problem.

The third term is a term that controls the use of the minimum displacement of the joint angle to prevent unnecessary movement in the singular posture.

로 표시된다.As described above, the amount of change in the joint angle at the i th time of the third term is

Is displayed.

FIG. 4 is an exemplary diagram for describing unnecessary motions occurring in a singular posture of a humanoid robot, and FIG. 4A is a screen for monitoring unnecessary motions occurring in a singular posture before adding a third term in the objective function of the present invention. 4b is a screen that monitors the removal of unnecessary motion that occurs in a singular posture after adding the third term. Referring to the drawings, as shown by the dotted line in FIG. 4A, the unnecessary phase value existed during the operation before adding the third term, but after adding the third term, the phase value disappeared to perform the same operation with the minimum joint angle change. It can be done.

By constructing the objective function as the above terms, it is possible to mimic the human arm motion using the minimum joint displacement while minimizing the difference in the position of the hand and the upper arm.

However, since humanoid robots have limited degrees of freedom, the imitation of human arm motion by the robot must be within the range between the minimum and maximum displacement variation of the arm joint of the humanoid robot and between the minimum and maximum displacement velocity variations. Restricted

The driving limit angle of the joint is determined by Equation 4 as follows.

와

는 팔 관절의 최소 구동각과 최대 구동각이다. 이 수학식 4를 수학식 1에 대입하면 다음과 같이 최소 변위 변화량과 최대 변위 변화량을 나타내는 수학식 5를 얻을 수 있다. here,

Wow

Is the minimum and maximum driving angle of the arm joint. Substituting Equation 4 into Equation 1, Equation 5 representing the minimum displacement change amount and the maximum displacement change amount can be obtained as follows.

이고

이다. Equation 6 below shows the limit value of the angular velocity of each joint. here,

ego

to be.

과

는 각 관절의 변위 속도의 최대값과 최소값이다.here,

and

Is the maximum and minimum values of the displacement velocity of each joint.

으로 표현할 수 있다. 이 식을 수학식 6에 대입하면, 다음과 같은 최소 관절 변위속도와 최대 관절 변위속도를 나타내는 수학식 7이 얻어진다. The velocity of the joint at t _i is obtained using the Backward Difference Method (BDM),

It can be expressed as Substituting this equation into Equation 6, Equation 7 representing the following minimum and maximum joint displacement speeds is obtained.

이고,

이다.

는 인간 동작의 시간 간격이다. here,

ego,

to be.

Is the time interval of human motion.

Equations 5 and 7 limit the optimal amount of joint displacement of the objective function at time t _i .

Therefore, the human arm motion every time is realized by obtaining the optimum joint variation amount that minimizes the objective function of Equation 2 while satisfying the joint displacement amounts of Equations 5 and 7. This optimization problem is a quadratic programming problem that can be solved using well-known algorithms such as sequential quadratic programming (SQP).

Next, the curve of the joint trajectory is calculated by using an interpolation method using cubic splines for the optimal amount of joint displacement obtained for each time by calculating the objective function (S40). At this time, the trajectory of the joint is formed by performing a curve on the captured time trajectory.

Next, the joint position and the velocity trajectory are determined at each time (S50). Joint position and velocity trajectories can be obtained using curved joint trajectories and first-order differential trajectories.

Finally, the humanoid robot controls to perform the switched operation (S60). At this time, the displacement and velocity trajectory of each joint is used as the target trajectory of the motion controller such as proportional-integral-derivative to control the humanoid robot to perform the changed motion.

The above embodiments are only for explaining the present invention, and the scope of the present invention is not limited to the embodiments, and may be modified or modified by those skilled in the art within the scope of the present invention defined by the appended claims. .

According to the method for controlling the reproduction of the human arm motion of the humanoid robot of the present invention as described above, by converting the human arm motion into a reproducible motion of the humanoid robot, the humanoid robot can reproduce the smooth and flexible motion as the human motion.

In particular, by obtaining the optimized motion in consideration of the displacement variation and displacement velocity of the arm joint of the humanoid robot, it is possible to reproduce the delicate human arm movement regardless of the size of the humanoid robot and the degree of freedom of the arm (more than 4 degrees of freedom). Can be.

Claims (10)

Obtaining a visual trajectory capturing a human arm motion; Calculating an optimal joint displacement amount by calculating an objective function for minimizing the difference between a human arm motion and an arm of a humanoid robot for each time value of the captured time trajectory, wherein the objective function is configured to capture the human arm motion The first claim for minimizing the positional difference between the plurality of markers attached to the human hand of the human eye and the virtual points of the humanoid robot corresponding to the plurality of markers, and the phase difference between the upper arm of the human upper arm and the upper arm of the humanoid robot. A second term and a sum of a third term for minimizing the amount of change in the joint value of the human arm and the arm of the humanoid robot; And controlling the humanoid robot to perform a switched motion by using the obtained optimal amount of joint displacement for each time. The method of claim 1, And considering the difference between the arm length of the human arm and the arm length of the humanoid robot to equalize the boundary between the arm workspaces of the human and humanoid robots. The method of claim 1 or 2, wherein the optimum amount of joint displacement, The human displacement ranges between the minimum displacement variation and the maximum displacement variation of the arm joint of the humanoid robot, and the minimum displacement velocity variation and the maximum displacement velocity variation of the arm joint of the humanoid robot. How to control arm motion reproduction. The method of claim 1 or 2, wherein the controlling of the humanoid robot to perform a switched operation comprises: Performing a curve on the captured time trajectory time range of the joint trajectory using the optimum amount of joint displacement; Determining the displacement and velocity trajectory of each joint using the curved joint trajectory, And controlling the humanoid robot to reproduce the human arm motion by using the determined displacement and velocity trajectory of each joint as a target trajectory of the motion controller. The method of claim 4, wherein in the process of performing the curve, A method of controlling the reproduction of a human arm motion of a humanoid robot in which an optimal amount of joint displacement obtained at each time is performed by using an interpolation method using cubic splines. delete The method of claim 1, wherein the plurality of markers attached to the human hand controls human arm motion reproduction of a humanoid robot located at the wrist, thumb, and ring finger. The method of claim 1, wherein the first term, The matrix of equation (8)

) And the column matrix converted from the matrix

) To reproduce the human arm motion of the humanoid robot calculated by multiplying by the weight (W) for the arm motion. (Equation 8)

here,

And

Is the locus measured from markers located on the human wrist, thumb, and ring finger,

,

And

Is the trajectory of the point defined in the wrist, thumb and ring finger of the humanoid robot. The method of claim 1, wherein the second term, And multiplying a square of the magnitude of the vector S _elbow representing the direction vector difference between the human upper arm and the upper arm of the humanoid robot by a predetermined weight (W ₁₀ ). According to claim 1, The amount of change in the joint value of the arm of the human arm and the humanoid robot, A method of controlling the reproduction of the human arm motion of a humanoid robot, which is calculated by multiplying a predetermined weight (W ₁₁ ) by a square of the size of a matrix representing the amount of change in the joint value between i-1 th time and i th time in each joint.

Images