KR20210042644A - Control system and method for humanoid robot - Google Patents

Abstract

Disclosed are a control system and a method for a humanoid robot. According to the present invention, the method for the humanoid robot photographs a user through two lenses apart from each other, estimates the posture of the user from each of two images photographed from two different points of view, checks the three-dimensional position of each joint of the user from the estimated postures, removes noise, and corrects the positions of the corresponding joints. Afterwards, the method uses the three-dimensional position information of each joint, generates an angle of each motor included in the humanoid robot for the corresponding humanoid robot to show the same posture as that of the user, provides the angle to the humanoid robot, and thereby controls the posture of the humanoid robot. Accordingly, the present invention is able to rapidly control the humanoid robot in a stable manner from a remote place through a relatively simple calculation even without using a depth sensor.

Description

Humanoid robot control system and method {Control system and method for humanoid robot}

The present invention relates to a robot control system and method, and more particularly, to a humanoid robot control system and method.

A technology that attempts to save lives and perform dangerous tasks by putting a robot in a space that is difficult for humans to access or a space where humans can work dangerous is being actively researched, but has not yet been commercialized. This is because if the robot fails to accurately control its movement, the robot may be damaged or the robot itself may become a risk factor.

Currently, the control accuracy and speed of a remote-controlled robot are in a trade-off relationship, so either one must be selected when controlling the robot, and the universal remote-controlled robot places a greater weight on the accuracy of the control. For this reason, most of the robot remote control has the disadvantage that the speed is very slow.

This is because a lot of multiple verification processes are required to control the robot more accurately. In order to control a robot in real time, it is necessary to perform accurate calculations for motor control, but current robot control algorithms have a large amount of overall calculation, which makes real-time control difficult.

The problem to be solved by the present invention is to provide a humanoid robot control system and method capable of controlling a robot in real time using an algorithm with a small amount of computation, in particular, a humanoid robot having a human joint structure.

A humanoid robot control system according to a preferred embodiment of the present invention for solving the above-described problems, by using two lenses spaced apart from each other, respectively photographing the postures of the user who wants to control the robot, the first image and the second image A photographing unit to generate a; A posture information generator for generating posture information of the user from each of the first image and the second image, and generating position information of each joint of the user from the posture information of the user; A motor control signal generator for generating a motor control signal for controlling an angle of each motor included in the humanoid robot so that the humanoid robot to be controlled displays the same posture as the user by using the position information of each joint of the user; And a communication unit for transmitting the motor control signal to the humanoid robot.

In addition, the photographing unit may generate a first image (left image) and a second image (right image) photographing a user by two lenses being spaced apart from each other by a predefined distance in the horizontal direction.

In addition, the posture information generation unit estimates the posture of the user expressed in the first image and the second image, and provides position information of each joint of the user in the first image and the position information of each joint of the user in the second image. A first posture estimating unit and a second posture estimating unit respectively generating position information; A first noise removing unit and a second noise removing unit for removing noise from each of the position information of the joints of the first image and the second image; And correcting the position information of the corresponding joint at the current viewpoint using the position information of the joints corresponding to each other from which the noise of the first image and the second image has been removed, and the position information of the corresponding joints determined at a previous time point. It may include a position correction unit for generating the final joint position information of the joint.

In addition, the first noise removing unit and the second noise removing unit may perform noise removal by performing kalman filtering on position information of each joint.

In addition, the position correction unit performs median filtering on joint position information corresponding to each other from which noises of the first image and the second image are removed, and position information of the corresponding joint determined at a previous time point. Final position information may be generated and output to the motor control signal generator.

In addition, the motor control signal generator generates an angle value of each motor included in the humanoid robot so that the humanoid robot displays the same posture as the user by using the position information of each joint of the user input from the posture information generator. It may include a joint motor angle generator.

In addition, the motor control signal generation unit performs IIR (Infinite Impulse Response) filtering by using the angle value of each motor input from the joint motor angle generation unit and the angle value of each corresponding motor at a previous point in time to perform a joint motor angle value. It may further include an angle correction unit for correcting.

Meanwhile, a humanoid robot control method according to a preferred embodiment of the present invention for solving the above-described problems is a humanoid robot control method performed in a humanoid robot control system, comprising: (a) using two lenses spaced apart from each other, a robot Generating a first image and a second image by respectively photographing a posture of a user who wants to control the user's posture; (b) generating the user's posture information from each of the first image and the second image, and generating location information of each joint of the user from the user's posture information; (c) generating a motor control signal for controlling an angle of each motor included in the humanoid robot so that the humanoid robot to be controlled displays the same posture as the user by using the position information of each joint of the user; And (d) transmitting the motor control signal to the humanoid robot.

In addition, in the step (a), the two lenses are spaced apart from each other by a predefined distance in the horizontal direction, so that a first image (left image) and a second image (right image) photographed by the user may be generated. .

In addition, the step (b), (b1) by estimating the posture of the user expressed in the first image and the second image, the position information of each joint of the user of the first image and the user of the second image Generating location information of each joint of each; (b2) removing noise from each of the position information of the joints of the first image and the second image; And (b3) correcting the position information of the corresponding joint at the current time point using the position information of the joints corresponding to each other from which the noise of the first image and the second image has been removed and the position information of the corresponding joints determined at a previous time point. Thus, it may include the step of generating the final joint position information of each joint.

In addition, in step (b2), noise removal may be performed by performing kalman filtering on position information of each joint.

In addition, in step (b3), median filtering is performed on the joint position information corresponding to each other from which noises of the first image and the second image are removed and the position information of the corresponding joint determined at a previous time point. The final position information of the joint can be generated.

In addition, step (c), using the position information of each joint of the user generated in step (c1) (b), the angle of each motor included in the humanoid robot so that the humanoid robot displays the same posture as the user. It may include generating a value.

In addition, in step (c), the joint motor is filtered by performing IIR (Infinite Impulse Response) filtering using the angle value of each motor generated in step (c2) and the angle value of each corresponding motor at the previous point in time. It may further include the step of correcting the angle value.

The present invention photographs a user through two lenses spaced apart from each other, estimates the user's posture from two images taken at different points of view, checks the three-dimensional position of each joint of the user from the estimated posture, and reduces noise. After removal, correct the position of the corresponding joint. After that, using the 3D position information of each joint, the humanoid robot's posture is controlled by generating the angle of each motor included in the humanoid robot and providing it to the humanoid robot side so that the corresponding humanoid robot shows the same posture as the user. do. Accordingly, the present invention can quickly and stably control a humanoid robot remotely through a relatively simple operation without using a depth sensor.

1 is a diagram showing the overall configuration and connection relationship of a humanoid robot control system according to a preferred embodiment of the present invention.
2 is a block diagram showing a detailed configuration of a humanoid robot control system according to a preferred embodiment of the present invention.
3 is a diagram illustrating an example of estimating a user's posture using the Human Pose Estimation (HPE) method applied to the present invention.
4 is a diagram illustrating the performance of a humanoid robot control system according to a preferred embodiment of the present invention.
5 is a flowchart illustrating a method of controlling a humanoid robot according to a preferred embodiment of the present invention.

Hereinafter, exemplary embodiments of the present invention will be described with reference to the accompanying drawings.

1 is a diagram showing the overall configuration and connection relationship of a humanoid robot control system according to a preferred embodiment of the present invention.

Referring to FIG. 1, a humanoid robot control system (hereinafter, abbreviated as "robot control system") according to a preferred embodiment of the present invention, when a user 100 who wants to control the humanoid robot 300 takes a posture. , By photographing this, a motor control signal for controlling the motor included in each joint of the humanoid robot is generated so that the humanoid robot 300 as a control target takes the same posture as the user, and is used as a humanoid robot through wireless communication or wired communication. send.

The humanoid robot 300 receives a motor control signal and drives the motor according to the motor control signal to take the same posture as the user's posture.

The humanoid robot control system 200 according to a preferred embodiment of the present invention for implementing such an operation includes a photographing unit 210, a posture information generation unit 220, a motor control signal generation unit 230, and a communication unit 240. Consists of including.

The photographing unit 210 generates a first image (eg, left image) and a second image (eg, right image) by photographing the posture of the user who wants to control the robot using two lenses spaced apart from each other. , These images are output to the posture information generation unit 220.

The posture information generation unit 220 estimates the user’s posture information from each of the first image and the second image input from the photographing unit 210, checks the user’s joint position from the user’s posture information, and Generates and outputs 3D position information of the joint. In a preferred embodiment of the present invention to be described below, as shown in FIG. 1, an example of generating a control signal for a humanoid robot having 8 joints will be described.

The motor control signal generation unit 230 is included in the humanoid robot so that the corresponding humanoid robot displays the same posture as the user by using the 3D position information of each joint of the user input from the posture information generation unit 220. A motor control signal for controlling the angle of each motor is generated and output to the communication unit 240.

The communication unit 240 transmits the motor control signal to the humanoid robot through a predetermined wireless or wired communication method.

Meanwhile, the communication unit 310 of the humanoid robot 300 receives a motor control signal and outputs it to the robot control unit 320, and the robot control unit 320 corresponds to each motor control signal input from the communication unit 310. By driving each motor, the humanoid robot takes the same posture as the user.

2 is a block diagram showing a detailed configuration of a humanoid robot control system 200 according to a preferred embodiment of the present invention.

Referring to FIG. 2, the photographing unit 210 of the present invention includes two lenses spaced apart from each other by a predefined distance L, and a first image captured at different viewpoints through each lens (for convenience, "left side An (L) image") and a second image (referred to as a "right (R) image" for convenience) are output to the posture information generator 220, respectively. In a preferred embodiment of the present invention, it has been described that the two lenses are spaced apart from each other by a distance L in the horizontal direction, but may be disposed vertically or diagonally from each other.

The posture information generation unit 220 includes a first posture estimating unit 221-1, a second posture estimating unit 221-2, a first noise removing unit 223-1, and a second noise removing unit 223-2. ), and a position correction unit 225.

The first posture estimating unit 221-1 and the second posture estimating unit 221-2 respectively receive a first image and a second image from the photographing unit 210, and are expressed in the first image and the second image. The user's posture is estimated and the position information of each joint of the user is output to the first noise removing unit 223-1 and the second noise removing unit 223-2, respectively.

The first posture estimating unit 221-1 and the second posture estimating unit 221-2 according to a preferred embodiment of the present invention may estimate a user's posture using a known posture estimation algorithm. However, in this embodiment, "A. Kanazawa, MJ Black, DW Jacobs, and J. Malik, "End-to-end recovery of human shape and pose," IEEE Conference on Computer Vision and Pattern Recognition, Jun. 2018.』, the user's posture was estimated and the position of each joint was estimated using the HPE (Human Pose Estimation) method. 3 is a diagram illustrating an example of estimating a user's posture using the Human Pose Estimation (HPE) method applied to the present invention.

로 표현된다. 이 때 V는 좌측 영상(L) 또는 우측 영상(R)(V={L,R})을 나타내고, j는 각 관절의 식별번호를 나타낸다(

).When this HPE method is applied to a humanoid robot having eight joints shown in FIG. 1(a), the first posture estimating unit 221-1 and the second posture estimating unit 221-2 output time t , The location information of the j-th joint is

It is expressed as At this time, V represents the left image (L) or the right image (R) (V={L,R}), and j represents the identification number of each joint (

).

를 입력받고, 제 2 노이즈 제거부(223-2)는 제 2 자세 추정부(221-2)로부터 우측 영상(제 2 영상)에서 추정된 사용자 관절의 위치 정보

를 입력받아, 해당 관절의 위치 정보에 포함된 노이즈를 제거한 후, 노이즈가 제거된 관절의 위치 정보를 위치 보정부(225)로 출력한다. On the other hand, the first noise removing unit 223-1 is the position information of the user joint estimated from the left image (first image) from the first posture estimating unit 221-1

Is input, and the second noise removing unit 223-2 determines the position information of the user's joint estimated from the right image (the second image) from the second posture estimating unit 221-2

Is received, noise included in the position information of the corresponding joint is removed, and then the position information of the joint from which the noise is removed is output to the position correction unit 225.

In this case, the first noise removing unit 223-1 and the second noise removing unit 223-2 may remove noise included in the joint position information by applying various noise filtering algorithms. In a preferred embodiment, the Kalman filter is applied to position information of each joint as shown in Equation 1 below to remove noise.

는 시간 t 에서의 좌측 영상 및 및 우측 영상의 각 관절의 위치 정보를 나타내고,

는 칼만 필터링을 통해서 노이즈가 제거된 관절의 위치 정보를 나타낸다.In Equation 1 above

Represents the location information of each joint in the left image and the right image at time t,

Represents the location information of the joint from which noise has been removed through Kalman filtering.

) 및 이전 시점(t-1)에 위치 보정부(225)에서 결정된 대응되는 각 관절의 위치 정보를 이용하여 현재 시점(t)의 해당 관절의 위치 정보를 보정하여, 최종 관절 위치 정보를 생성하여 모터 제어 신호 생성부(230)로 출력한다. The position correction unit 225 corresponds to each other of the first image (left image) and the second image (right image) input from the first noise removing unit 223-1 and the second noise removing unit 223-2. Position information of each joint (

) And the position information of the corresponding joint determined by the position correction unit 225 at the previous time point (t-1), correcting the position information of the corresponding joint at the current time point (t), and generating the final joint position information. Output to the motor control signal generator 230.

To this end, the position correction unit 225 according to a preferred embodiment of the present invention uses a median filter according to Equation 2 below.

은 칼만필터 적용 후 미디안필터를 적용시켜 구해진 관절의 위치 정보를 의미한다. As can be seen in Equation 2, the median filter refers to a filter that returns an intermediate value for each dimension of an input value,

Denotes the position information of the joint obtained by applying the Kalman filter and then applying the median filter.

)와 우측 영상의 관절 위치 정보(

) 외에, 이전 시간(t-1)에 미디언 필터를 통해서 얻은 최종적인 관절의 위치 정보(

)를 다시 미디안필터에 적용하는 것으로, 이는 좌측 영상에서 추정된 관절의 위치 정보와 우측 영상에서 추정된 관절의 위치 정보가 서로 다른 위치를 나타낼 때 생기는 사용자 자세의 글리치(glitch)를 제거해주는 역할을 한다. 또한, 두 자세가 동일한 형태로 변화할 때, 좀 더 느리게 변화를 유도하는 것으로 볼 수 있으며, 이는 안정적인 로봇의 동작 제어에 기여한다.Note that in Equation 2, the joint position information of the left image (

) And joint position information in the right image (

), the final joint position information obtained through the median filter at the previous time (t-1) (

) Is again applied to the median filter, which removes glitches in the user posture that occur when the position information of the joint estimated from the left image and the position information of the joint estimated from the right image indicate different positions. do. In addition, when the two postures change into the same form, it can be seen that the change is induced more slowly, which contributes to the stable motion control of the robot.

On the other hand, the motor control signal generation unit 230 uses the 3D position information of each joint of the user input from the position correction unit 225, so that the humanoid robot 300 as a control target represents the same posture as the user, A motor control signal for controlling the angle of each motor included in the robot 300 is generated and output to the communication unit 240.

To this end, the motor control signal generation unit 230 includes a joint motor angle generation unit 231 and an angle correction unit 233.

First, the joint motor angle generator 231 calculates the angle of each motor included in the humanoid robot so that the corresponding humanoid robot displays the same posture as the user using the input 3D position information of each joint. Output to the government 233. To this end, the joint motor angle generator 231 according to a preferred embodiment of the present invention includes the angle values Q of each motor included in the humanoid robot LIMS shown in FIG. 1(b) according to Equation 3 below. _t ).

In Equation 3, J2Q() represents a Q operation function specialized for humanoid robots.

Here, Φ1 to Φ3 denote motors installed on the left and right shoulders of the LIMS robot shown in Fig. 1(b), and Φ4 denotes motors installed on the left and right elbows of the LIMS robot shown in Fig. 1(b), Each motor is driven according to the angle values of each motor included in _{Q t.} Since the method of calculating the angle of the motor using the Q operation function specialized for humanoid robots is well known, a detailed description will be omitted.

_{On the other hand, the angle correction unit 233 uses the angle values (Q t} ) of each motor at the current time t input from the joint motor angle generator 231 and the angle values of each motor determined at the previous time t-1. By performing correction, noise and errors are corrected, and a motor control signal to finally control the motors included in the humanoid robot is generated and output to the communication unit 240.

_{The angle values (Q t} ) of each motor at the current time t input from the joint motor angle generator 231 include discontinuous angle values that can cause serious problems by generating a sudden displacement when controlling a humanoid robot. can do. _{Accordingly, the angle correction unit 233 calculates the angle values Q t} of each motor at the current time t input from the joint motor angle generator 231 to ensure the continuous motion of the motor, and the previous time t- Correct using the angle values of each motor determined in 1.

In a preferred embodiment of the present invention, the angle correction unit 233 corrects the angle of each motor by removing discontinuous motor angle values using an Infinite Impulse Response filter (IIR) as shown in Equation 4 below.

이 휴머노이드 로봇의 모터 제어 신호로서 통신부(240)를 통해서 휴머노이드 로봇으로 전송된다. In a preferred embodiment of the present invention, the attenuation coefficient α of Equation 4 is set to 0.05,

As a motor control signal of the humanoid robot, it is transmitted to the humanoid robot through the communication unit 240.

4 is a diagram illustrating the performance of a humanoid robot control system according to a preferred embodiment of the present invention.

)의 각도 변화를 도시하였다. 상기한 종래의 HPE(Human Pose Estimation) 방식에 따라서만 계산된 각도는 파란색으로, 본 발명의 제 1 및 제 2 노이즈 제거부(223-1,223-2)에서 칼만 필터링되고 위치 보정부(225)에서 미디안 필터링되었으나 각도 보정부(233)에서 IIR 필터링되지 않은 각도는 하늘색으로, 본 발명에 따라서 최종적으로 생성된 각도는 빨간색으로 각각 도시하였다.Referring to the example shown in Fig. 4, the right fourth motor (

) Of the angle change is shown. The angle calculated only according to the conventional HPE (Human Pose Estimation) method is blue, and Kalman is filtered by the first and second noise removal units 223-1 and 223-2 of the present invention, and the position correction unit 225 Angles that are median filtered but not IIR filtered by the angle correction unit 233 are shown in light blue, and angles finally generated according to the present invention are shown in red.

Referring to FIG. 4, it can be seen that the angle of the motor changes rapidly in the case of the prior art, and it can be seen that the angle changes naturally and continuously in the case of the present invention.

In addition, in the lower part of FIG. 4, an image of a user photographed at a time point t=1190 and a time point t=1575, respectively, and an initial pose of a humanoid robot controlled using a posture estimated according to the prior art at each time point, and It shows the posture (Refined Pose) of the humanoid robot controlled according to the present invention. As shown, it can be seen that the posture of the humanoid robot controlled according to the present invention is much more accurate than that of the prior art, and is suitable for remote control of the humanoid robot.

So far, the humanoid robot control system 200 according to a preferred embodiment of the present invention has been described.

5 is a flowchart illustrating a method of controlling a humanoid robot according to a preferred embodiment of the present invention.

Hereinafter, a method for controlling a humanoid robot according to a preferred embodiment of the present invention will be described with reference to FIG. 5. However, since all humanoid robot control methods described below are performed by the above-described humanoid robot control system 200, their functions are the same. Therefore, unless otherwise stated to the contrary, the functions described with reference to FIGS. 1 to 4 are applied as they are in the humanoid robot control method of the present invention. Hereinafter, in order to avoid duplication of description, only the overall flow of the method is It will be described, and specific functions should refer to the functions of the humanoid robot control system 200 described above.

First, in the humanoid robot control method of the present invention, the humanoid robot control system 200 generates a first image (left image) and a second image (right image) by photographing a user with two lenses spaced apart from each other inside. Do (S510).

Then, the humanoid robot control system 200 estimates the user's posture from each of the first image and the second image, and estimates the position information of each joint of the user estimated from the first image and the user's angle estimated from the second image. The position information of the joint is generated (S520).

Next, the humanoid robot control system 200 removes noise included in the joint position information of the first image and the second image (S530). To this end, it is as described above that the present invention performs Kalman filtering on each joint position information.

Thereafter, the humanoid robot control system 200 uses the position information of the joints corresponding to each other acquired from the first image and the second image, and the position information of the joints corresponding to each other determined at a previous point in time. Final joint position information is generated by correcting the corresponding position information (S540). In this process, it is as described above that the preferred embodiment of the present invention applies the median filter.

Next, the humanoid robot control system 200 calculates the angle of each motor included in the humanoid robot so that the humanoid robot to be controlled displays the same posture as the user by using the final 3D position information of each joint (S550). .

After that, in order to prevent rapid movement of the robot, the motor control signal that will finally control the motors included in the humanoid robot by correcting the angle values of each motor at the current time using the angle values of each motor determined at the previous time. Is generated and transmitted to the humanoid robot (S560).

The humanoid robot receiving the motor control signal is controlled to take a posture taken by the user by driving each motor included in the humanoid robot according to the motor control signal (S570).

So far, the present invention has been looked at around its preferred embodiments. Those of ordinary skill in the art to which the present invention pertains will be able to understand that the present invention may be implemented in a modified form without departing from the essential characteristics of the present invention. Therefore, the disclosed embodiments should be considered from an illustrative point of view rather than a limiting point of view. The scope of the present invention is shown in the claims rather than the above description, and all differences within the scope equivalent thereto should be construed as being included in the present invention.

100: user
200: Humanoid robot control system
210: photographing department
220: posture information generation unit
221-1: first posture estimating unit 221-2: first posture estimating unit
223-1: first noise removing unit 223-2: second noise removing unit
225: position correction unit
230: a corner control signal generation unit
231: joint motor angle generation unit 233: angle correction unit
240: Ministry of Communications
300: Humanoid robot
310: Ministry of Communications
320: robot control unit

Claims (14)

A photographing unit for generating a first image and a second image by photographing a posture of a user who wants to control the robot using two lenses spaced apart from each other;
A posture information generator for generating posture information of the user from each of the first image and the second image, and generating position information of each joint of the user from the posture information of the user;
A motor control signal generator for generating a motor control signal for controlling an angle of each motor included in the humanoid robot so that the humanoid robot to be controlled displays the same posture as the user by using the position information of each joint of the user; And
And a communication unit for transmitting the motor control signal to a humanoid robot. The method of claim 1, wherein the photographing unit
A humanoid robot control system, characterized in that the two lenses are spaced apart from each other by a predefined distance in a horizontal direction to generate a first image (left image) and a second image (right image) photographing a user. The method of claim 1, wherein the posture information generation unit
A first for estimating the user's postures expressed in the first image and the second image, and generating position information of each joint of the user in the first image and position information of each joint of the user in the second image, respectively. A posture estimating unit and a second posture estimating unit;
A first noise removing unit and a second noise removing unit for removing noise from each of the position information of the joints of the first image and the second image; And
By correcting the position information of the corresponding joint at the current time point by using the position information of the joints corresponding to each other from which the noise of the first image and the second image has been removed and the position information of the corresponding joints determined at a previous time point, each joint Humanoid robot control system comprising a position correction unit for generating the final joint position information of. The method of claim 3,
The first noise removing unit and the second noise removing unit
A humanoid robot control system, characterized in that noise removal is performed by performing kalman filtering on position information of each joint. The method of claim 3, wherein the position correction unit
By performing median filtering on the joint position information corresponding to each other from which noise of the first image and the second image has been removed, and the position information of the corresponding joint determined at a previous point in time, the final position information of the corresponding joint is generated. Humanoid robot control system, characterized in that output to the motor control signal generating unit. The method of claim 1, wherein the motor control signal generation unit
Including a joint motor angle generator for generating an angle value of each motor included in the humanoid robot so that the humanoid robot represents the same posture as the user by using the position information of each joint of the user input from the posture information generator. Humanoid robot control system characterized by. The method of claim 6, wherein the motor control signal generator
Further comprising an angle correction unit for correcting the joint motor angle value by performing IIR (Infinite Impulse Response) filtering using the angle value of each motor input from the joint motor angle generator and the angle value of each motor at a previous time point. Humanoid robot control system, characterized in that.
As a humanoid robot control method performed in a humanoid robot control system,
(a) generating a first image and a second image by photographing a posture of a user who wants to control the robot using two lenses spaced apart from each other;
(b) generating the user's posture information from each of the first image and the second image, and generating location information of each joint of the user from the user's posture information;
(c) generating a motor control signal for controlling an angle of each motor included in the humanoid robot so that the humanoid robot to be controlled displays the same posture as the user by using the position information of each joint of the user; And
(d) transmitting the motor control signal to the humanoid robot. The method of claim 8, wherein step (a) is
A method for controlling a humanoid robot, characterized in that two lenses are spaced apart from each other by a predefined distance in a horizontal direction to generate a first image (left image) and a second image (right image) photographing a user. The method of claim 8, wherein step (b) is
(b1) By estimating the posture of the user expressed in the first image and the second image, position information of each joint of the user in the first image and position information of each joint of the user in the second image are respectively generated. The step of doing;
(b2) removing noise from each of the position information of the joints of the first image and the second image; And
(b3) By correcting the position information of the corresponding joint at the current point of view using the position information of the joints corresponding to each other from which the noise of the first image and the second image has been removed, and the position information of the corresponding joints determined at a previous point in time. , Humanoid robot control method comprising the step of generating the final joint position information of each joint. The method of claim 10, wherein the step (b2) is
A method for controlling a humanoid robot, characterized in that noise removal is performed by performing Kalman filtering on position information of each joint. The method of claim 10, wherein step (b3) is
Generating final position information of a corresponding joint by performing median filtering on joint position information corresponding to each other from which noise of the first image and the second image has been removed, and position information of a corresponding joint determined at a previous time point. Humanoid robot control method characterized in that. The method of claim 8, wherein step (c) is
(c1) using the position information of each joint of the user generated in step (b), including the step of generating an angle value of each motor included in the humanoid robot so that the humanoid robot represents the same posture as the user. Humanoid robot control method characterized in that. The method of claim 13, wherein step (c) is
(c2) Compensating the joint motor angle value by performing IIR (Infinite Impulse Response) filtering using the angle value of each motor generated in step (c1) and the angle value of each motor corresponding to the previous time point. Humanoid robot control method, characterized in that.

Images