KR20230087303A - Robot arm joint friction estimation device and estimation method - Google Patents

Abstract

The present invention relates to a robot arm joint friction estimator and estimation method capable of improving the control precision of a robot arm by accurately estimating the friction torque generated in the robot arm in real time,
The robot arm joint friction estimator according to an embodiment of the present invention,
a data collection unit that collects an input data set in which friction torque values and joint velocity values of the robot arm in an arbitrary posture without external force are matched; a recursive linear regression model unit for estimating a friction parameter by driving a recursive linear regression model using the input data set as an input value; and a friction torque estimating unit that calculates an estimated friction torque using the estimated friction parameter and transmits the calculated estimated friction torque to a control unit that controls an operation of the robot arm.

Description

Robot arm joint friction estimation device and estimation method {Robot arm joint friction estimation device and estimation method}

The present invention relates to an apparatus and method for estimating joint friction of a robot arm capable of improving control precision of a robot arm by accurately estimating friction torque generated in a robot arm in real time.

Recently, various robots are being developed. Among them, a multi-joint structured robot arm having joints and having a high degree of freedom is widely used in living and industrial fields. A robot arm may be used to support or carry a workpiece.

The frictional force generated from the “joints” of these robot arms causes non-linear “frictional” phenomena such as robot position and speed errors, response performance deterioration, and stick-slip motion, which is the main cause of deteriorating the performance of robot precision control and joint torque-based control.

The deterioration in the control performance of the robot due to the friction phenomenon can be improved by estimating the friction force of the robot's joint and appropriately compensating for it, and this technology is called the "friction" compensation technology of the robot arm.

In order to predict the frictional force generated from the joints of the robot, a mathematical model of the frictional force must be constructed, and in order to construct a mathematical model of the frictional force, the process of estimating the frictional force is required.

Korean Patent Publication No. 10-2017-0086862

An object of the present invention is to provide a robot arm joint friction estimator and method capable of improving control precision of a robot arm by accurately estimating friction torque generated in a robot arm in real time.

The robot arm joint friction estimator according to an embodiment of the present invention,

a data collection unit that collects an input data set in which friction torque values and joint velocity values of the robot arm in an arbitrary posture without external force are matched; a recursive linear regression model unit for estimating a friction parameter by driving a recursive linear regression model using the input data set as an input value; and a friction torque estimating unit that calculates an estimated friction torque using the estimated friction parameter and transmits the calculated estimated friction torque to a control unit that controls an operation of the robot arm.

In the robot arm joint friction estimator according to an embodiment of the present invention, the data collection unit may process at least one of the input data sets having the same joint velocity value among the input data sets as noise.

In the robot arm joint friction estimator according to an embodiment of the present invention, the recursive linear regression model unit may drive a least squares recursive linear regression model using a predictive model of Equation (a) below.

Equation (a):

는 마찰토크,

는 관절속도,

는 쿨롬 마찰계수,

,

,

는 점성계수임)(here,

is the friction torque,

is the joint velocity,

is the Coulomb friction coefficient,

,

,

is the viscosity coefficient)

,

,

,

)를 갱신할 수 있다. In the robot arm joint friction estimator according to an embodiment of the present invention, the recursive linear regression model unit reflects the input data set collected in real time by the data collection unit to reflect the friction parameter (

,

,

,

) can be updated.

,

,

,

)를 갱신할 수 있다.In the robot arm joint friction estimator according to an embodiment of the present invention, the recursive linear regression model unit drives a least squares recursive linear regression model using the prediction model of the following equation (b) to drive the friction parameter (

,

,

,

) can be updated.

Equation (b):

는 i번째 데이터 수집 주기에서의 마찰 파라미터 행렬,

는 (i-1)번째 데이터 수집 주기에서의 마찰 파라미터 행렬,

는 i번째 데이터 수집 주기에서의 마찰토크 행렬,

는 i번째 데이터 수집 주기에서의 관절속도

을 포함하는 행렬임)(here,

Is the friction parameter matrix at the ith data collection period,

Is the friction parameter matrix in the (i-1)th data collection period,

is the friction torque matrix in the ith data collection period,

is the joint velocity at the ith data collection period

is a matrix containing)

Robot arm joint friction estimation method according to an embodiment of the present invention,

A data collection step of collecting an input data set in which friction torque values and joint speed values of the robot arm in an arbitrary posture without an external force are matched; a friction parameter estimation step of estimating a friction parameter by driving a recursive linear regression model using the input data set as an input value; and a friction torque estimation step of calculating an estimated friction torque using the estimated friction parameter.

In the robot arm joint friction estimation method according to an embodiment of the present invention, in the data collection step, at least one of the input data sets having the same joint velocity value may be subjected to noise processing.

In the robot arm joint friction estimating method according to an embodiment of the present invention, in the friction parameter estimating step, a least square recursive linear regression model may be driven using a predictive model of Equation (a) below.

Equation (a):

는 마찰토크,

는 관절속도,

는 쿨롬 마찰계수,

,

,

는 점성계수임)(here,

is the friction torque,

is the joint velocity,

is the Coulomb friction coefficient,

,

,

is the viscosity coefficient)

,

,

,

)를 갱신할 수 있다. In the method for estimating joint friction of a robot arm according to an embodiment of the present invention, the friction parameter estimating step reflects the input data set collected in real time in the data collection step to reflect the friction parameter (

,

,

,

) can be updated.

,

,

,

)를 갱신할 수 있다.In the robot arm joint friction estimation method according to an embodiment of the present invention, in the friction parameter estimation step, the friction parameter (

,

,

,

) can be updated.

Equation (b):

는 i번째 데이터 수집 주기에서의 마찰 파라미터 행렬,

는 (i-1)번째 데이터 수집 주기에서의 마찰 파라미터 행렬,

는 i번째 데이터 수집 주기에서의 마찰토크 행렬,

는 i번째 데이터 수집 주기에서의 관절속도

을 포함하는 행렬임) (here,

Is the friction parameter matrix at the ith data collection period,

Is the friction parameter matrix in the (i-1)th data collection period,

is the friction torque matrix in the ith data collection period,

is the joint velocity at the ith data collection period

is a matrix containing)

Other specific details of implementations according to various aspects of the present invention are included in the detailed description below.

According to an embodiment of the present invention, the control precision of the robot arm can be improved by accurately estimating and updating the friction torque generated in the robot arm in real time.

1 is a flowchart conceptually illustrating a process of calculating friction torque in a conventional method and reflecting it to robot arm control.
2 is a flowchart conceptually illustrating a process of estimating friction torque in the apparatus for estimating joint friction of a robot arm according to an embodiment of the present invention and reflecting it to control of a robot arm.
3 is a schematic schematic diagram of a robot arm.
4 is a block diagram illustrating an apparatus for estimating joint friction for a robot arm according to an embodiment of the present invention.
5 is a diagram showing equations related to real-time update of friction parameters of an apparatus for estimating joint friction of a robot arm according to an embodiment of the present invention.
6 is a flowchart illustrating a method for estimating joint friction of a robot arm according to an embodiment of the present invention.
7 is a diagram illustrating a computing device according to an embodiment of the present invention.

Since the present invention can apply various transformations and have various embodiments, specific embodiments will be exemplified and described in detail in the detailed description. However, it should be understood that this is not intended to limit the present invention to specific embodiments, and includes all transformations, equivalents, and substitutes included in the spirit and scope of the present invention.

Terms used in the present invention are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly dictates otherwise. In the present invention, terms such as 'include' or 'having' are intended to designate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, but one or more other features It should be understood that the presence or addition of numbers, steps, operations, components, parts, or combinations thereof is not precluded.

First, with reference to FIG. 1, a process of calculating friction torque in the conventional method will be described. 1 is a flowchart conceptually illustrating a process of calculating friction torque in a conventional method and reflecting it to robot arm control.

Referring to FIG. 1 , conventionally, friction torque is calculated by applying a predetermined friction parameter to a friction model, and the calculated friction torque is transmitted to a controller. Then, the control unit controls the operation of the robot arm by reflecting this to the robot dynamics.

However, since the joint friction generated in the joints of the robot arm is a physical property that varies depending on operating environment and conditions such as operating time and operating temperature, robot arm control based on friction torque calculated using predetermined fixed friction parameters. has a problem of not showing the actual expected operating characteristics. That is, there is a problem of limiting the performance of functions (direct teaching, collision detection, stiffness/force control, position control, etc.) of the robot dynamics-based collaborative robot due to inaccurately calculated joint friction.

Accordingly, the inventors of the present invention made it possible to improve the control precision of the robot arm by accurately estimating the frictional torque generated in the robot arm in real time.

Hereinafter, an apparatus for estimating joint friction for a robot arm according to an embodiment of the present invention will be described with reference to FIGS. 2 to 4 . 2 is a flowchart conceptually illustrating a process of estimating friction torque in a robot arm joint friction estimator and reflecting it in robot arm control according to an embodiment of the present invention, and FIG. 3 is a schematic diagram of a robot arm, 4 is a block diagram illustrating an apparatus for estimating joint friction for a robot arm according to an embodiment of the present invention.

)의 함수로 표현되는 마찰모델을 이용하여 실시간으로 가변하는 마찰 파라미터를 추정한다.Referring to FIG. 2, the robot arm joint friction estimator according to an embodiment of the present invention does not use a predetermined fixed friction parameter, but joint velocity (

) is used to estimate friction parameters that change in real time using a friction model expressed as a function of

Equation (1):

는 마찰토크,

는 관절속도,

는 쿨롬 마찰계수(Coulomb friction coefficient),

는 점성계수(Viscous coefficient)이다. i값은 1, 2, 3이나 이에 한정되지 않고, n(≥4)이 될 수 있다.here,

is the friction torque,

is the joint velocity,

is the Coulomb friction coefficient,

is the viscosity coefficient. The value of i is not limited to 1, 2, or 3, and may be n (≧4).

Expanding Equation (1) gives the following Equation (2).

Equation (2):

On the other hand, the general dynamics equation of the robot arm is expressed as a link-side dynamics equation such as Equation (3) below and a motor-side dynamics equation such as Equation (4).

Equation (3):

Equation (4):

는 링크 관성 토크,

는 링크 코리올리 및 구심력 토크,

는 링크 중력 토크,

는 관절 토크,

는 외력 토크,

는 모터 관성 토크,

은 모터 토크이다.

은 하기 식 (5)와 같이 모터의 전류를 측정하여 산출될 수 있다. here,

is the link inertial torque,

is the link Coriolis and centripetal torque,

is the link gravity torque,

is the joint torque,

is the external force torque,

is the motor inertial torque,

is the motor torque.

can be calculated by measuring the current of the motor as shown in Equation (5) below.

Equation (5):

는 모터 토크 상수(Nm/A)이고,

는 모터 전류(A)이다.here,

is the motor torque constant (Nm/A),

is the motor current (A).

는 관절 토크 센서에 의해 측정될 수 있으므로, 식 (4)만을 이용하여 마찰 토크를 산출할 수 있다.If there is a joint torque sensor,

Since can be measured by the joint torque sensor, the friction torque can be calculated using only Equation (4).

의 값을 알 수 없으므로, 식 (3)과 식 (4)를 합하여 하기 식 (6)을 도출한다.If there is no joint torque sensor,

Since the value of is unknown, the following equation (6) is derived by combining equations (3) and equations (4).

Equation (6):

= 0 이므로, 하기 식 (7)과 같다.In Equation (6), when there is no external force,

= 0, it is the same as the following formula (7).

Equation (7):

은 모터 전류를 측정하여 얻을 수 있고,

는 엔코더의 측정값으로 얻을 수 있고,

은

를 미분하여 얻을 수 있으며,

은 엔코더 측정값을 두번 미분하여 얻을 수 있다. 그러나, 엔코더 측정값을 두번 미분하여

을 구할 경우, 노이즈가 심하게 되어 부정확하게 되므로, 로봇아암을 제어하는 제어부에서 산출하는 모션 계획값(

)으로

을 대체한다. 모션 계획값은, 작업자가 입력한 타겟 위치, 관절 속도(통상 최대 속도), 관절 가속도값(통상 최대 가속도)을 수행하기 위해 로봇아암의 제어부가 자체적으로 계획하여 제어하는 값이다. 이러한 모션 계획값은 현재 로봇아암 분야에서는 잘 알려진 기술 내용이므로 상세한 설명은 생략한다.here,

can be obtained by measuring the motor current,

can be obtained from the measured value of the encoder,

silver

can be obtained by differentiating

can be obtained by differentiating the encoder measurement twice. However, by differentiating the encoder measurement twice,

If is obtained, the noise becomes severe and inaccurate, so the motion planning value calculated by the control unit that controls the robot arm (

)by

replace The motion planning value is a value that the control unit of the robot arm plans and controls by itself in order to perform the target position, joint speed (normally maximum speed), and joint acceleration value (normally maximum acceleration) input by the operator. Since these motion planning values are well-known technical contents in the current robot arm field, a detailed description thereof will be omitted.

)을 반영하면, 하기 식 (8)과 같다.Motion planning value in Equation (7) (

) is reflected in the following equation (8).

Equation (8):

은 모터 전류를 측정하여 얻을 수 있고,

는 엔코더의 측정값으로 얻을 수 있고,

은

를 미분하여 얻을 수 있으며,

은 모션 계획값으로부터 얻을 수 있다. 따라서, 관절 토크 센서가 없는 경우에도 마찰토크

의 값을 산출할 수 있게 된다.In equation (8),

can be obtained by measuring the motor current,

can be obtained from the measured value of the encoder,

silver

can be obtained by differentiating

can be obtained from the motion planning value. Therefore, even when there is no joint torque sensor, the friction torque

value can be calculated.

는 관절 토크 센서의 유무에 따라 식 (4) 또는 식 (8)로부터 산출되고,

은 엔코더 측정값

를 미분하여 얻을 수 있다. 본 발명에서는

및

을 제외한 나머지값 (

,

,

,

)들을 추정하며, 추정 대상인 (

,

,

,

)들을 마찰 파라미터라 한다.Referring back to equation (2), in equation (2),

Is calculated from Equation (4) or Equation (8) depending on the presence or absence of the joint torque sensor,

is the encoder reading

can be obtained by differentiating In the present invention

and

The rest except for (

,

,

,

) are estimated, and the estimated target (

,

,

,

) are called friction parameters.

,

,

,

)를 이용하여 추정 마찰토크를 산출하고, 산출된 추정 마찰토크를 로봇아암의 동작을 제어하는 제어부로 전송한다. 제어부는 이를 로봇 동역학에 반영하여 로봇아암의 동작을 제어한다. 도 3 및 도 4를 참조하여 이에 대해 설명한다.And, the estimated friction parameter (

,

,

,

) is used to calculate the estimated friction torque, and the calculated estimated friction torque is transmitted to the control unit that controls the operation of the robot arm. The control unit controls the operation of the robot arm by reflecting it to the dynamics of the robot. This will be described with reference to FIGS. 3 and 4 .

는 관절부(11)에서 감속기 등의 기계적 요소에서 발생하는 관절부(11)에 대한 마찰토크를 의미하고,

은 관절부(11)에 대응하는 구동모터에서 생성되는 모터토크를 의미한다. 도 3에는 일축 로봇아암이 도시되어 있으나, 본 발명은 이에 한정되지 않고 다축 로봇아암도 적용 가능하다.Referring to FIG. 3 , in general, the robot arm 10 includes one or more joint parts 11, and a drive motor (not shown) for operation is disposed on the joint part 11. in Figure 3

Means the friction torque for the joint portion 11 generated from mechanical elements such as reducers in the joint portion 11,

Means motor torque generated by a driving motor corresponding to the joint part 11 . Although a single-axis robot arm is shown in FIG. 3, the present invention is not limited thereto and may also be applied to a multi-axis robot arm.

Referring to FIG. 4 , the robot arm joint friction estimator 100 according to an embodiment of the present invention includes a data collection unit 110, a recursive linear regression model unit 120, and a friction torque estimation unit 130. do.

)과 관절속도값(

)을 매칭한 입력 데이터 세트(

,

)를 수집한다. 예시적으로 구하고자하는 마찰 파라미터는 4개일 수 있으며, 데이터 수집부(110)는 최소한 4개의 입력 데이터 세트를 수집할 수 있다. 물론, 식 (1)에서 i가 n(≥4) 일때는, 데이터 수집부(110)는 최소한 n개의 입력 데이터 세트를 수집한다. 마찰토크값(

)은 토크 센서에 의해 측정되는 값이거나 또는 관절을 구동하는 모터의 전류값으로부터 산출되는 값이고, 관절속도값(

)은 엔코더 측정값

를 미분하여 획득되는 값이다. The data collection unit 110 is the friction torque value of the robot arm in an arbitrary posture without external force (

) and joint velocity values (

) to match the input data set (

,

) are collected. Illustratively, there may be four friction parameters to be obtained, and the data collection unit 110 may collect at least four input data sets. Of course, when i is n (≧4) in Equation (1), the data collecting unit 110 collects at least n input data sets. Friction torque value (

) is a value measured by a torque sensor or a value calculated from a current value of a motor that drives a joint, and a joint velocity value (

) is the encoder measurement

is a value obtained by differentiating

,

)가 많을수록 관절속도값(

)에 대한 추정 마찰토크가 정확하게 산출될 수 있다. 입력 데이터 세트(

,

)에 관절속도값(

)이 일정(즉, 관절가속도값이 0)한 등속 데이터가 포함되어도 무방하나, 관절속도값(

)이 같으면 실질적으로 중복 데이터가 되므로, 등속 데이터를 필터링하여 제거하는 것이 바람직하다. 즉, 선택적으로, 데이터 수집부(110)는 수집된 입력 데이터 세트 중에서 관절속도값이 동일한 입력 데이터 세트 중 적어도 어느 하나는 노이즈 처리할 수 있다.An input data set collected by the data collection unit 110 (

,

), the higher the joint speed value (

) can be accurately calculated. input data set (

,

) to the joint velocity value (

) is constant (that is, the joint acceleration value is 0), it is okay to include constant velocity data, but the joint velocity value (

) is substantially redundant data, it is preferable to filter and remove constant velocity data. That is, selectively, the data collection unit 110 may process at least one of the input data sets having the same joint velocity value among the collected input data sets as noise.

,

)를 입력값으로 재귀선형회귀 모델을 구동하여 마찰 파라미터(

,

,

,

)를 추정한다.The recursive linear regression model unit 120 is an input data set (

,

) as an input value to drive a recursive linear regression model, and the friction parameter (

,

,

,

) is estimated.

A recursive linear regression model is a regression analysis technique that models a linear correlation between a dependent variable (y) and one or more independent variables (or explanatory variables, x) and repeatedly applies data added in real time. . Recursive linear regression models have several use cases, but in the present invention, they can be used for predicting friction parameters that change in real time. If your goal is to predict a value, use linear regression to develop a predictive model that fits your data. The developed linear regression equation can be used to predict y for x values without y. Linear regression models are usually built using the least square method. Of course, linear regression models can be built with other techniques besides the least squares method. Since this linear regression model is currently known, a detailed description thereof will be omitted.

,

,

,

)이다.In an embodiment of the present invention, the predictive model may be expressed as Equation (2), and the prediction target is a friction parameter (which is each coefficient of Equation (2))

,

,

,

)am.

Equation (2):

는 마찰토크,

는 관절속도,

는 쿨롬 마찰계수,

,

,

는 점성계수(Viscous coefficient)이다.

is the friction torque,

is the joint velocity,

is the Coulomb friction coefficient,

,

,

is the viscosity coefficient.

,

)의 개수가 n개인 경우에 대해, 식 (2)는 하기 식 (9)와 같이 행렬식으로 표시될 수 있다.Collected input data set (

,

For the case where the number of ) is n, Equation (2) can be expressed as a determinant as shown in Equation (9) below.

Equation (9):

,

,

,

)를 제외한 나머지 값들은 측정값 또는 입력값들로부터 산출되는 값이므로, 재귀선형회귀 모델부(120)는 식 (9)로 표시되는 최소 자승법 재귀선형회귀 모델을 구동하여 마찰 파라미터(

,

,

,

)를 추정할 수 있다.In equation (9), the friction parameter (

,

,

,

Since the remaining values except for ) are values calculated from measured values or input values, the recursive linear regression model unit 120 drives the least squares recursive linear regression model represented by Equation (9) to drive the friction parameter (

,

,

,

) can be estimated.

,

,

,

)를 이용하여 추정 마찰토크를 산출한다. 그리고, 산출된 추정 마찰토크를 로봇아암의 동작을 제어하는 제어부로 전송한다. 제어부는 이를 로봇 동역학에 반영하여 로봇아암의 동작을 제어한다.The friction torque estimator 130 is a friction parameter (estimated by the recursive linear regression model unit 120).

,

,

,

) to calculate the estimated friction torque. Then, the calculated estimated friction torque is transmitted to the control unit that controls the operation of the robot arm. The control unit controls the operation of the robot arm by reflecting it to the dynamics of the robot.

,

,

,

)를 추정하는데, 입력 데이터 세트(

,

)가 많아질수록 처리해야할 연산량이 많아져서 실시간으로 마찰 파라미터(

,

,

,

)를 추정하기 어려울 수 있다.The recursive linear regression model unit 120 drives the least squares recursive linear regression model represented by Equation (9) to determine the friction parameter (

,

,

,

) to estimate the input data set (

,

) increases, the amount of calculations to be processed increases, so the friction parameter (

,

,

,

) can be difficult to estimate.

,

,

,

)에 현재 수집 주기에서 수집된 입력 데이터 세트를 반영하여 실시간으로 마찰 파라미터(

,

,

,

)를 신속하게 갱신할 수 있다. Accordingly, the recursive linear regression model unit 120 is the friction parameter estimated in the previous collection period in the data collection unit 110 (

,

,

,

) in real time by reflecting the input data set collected in the current collection period (

,

,

,

) can be updated quickly.

That is, the recursive linear regression model unit 120 first estimates the friction parameter using the input data set collected in the first data collection period, and first estimates the friction parameter using the input data set collected in the second data collection period. Secondary estimation (primary update) of friction parameters is performed.

In this way, the recursive linear regression model unit 120 may estimate the (n-1)th updated friction parameter using the input data set collected in the nth data collection period. (nth order estimated friction parameter)

This will be described with reference to equations in FIG. 5 .

First, Equation (9) is expressed as follows.

,

Equation (9'):

,

Here, i is 1, 2, 3 ?? n, which means a data collection period of the data collection unit 110.

,

,

,

)의 행렬

는 하기 식 (10)으로 정리된다.Referring to (a) of FIG. 5, the friction parameter to be obtained (

,

,

,

) matrix

is summarized by the following formula (10).

Equation (10):

를 행렬

라고 한다.A matrix that is part of equation (10)

matrix

It is said.

는 하기 식 (11)로 정리된다.Referring to (b) of FIG. 5, the matrix

is summarized by the following formula (11).

Equation (11):

는 하기 식 (12)로 정리된다.Referring to (c) of FIG. 5, the friction parameter matrix in the nth data collection period

is summarized in the following formula (12).

Equation (12):

When equation (11) is substituted into equation (12), it is organized into the following equation (13). In Equation (13), the subscript n, which means the data collection period, is expressed by replacing it with i.

Equation (13):

)으로부터, i번째 데이터 수집 주기에서의 마찰 파라미터값(

)을 예측하여 갱신할 수 있음을 알 수 있다.Referring to Equation (13), the friction parameter value in the (i-1)th data collection period (

), the friction parameter value at the ith data collection period (

) can be predicted and updated.

5 and Equations (10) to Equations (13) are summarized as Equation (14) below. (Subscript n is expressed by replacing it with i)

Equation (14):

가 추정(또는 갱신)하고자 하는 마찰 파라미터로 구성되는 행렬로서,

와 같이 표시될 수 있다. 따라서, 로봇아암 구동시 미리 설정된 주기 마다 실시간으로 마찰 파라미터(

,

,

,

)가 갱신될 수 있음을 알 수 있다.here,

As a matrix composed of friction parameters to be estimated (or updated),

can be displayed as Therefore, when the robot arm is driven, the friction parameter (

,

,

,

) can be updated.

In this case, when estimating the primary friction parameter, the initial condition may be set as follows.

,

Initial conditions:

,

In the present invention, the robot arm may include the robot arm joint friction estimator 100 according to one embodiment of the present invention described above. The robot arm joint friction estimator 100 may be a physical component of the robot arm or a component of an external controller that controls the robot arm.

Next, a robot arm joint friction estimation method according to an embodiment of the present invention will be described with reference to FIG. 6 . 6 is a flowchart illustrating a method for estimating joint friction of a robot arm according to an embodiment of the present invention.

Referring to FIG. 6 , the robot arm joint friction estimation method according to an embodiment of the present invention includes a data collection step ( S110 ), a friction parameter estimation step ( S120 ), and a friction torque estimation step ( S130 ).

)과 관절속도값(

)을 매칭한 입력 데이터 세트(

,

)를 수집한다. 예시적으로 구하고자하는 마찰 파라미터는 4개일 수 있으며, 데이터 수집부(110)는 최소한 4개의 입력 데이터 세트를 수집할 수 있다.In the data collection step (S110), the friction torque value of the robot arm in an arbitrary posture without external force (

) and joint velocity values (

) to match the input data set (

,

) are collected. Illustratively, there may be four friction parameters to be obtained, and the data collection unit 110 may collect at least four input data sets.

Optionally, in the data collection step (S110), some of the collected data may be filtered to increase processing efficiency. That is, selectively, at least one of the input data sets having the same joint velocity value among the input data sets collected in the data collection step ( S110 ) may be subjected to noise processing.

,

)를 입력값으로 재귀선형회귀 모델을 구동하여 마찰 파라미터(

,

,

,

)를 추정한다. 이때, 하기 식 (2)의 예측 모델을 이용하여 이용하여 최소자승법 재귀선형회귀 모델을 구동한다.Next, in the friction parameter estimation step (S120), the input data set (

,

) as an input value to drive a recursive linear regression model, and the friction parameter (

,

,

,

) is estimated. At this time, a least square recursive linear regression model is driven using the predictive model of Equation (2) below.

Equation (2):

는 마찰토크,

는 관절속도,

는 쿨롬 마찰계수,

,

,

는 점성계수(Viscous coefficient)이다.

is the friction torque,

is the joint velocity,

is the Coulomb friction coefficient,

,

,

is the viscosity coefficient.

는 관절 토크 센서의 유무에 따라 식 (4) 또는 식 (8)로부터 산출되고,

은 엔코더 측정값

를 미분하여 얻을 수 있다.

Is calculated from Equation (4) or Equation (8) depending on the presence or absence of the joint torque sensor,

is the encoder reading

can be obtained by differentiating

Equation (2) can be converted into a least square recursive linear regression model prediction equation such as Equation (9).

Equation (9):

,

,

,

)를 제외한 나머지 값들은 측정값 또는 입력값들로부터 산출되는 값이므로, 식 (9)로 표시되는 최소 자승법 재귀선형회귀 모델을 구동하여 마찰 파라미터(

,

,

,

)를 추정할 수 있다.In equation (9), the friction parameter (

,

,

,

Since the remaining values except for ) are values calculated from measured values or input values, the friction parameter (

,

,

,

) can be estimated.

,

,

,

)를 갱신할 수 있다. 이는 도 5의 수식들과 식 (9) ~ 식 (13)을 참조하여 설명하였으므로, 반복 설명은 생략하며, 도 5의 수식들과 식 (9) ~ 식 (13)를 정리하면 하기 식 (14)과 같다. (아래 첨자 n은 i로 대체하여 표현함)On the other hand, in the friction parameter estimation step (S120), the friction parameter (

,

,

,

) can be updated. Since this has been described with reference to the equations of FIG. 5 and equations (9) to (13), repeated explanations are omitted, and the equations of FIG. 5 and equations (9) to (13) are summarized in the following equation (14). ) is the same as (Subscript n is expressed by replacing it with i)

Equation (14):

가 추정(또는 갱신)하고자 하는 마찰 파라미터로 구성되는 행렬로서,

와 같이 표시될 수 있다. 따라서, 로봇아암 구동시 미리 설정된 주기 마다 실시간으로 마찰 파라미터(

,

,

,

)가 갱신될 수 있음을 알 수 있다.here,

As a matrix composed of friction parameters to be estimated (or updated),

can be displayed as Therefore, when the robot arm is driven, the friction parameter (

,

,

,

) can be updated.

,

,

,

)를 이용하여 추정 마찰토크를 산출한다.In the friction torque estimation step (S130), the estimated friction parameter (

,

,

,

) to calculate the estimated friction torque.

According to the apparatus and method for estimating joint friction of a robot arm according to an embodiment of the present invention as described above, the control precision of the robot arm can be improved by accurately estimating and updating the friction torque generated in the robot arm in real time.

7 is a diagram illustrating a computing device according to an embodiment of the present invention. The computing device TN100 of FIG. 7 may be the robot arm joint friction estimator described in this specification. Alternatively, the computing device TN100 may be a device for performing the robot arm joint friction estimation method described in this specification.

In the embodiment of FIG. 7 , the computing device TN100 may include at least one processor TN110, a transceiver TN120, and a memory TN130. In addition, the computing device TN100 may further include a storage device TN140, an input interface device TN150, and an output interface device TN160. Elements included in the computing device TN100 may communicate with each other by being connected by a bus TN170.

The processor TN110 may execute program commands stored in at least one of the memory TN130 and the storage device TN140. The processor TN110 may mean a central processing unit (CPU), a graphics processing unit (GPU), or a dedicated processor on which methods according to embodiments of the present invention are performed. Processor TN110 may be configured to implement procedures, functions, methods, and the like described in relation to embodiments of the present invention. The processor TN110 may control each component of the computing device TN100.

Each of the memory TN130 and the storage device TN140 may store various information related to the operation of the processor TN110. Each of the memory TN130 and the storage device TN140 may include at least one of a volatile storage medium and a non-volatile storage medium. For example, the memory TN130 may include at least one of read only memory (ROM) and random access memory (RAM).

The transmitting/receiving device TN120 may transmit or receive a wired signal or a wireless signal. The transmitting/receiving device TN120 may perform communication by being connected to a network.

Meanwhile, the present invention may be implemented as a computer program. The present invention can be combined with hardware and implemented as a computer program stored in a computer-readable recording medium.

Methods according to embodiments of the present invention may be implemented in the form of programs readable by various computer means and recorded on computer-readable recording media. Here, the recording medium may include program commands, data files, data structures, etc. alone or in combination.

Program instructions recorded on the recording medium may be those specially designed and configured for the present invention or those known and usable to those skilled in computer software.

For example, recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CDROMs and DVDs, and magneto-optical media such as floptical disks. optical media), and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like.

Examples of the program command may include a high-level language that can be executed by a computer using an interpreter, as well as a machine language generated by a compiler.

These hardware devices may be configured to act as one or more software modules to perform the operations of the present invention, and vice versa.

Although one embodiment of the present invention has been described above, those skilled in the art can add, change, delete, or add components within the scope not departing from the spirit of the present invention described in the claims. The present invention can be variously modified and changed by the like, and this will also be said to be included within the scope of the present invention.

110: data collection unit
120: recursive linear regression model unit
130: friction torque estimation unit

Claims (11)

a data collection unit that collects an input data set in which friction torque values and joint velocity values of the robot arm in an arbitrary posture without external force are matched;
a recursive linear regression model unit for estimating a friction parameter by driving a recursive linear regression model using the input data set as an input value; and,
a friction torque estimation unit that calculates an estimated friction torque using the estimated friction parameters and transmits the calculated estimated friction torque to a control unit that controls an operation of the robot arm;
Including, robot arm joint friction estimator.
The method according to claim 1, wherein the data collection unit,
At least one of the input data sets having the same joint velocity value among the input data sets is subjected to noise processing, the robot arm joint friction estimator.
The method according to claim 1, wherein the recursive linear regression model unit,
A robot arm joint friction estimator for driving a least square recursive linear regression model using the predictive model of equation (a) below.
Equation (a):

(here,

is the friction torque,

is the joint velocity,

is the Coulomb friction coefficient,

,

,

is the viscosity coefficient)
The method according to claim 3, wherein the recursive linear regression model unit,
Friction parameters in real time by reflecting the input data set collected in real time by the data collector

,

,

,

), Robot arm joint friction estimator.
The method according to claim 4, wherein the recursive linear regression model unit,
The friction parameter (

,

,

,

), Robot arm joint friction estimator.
Equation (b):

(here,

Is the friction parameter matrix at the ith data collection period,

Is the friction parameter matrix in the (i-1)th data collection period,

is the friction torque matrix in the ith data collection period,

is the joint velocity at the ith data collection period

is a matrix containing)
performed by a computing device;
A data collection step of collecting an input data set in which friction torque values and joint speed values of the robot arm in an arbitrary posture without an external force are matched;
a friction parameter estimation step of estimating a friction parameter by driving a recursive linear regression model using the input data set as an input value; and,
a friction torque estimation step of calculating an estimated friction torque using the estimated friction parameter;
Including, robot arm joint friction estimation method.
The method according to claim 6, wherein the data collection step,
A method for estimating robot arm joint friction, wherein at least one of the input data sets having the same joint velocity value among the input data sets is subjected to noise processing.
The method according to claim 6, wherein the friction parameter estimating step,
A robot arm joint friction estimation method for driving a least square recursive linear regression model using the predictive model of equation (a) below.
Equation (a):

(here,

is the friction torque,

is the joint velocity,

is the Coulomb friction coefficient,

,

,

is the viscosity coefficient)
The method according to claim 8, wherein the friction parameter estimating step,
In real-time by reflecting the input data set collected in real-time in the data collection step

,

,

,

), a method for estimating robot arm joint friction.
The method according to claim 9, wherein the friction parameter estimating step,
The friction parameter (

,

,

,

), a method for estimating robot arm joint friction.
Equation (b):

(here,

Is the friction parameter matrix at the ith data collection period,

Is the friction parameter matrix in the (i-1)th data collection period,

is the friction torque matrix in the ith data collection period,

is the joint velocity at the ith data collection period

is a matrix containing)
A robot arm comprising the robot arm joint friction estimator according to any one of claims 1 to 5.

Images