JPS6148003A - Controller for robot of multi-joint type - Google Patents

Abstract

PURPOSE:To control the locus of instruction value by simple calculation but with high accuracy by determining directly the compensation magnitude of acceleration in the operation coordinates by the adaptive control operation to assume the signal as the compensation magnitude for the acceleration instruction value determined by the linear control operation. CONSTITUTION:The linear control operation is performed for each operation coordinate at a linear controller 1 by using a locus-instruction value Xr for a multi-joint type robot written in the terms of operation coordinates the response of location X and the response of velocity dX/dt of robot main body 4, thereby calculating acceleration instruction values d<2>Xr1/dt<2> for each operation coordinate are derived as the result. The acceleration instruction value d<2>Xr/dt<2> for robot 4 is determined by adding said accelerator instruction value d<2>Xr1/dt<2> and adaptive control signal d<2>Xra/dt<2>, where said adaptive control signal d<2> Xra/dt<2> is obtained for each operation coordinate by executing model standard controlling by using response XM dXM/dt of a standard model part at adaptive controller 7, and response X and dX/dt of the robot and the deviation (e) and de/dt and locus instruction value Xr. The result is used as the compensation signal for the result of linear control processing d<2>Xr1/dt<2>.

Description

[Detailed Description of the Invention] [Field of Application of the Invention] The present invention relates to a control device for an articulated robot that is widely used as an industrial robot, and particularly to a control device for controlling the trajectory of an articulated robot with high precision. The present invention relates to a control device suitable for.

[Background of the invention]

Generally, articulated robots exhibit strong nonlinear drive characteristics, making it difficult to control their trajectory with high precision. Conventionally, as a method for controlling the trajectory of this type of articulated robot with high precision, there is a procedure of 13th
International Symposium
II Robots
A in (April 1983), 13alestr.
“Adaptive Control” by ino et al.
of Manipulators in t
There is a method proposed in the document entitled "be Ta5k QrientedSpace". This method performs adaptive control in the work coordinate system in which the trajectory command of the articulated robot is written, and as a result, the compensation torque in the work coordinate system is This compensation torque is multiplied by a transposed matrix of the Jacobian matrix representing the relationship between work coordinates and robot joint angle coordinates to convert it into a joint angle drive torque compensation signal.
This signal compensates for nonlinearity in robot drive characteristics and parameter fluctuations of the driven object, thereby increasing the accuracy of robot trajectory control.

However, in this method, the compensation torque for each work coordinate axis is determined by adaptive control, and this result is converted into a drive torque compensation amount by coordinate transformation, and the result is fed back to the result of the linear control system. Generally, in model-based adaptive control, the error between the position and velocity response of a reference model and the robot's response is multiplied by an adaptive control gain, and an adaptive control signal is determined based on the result. In the conventional method, it is necessary to calculate the compensation torque in the work coordinate system using adaptive control, but this requires knowing constants such as the mass of the robot mechanism, making it difficult to design the adaptive control gain. There was a problem. In addition, in the conventional method, it is necessary to convert the compensation torque in the work coordinate system determined by adaptive control into the drive torque compensation amount of the robot joint angle, and there is a problem that additional calculation time is required for this coordinate conversion. There were also points. For example, if the degree of freedom of the robot's work coordinate system is n, this coordinate transformation requires 02 multiplications. Furthermore, this method has the problem that the drive torque of the robot's joints cannot be directly compensated for by adaptive control, and calculation errors associated with coordinate transformation have a sensitive effect on the robot's drive characteristics.

[Purpose of the invention]

The present invention has been made in view of the problems of the conventional method described above, and an object of the present invention is to
An object of the present invention is to provide a control device that highly accurately controls the trajectory of an articulated robot according to command values.

[Summary of the invention]

The present invention directly determines the amount of compensation for acceleration in the work coordinate system through adaptive control calculations, and uses this signal as the amount of compensation for the acceleration command value determined by linear control calculations. to facilitate the design of
This eliminates the need to convert the compensation torque in the work coordinate system to the drive torque compensation amount in the joint angle.

[Embodiments of the invention]

Hereinafter, the present invention will be described in detail based on FIG. 111. FIG. 1 shows the configuration of a control system for an articulated robot according to an embodiment of the present invention.

In general, work commands for articulated robots are given by specifying the position and orientation of several robot nodes marked at the tip of the robot. Normally, such work commands are given by describing the position of the robot node in a stationary orthogonal coordinate system and the posture of the robot node in terms of Euler angles relative to the position.

Generally, a coordinate system that describes work instructions for such an articulated robot is called a work coordinate system. In the present invention, highly accurate trajectory control is achieved by performing adaptive control in this work coordinate system.

In FIG. 1, the linear control unit 1 calculates the trajectory command value X for the articulated robot 4 described in the work coordinate system, and the position and velocity responses X and X of the robot 4 in the work coordinate system.
A linear control calculation is performed for each work coordinate axis using
get. The acceleration command value X for the articulated robot 4 is determined by adding this acceleration command value X and an adaptive control signal Xra, which will be described later. Here it acts as an adaptive compensation signal. In the drive torque calculation section 2,
The acceleration command value X determined by the control process, each joint angle of the articulated robot 4, and the torque U required to drive the robot are calculated. Based on this driving torque command value U, the joint driving unit 3 drives each joint of each articulated robot 4 with the driving torque U. Here, each joint angle 1 of the articulated robot 4 and the detected value θ of its angular velocity are
, θ are converted by the coordinate conversion unit 5 into detected values X and X of the position and velocity in the work coordinate system. As mentioned above, the detected values X and X are the digit 16 of the articulated robot 4. and is fed back to the linear control unit 1 as a speed response.

On the other hand, adaptive control of the articulated robot 4 is performed by a reference model section 6 and an adaptive control section 7.

The reference model section 6 defines desirable response characteristics of the articulated robot 4 in each work coordinate axis. Therefore, by inputting the trajectory command value X into the reference model section 6,
Desired position and velocity responses for each work coordinate axis xM, xM
is obtained. In this embodiment, adaptive control is performed so that the response of the reference model unit 6 and the articulated robot 4 match.

In the adaptive control section 7, the response XM of the reference model section + the response X of the robot, the deviation e from X, e, and the trajectory command value x 1
, robot responses X, X are used to perform model norm adaptive control, and as a result, an adaptive control signal Xt for each work coordinate axis is obtained. As already mentioned, this adaptive control signal xra becomes a compensation signal for the acceleration command value Xrj found as a result of the linear control section 1.

As described above, in the present invention, adaptive control is executed using the response of the reference model in each work coordinate axis as a standard, and as a result, the acceleration compensation amount for each work coordinate axis is calculated, so the shift torque compensation amount is calculated by adaptive control. Compared to calculating, the adaptive control gain can be calculated by 1 without knowing the robot mechanism constants.
Since the jiI adjustment can be made, it becomes easy to design the adaptive control law. Furthermore, since the results of the adaptive control can be directly fed back to the results of the linear control calculation, no extra coordinate transformation is required.

Next, detailed examples will be explained using FIG. 2 and subsequent figures. FIG. 2 shows the mechanism of the articulated robot applied in this embodiment. As an industrial robot, a robot mechanism with 5 or 6 degrees of freedom is used to control the position and posture of a robot hand, but in this example, for ease of explanation, a robot mechanism with 3 degrees of freedom is used. . In FIG. 2, the robot mechanism includes a first link 401, a second link 4
02 and a third link 403, and controls the locus of the tip position tffi,P of the third link 403 according to the work command. Each link of the robot is driven by a motor (not shown), and its rotation angle is θtsuta and θ.
It is expressed as 2・θ3. Also, xyz for robot 4
An orthogonal coordinate system is set as shown in FIG. 2, and this is used as a work coordinate system. That is, the robot tip position P is described in an xyzW coordinate system, and the locus of point P is controlled in this coordinate system.

FIG. 3 shows the configuration of the linear control section 1 in this embodiment.

The linear control unit performs linear control by feeding back the position and velocity for each xyz coordinate axis, and determines the acceleration command value Xrj Hyrj + Zrj for each axis. Here, 1
01, 103, 105 are the transfer functions G, Gt, GtA, G2 .
102, 104°106 is the transfer function Gaff1°G, a, G, of the compensation element of the speed control system.

In this embodiment, each joint angle of the robot mechanism shown in FIG. 2 and its angular velocity θl, θ2, . θ3. θ1. θ2°θ3 is detected and fed back to the control system. Therefore, from these detected values, the position and speed of the two points at the tip of the robot x+)'l
It is necessary to calculate Zl x+ >'+ z. Now, the first robot. Second. Assuming that the length of the third link is At + 12+ As, the calculation formula is (1
) is given as Eq.

X=(t, Sinθ2+ t3stn (θ2+θ3)
Hillθly = [tzsLnθ2+ L 3 S
in (θ2+θ3)) animals θIZ, = LH+ 12c
OSθ2 + t3 CO5 (θ2+03)...
・・・・・・(1) X=J (θ)θ ・・・・・・・・・
・(2) Here, the intersection ""('x+ Y + Q)・
,b=(ha.

i2. θ3 ), and J (3× in Noke Jacobi 77 matrix
It has 3 elements. Let each element of this matrix be Ja, (α.

β=1.2.3), it can be expressed by equation (3).

J ++= (t2sinθ2 + t3 sln (
θ2+03)IcO5θIJ 12= (,!2cos
θ2-1-73cos(θ2+θ3) l 5IfJ
θl, L3=A3cos(θ2+θ3) si
Mouth θ 1J 21 = −(72 sin θ2+ t3
Sin (θ2+03)lsinθ1J22= (t2
coSθz+13anδ(θ2+03) ) C[l
Fiθ1J 23= tsclJs (θ2+03)c
osθ1J31=O J3□=-12sroθ2-13srn (θ2+θ3
J33=-t3sin (θ2+03)...
(3) Through the above calculations, the robot joint angle θ and angular velocity i are also calculated at the position T11 of the robot tip P point. A conversion to x and velocity X is performed. This calculation is executed by the coordinate transformation section 5 in FIG.

The acceleration command value XrA determined by the linear control unit 1 is based on the adaptive control signal X1. The acceleration command value X in each work coordinate axis is compensated by X. The drive torque calculation unit 3 determines a drive torque command value U based on a mathematical model of 44 target robots. The calculation is performed as follows.

1. Robot) If the link mass of the l& structure and its moment of inertia are known, the equation of motion of the target robot mechanism can be derived as follows.

R (θ) θ + C (θ, θ2) + τ, (θ) - u ・・
...(4) Here, θ is the robot joint angle and θ=(
θl.

θ2. θ3)' (t represents transposition), u is the driving torque of each joint "" (uI + u2 + u3)
'is. Still, B, (θ) is 3× which is realized as the inertia matrix.
3 matrix, C(θ, θ) is the apparent force caused by the relative motion of each link, and 3×1 vector, τ, (θ) is 1
It is a 3×1 vector in Katoruk. Using this equation, the driving torque U for operating the robot mechanism at the joint angle quick reading θ can be calculated.

On the other hand, the results of the control calculations in the work east coordinate axes described above are obtained as acceleration command values X in each work coordinate axes. Therefore, the relationship between the acceleration X on the work coordinate axis and the robot joint angular acceleration θ is determined, and using this relational expression and equation (4), the driving torque U required to operate the robot with the acceleration command value seek. As mentioned above, the relationship between the speed X at the working hip joint and the joint angular velocity θ can be expressed by equation (2), and by further differentiating this with respect to time, the following relational equation can be obtained.

X=J(θ)θ+J(θ)θ・・・・・・・・・
(5) From this, the joint angular acceleration command value θ of the robot corresponding to the acceleration command value×2 is given by the following equation.

'tj,=J-1<θ)('X,-J(θ)δ)...
(6) Here, J-'(θ) is the inverse matrix of the Jacobian matrix J(θ). On the other hand, from equation (4),
Driving torque U corresponding to angle addition degree command value θ,
is U, -RVL+c(θ, j2)-tr, (θ)
・・・・・・・・・(7) Calculate it like this. Therefore, using equations (6) and (7), we can move the robot with the acceleration command value
The j driving torque U of the temple is ur −rt (θ+J”<o)< disaster, −:r
<o)j ++c <o, i>z >+r, (θ)
It can be determined by (8). The actual torque calculation unit 2 calculates the joint and drive torque command values U according to equation (8) from the acceleration command value X and the detected values of the robot joint angle and angular velocity of 0.0. Based on this command value u2, the joint drive unit 3 controls the motion of the multi-joint robot 4 by controlling the torque of the motors for driving each joint of the robot.

As described above, the drive torque calculation unit 2 calculates the drive torque assuming that the mechanical constants of the target articulated robot are known. Generally, these mechanical constants vary depending on the sitting state of the robot, and load disturbances other than the drive torque shown in equation (8) also act on each link drive system. These have a negative effect on accurately controlling the trajectory of the robot according to commands. Therefore, such 5
Model-based adaptive control is applied to compensate for imperfections in the rumble torque calculation unit 2.

This model norm adaptive control is composed of a norm model section 6 and an adaptive control section 7 shown in FIG.

The details are shown in FIG.

The part surrounded by a broken line in FIG. 4 is the reference model part 6. As a standard model, the locus of the robot is controlled for each xyz work and each industrial coordinate axis, and the response of the position control system for each axis is set to be a quadratic system. At this time, the response x it of the model to the command direct x to the 1st digit of the X coordinate axis is expressed by the following transfer function.

Xsl a, xGgx 8, 5□. , 8. +. pRGmX °−°゛−
-----(91 Here, Gull r Q * O is the compensation element 60 of the position control system of the X coordinate axis in the standard model 6
1 and the compensation element 602 of the speed control system, which is the transfer function of the compensation element 101, 102 of the linear control section in FIG.
Set to the same constant as . Further, 603 and 504 are integral elements, and in this reference model, the speed response and position response x are determined according to the acceleration command value xM determined by the linear control law.
This shows that M can be determined by a linear relationship. In addition, other work coordinate axes (y-axis.

2 axes) are also set in the same way as the X coordinate axis.

In the present invention, model-norm adaptive control is executed to follow the response of such a reference model. The speed response of the standard model part is XM = (XM + Yw +z M)L
, the position response is `` '' (xM + >'M, zy)
”, and the response of robot 4 is x = (x+ )'+ z)
',x'' (X+lYtZ)t, the state deviation between the two is calculated as follows.

In model-based adaptive control, an adaptive control signal is generated to make this state deviation zero. Regarding the design method of this adaptation side, for example, IdY, D, Landan "4, "Ad
aptive Control, The Mode
几efference Approach J (Ma
rcel I) Ekker Inc.). According to this, when the state deviation is described by the α0 formula,
Adaptive control signal X,.

is calculated as follows. Now, when , it is calculated by ′ ・. Here, ΔKt+=fMV(NXr)'dτ+M'v(Nxr)
t......a4 In particular, D is called an adaptive gain matrix. This adaptive control signal Xt&'' (xral yra l z,, )t
is the acceleration command value X y j” determined by the linear control law.
Perform the adaptation side by adding to (X r A HY t I H2rj )'.

Here, the stability on the adaptive side can be expressed as 4 based on the popov stability condition. By selecting an adaptive gain matrix within a range that satisfies the stability of the adaptive side, it is possible to construct an adaptive control system that asymptotically reduces the deviation between the response of the reference model and the robot to zero.

As mentioned above, according to the present invention, since the acceleration compensation amount is determined by adaptive control in the work coordinate system, the design of the adaptive control system is easier than when calculating the torque compensation amount. In addition, the results of adaptive control can be directly fed back to the linear control system, eliminating the need for calculations such as coordinate transformation.
Highly accurate trajectory control can be performed using simple adaptive control calculations.

Now, in the present invention, the imperfection of the drive torque calculation unit 2 that calculates each joint drive torque based on the result of linear control calculation using the mechanical constants of the robot is corrected by an adaptive control law,
This makes it possible to control the trajectory of an articulated robot with high precision. Therefore, even if the torque calculation unit 2 does not faithfully calculate the mathematical model of the robot shown in equation (8), highly accurate trajectory control is possible by executing adaptive control. Generally 1. In the driving torque calculation formula (8), the terms j(σ)δ and C(θ (jz) are proportional to the square of the joint angular velocity θ of the robot. In normal robot motion, the maximum value of the joint angular velocity is suppressed below a certain value, the square terms of these joint angular velocities are sufficiently small compared to the other terms in equation (2).

Therefore, in the torque calculation unit 2, the square term j of the joint angular velocity
(θ)I! Even if the driving torque is calculated by ignoring i, C(θ, d2), the error caused by this is sufficiently small, and the adaptive control unit 7
Accurate trajectory control can be realized by compensation.

At this time, the calculation of each joint drive torque in the torque calculation unit 2 is executed based on the following equation.

u,=几(θ)J-1(θ)Xr+τ1(θ)...
...a9 Here, FL(0) is an inertia matrix, J-'(θ) is an inverse Jacobian matrix, and x1 is an adaptive fli control signal determined by the adaptive control unit 7 based on the calculation result of the linear control unit 1. The acceleration command value τ1(θ) in the work coordinate axis obtained by adding τ1(θ) is a gravitational torque component acting on each link mechanism. 3 shown in Figure 2
In the degree-of-freedom robot mechanism, the inertia matrix 几(θ) is a 3×3 matrix, and the gravitational torque τ,(θ) is a 3×1 vector, which can be expressed as follows.

Then, each element becomes as shown in equation 09.

R eye=11+a2sin”θ2+ akin2(θ2+
03) +2 a, sinθ2sin (θkoR22=
I 2+ I 3+ a 2+ a 3+ 2 a 4
CO1iθ3R, 33” I 3+ 83 TLzs=R3z=Is+a3+a4rtysθ3RI
2 =R+3=R121''R131=0...
...αη Also, each element can be expressed by (11 formulas).

τ, grave=O rt2= −(a5slII02-1−a6s+n (
θ2+θ3)) gτtm= −a@sin (θ2+θ
3) g......α Yu At this time, I+ is the moment of inertia around the pivot axis of the first link, and I+ (i=2.3) is the moment of inertia around the center of gravity of the second and third links. Moment of inertia, aj (j=2.
...6) is a parameter calculated from constants of the robot mechanism as shown in the following equation.

az= ma11X+ (ms +mv) t,”
a a = ms tt3+mv A3a a = (
mstg3+mwt3) Ax2s=m2t,z-)
-(m3+m1F) t2a 6 =I'l'13 Z
13 + In, Z 3・・・・・・・・・(a) Here, m+ is the mass of the i-th link, A+ is the length of the i-th link, t, and I are the driving end of the i-th link. The distance from 75- to its center of gravity, mv, is the robot load applied to the tip of the third link. Also, (in equation 19, g represents gravitational acceleration.

The inverse Jacobian matrix J-10) is determined as follows by calculating the inverse matrix of the Jacobian matrix J(θ) shown in equation (3).

Then, each element becomes as shown in equation (c).

z2sinθz+A3s+n(θ2+03)J113=
0 ・・・・・・・・・(e) Based on the above calculation formula, the torque calculation in this example is as follows:
Compensating for the inertia matrix and gravitational torque...
...here 1. Drive torque command value U, u, = [13, .

utz, Llr3), the acceleration command value x 7 found as a result of the control calculation is X r = (xv + Yr + Zr
]'.

As described above, the torque calculation unit 2 performs torque calculation consisting only of inertia compensation and gravitational torque compensation based on the acceleration command value X obtained as the calculation result of linear control and adaptive control in the work coordinate system of the articulated robot. By executing the following and driving each joint of the robot according to the results, trajectory control of an articulated robot can be performed with simpler calculations. Furthermore, since the torque calculation formula can be simplified, there is also the advantage that the design of the torque calculation section 2 in the robot control device is facilitated.

In the embodiment described above, each element of the inertia matrix R(θ) that changes as a function of each joint angle of an articulated robot is calculated based on the detected value of each joint angle, and this is used to drive each joint. Determine the torque. On the other hand, current industrial robots are driven by transmitting the output torque of the motor, which is the drive torque generation source, to each link part through a reduction mechanism with a high reduction ratio (about 1/100 to 1/200). There is. The configuration of the drive section at this time is shown in FIG. The rotation angle of the motor 301 is determined by the reduction ratio (1/N++) by the reduction mechanism 302.
The rotation angle of the link mechanism 401 is obtained by decelerating the rotation angle by .

Therefore, the moment of inertia of the link mechanism that makes up the robot is 2 times the reduction ratio when viewed from the motor shaft that drives it.
and decreases in proportion to 1/NH2.

Since each joint drive motor 301 has a unique moment of inertia at its rotation axis, the equivalent moment of inertia J of the robot joint drive system viewed from the motor axis can be expressed as follows.

Jwa* = Jwa + Jt/N)I2
(c) Here, J is the moment of inertia of the motor rotating shaft, and Jt is the moment of inertia of the link mechanism connected thereto. Therefore, when the drive torque of the motor 301 is transmitted to the link mechanism 401 via the reduction mechanism 302 with a high reduction ratio, the equivalent moment of inertia J□ seen from the motor shaft changes as the robot joint angle θ changes. However, it does not change much. For example, in an example of a current industrial robot, the reduction ratio is l/1.
Even if the robot joint angle changes to the maximum at 28,
The change in the equivalent moment of inertia J, converted to the motor shaft is within 1/2 of the moment of inertia J, , of the motor rotation shaft. Therefore, even if the drive torque calculation unit 2 calculates the cooperative torque by assuming that each element of the inertia matrix R(θ) converted to the motor axis remains constant regardless of changes in the robot joint angle θ, the error associated with this is small. , it can be sufficiently compensated by the adaptive control processing of the adaptive control section 7 described above. According to this method, a constant value obtained by adding an appropriate correction value to the motor-specific moment of inertia can be used as the value of the equivalent inertia matrix used for torque meter q4, so each element of the inertia matrix can be calculated in real time. There is an advantage that there is no need to do this, and the calculation time of the drive torque calculation section 2 can be further shortened.

In the embodiments described above, the gravitational torque component acting on the robot link mechanism is calculated using constants such as the mass of the robot mechanism and compensated for in a positive manner. However, as already mentioned, each joint angle of the robot is driven by a drive motor via a speed reduction mechanism.

Therefore, when viewed from the motor shaft, the gravitational torque component used for the link mechanism decreases by the reduction ratio and becomes a load torque for the motor. Now, the gravitational torque τ11 acting on the i-th IJ link of the robot has a motor shaft conversion value of τ
11′, τ, / = τ−+/N old −H+−−−
--(25;) Here, 1/N old is the reduction ratio of the i-th IJ link drive shaft.In the current robot drive system, there is a limit to the output torque of the motor, so 1/1
A speed reduction mechanism of about 1/200 to 1/200 is used.

Therefore, the gravitational torque component can be compensated using the adaptive control law when converted to the motor shaft, so high-precision trajectory control can be achieved without having to calculate it from the robot mechanism constants and the joint angles at that time and compensate forward. .

Further, at this time, the driving torque can be calculated with the inertia matrix being constant. The configuration of the drive torque calculation section at this time is shown in the sixth section.
As shown in the figure. Acceleration command value in the robot's work coordinate axis (value compensated by adaptive control) Xr + )'r
*z, r are the inverse Jacobian matrix J” (θ
), and the robot's joint acceleration command values θrl, ``r2.θ13'' are obtained.

Add to this value the reciprocal of the reduction ratio of each drive shaft, N)II, N
*By multiplying by 2°N ++ 3, the acceleration command value θ1. θ□2. Convert to θ, n3.

In addition to this, the equivalent moment of inertia (constant value) converted to the motor shaft is
Jet I Je2, multiplied by Ls, the motor drive torque command value τ,...! ,τ,,2. τ, a3 are obtained.

In this way, even if the calculation of the drive torque calculation unit 2 is simplified, the incompleteness of the torque calculation calculation can be compensated for by the adaptive control processing. is obtained.

In the embodiment described above, the position X and the speed X in the work coordinate axis of the articulated robot are used as the state variables of the controlled object.
, and these responses are the responses xM, xM of normative model 6
Model-norm adaptive control is performed to match. At this time,
The robot's response has interaction between each work coordinate axis. For this reason, the adaptive control unit 7 executes adaptive control calculations taking these interactions into consideration, as shown in equations (10 to (14). However, as the reference model 6, as shown in FIG. Since a model is used that responds independently to each work coordinate axis, trajectory control that compensates for interactions can be executed by individually performing adaptive control for each work coordinate axis.

FIG. 6 shows the configuration of the adaptive control section at that time. The hand position of a three-degree-of-freedom robot as shown in FIG. 2 is controlled in an xyz orthogonal coordinate system. Here, it is assumed that the reference model section 6 is the same as that shown in FIG. The adaptive control law is executed independently for each work coordinate axis x+Y+z, and each result is an adaptive control signal x, , y, □ for each work coordinate axis.
It becomes 2□. At this time, the adaptive control law is executed as follows. Now, consider the X-axis as the work coordinate axis, and let x and x be the state variables to be controlled. The state deviation between the two is as follows, so the adaptive gain matrix is D=[d18°d2x
] By doing so, we get the excerpt, ・・・・・・・・・(2'l). Now, in the adaptive control calculation in equations (121 to (14)), F=f, )O, F'=f, '>O.

β engineering) 0. M=m〉0゜M'=llll,'> OI N= 1---
(If z di, ΔKp= f f, v, [x, βxx] dr+f,'
v* [x, β8 delusion]・・・・・・・・・(21) Δl(υ=: f m, v , x, d ++m,
'vxx, =----(3o), so
Adaptive control signal xF in the X coordinate axis.

is Xrm"'()°f x V z X dτ) X+f
t'VXX20βz (f f, v!xdt)X+βx
f! 'vxx2+ (,/' rn
r d T ) X y +In x' V x X
The total η is calculated by r''... (31) where (x+ x>'' is the robot's position and speed response on the X coordinate axis, and
This is the trajectory command value on the coordinate axes. In addition, other adaptive constants are defined as (z8X), and the adaptive control gain dI
x, d2! Finally, a selection is made so that the adaptive control is executed stably. In the X-axis adaptation law 717 in FIG.
The adaptive control calculation shown in 31c is executed. Adaptive control calculations are also performed for the other work coordinate axis Y+'' in the same manner as for the X coordinate axis. These blocks are shown at 718 and 719 in FIG. 6, respectively.

According to this embodiment, since the design of the adaptive control law is performed independently for each work coordinate axis, the number of state variables used in the adaptive law is small.
It has the advantage of being easy to design. - This also reduces the time required for adaptive control calculations, making it an effective method for executing these calculations in real time. Furthermore,
This method is useful when the number of work coordinate axes to be controlled increases (for example,
This is particularly effective even in cases where the posture of the wrist is controlled in addition to the position of the hand, such as in a six-degree-of-freedom robot mechanism, since adaptive control laws can be designed independently for each work coordinate axis.

In the several embodiments described above, the case has been described in which the 3-degree-of-freedom robot 'tR4'ip as shown in FIG. Furthermore, the present invention can be similarly applied to robots having a large number of control axes.

[Effect of abandonment]

As described above in detail, according to the present invention, acceleration compensation for each work coordinate axis, l'
By calculating N7 and directly feeding back this adaptive control signal to the linear control system, the trajectory of the robot is controlled with high precision, making it easy to design the adaptive control gain and easily execute the necessary adaptive control calculations. There is an advantage.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing the configuration of a control device according to the present invention, Fig. 2 is a mechanical diagram of an articulated robot to be controlled in an embodiment of the present invention, and Fig. 3 is a linear control diagram in the above embodiment. FIG. 4 is a block diagram showing the configuration of the reference model unit and adaptive control unit in the embodiment, FIG. 5 is a schematic diagram of the configuration of the robot drive unit, and FIG.
The figure is a schematic diagram of the configuration of the drive torque calculation section, and FIG. 7 is a block diagram showing the configuration of the adaptive control section in another embodiment. DESCRIPTION OF SYMBOLS 1... Linear control part, 2... Drive torque calculation part, 4.
・・Articulated robot, 6・Standard model part, 7・・
・Adaptive control unit.

Claims (1)

[Claims] 1. Calculate the acceleration command value for each work coordinate axis by executing a linear control law from the trajectory command value described in the work coordinate system and the position and velocity feedback values for each work coordinate axis of the robot. linear control means; drive torque calculation means for calculating each joint drive torque of the robot from the acceleration command value and detected values of position and velocity at each joint of the robot; In a control device for an articulated robot equipped with a joint driving means for driving joints, a reference model means for defining a response in the work coordinate system of the robot is provided, and a reference model means for specifying a response in the robot's working coordinate system is provided so that the response of the reference model unit matches the response of the robot. 1. A control device for an articulated robot, comprising: an adaptive control means for executing an adaptive control law, and corrects an acceleration command value, which is a calculation result of the linear control section, according to a calculation result of the adaptive control section. 2. A control device for an articulated robot according to claim 1, wherein the drive torque calculation means performs only inertia compensation and gravitational torque compensation. 3. A control device for an articulated robot according to claim 1, wherein the drive torque calculation means calculates the drive torque assuming that the inertia matrix is constant without performing gravitational torque compensation. 4. The control device for an articulated robot according to claim 1, wherein the adaptive control section independently executes the adaptive control law for each work coordinate axis.