JPS62154004A - Control system for robot arm - Google Patents

Abstract

PURPOSE:To remove the resonance of an arm and to attain a stable quick control by applying an exact model matching method to the control of a robot arm by servomotor. CONSTITUTION:According to a positive command i(t) being an input signal, an application control error e(t) between an output ym(t) and an output y(t) of an encoder in a model A is obtained. The application control error e(t) is an error dynamic including an unknown parameter theta(t) and a numeric matrix denoted by a state space parameter, and the output e' of the identifier C of said error is obtained from a torque command u(t) and the output y(t) of the encoder. An identification error epsilon(t) is obtained from e'(t)-e(t), and inputted to an application rule computer D, which appropriately controls plural variables k'(t) and h'(t), and feeds back a control signal to a motor amplifier load system B denoted as a plant.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a control method for a robot arm, and particularly to a control method for a robot arm that controls the robot arm at high speed without vibration.

(Prior art) The arrangement of a robot arm driven by a DC servo motor or the like is shown in Figure 6, where Jm
is the motor inertia, JL is the load inertia,
K is the low rigidity spring coefficient that exists between the motor and the load, and C is the friction. The present invention shows that when driving such a low-rigidity load, stable control can be performed by estimating the speed and position of the tip of the mechanical load using an observer and feeding this back to the control system. The applicant proposed this in Japanese Patent Application Laid-Open No. 60-200788. That is, when controlling a load connected by a low-rigidity spring system with a servo motor, a control system with poor damping is formed depending on the value of the spring coefficient. Although it is necessary to install a highly accurate sensor, by estimating these values using an observer, such a sensor is not necessary and the accuracy of control can be improved.

(Problems to be Solved by the Invention) In this way, when driving the load inertia via a low-rigidity spring system, when the resonance frequency of the arm is fo, fo = (1/2π) 77π ...f expressed by (1). The value of is a low value of 10 Hz or less for a normal robot, and the resonant frequency band of the servo system is 20 to 30 Hz.
Therefore, in the conventional control method, there was a problem in that the resonance of the arm could not be eliminated. Therefore, the present invention
This is an attempt to solve these problems of the prior art.

(Means for Solving the Problems) The present invention provides a robot arm control system in which a servo motor connected to a robot arm with a low rigidity load is controlled using a servo amplifier. An observer is configured for the system including the state variables from this servo amplifier to the arm resonance system, and the observed values of the observer are configured to provide state feedback to the servo amplifier, and the robot arm is approximated by The above problems of the prior art are solved by providing a robot arm control method characterized by eliminating resonance.

(Operation) The present invention configures an observer for the system including the resonance system of the robot arm with a low rigidity load, and performs control to stabilize the entire system including the arm resonance system by state feedback. Therefore, arm resonance can be eliminated and stable high-speed control can be performed.

(Example) The present invention will be explained below with reference to the drawings. Figure 2 is a block diagram that approximates the control system of the system including the arm resonance system.
In the figure, U is the current command of the input signal, Kt is the torque constant, Jm is the motor inertia, wm is the motor speed, K
r is an integral constant, 0m is a motor position detected by a rotary encoder, KL is a low-rigidity spring constant, JL is a load inertia, C is friction, W is a load speed, and θ or a load position.

In such a control system, since the servo amplifier is equipped with a sufficiently high-speed motor current control loop, it is assumed that the transfer function of the servo amplifier can be approximated by a torque constant Kt, so that the following equation holds true.

(0m(s)/u(s))=((KrKt)mu1ml
X(sz +2ζwo s+w(,” )/(s'
+2ζw(,s 3+(++a) WO” s ”+
2αζw o 3s)
-(2) However, wo=fi〒U7GO 了-・(3) α
=JL/Jm...(4)ζ=(C/
2) Xi/π]ζ ... (5).

In equation (2), Wo, α, and ζ are parameters that vary depending on the posture of the arm, so in the present invention, a desirable system model is set to perform stable control according to the exact model matching method. exact model match
ng is, for example, "Measurement and Control" June 1980 issue ~ 12
As introduced in the monthly issue, it is a method of designing control systems, and the reference model when performing adaptive control is a transfer function whose properties completely satisfy the given specifications. Assume that it is a closed-loop system, ie, a desired transfer function for a plant controller. And exact m
Odel matching is defined as a calculation method for finding a controller that matches a closed-loop transfer function to a reference model.

In the present invention, as a model of a desirable system using this method,
For example, let the normative model t d(s) be t d(s) =
4,000/((g+40)(s+Ioo))
... (8) is defined. This assumes a system having a step response with a time constant of 25 m5 as shown in FIG.

Incidentally, the control system shown in FIG. 2 can eliminate the resonance of the arm by performing adaptive control using the system shown in FIG. 4. In other words, an observer is configured for the system including the arm resonance system, and control is performed to stabilize the entire system including the arm system by state feedback.Here, the load inertia JL due to the arm posture is To respond to fluctuations, observers are configured with adaptive observations.

Now, exact mode1m in the frequency domain
Since the attaching system is generally represented by a block diagram as shown in FIG. 5, in the present invention, a specific control system for multivariable adaptive control as shown in FIG. 1 is constructed. That is, FIG. 5 shows that the pole arrangement in the frequency domain is the sum of the observer and state feedback, and k(s), h(s), and q(s) represent polynomials.

Next, the block diagram of FIG. 1 will be explained.

The adaptive control error e (t) between the output y m (t) of the reference model A obtained using equation (6) and the encoder output y (t) is determined for the position command 1 (t) that is the input signal.

This adaptive control error e (t) is the unknown parameter θ (1)
and the false Zhao dynamics including the numerical matrix Bx° expressed by state space parameters, and the output of the identifier C is
It is obtained from the torque command u (t) and the encoder output y(1). In addition, the identification error ε(1) is “i; (t
) −e (t) and input into the adaptive law calculator. One adaptive law needle has multiple variables i<(t), h
(t) is appropriately controlled and a control signal is fed back to the motor amplifier load system B represented as a plant.

That is, scalar adaptive control is performed on the motor amplifier load system. Note that in FIG. 1, each parameter is set as follows.

k (t) k [kz, kt, k
o] −(7)h(t)” [hs
, h2, hl, hol...(8)□
θ(t) = [-kg, -ζ, -'';:o, -73°-h
2+-hl,-hol・(9)w(t)=
[[P2 /+(P+40)zCP+100))]
u(t).

[p / ((P, 40)2 (P+100)+1
u(t).

[1/((p+to)2 (p+1oo))] u
(t).

[P3/+(P+40)” (P+1oo))]
y(t).

[P/+(P+40)” (P+100)11 Y
(t).

[1/+(P+40)z(P+100))] Y (
t) ]...(lO) + U (tJ X 11/()'+4117
+ l W (tJ −Y ml(tJ・・
・(zl) From this, the parameter vector is 0 (t) = −r [(1/(P+40)21 w
(t)] e (t) / [C+ [(1/(P+40
)2) w(t)]X [(1/CP+40)2)
W(t)] ...(12).

In addition, the differential coefficient S is the Laplace transform operator P
It is replaced with

(Effects of the Invention) As described above, according to the present invention, the robot 7 using the servo motor
- Exact model matc is used to control the system.
Since adaptive control based on hing's method is used, lO
The resonance system of the robot arm below H2 can also be controlled, and high-speed response of the control system can be realized.

[Brief explanation of drawings]

Fig. 1 is a block diagram of the present invention, Fig. 2 is a block diagram approximating the system including the resonance system of the load arm, Fig. 3 is a characteristic diagram,
FIG. 4 is a block diagram explaining the present invention in detail, FIG. 5 is a block diagram explaining control by exact modsl matching, and FIG. 6 is an explanatory diagram explaining the arrangement of the motor and the load. A... Normative model, B... Motor amplifier load system, C
...Error Guinemics Identifier, D...Adaptive Law Calculator.

Claims (2)

[Claims] (1) In a robot arm control method that uses a servo amplifier to control a servo motor connected to a robot arm with a low rigidity load, the mechanical system of the robot arm is approximated by a resonance model, and from this servo amplifier to the arm resonance system. Configure an observer for the system including state variables,
1. A control method for a robot arm, characterized in that resonance of the robot arm is eliminated by providing state feedback to a servo amplifier based on observed values from the observer. (2) The robot arm control method according to claim (1), wherein the observer is an adaptive observer.