JPH04109304A - Adaptive pi control system - Google Patents

Abstract

PURPOSE:To attain control with a high robust property even in a plant with large load variation by determining the gain of a feedforward loop and that of a feedback loop by the operation of a load. CONSTITUTION:The gain of the feedforward loop is changed and converged upon a real value at first. Then, the gain of the feedback loop is determined by utilizing the feedforward gain. Said calculation is executed by a DSP 11, which finds out a current command value I from an angular velocity command Yr and an angular speed feedback value Y based upon these gain values and applies the current command value I to a servo amplifier 21 to control a servo motor 22. Since the feedforward loop is formed to determine the feedforward loop gain and the feedback loop gain and operate a robot or the like, the robust control of the robot can be attained even when the load such as the inertia of the robot is charply changed.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to an adaptive PI control method applying adaptive control.
In particular, it relates to an adaptive PI control method for controlling loads such as robots that have large load fluctuations.

[Background Art] In plants such as robots, systems are generally controlled using a PI control loop with an identified gain. Further, the parameters of these systems are adjusted manually by trial and error.

What is important in the PI control method normally used in robots and the like is the quick response of the system and the robustness that allows it to respond to parameter fluctuations such as inertia fluctuations.

[Problem to be solved by the invention]

However, when a servo system is configured using the PI control method, there is only a feedback loop, so it is necessary to increase the feedback gain in a controlled object that requires quick response. However, if the feedback gain is increased too much, the stability of the servo system will be impaired and there will be a risk of oscillation, etc. Therefore, responsiveness can only be achieved to a certain extent. Control using a feedforward loop is a method of increasing responsiveness while maintaining stability, but since there is no feedback from the target, this method is extremely susceptible to parameter fluctuations of the controlled target.

Further, even for Ai'J targets where quick response is not an issue, it takes time and effort to adjust the parameters of the feedback loop, and it is extremely difficult to set them to optimal values in actual applications.

The present invention has been made in view of the above points, and is an adaptive P that is provided with an adaptive feedforward loop.
The purpose is to provide an I control method.

[A means of promise to solve problems]

In order to solve the above problems, the present invention provides an adaptive PI control method in which the motor is controlled by PI, and operates the plant so that the feedforward loop becomes the yarn threading of the plant. , an adaptive PI control system is provided, characterized in that the gain of the feedforward loop is determined, and the gain of the feedback loop is automatically determined from the gain of the feedforward loop.

Further, there is provided an adaptive PI control method characterized in that the gain of the feedforward loop and the gain of the feedback loop are determined online in accordance with the operation of the plant to operate the plant.

Furthermore, an adaptive P'I control method is provided, wherein the gain of the feedforward loop and the gain of the feedback loop are determined by operating the plant for a certain period of time, and are fixed while operating the plant.

[Effect]

Since the feedforward loop is used as an adaptive loop for threading the plant, a servo system with robustness and adaptability can be obtained even if there are variations in the parameters of the controlled object. Since the feedback loop gain is determined after determining the feedforward loop gain, a desired feedback loop gain can be obtained. In addition, when used as an automatic adjustment loop, it saves manual labor and allows the optimum servo parameters to be set according to the application.

Furthermore, if the plant is operated while determining the feedforward loop gain and the feedback loop gain online, it is possible to operate the plant in a way that always responds to the load.

Further, by operating the plant for a certain period of time and fixing the gain of the feedforward loop and the gain of the feedback loop, the plant can be operated safely even when large disturbances act on it.

〔Example〕

FIG. 2 is a hardware configuration diagram of a robot system that is an example of a plant for implementing the present invention. The host processor 9 is a processor that controls the entire robot. A robot position command is written from the host processor 9 to the shared RAM 10. Note that the ROM, RAM, etc. coupled to the host processor 9 are omitted.

DS that controls the servo motor 22 built into the robot
P (Digital Signal Processor) 11 is R
, according to the system program of 0M12, the shared RAM
The position command of l0 is read at regular intervals.

The DSP II calculates the position deviation based on the difference between this position command and the position feedback from the pulse fooder 23 built into the servo motor 22. Y'vru. Furthermore, this positional deviation is differentiated to calculate the speed deviation.

As will be described later, the gain of the feedforward loop is determined from these velocity deviations, and further, the gain of the feedback loop is calculated from the feedforward gain.

The input torque T of the current loop, that is, the current command value, is calculated from these gains and speed feedback, etc., and a PWM waveform for driving the servo motor 22 is generated from this, and the PWM waveform is sent via the DSL (digital servo LSI) 14. It is sent to the servo amplifier 21.

The servo amplifier 21 receives the PWM command and drives the servo motor 22 in response to the PWM command. Servo motor 22 drives arm 26 via a reduction gear. The servo motor 22 has a built-in pulse coder 23, which sends feedback pulses to D E L
Return to DSPll via 1.4. DSPl, 1
calculates the angular velocity of the servo motor, etc. from this feedback pulse.

Consider a motor model that can be expressed by the following equation with the servo motor 22 as the controlled object.

Y/l=Kt/J *S-! -A (transfer function representation) If this is expressed in state variable representation, the following equation is obtained.

χ1 ” -1 (A/J) *X 1+ (Kt/
J) *1 ¥−χ1 Here, each coefficient has the following meaning.

j Nibrant inertia Kt:) Lux constant A: Dynamic friction coefficient 1: Current command value Y: Angular velocity feedback ) *X 1 + (Kt/J) *1 (1a) N, a>+ (2>*q Considering 1 and solving for current command value 1, 1= (J/KT) * (A, /J -Q 1.) *
X1+ (J/Kt)* (S+q 1)Yku3) Equation (3) is an interconnection expression using state feedback.

Rewrite this as follows.

1=to 1*X1+G* (S+ql)Y However, K1,
G is, K1= (J/KT)* (A/J-q 1)G=
J/Kt, and cannot be directly known.

FIG. 3 is a diagram showing the relationship between the plant and the reverse system. That is, if the inverse system 32 is connected to the plant 31, the transfer function of the entire system becomes 1, and control that is not affected by plant parameter fluctuations becomes possible.

Next, consider the problem of estimating G in equation (4). (
4) Divide both sides of the equation by (S+ql>) and get I/(S+q
If 1>=V X 1/ (S+q 1)-ξ1, then V=1*ξ1+G*Y (5). Here, the absolute model for system identification of the plant is set to Vp=1p*ξ1+Gp*Y. Here, the subscript p indicates an estimated value. Coefficients without the subscript p are actual parameters of the plant.

ε=V−Va, Lyapunov function Jr, Jr=0.
5*ε2.

δJr/δ is set to 1p−−ε*ξ1 δJr/δGp−−ε*Y From this, dK1p/ctt−β*ε*ξ1 dGp/dt2β*ε*Y (6), and β≧0. Old, d J r / d 1 (δJr/δK 1 p) * (dK 1. p/
d, t) + (δJr/δGp) * (dGp/d
t)β*ε2*(ξ]2+Y2)≦D, Jr monotonically decreases and approaches the minimum value O, V=Vp, and the parameter converges to the true value.

Here, as an example of a plant, consider a servo motor speed control system. FIG. 1 is a block diagram of a servo motor speed control system. The angular velocity command Yr is manually input to the adder 44 and the transfer function 41. The transfer function 41 constitutes a feedforward loop, S is a Laplace operator, and ql is a constant. The output of the transfer function 41 is input to the transfer function 42, multiplied by a gain Gp, and the output is input to the adder 43.

Adder 43 adds the outputs from transfer function 47 and sends the output to adder 46 .

On the other hand, the adder 44 subtracts the angular velocity feedback Y from the angular velocity command Yr, and outputs the output to the transfer function 45. Transfer function 45 constitutes a feedback loop. Here, K2 is the integral gain of the feedback loop,
K3 is a proportional gain.

The output (Xdis) of the transfer function 45 is sent to an adder 46 and added to the output of the adder 43 of the feedforward loop. The output of adder 46 is input to a transfer function 48 representing a servo motor. Here, j is the inertia of the system, S
is the Laplace operator, A is the dynamic friction coefficient, and Kt is the torque constant. The output of transfer function 48 is angular velocity Y. Also,
The same angular velocity XI (-Y) is output to the transfer function 47, multiplied by a gain KIP, and sent to the adder 43.

Here, Xdis input from the feedback loop
If we rewrite it as a disturbance, we get Figure 4.

In the system shown in FIG. 4, equation (5) can be rewritten as the following equation by adding Xdis.

V = 1 * ξ 1 + G * Y + CXd i s/ (S + q 1)] Vp = 1
Make it p*ξ1+Gp*Y.

Next, considering Yr to electric current ■, Yr/I
=1/C(Gp*S)+Gp*q 1+ becomes 1al. If we assume that these are equal so that this and the transfer function of the controlled object, Kt/(J*S+A), are the same, we get J/
K t = G p (9a
) A, / K t = G p * q 1. +
K 1 p (9b). Now, considering a feedback only system, Y/I = (2*Kt to K3*Kt+)/CJ*S 2
”, (3*Kt for A+) 2*Kt for S+] Here, if the transfer function of the feedback loop that the adjuster wants to give is 1/(T*S+1) (T: time constant), K 2 = A/ (K t *T) (ql ・Gp+K ], p) * (1/T) (10a
) K3=A/(Kt*J)=Gp/T (10b) That is, first, the gain of the feedforward loop is changed according to equation (6) to converge to the true value. and,
Using that feedforward gain, (10a)
The gain of the feedback loop is determined from the equation and equation (10b). These calculations are performed by the DSPII shown in Fig. 2, and the DSPII calculates the current command value 1 from the angular velocity command Yr and the angular velocity feedback Y using these gains.
is calculated, a current command value is given to the servo amplifier 21, and the servo motor 22 is controlled.

Further, the gain of the feedforward loop and the gain of the feedback loop can be determined at regular intervals to control the servo motor.

However, this type of control may be dangerous because gain convergence may not be successful if the disturbance is large. For this reason,
Operate the robot for a certain period of time, determine the gain of the feedforward loop and the gain of the feedback loop,
It is also possible to operate with these gains fixed.

In this way, a feedforward loop is established,
Since the feedforward loop gain and the feedback loop gain are determined to operate the robot, etc., robust control is possible even if the load such as the inertia of the robot fluctuates significantly.

In the above description, the speed control loop of a robot was used as an example of a plant, but the present invention can also be applied to control of other plants with large load fluctuations.

[Effects of the Invention]- As explained above, in the present invention, a feedforward loop is provided in the motor control loop, and the gain of the feedforward loop and the gain of the feedback loop are determined by driving the load. This enables highly robust control even in plants with large load fluctuations.

In addition, the plant is operated by determining the feedforward loop gain and feedback loop gain online, making it possible to always control the plant with the optimal gain even if plant parameters fluctuate. Become.

Further, since the motor is operated for a certain period of time and the gain of the feedforward loop and the gain of the feedback loop are fixed, safe operation is possible even when a large disturbance occurs.

In particular, by determining this gain every time the application is changed, it becomes possible to set the optimum parameters for the application.

[Brief explanation of the drawing]

Figure 1 is a block diagram of the servo motor speed control system, Figure 2 is a hardware configuration diagram of a robot system that is an example of a plant for implementing the present invention, and Figure 3 is the relationship between the plant and threading. FIG. 4 is a block diagram obtained by rewriting FIG. 3 by using the input from the feedback loop as a disturbance. DSP (Digital Signal Processor) Servo amplifier servo motor arm transfer function (feedback loop) Figure 1

Claims (5)

[Claims] (1) In an adaptive PI control method for PI controlling a motor, a feedforward loop is provided, the plant is operated, and the gain of the feedforward loop is determined so that the feedforward loop becomes a reverse system of the plant. , an adaptive P characterized in that the gain of the feedback loop is automatically determined from the gain of the feedforward loop.
I control method. (2) The adaptive PI control method according to claim 1, wherein the gain of the feedforward loop and the gain of the feedback loop are determined online as the motor operates, and the motor is operated. (3) The adaptive PI control method according to claim 1, wherein the gain of the feedforward loop and the gain of the feedback loop are determined by operating the motor for a certain period of time, and are fixed while operating the motor. (4) The adaptive PI control system according to claim 1, wherein the motor is a motor for a robot, and the feedback loop is a speed control loop. (5) The feedback loop is a first-order delay system,
2. The adaptive PI control method according to claim 1, wherein the proportional gain and the integral gain of the feedback loop are determined from the gain of the feedforward loop by giving a first-order delay time constant.