JPH0327402A - Controller - Google Patents

Abstract

PURPOSE:To obtain sufficient control performance by calculating the deviation of an optimized evaluation function from an optimum value and weighting and averaging the outputs of control units in accordance with the deviation so as to control an object to be controlled. CONSTITUTION:Target values and the output state of an objective plant 13 are respectively inputed to respective controllers 1 by using N-number of controllers which are optimumly designed for different N-number of target values so as to obtain control outputs. Respective obtained control outputs are inputted to a weighted average operation means 11. The weight of the weighted average operation means 11 can be obtained by using the target values and by a weight distribution deciding means 12. Then, the objective plant 13 is controlled by the operation result of the weighted average operation means 11. When the target values are not those which are set, the object to be controlled is controlled by a result obtained by weighting and averaging the outputs of the control units for respective target values which are distributed by the values corresponding to the deviation from the optimum value of the evaluation function. Thus, sufficient control performance can be obtained even if different control algorithms are required for plural target values.

Description

[Detailed Description of the Invention] Industrial Field of Application This invention G1 relates to an automatic control device, in particular a control device that is effective for a controlled object that has a plurality of target values and whose control characteristics vary greatly depending on each target value. It is. An example of a controlled object with such characteristics is a robot or manipulator, which is an example of conventional technology.The conventional technique involves linearly approximating the controlled object and constructing a control device using the approximation result. For example, the method described in Chapter 1 of "Introduction to System Control Theory" by Kogo and Mita. In other words, the control target near the control target value is determined by the amount of minute change in the target value in response to a minute change in the control amount.

Problems to be Solved by the Invention When performing such linear approximation, the ratio of the amount of change changes greatly depending on the target value. Therefore, even if the control device is designed for one target value and sufficient control performance cannot be obtained depending on the target value, the present invention provides a means for solving the problem. Design each control unit for the target value of The output is weighted and averaged, and the controlled object is controlled based on the average result.

Effects Regarding several target values (No. 1) The evaluation function of the control unit is designed to be the optimal value. There is no deviation from the optimal value of the evaluation function at the target value, and the performance as designed is obtained. In addition, when the target value is not the set value, control is performed using the result of a weighted average of the outputs of the control units for each target value, which are distributed according to the deviation from the optimal value of the evaluation function.

At this time, since the evaluation function changes continuously as it deviates from the target value, smooth and accurate control is achieved with respect to the continuous control target connecting each target value. can be done.

Embodiment A control device according to an embodiment of the present invention will be explained based on the drawings. FIG. 1 is a diagram showing the configuration of the device. In other words, by using N controllers 1 that are optimally designed for N different target values, the target value and the target plant 13 are
The output state of is inputted to each controller 1 to obtain a control output. Each of the obtained control outputs is input to the weighted average calculation means 1l. The weight of the weighted average calculating means 11 is obtained by the weight distribution determining means 12 using the target value. The specific method of weight distribution will be described later. The target plant l3 is controlled based on the calculation result of the weighted average calculation means 1l. Furthermore, the control result of the target plant l3, that is, the plant output, is inputted again to each controller l to control the closed loop control system.

Next, the determining method of the weight distribution determining means 12 will be explained. Second
Figure (a) shows an example of the concept of an evaluation function in designing a control device. This shows the response when the state (output) of plant 13 returns to the target value after a deviation. As a method to show this control performance, the target value y ``●
There is a time integral value of the square of the deviation (deviation) of the output y of the plant l3 with respect to t. FIG. 2(a) shows the time integral value of the square of the deviation. FIG. 2(b) shows how the evaluation function changes with respect to a change in the target value, and also shows a method for calculating weight distribution. In the l-th controller 1 designed for the target value X1, as the target value X changes, the evaluation function Jl changes, for example, like a curve Jl(X). According to the curve Jl(x)l, when x=xl, the minimum value J1-+n is obtained. Similarly, with the second controller 1 designed for the target value X2, the Cyo evaluation function curve J2(X) is obtained, hx=x2
When , the minimum value J2-1n is taken. Next, a straight line JIIIn(X) connecting Jl-+n and J2-+n is created. This straight line hypothetically corresponds to the design value when another controller is provided between the two target values XI and X2. The deviations between this straight line and the evaluation functions of the first and second controllers l are defined as ΔJ1 and ΔJ2, respectively, and the weights of the outputs of these controllers l are determined by the ratio of these values. If the outputs of the two controllers 1 are C1 and C2,
The weighted average calculation result, that is, the control output C is as follows.

C=(△J2*C1+△JIC2)/(AJl+△J2
) Therefore, at x=xl, △J 1= 0
becomes ``'QS C-Cl.Well, L x is xi
As it deviates from X2 and approaches X2, ΔJl becomes larger, ΔJ2 becomes smaller, and the weight of c2 gradually becomes higher. This change is continuous and mainly indicates a gradual switch from the first to the second controller operating effectively.

FIG. 3 shows how the weight of each controller 1 is switched depending on the target value. In Fig. 3 (above), when △J is close to zero, control is performed only by the output from the l-th controller 1. At the target value Xl, the evaluation of this controller 1 is shown. Since the deviation from the optimal value of the function is zero, the weight of the first controller 1 is l1 and the weight of the other controllers 1 is zero.When the target value approaches X2, the optimal value of the evaluation function of the second controller l The deviation from the second controller 1 becomes smaller, and the weight of the second controller 1 starts to increase.

When the target value becomes X2, the deviation of the evaluation function of this controller l from the optimum value becomes zero.The weight of the second controller 1 becomes l, and the weights of the other controllers 1 become zero. As the target value deviates from X2, the deviation of the evaluation function of the second controller 1 becomes larger, so the weight of this controller 1 becomes smaller again, and the weights of the other controllers 1 increase.

In the above example, it can be easily inferred that the force exchange in which the target value is set in one dimension can be carried out even if the target value is N-dimensional (N is an integer).Also, in the above example Now, the force t explained using a plurality of controllers can also be implemented by time-sharing processing using a digital computer.

Effects of the Invention As explained above, the present invention provides a control device that can obtain sufficient control performance even when different control algorithms are required for multiple target values, unlike conventional control devices. Yes, the effect is amazing

[Brief explanation of drawings]

Figure 1 is a block diagram showing the configuration of a control device in an embodiment of the present invention. Figure 2 (a) and (b) are blocks showing the method for determining the weight distribution of each controller. FIG. 3 is a diagram showing an example of changes in weights. 1... Controller, 11... Weighted average calculation hand R 13... Target plant.

Claims (1)

[Claims] For some of the target values, the control unit is configured to optimize each evaluation function for each target value, and for the target values other than the optimized target values. In this case, the deviation from the optimum value of the optimized evaluation function is calculated, the output of each control unit is weighted averaged according to the obtained deviation, and the controlled object is controlled based on the average result. A control device characterized by