JPH036702A - Control parameter deciding method for fuzzy control - Google Patents

Abstract

PURPOSE:To effectively decide a control parameter by setting the control parameter at plural different levels to perform the control and evaluating each control result by a prescribed evaluation rule. CONSTITUTION:A control parameter deciding device 2 sets the control parameter of a fuzzy controller 1 at plural different levels. Based on this control parameter, the controller actually controls the holding power of a robot hand 3 in terms of the holding subjects having various levels of hardness. Then the device 2 obtains each control performance, e.g., the focusing time or the overshoot value, etc., from each control result. This obtained control performance is applied to a prescribed evaluation function for acquisition of each evaluation value. These evaluation results are compared with each other so that a control parameter is selected for acquisition of the best evaluation result. In such a way, the control parameter is effectively decided.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention provides a membership function in fuzzy control,
This invention relates to a method for determining control parameters for fuzzy control laws and other fuzzy controls.

[Prior Art] As a method of controlling a certain physical quantity to a desired value, methods such as PID control and a method in which the controlled object is expressed using a precise mathematical formula and controlled based on the mathematical model have been known. It is being

However, in cases where the controlled object is complex or nonlinear, these controls often fail to provide appropriate control, or the control output is often inferior to that of humans. For example, in the case of certain types of process control, much must be relied on the intuition and experience of skilled operators.

In recent years, as a method to solve such problems, computer processing has been made possible by quantifying human experiential knowledge using fuzzy sets.

Many attempts have been made to use fuzzy control.

Regarding fuzzy control, see “Measurement and Control J, Vol. 22. No. 1. pp84-
86, “Nikkei Mechanical, 198B, No. 6.18,
pp54-65, “Nikkei Electronics J, 19
84.12.3, pp165192 "Factory Automation"'89.1.1)l) 33-38 (1989), etc.

[Problems to be Solved by the Invention] Incidentally, in order to perform fuzzy control, it is necessary to determine control parameters such as membership functions and fuzzy control laws according to control conditions. Furthermore, it is often necessary to adjust and re-determine control parameters that have been determined by degrees depending on changes in control conditions, etc.

However, at present, methods for determining and adjusting these control parameters have not yet been established. Therefore, when actually performing fuzzy control, there is a problem in that a lot of time and effort are spent on determining the control parameters.

The present invention was made against the above-mentioned background, and an object of the present invention is to provide a control parameter determination method for fuzzy control that can determine control parameters extremely efficiently.

[Means for Solving the Problems] The present invention solves the above problems by having the following configurations.

(1) A control parameter determination method for fuzzy control in which a control law is created using a fuzzy set and control is performed based on this control law, the control parameter being set to a plurality of different values to perform control. and evaluating each of the control results according to a certain evaluation rule, comparing these evaluation results, and determining control parameters based on the comparison results.

(2) In the control parameter determination method for fuzzy control according to claim 1, the evaluation rule is an evaluation function predetermined according to a control purpose, and the control evaluation value (control satisfaction degree)
A configuration characterized in that a control evaluation value is obtained from the control performance value obtained from each of the control results using an evaluation function expressed as a function of control performance, and the magnitude of the control evaluation value is compared. .

[Operation] The above configuration (1) performs control by setting the control parameter to a plurality of different values, evaluates each control result using a certain evaluation rule, compares these evaluation results, and uses the comparison result as the result. The control parameters are determined based on the

That is, a plurality of control results are evaluated, the evaluation results are compared, and, for example, a control parameter that provides the best evaluation result is selected, and this is determined as the desired control parameter.

Therefore, for example, if the evaluation rule is defined as in configuration (2), control parameters can be determined extremely efficiently. Furthermore, if a microcomputer or the like is used, this determination step can be relatively easily automated.

[Example] Fig. 1 is a schematic explanatory diagram of a control parameter determination method for fuzzy control according to an embodiment of the present invention, Fig. 2 is an explanatory diagram of a control law used in one embodiment, and Fig. 3 is an illustration of an embodiment. 4 is a schematic diagram of the configuration of the robot hand, FIG. 5 is an explanatory diagram of the evaluation function, and FIG. 6 is a flowchart showing the control parameter determination procedure. Hereinafter, one embodiment will be described in detail with reference to these drawings.

This example uses the control parameters gripping force deviation (E) and deviation change amount (CE ) is an example of determining the respective membership functions (ER and CER).

First, the outline of this embodiment will be explained with reference to FIG. In FIG. 1, reference numeral 1 indicates a fuzzy control device, reference numeral 2 a control parameter determination device, and reference numeral 3 a robot hand.

The fuzzy control device 1 evaluates the state by inputting the gripping force F of the robot hand 3 (see FIG. 4) for gripping objects having various hardnesses, and uses the internal fuzzy control law ( Fuzzy inference is performed using a membership function (see FIG. 2) and a membership function (see FIG. 5), and a control command (amount of change in motor speed) DU corresponding to each grasped object is output. The robot hand 3 that receives the control command DU generates a gripping force F. Note that this fuzzy control device 1 is constructed using a well-known microcomputer or the like.

Here, the control law used in this example is shown in FIG. Second
The control law shown in the figure is a verbal control law that uses the gripping force deviation E and the change CE of the gripping force deviation as input information, and outputs the increase/decrease DU of the control amount.

Here, E, CE, and DU are fuzzy variables (control parameters). Furthermore, the control law is interpreted as follows. For example, the control law defined in the 1st row (E=PB) and 4th column (CE=Z) in FIG. 2 is If(E=PB) and (C
It represents the fuzzy control law E=Z)then DU=PB. This control law then states, ``If the deviation (
If E) is positive and large (PB) and the amount of change in deviation (CE) is about zero (Z), the control law states that the control amount (DU) should be greatly increased (PB)J.

In addition, FIG. 3 shows the ranges of the above-mentioned fuzzy variables E, CE, and DU, respectively, from -ER to +ER1CER to +C
From ER5-DUR to +DUR, N B (Neg
active Bi(1) to P B (Positive
It shows seven membership functions up to e Big. Note that, as described above, this embodiment is an example in which ER and CER are determined among these fuzzy control parameters.

Further, the control parameter determining device 2 is in communication with the fuzzy control device 1, and sets the control parameters in the fuzzy control device 1 to desired values. In this case, the control parameters are determined as follows.

■ The control parameter of the fuzzy control device 1 is set to a certain value (initial value = initial vertex), and using this control parameter, the fuzzy control device 1 actually adjusts the gripping force for objects to be gripped with various hardnesses. Take control.

(2) As the control parameter, several values close to the specific value (initial vertex) are selected, and the same gripping force control as above is performed for each value.

■ From the control results of each control above, each control performance,
For example, the convergence time, overshoot 1, etc. are determined.

(2) Apply the obtained control performance to the evaluation function (see Figure 5) to obtain each evaluation value.

■ Compare the evaluation value of the initial vertex and the maximum value of the evaluation values of points in the vicinity of the initial vertex, and if they are the same or the maximum value is larger, then the point giving this maximum value is the center. Select several points in the vicinity, perform control for each, compare the same as above based on the control results, repeat this until the maximum value becomes smaller, and when the maximum value becomes smaller, that point becomes the new peak.

■ After that, compare the value of the new vertex with the value of the previous vertex, and if the value of the new vertex is large, set the value of the new vertex significantly smaller, and then In the same way, find vertices one after another, and when the value of a new vertex becomes smaller than the value of the previous vertex, determine the value of the previous vertex as the best control parameter value and end. . Incidentally, the above-mentioned steps (1) and (2) constitute evaluation rules in the present invention.

Here, the evaluation function used in this example is shown in Fig. 5(a).
, (b), (c), and (d), the horizontal axis shows the average focusing time (m(ST)), the maximum focusing time (max(ST)), and the overshoot. The average (m(O3)) and the maximum value of overshoot (max(O3)) are taken, and the vertical axis represents the satisfaction level of the focusing time (h) and the satisfaction level of the maximum focusing time (m(ST)), respectively. hmax(ST)), satisfaction with overshoot 1 ((h), m(O8), satisfaction with the maximum value of overshoot (hmax(O3)), each of the illustrated curves a, bc, d
It is expressed as

From this evaluation function, the evaluation value H is determined as follows.

From the control result by the fuzzy control device 1, the m
(S season, max(ST), m(O3), max(O
3), and apply these to each of the evaluation functions described above to obtain hm (8 guns), hmax (8 guns), h
Find each value of m(O3) −hmax(O8). Then, these values are entered into the following evaluation formula to calculate the evaluation value H.

"" (hm (8 guns) max (8 guns)
III (O3) + h 10 (h) + h l11aX (O8) ) / 4- (1) As shown in FIG. ) It is to be grasped with operating fingers 32 and fixed fingers 33.

The operating finger 32 is fixed to a movable piece 35 on which a ball screw mechanism or the like is formed so that it can be moved up and down by the rotation of a threaded rod 34. The threaded rod 34 is coupled to the shaft of a stepping motor 37 via a coupling 36. A strain gauge 38 for detecting the gripping force F is attached to the operating finger 32.

Thereby, by controlling the stepping motor 37 while detecting the gripping force F using the strain gauge 38, the object 31 to be gripped is gripped with a desired gripping force.

Next, the control parameter determination procedure executed by the control parameter determination device 2 will be described in more detail with reference to FIG.

When the control parameter determination operation is started (61), initialization is performed (step 62). Here, the initial values of ER and CER, which are the control parameters to be determined, are set values. Here, in order to perform sufficiently safe control, the initial value is set to ER=20°CER=1.

Next, four values close to the set value are set around the set value (63). That is, if the control parameters ER and CER are now expressed as points in a coordinate system with ER on the vertical axis and CER on the horizontal axis, the point P(0) indicating the set value is p(n). = (ER(n), CER(n
)). In this case, centering on this point P, p
l(n) = (ER(n>, CER(n) 1
)F2(n) = (ER(n) 1 , CER
(n) )F3(') = (ER('), CER
(') +1)F4(') = (ER(n) +1.
CER(n)) Next, the control performance is determined from each control result using the five control parameters described above and applied to the evaluation function, and hm(S□), hm
Determine ax(ST), hm(O3), and hmax(O8), and insert these into the equation (1), respectively, to determine the evaluation value HP4(n) at each point (64).

Next, evaluate values H(n) of four points near the set value.
(n) (n) (n), H, H,
Maximum value of H PI F2 F3 F4, (
n) and the evaluation value H of the point of the set value to determine its magnitude (65).

Maximum value of F3(n) (r+) and evaluation values H and H-
HF4P”)(n)
(n) If the maximum values of H and H are greater (P3 F4) (Y), the process proceeds to step 651. On the other hand, if the evaluation value H is larger (Y), step 6
Proceed to step 6.

In the step 651, the n is incremented by 1 (n=n+1>, and then the evaluation is 4(n-1).
) respectively. Point P1 (n-1) (n-1)
(n-1) (n-1). Nakauriki, Lamae P
2 ・F3 ・F4 Select the point where the evaluation value is maximum, and set the above set value P(1)
is replaced with the point at which this evaluation value shows the maximum and set as a new set value (652), and the process returns to step 63.

(n) In step 65, if the evaluation value H is larger (Y), i is incremented by 1 (i=i+1) in step 66, and then this new vertex P1 is
A comparison is made with the previously obtained vertex P1-1 (68), and P
If i-1>P (Y), this vertex P1-1 is determined as the best vertex 69>, and the control parameter determining operation is ended (70). That is, the values of ER and CEH indicated by the coordinate position values of this vertex are determined as the best control parameters.

Meanwhile, in step 68, P! −1<Pi and P(n
) (N), the value of H is set small and the process returns to step 64 <681>. That is, this continues the search for new vertices.

7 and 8 are diagrams showing an example of the results of actually determining the control parameters by the method of the above-mentioned embodiment. FIG. 7 is an explanatory diagram of the vertex search process, and FIG. It is an explanatory diagram.

In this example, the gripping force that is the control target is 209, and the object to be gripped in each control is based on stiffness.
> is 0.05, 0.1, 0.2, 0.3, 0
．． 4.0.50.6, 0.7, 0.8, 0.9
, 1.0 is used.

In this example, the value of each vertex in the vertex search process is Po (initial value - before determination) = -1,6i.

Pl =0.36. P2 (best vertex) =0.
52. P3 = 0.00. Moreover, the coordinates of the best vertex, that is, the control parameter determination value is ER=1
8, CER=3.

As shown in FIG. 8, the focusing time at the best vertex (vertex P2 after determination) is greatly improved compared to other vertices, while the overshoot is It is within the permissible range.

As mentioned above, in the embodiment, "the value of a vertex given as an initial value is compared with the maximum value of the values of several points in the vicinity of this vertex, and it is determined whether they are the same or the maximum value is better. If it is large, select several points in the vicinity of the point that gives this maximum value, compare it in the same way as above, repeat this until the maximum value becomes smaller, and when the maximum value becomes smaller, Set that point as a new vertex. Next, compare the value of this new vertex with the value of the previous vertex, and if the value of the previous vertex is greater, select this vertex as the best vertex. On the other hand, if the value of the later vertex is larger, the value of the later vertex is set significantly smaller, several points near this point are selected again, and the next vertex search is performed in the same way as above. ” to determine the best control parameters.

Therefore, efficient calculations are possible, the program can be made relatively simple, and the device that executes this procedure can be configured using an IC or the like with relatively small capacity.

In the above embodiment, the operation of finding a vertex from among several points is repeated one after another, and each time a comparison is made with the previous vertex, thereby simplifying the program and the hardware configuration. However, the present invention is not limited to the above-mentioned embodiment, but includes all methods that compare the evaluation values at each point using some method to find the best score. Effects of the Invention As detailed above, the present invention performs control by setting control parameters to a plurality of different values, evaluates each control result using a certain evaluation rule, and compares these evaluation results. Control parameters are determined based on the comparison results, thereby providing a control parameter determination method for fuzzy control that can determine control parameters extremely efficiently.

[Brief explanation of drawings]

FIG. 1 is a schematic explanatory diagram of a control parameter determination method for fuzzy control according to one embodiment of the present invention, FIG. 2 is an explanatory diagram of a control law used in one embodiment, and FIG. 3 is a diagram used in one embodiment. An explanatory diagram of the membership function, Fig. 4 is a schematic diagram of the robot hand, Fig. 5 is an explanatory diagram of the evaluation function, Fig. 6 is a flow diagram showing the control parameter determination procedure, and Fig. 7 is an explanatory diagram of the vertex search process. , FIG. 8 is an explanatory diagram of the control results. DESCRIPTION OF SYMBOLS 1...Fuzzy control device, 2...Control parameter determination device, 3...Robot hand, 31...Gripped object, 32...Moving finger, 33...Fixed fingers 1, 34...
- Threaded rod, 35... Movable piece, 36... Coupling, 37... Stepping motor, 38... Strain gauge. Explanatory diagram 3 of the Menhesi 17° function used in one example
Figure 1: Explanatory diagram of each control used in the embodiment Figure 2: Figure 8

Claims (2)

[Claims] (1) A control parameter determination method for fuzzy control in which a control law is created using a fuzzy set and control is performed based on this control law, the control parameter being controlled by setting the control parameter to a plurality of different values. A control parameter determination method for fuzzy control, comprising: evaluating each of the control results according to a certain evaluation rule, comparing these evaluation results, and determining a control parameter based on the comparison result. (2) In the control parameter determination method for fuzzy control according to claim 1, the evaluation rule is an evaluation function predetermined according to a control purpose, and the control evaluation value (control satisfaction degree)
A fuzzy method characterized in that a control evaluation value is obtained from the control performance value obtained from each of the control results using an evaluation function expressed as a function of control performance, and the magnitude of the control evaluation value is compared. Control parameter determination method.