JPH0869379A - Method and device for generating rule - Google Patents

Abstract

PURPOSE: To effectively use an existing rule group by facilitating the generation of a rule and correction work by generating a new rule group by synthesizing specific rule groups with common input variables from the existing rule groups. CONSTITUTION: This rule generator includes a rule selector 1, a synthesizer 2, a rule setter 3, memory 4, 5 and a fuzzy inference part 6, etc. The rule selector 1 selects two or more rule groups set as synthesis objects designated from the outside out of (n) existing rule groups, and supplies them to the synthesizer 2. The synthesizer 2 executes a synthesizing arithmetic operation based on a prescribed synthesizing rule stored in the memory 5 for synthesizing rule on the rule group selected by the rule selector 1, and generates the new rule group. In such a case, for example, two rule groups are selected, and fuzzy rules whose antecedent parts coincide with each other are sampled from the rule groups, and the new rule group can be generated by executing prescribed addition or subtraction on them.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a rule generation method for generating a new fuzzy rule using an existing rule group and an apparatus using the rule generation method.

[0002]

2. Description of the Related Art Generally, in fuzzy control, when an input signal is applied to a fuzzy inference unit, a predetermined rule group is activated, and a control operation of a controlled object is determined by inference based on this rule group. The rule group includes a lot of fuzzy rules that are divided into roles to achieve a predetermined control purpose, and these fuzzy rules are usually expressed in a form called if-then rule. Conventionally, the generation of this fuzzy rule is created for each application, and whenever the application is changed, a new fuzzy rule is generated or modified each time.

[0003]

However, it is troublesome for the operator to generate and modify the rule each time the application is changed, and the work load is extremely heavy. Moreover, the existing rule group is only activated for a specific application and is not effectively used for generating a new rule group. The present invention has been made focusing on the above problem, and when a plurality of rule groups for another application already exist, the existing rule group is used to create a rule for a new application. By creating groups, we aim to make effective use of existing rules and to facilitate the work of creating and modifying rules.

[0004]

According to a first aspect of the present invention, there is provided a method for generating a rule group, wherein an input variable is designated from a plurality of rule groups stored in a memory. After selecting two or more rule groups by the selecting means, the selected two or more rule groups are read from the memory by the calculating means and a combination operation is executed to generate a new rule group. To do.

According to a second aspect of the present invention, there is provided a rule generation device using the above method, wherein a memory for storing a plurality of rule groups and two or more rule groups designated by an external source having a common input variable are provided. A selection unit for selecting from the memory and an operation unit for generating a new rule group by reading two or more rule groups selected by the selection unit from the memory and executing a composition operation.

In the invention of claim 3, the arithmetic means extracts a fuzzy rule whose antecedent part coincides with the rule group selected by the selecting means, and executes a synthetic operation to generate a new rule group. To do.

[0007] In the invention of claim 4, the rule group is composed of a plurality of fuzzy rules which have two fuzzy variables in the antecedent part and can be expressed in a matrix.

[0008]

[Action] For a new application, two or more rule groups having common input variables are combined from the existing rule groups to generate a new rule group, so that it is easy to create and modify rules. In addition, the existing rules can be effectively used.

[0009]

1 shows a schematic construction of a rule generation device according to an embodiment of the present invention. A rule selector 1, a synthesizer 2, a rule setter 3, a memory 4,5, and a fuzzy inference unit 6 are shown. Etc. are included. The rule selector 1 selects, from the n existing rule groups stored in the memory 4 for storing rule groups, two or more rule groups to be combined specified externally and gives them to the combiner 2. . The synthesizer 2 generates a new rule group by executing a synthesis operation based on a predetermined synthesis rule stored in the synthesis rule storage memory 5 for the rule group selected by the rule selector 1. The rule setter 3 sets the new rule group generated by the synthesizer 2 in the fuzzy inference unit 6.

In the case of this embodiment, the rule selector 1 selects two rule groups, and the synthesizer 2 extracts the fuzzy rules in which the antecedents coincide with each other from these two rule groups, and performs a predetermined addition / subtraction. A new rule group is generated by performing addition or subtraction based on rules.

2 (1) and 2 (2) show the rule selector 1
The two rule groups selected by are shown in the form of a table. The first and second rule groups both define what kind of operation is to be performed with respect to the two types of inputs θ ₁ and θ ₂ , and for example, the first rule group (see FIG. (Shown), when one input θ ₁ is P _M and the other input θ ₂ is P _M , it means that an output u is set to Z _R. Here, P _M and Z _R are fuzzy labels, and P _L (Positive Large) and P _M (Positive Mi)
dium), P _S (Positive Small), Z _R (Zero), N _S (Neg
_{ative Small), N M (Negative} Midium), N L (Negativ
e Large).

FIG. 3 shows a first rule group on the left side and a second rule group on the right side as if-then statements as a list. According to this list, the first rule group includes nine fuzzy rules, and the second rule group includes seven fuzzy rules. When each of the first and second rule groups is selected by the rule selector 1, the synthesizer 2 reads the data related to the fuzzy rules included in each rule group from the memory 4 and is shown in FIGS. A combination operation is executed according to the addition / subtraction rule.

Here, the addition / subtraction rule means a rule for executing a synthetic operation by addition or subtraction,
4 and 5 show these rules in tabular form. In FIG. 4, a label sequence a on the left end and a label sequence b on the top end are labels in which the labels of the output u to be added or subtracted are arranged, and the composite operation by addition is performed in FIG. 5 and the result of the combining operation by subtraction are shown by labels in FIG. For example, in FIG. 4, when a composite operation by addition is executed for the label P _M and the label P _M , the label of the operation result becomes P _L.

Assuming now that the synthesizer 2 executes a synthetic operation by addition for each of the first and second rule groups, first, a fuzzy rule whose antecedent part matches between these rule groups is the object of the synthetic operation. Is extracted as. In the case of the first and second rule groups shown in FIG. 3, 1 of the first rule group
The first fuzzy rule of the second fuzzy rule and the first fuzzy rule of the second rule group match the antecedent part, and so on.
The 5th and 4th, the 6th and 5th, and the 9th and 7th have the antecedent part respectively, and these 5 sets of fuzzy rules are extracted.

For example, when the first fuzzy rule of the first rule group and the first fuzzy rule of the second rule group are subjected to a combining operation based on the addition / subtraction rule of FIG. 4, the output u of each fuzzy rule is represented. Since the labels are Z _R and P _M , a new label P _M is determined as the result of the combining operation. When similar composition operations are executed for the remaining four sets, new labels P _M , Z _R , N _M , and P _L are obtained.

FIG. 6 is a table showing a new rule group generated by executing these composition operations. In the figure, the shaded area is the label obtained by the combining operation. The other labels are related to the fuzzy rules whose antecedent parts do not match. In this case, the label of the output u of each fuzzy rule is set as it is.

The new rule group generated by using each of the first and second rule groups is expressed by the if-then statement as follows. 1. if θ ₁ = P _M , θ ₂ = N _S then P _M 2. if θ ₁ = P _S , θ ₂ = P _S then Z _R 3. if θ ₁ = N _S , θ ₂ = N _S then Z _R 4. if θ ₁ = N _M , θ ₂ = P _S then N _M 5. if θ ₁ = P _S , θ ₂ = N _S then P _S 6. if θ ₁ = N _S , θ ₂ = P _S then N _S 7. if θ ₁ = P _M , θ ₂ = P _M then P _M 8. if θ ₁ = P _S , θ ₂ = N _M then P _M 9. if θ ₁ = Z _R , θ ₂ = Z _R then Z _R 10. if θ ₁ = N _S , θ ₂ = P _M then N _M 11. if θ ₁ = N _M , θ ₂ = N _M then P _L Of the rule groups, the 1st to 4th fuzzy rules are related to the 1st rule group, and the 5th to 6th fuzzy rules are the 2nd 7 to 1
The first fuzzy rule is based on a synthetic operation.
The rule group generated in this way is used by the rule setter 3
Sent to and the settings for the fuzzy inference section are made.

In the above-described embodiment, a new rule group is generated by executing the combining operation by addition, but it is also possible to execute the combining operation by subtraction according to the addition / subtraction rule shown in FIG. FIG. 7 is a table showing a new rule group generated by executing the combining operation by this subtraction. In FIG. 7, the shaded portion is the label obtained by the combining operation.

In the above embodiment, two rule groups are combined to generate a new rule, but the present invention is not limited to this.
Of course, it is also possible to combine three or more rule groups.
Further, in the above-mentioned embodiment, the synthetic operation by addition and subtraction is used, but other operation methods such as MAX operation and MIN operation can be introduced as required.

[0020]

As described above, according to the present invention, from the plurality of rule groups stored in the memory, two or more rule groups having the same input variable and designated from the outside are selected, and the selected two rule groups are selected. Since the above rule groups are read from the memory and the combining operation is executed to generate a new rule group,
A new rule group can be easily generated for a new application by a simple arithmetic process. As a result, the task of generating and modifying rules is simplified and speeded up, and the existing rule group can be effectively used.

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of a rule generation device according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing, in the form of a table, two rule groups to be combined.

FIG. 3 is an explanatory diagram showing the rules shown in FIG. 2 in the form of a list with if-then statements.

FIG. 4 is an explanatory diagram showing addition / subtraction rules in the form of a table.

FIG. 5 is an explanatory diagram showing addition / subtraction rules in the form of a table.

FIG. 6 is an explanatory diagram showing a combined new rule group in the form of a table.

FIG. 7 is an explanatory diagram showing a combined new rule group in the form of a table.

[Explanation of symbols]

 1 rule selector 2 synthesizer 4 memory

Claims (4)

[Claims] 1. The selecting means selects, from the plurality of rule groups stored in the memory, two or more rule groups which have a common input variable and which are designated from the outside, and then select the two or more rule groups. A rule generation method characterized in that a new rule group is generated by reading a group from the memory by means of a calculation means and executing a combination calculation. 2. A memory for storing a plurality of rule groups, a selection means for selecting from the memory two or more rule groups designated by the outside, which have a common input variable, and two or more selected by the selection means. And a calculation unit for generating a new rule group by reading the rule group from the memory and executing a combination calculation. 3. The calculation unit generates a new rule group by extracting a fuzzy rule whose antecedent part matches with respect to the rule group selected by the selection unit and executing a combination calculation. The described rule generator. 4. The rule generation according to claim 2 or 3, wherein the rule group is composed of a plurality of fuzzy rules having two fuzzy variables in the antecedent part and which can be expressed in a matrix form. apparatus.