JPH05127909A - Fuzzy arithmetic unit - Google Patents

Abstract

PURPOSE:To obtain a fuzzy arithmetic unit which simultaneously realizes the flexibility of a function and the speed-up of inference arithmetic speed by inputting an antecedent part operation result to a consequent part arithmetic circuit so as to execute a consequent part operation. CONSTITUTION:In an inference process called as a MAX-MIN operation in fuzzy inference, the fuzzy arithmetic unit is provided with an MIN arithmetic circuit 1 executing an MINMUM operation the consequent part arithmetic circuit 2 executing the operation called as a consequent part in fuzzy inference and a MAX arithmetic circuit 3 executing a MAXIMUM operation. The operation result of a part called as the antecedent part in fuzzy inference is given from the external antecedent part arithmetic part 4. It obtains an antecedent part operation result mainly from a software processing. The antecedent part arithmetic part 4 can be either software or hardware, and it is suitably selected in accordance with the content of an inference operation. The consequent part operation in fuzzy inference expressed by an IF-THEN rule is executed by the consequent part arithmetic circuit 2.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a fuzzy computer which performs information processing based on fuzzy inference, and a fuzzy arithmetic unit which can be used as a fuzzy controller.

[0002]

2. Description of the Related Art Conventionally, many proposals for fuzzy computers and fuzzy controllers have been made in various documents. For example, JP-A-2-93905, JP-A-2-93904, and JP-A-2-939.
No. 02 publication and the like.

The fuzzy controller found in the above literature uses a software method to realize the fuzzy controller.
It can be roughly divided into two methods using dedicated hardware.

By realizing the fuzzy inference function by software, it is possible to construct an extremely flexible system. Also, by implementing the fuzzy inference function with a hardware circuit, high-speed inference operation becomes possible.

However, the software fuzzy inference has a drawback that the inference operation speed becomes extremely slow when the inference mechanism is complicated or the number of data to be processed becomes large. In addition, the fuzzy inference by hardware has a drawback that it is not flexible and cannot be easily dealt with even when the input / output variable is changed.

Therefore, conventional fuzzy computers and fuzzy controllers have been constructed by either software or hardware depending on the purpose of use or the controlled object.

[0007]

As described above, in the conventional fuzzy computer, fuzzy controller, etc., the fuzzy inference mechanism is constructed by either software or hardware, so that the inference speed can be increased. If it is attempted, the flexibility of the function is reduced, and if the flexibility of the function is pursued, the inference speed is reduced. The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a fuzzy arithmetic unit that simultaneously realizes flexibility of functions and high speed of inference arithmetic.

[0008]

In order to achieve the above object, the present invention provides a fuzzy arithmetic unit for executing fuzzy inference expressed by IF-THEN rule having an antecedent part and a consequent part. A consequent part arithmetic circuit for inputting the arithmetic result of the part, calculating the inference condition of the consequent part, and outputting a predetermined inference result as the calculation result is provided.

[0009]

According to the present invention having the above-described means, the result of the antecedent operation is input to the operation circuit of the antecedent and the operation of the antecedent is performed. That is, at least the consequent operation in the fuzzy inference expressed by the IF-THEN rule is executed by the consequent operation circuit, so the inference operation speed can be increased, and when the antecedent is complicated, The calculation of the conditional part can be processed by software, and flexibility can be realized.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 shows a fuzzy arithmetic unit according to an embodiment of the present invention.

The fuzzy arithmetic unit according to the present embodiment includes a MIN arithmetic circuit 1 for executing a MINMUM operation in an inference process called MAX-MIN operation in fuzzy inference, and an operation called a consequent part in fuzzy inference. It is provided with a condition part arithmetic circuit 2 and a MAX arithmetic circuit 3 which executes a MAXIMUM operation.

In this embodiment, the operation result of the part called the antecedent part in fuzzy inference is given from the external antecedent part operation part 4. The antecedent part arithmetic unit 4 obtains the antecedent part arithmetic result mainly by software processing. The antecedent operation unit 4 may be either software or hardware, and is appropriately selected according to the content of the inference operation.

In the present embodiment, the circuit for executing the MAX-MIN operation of the consequent part is configured as shown in FIG. 2, and the parallel operation is performed for each label (fuzzy variable) per consequent part operation of one rule. It is supposed to be done.

That is, the MIN calculation circuits 1 are provided by the number corresponding to the number of divisions n of the membership function of the consequent part, and the conformance ω of each rule is input with the calculation result of the antecedent part of each rule as an input. To calculate. And the consequent part arithmetic circuit 2
Is provided with membership function generating circuits 10-1 to 10-n that generate a membership function of the consequent part of each rule, and outputs the input from the corresponding MIN operation circuits 1-1 to 1-n. Goodness of fit μi (i = 1
~ N) are output. The MAX unit 3 inputs the goodness of fit μi for each label of the consequent part, and outputs the logical sum of each label as the inference result as a MAXIMUM operation. Next, as an operation of this embodiment, a case of executing the rule shown in FIG. 5 will be described.

First, the calculation performed by the antecedent calculation unit 4 will be briefly described. For example, assuming that IN ₁ in the antecedent part of rule number 1 is 7 ° C., IN ₂ is −8 ° C., and the membership function is as shown in FIGS. 3A and 3B, PB = 0. 5, the calculation result of NB = 0.3 is obtained. Similarly, the antecedent operation for all input variables (m in this embodiment) is executed, and as a result, the operation result of rule number 1, that is, the input of the MIM operation circuit 1 is of n × m order as shown below. Become a matrix.

[0016]

[Equation 1]

The antecedent operation results shown in the above matrix are input to the MIM operation circuits 1-1 to 1-n, respectively, and the MINIMUM operation is applied to each of the inputs to obtain the rule conformance ω _q ^p. (P is the corresponding rule number; p =
1 to L and q are corresponding membership functions; q = 1 to
n) is required.

For example, in the case of rule 1, since the antecedent part calculation results are ω _PB and ω _NB , the smaller value of the two is the goodness of fit ω ₁ ^{1 of} rule ^1. Becomes ω _PB ＞ ω _NB
Then, ω ₁ ¹ = Ω _NB . The output at this time is

[0019]

[Equation 2] Is.

Each membership function generation circuit 10-1 to 10-1
The output represented by the above formula (2) is input to 10-n,
The suitability of each membership function in the consequent part is calculated.
In this embodiment, each membership function generation circuit 10-1
Membership functions μ _{1 to} μ _n are set to 10 to n, respectively.

For example, the membership function generating circuit 10
The membership function μ ₁ set to −1 is shown in FIG.
As shown in (a), the grade value is realized by 8 bits and the platform set is realized by 10 bits. The membership function generation circuit 10-1 includes a MIM calculation circuit 1-1.
Goodness of fit entered from ω ₁ ¹ Therefore, the shaded area shown in FIG. 4 (b) is output to the MAX operation circuit 3 as the fitness of the membership function. Similarly, the fitness of the membership function is obtained from the other membership function generating circuits 10-2 to 10-n as shown in FIG.
It is output to the arithmetic circuit 3.

[0022] In the MAX operation circuit 3, the fitness of each membership function μ ₁ ~μ _n, i.e. obtaining the logical sum of the hatched portion shown in FIG. 4 (b) for each membership function μ ₁ ~μ _n. Then, the same operation is executed for the rule 2, and the logical sum of the rule 1 and the operation result of the rule 2 is obtained. This is repeated up to rule L, and the logical sum of rule L is output as the inference result.

As described above, according to this embodiment, the operation result of the antecedent part, which is composed of either software or hardware, is input to the antecedent part arithmetic circuit 2 to execute the operation of the antecedent part. As a result, even if the antecedent part is complicated, it can be easily realized, and the flexibility is improved as compared with the case where the entire inference mechanism is configured by hardware. Moreover, since the calculation of the consequent part is executed by the consequent part calculation circuit 2, the inference calculation can be speeded up. In particular, since the MAX-MIN operation can be realized by a dedicated arithmetic circuit, an extremely high speed inference operation can be realized when the consequent operation is simplified.

[0024]

As described above in detail, according to the present invention, it is possible to provide a fuzzy arithmetic unit which simultaneously realizes flexibility of function and high speed of inference operation.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a fuzzy arithmetic unit according to an embodiment of the present invention.

FIG. 2 is a fuzzy arithmetic unit MAX-M according to an embodiment.
The block diagram of the part which performs IN operation.

FIG. 3 is a diagram showing a membership function of rule 1 in the antecedent calculation.

FIG. 4 is a diagram showing a membership function of rule 1 in the consequent operation.

FIG. 5 is a diagram showing a specific example of a rule.

[Explanation of symbols]

1 ... MIM arithmetic circuit, 2 ... Consequent part arithmetic circuit, 3 ... MAX
Arithmetic circuit, 4 ... Antecedent arithmetic circuit, 10 ... Membership function generating circuit.

Claims (1)

[Claims] 1. A fuzzy arithmetic unit for executing fuzzy inference expressed by an if-then rule having an antecedent part and a consequent part, inputs an operation result of the antecedent part, and inputs an inference condition of the antecedent part. And a consequent operation circuit that outputs a predetermined inference result as the operation result.