JPH089761Y2 - Fuzzy controller - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Field of Industrial Application) The present invention relates to a control device using fuzzy theory.

(Prior Art) In fuzzy control in which fuzzy theory is applied to process control, as shown in FIG. 8, when variables x ₁ and x ₂ are input to a fuzzy controller, a controlled variable y for controlling the process is obtained.
Is output. At this time, the causal relations that regulate the input / output relations, for example, when the arithmetic rules between fuzzy sets hold and only the input / output variables are ambiguous quantities, a mathematical model that uses a fuzzy number that is a normal number in a fuzzy set is used. use. In this model, as shown in FIG. 9, the fuzzy membership relationship is an isosceles triangle with a vertex having a membership value μ = 1. In this figure, a is the center and C is the ambiguity width. The fuzzy controller processes the input information by fuzzy theory based on membership functions and control rules.

(Problems to be Solved by the Invention) In a fuzzy control device that has been conventionally used, inference is performed using the same membership function even if the input sensitivity changes. Therefore, there is a drawback that the control becomes inaccurate.

The present invention has been made in view of the above circumstances, and an object thereof is to provide a fuzzy control device capable of performing appropriate control by considering the input sensitivity of an input value.

[Constitution of device]

(Means for Solving the Problems) In order to solve the above problems, the present invention is directed to inputting information for process control as an input value and determining the sensitivity of the input value, as shown in FIG. Using the discriminating means, the ambiguity selecting means for selecting the ambiguity used for inference based on the sensitivity, and the ambiguity selected by the ambiguity selecting means, the input value of the process control based on the fuzzy inference is used. It is a gist to have a fuzzy calculation means which outputs as a control amount.

(Operation) As shown in FIG. 2, in the sensitivity curve in which the input sensitivity β (xi) with respect to the input value xi shows the maximum sensitivity (100%) at the input value x ₀ , each point of the input value xi is Figure 3 shows the membership function as a fuzzy set.

Here, when the resolution is 100%, the width of the membership function is 0. That is, the membership function is in a state where there is no set width due to the straight line shown in FIG. When the membership function is a straight line, it means that the input sensitivity is 100%, the reliability of the input value is 1, and it is the same as the control using a normal information input means that does not use fuzzy.

The present invention has a range of ambiguity γ (xi) depending on the input sensitivity.
Focusing on the change of, the inference is performed based on the sensitivity of the input value based on the ambiguity.

(Embodiment) An embodiment of the present invention will be described below with reference to the drawings. This embodiment is an example of adjusting a brake of a vehicle based on information from an image pickup camera provided in the vehicle. FIG. 4 is an external view of this embodiment, FIG. 5 is a captured image, FIG. 6 is a diagram showing a range of ambiguity, and FIG. 7 is an algorithm for calculating an objective value from a control rule and a membership function. FIG.

The configuration of this embodiment will be described with reference to FIG. Reference numeral 1 is a vehicle, and 2 is an image pickup camera such as a CCD camera, which is attached to the front portion of the vehicle 1 and images the traveling direction. An object 3 is located at a distance a from the camera in the traveling direction of the vehicle 1.

 Next, the operation of this embodiment will be described.

First, the principle of this embodiment will be described. Conventionally, in a sensor input fuzzy controller, the detection value detected by the sensor unit is directly input. However, when the focus is set at a certain distance with a camera or the like, the resolution of the image is reduced at other distances, so that the detection accuracy of the object by image processing is reduced. This tendency increases as the object moves away from the focal length. For example, the detection error of an object captured at a focal length of 40 m is 0, whereas it is ± 0 at 100 m.
It becomes ± 0.1m at 3m and 10m. In this embodiment, the detected value is represented by a fuzzy set. That is, at 100 m, the detected value α has a range from α-0.3 to α + 0.3 centering on α,
When α has a reliability of 1, the reliability of α-0.3 or less is 0, and the reliability of α + 0.3 or more is also 0. (See FIG. 6).

Further, when an ultrasonic sensor is mounted on a vehicle to detect the distance to an object around the vehicle, the ultrasonic sensor exists between the vehicle and the object as the vehicle speed increases, that is, the relative speed between the vehicle and the object increases. The flow of air, which is the medium of ultrasonic waves, becomes faster, and the detection accuracy of the distance to the object decreases.

Therefore, in this embodiment, the fuzzy set (fuzzy set) having the detection value including the detection error value detected by the sensor unit as the information amount also has the detection accuracy (reliability) considering the sensor capability such as vehicle speed or resolution. ),
Accurate fuzzy inference can be performed as an input value that is faithful to the sensor detection value.

Now, the camera 2 attached to the vehicle 1 captures an image of the object 3 located at a distance a forward distance a. Fifth
FIG. (A) is an image captured in this way. Pay attention to the G-line of this image. This is for example this camera 1
Is the focus of. Then, as shown in FIG. 5 (b), the resolution β (d) in this image (where d is the front distance of the vehicle) is the maximum in the G-line, and this is set to 100%. The resolution decreases before and after this G-line. Here, when the membership function is taken with the input value being the resolution at d = a and this input value being the fuzzy number of the fuzzy set, a membership function of an isosceles triangle as shown in FIG. 6 (a) is obtained. At this time, the range of ambiguity γ
(A) is given by γ (a) = | 100−β (a) | × C. Here, C is a constant. From this formula β
When (a) becomes small, that is, the input sensitivity becomes low,
It can be seen that the range of ambiguity increases. This is the case relatively far from the G-line in FIG. 5 (a). On the other hand, it can be seen that the resolution becomes higher and the above-mentioned width becomes smaller as it gets closer to the G-line. Thus, it can be seen that the shape of the isosceles triangle of the membership function changes according to the resolution (that is, the input sensitivity), and the width of the ambiguity also changes accordingly.

 Now, determine the control rules as follows.

(1) if d is “Big” and Δd is “Big” t
hen C is “Small” (2) if d is “Small” and Δd is “Smal
l "then C is" Big "where Δd is the distance deviation per unit time, and C is the degree of braking.

The membership function for Δd at d = a is the sixth
It becomes as shown in FIG. 6B, and has an ambiguous width of γ (b). Here, given by b = (a'-a) / t, a'is an input value relating to the previous distance, and t is a sampling period. Further, γ (b) is represented by γ (b) = | 100−β (a) | × C ′, and C ′ is a constant. This γ (b) can also be changed according to the input condition (here, the sensitivity β (a)).

 The processing up to the output is performed as follows.

The antecedent part “d is Big” of the control rule (1) is added to the membership function for d when d = a shown in FIG.
Figure 7 (a) is obtained by superposing. Similarly, FIG. 7 (b) shows the membership function for Δd in FIG. 6 (b) superposed with the antecedent “Δd is Big” of the control rule (1). FIG. 7 (c) illustrates the consequent part "C is Small" of the control rule (1).

Since d is the distance and Δd is the distance deviation per unit time, Δd depends on the membership function of d. In FIG. 7 (a), the membership value μ (α ₁ ) at the intersection of the membership function of d and the membership function of “Big” according to rule ( ₁ ) becomes the maximum value,
The brake depression degree C takes a range (hatched portion) shown in FIG. 7 (c) from this μ (α ₁ ) according to the rule (1). Similarly, when C is "Big" according to the rule (2), it becomes as shown in FIG. 7 (f).

FIG. 7 (g) is obtained from FIGS. 7 (c) and 7 (f) thus obtained, and this inclined portion is calculated as the degree of brake depression.

Therefore, as can be seen from this embodiment, it is possible to accurately obtain the degree of brake depression by using the ambiguity width of the membership function according to the resolution of the image pickup camera.

In this embodiment, the image pickup camera is used as the one whose input sensitivity changes, but the present invention is not limited to this, and any device that changes the input sensitivity with respect to an input value such as an ultrasonic sensor can be used. It may be one. For a frequency sensor or the like, the sensitivity function β (v) according to the vehicle speed v may be set in advance.

[Advantage of the Invention] As described above, according to the present invention, since the inference is performed in consideration of the ambiguity according to the input sensitivity, the accurate fuzzy control can be performed.

[Brief description of drawings]

FIG. 1 is a claim correspondence diagram, FIG. 2 is a diagram showing an example of a sensitivity curve, FIG. 3 is a diagram showing a membership function at an input value corresponding to the sensitivity curve of FIG. 2, and FIG. FIG. 5 (a) is a diagram showing a captured image, FIG. 5 (b) is a diagram showing its resolution curve, and FIG. 6 is a membership function of d and Δd. FIG. 7 is a diagram showing a method of calculating a control amount, FIG. 8 is a diagram showing an input / output relationship of a fuzzy system, and FIG. 9 is a diagram showing a triangular membership function. 1 ... Vehicle 2 ... Imaging camera 3 ... Object

Claims (1)

[Scope of utility model registration request] 1. A sensitivity discriminating means for inputting information for process control as an input value and discriminating the sensitivity of the input value, and an ambiguity selecting means for selecting an ambiguity used for inference based on the sensitivity, A fuzzy control device comprising: a fuzzy computing means for outputting the input value as a control amount of the process control based on fuzzy inference using the fuzzy degree selected by the fuzzy degree selecting means.