Patent
1996-09-10
1998-08-04
MacDonald, Allen R.
395 61, G06G 700
Patent
active
057907560
DESCRIPTION:
BRIEF SUMMARY
The present invention refers to an analog fuzzy logic controller with a fuzzification circuit, an inference circuit and a defuzzification circuit, as known from EP 0 489 913 A1.
To an increasing extent, controllers based on fuzzy logic are used for controlling purposes. It is the basic idea of fuzzy logic to assign an input value or an input signal to one of a plurality of linguistic input variables and to determine the degree to which the input signal belongs to one or a plurality of linguistic variables. Depending on the degree to which one or a plurality of input signals belongs to, generally, a plurality of different linguistic variables, control operations of the system to be controlled are triggered. The range within which the input signal or value may vary is divided into a plurality of linguistic variables. If, for example, a control depending on the room temperature is to be made within a range from 0.degree. C. to 50.degree. C., this temperature range could be subdivided into the three linguistic variables "low", "medium" and "high", where the temperature range from 0.degree. C. to 25.degree. C. is assigned to the linguistic variable "low", while the temperature range from 20.degree. C. to 40.degree. C. is assigned to the linguistic variable "medium" and the temperature range from 25.degree. C. to 50.degree. C. is assigned to the linguistic variable "high". Each linguistic variable is associated with a membership function referred to as a semantic rule of the linguistic variable and describing the membership degree of an input signal, in this case a temperature input signal, to the respective linguistic variable. These membership functions are generally trapezoidal or triangular functions realized in a relatively simple manner by analog or digital circuits exhibiting such transfer behavior. It is more advantageous for the fuzzy logic control, however, if the respective membership functions are bell-shaped, since then the relation between the deviation of an input signal from the maximum and the degree of membership of the input signal to the linguistic variable would not be linear and, thus, weighted.
Bell-shaped and trapezoidal, as well as triangular membership functions for fuzzy controls are described, for example, in "Clevere Regler schnell entworfen", Prof. Dr. Ing. Karl Goser, Dipl.-Inf. Hartmut Surmann, Elektronik 6/1992, pages 60 to 68. It is generally known from EP 0 489 913 A1 to realize the membership functions in a fuzzy controller by analog circuits. Moreover, generating z- and s-shaped transfer functions by means of differential amplifier circuits is known from DE 42 43 506 C1.
Generally, 30 to 300 (syntactic) rules are realized in a fuzzy logic controller, depending on the complexity of the control. At least two linguistic variables, i.e. two membership functions, are required for each rule, if one assumes that each rule has an input and an output signal associated therewith. For, e.g., 300 realized rules in a fuzzy logic controller, it is not uncommon that up to 1000 membership functions must be realized by circuitry. Up to the present day, bell-shaped membership functions assigned to the linguistic variables have been realized in circuitry by linearizing the membership function portion by portion using triangular or trapezoidal functions and realizing the individual linear portions by circuitry afterwards. In contrast to the bell-shaped function, where membership degrees other than zero correspond to a wide range of input signals, a triangular or trapezoidal function that is to simulate the bell-shaped function has a comparatively narrow range in which the membership degree is different from zero. Therefore, the value range within which the input signal(s) can fall must be covered by more linguistic variables than is the case for bell-shaped semantic rules, where each linguistic variable is described by a bell-shaped function. This increases the complexity of the controller. Since more of such functions have to be realized by circuitry for triangular or trapezoidal membership fun
REFERENCES:
patent: 5245695 (1993-09-01), Basehore
patent: 5371832 (1994-12-01), Eichfeld et al.
patent: 5515476 (1996-05-01), Nishidai et al.
patent: 5561739 (1996-10-01), Muraji
(Leme et. al, Programmable CMOS Analog Function Synthesis, IEEE ISCS, vol. 2, pp. 1397-1400), May 1993.
(Tsukano et. al, A Design of Current-Mode Analog Circuits for Fuzzy Inference Hardware Systems, IEEE, vol. 2, pp. 1385-1388), May 1993.
Beck Klaus
Guo Shuwei
Peters Liliane
GMD-Forschungszentrum Informationstechnik GmbH
MacDonald Allen R.
White Keith B.
LandOfFree
Analog fuzzy logic controller for posterior alteration of fuzzy does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Analog fuzzy logic controller for posterior alteration of fuzzy , we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Analog fuzzy logic controller for posterior alteration of fuzzy will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-1187562