|
APPROACH TO THE AUTOMATIC GENERATION OF RULES FOR A FUZZY CONTROLLER IN THE PROCESS OF LEARNING WITH A TEACHER
|
|
|
J. Daeibal;N. Sergeev
|
|
|
||
|
|
|
|
1314-2704
|
|
|
||
|
English
|
|
|
20
|
|
|
2.1
|
|
|
|
|
|
||
|
The authors propose an algorithm for automatically constructing rules for control
subsystems, which can be attributed to fuzzy regulators. Numeric variables used to obtain information about the control object and control actions by which a human operator controls the object in the work are presented in the form of linguistic variables. The membership functions of these variables are constructed as a result of an expert survey. The example describes the operation of the simplest odd regulator with readymade rules. Next, the author's algorithm for automatically constructing rules in the learning process is sequentially presented. The controller is trained in the process of monitoring the real actions of the human operator. The challenge is to build a learning mechanism based on a fuzzy controller using production rules. An important feature of the presented learning mechanism is that at each step of the training the meaning of the changes that have occurred was understood. It is important to see that the changes that occur are expressed not only in the change of coefficients or any functions but are expressed in the form of quite understandable sentences of a natural language, for example. Such ?understandability? is necessary when the process of forming a decision or control action is multi-step or multi-stage, when the goal of the process is to form some chain of reasoning, or an autonomous object needs to ?understand? what is actually happening. For simplicity, suppose that we have some informative input quantity and some output quantity, which can be considered as a control action. The difference from the classical regulators is that we do not control (or maintain) the value of the output quantity by influencing the input value through feedback but respond to the value of the input quantity by changing some output quantity, which may not be directly related to the input value. For example, in response to an increase in vehicle speed, we increase the distance to the curb. In this case, to increase the level of travel safety. Those. according to the values of one quantity, we change the values of another quantity for the sake of changing some third. |
|
|
conference
|
|
|
||
|
||
|
20th International Multidisciplinary Scientific GeoConference SGEM 2020
|
|
|
20th International Multidisciplinary Scientific GeoConference SGEM 2020, 18 - 24 August, 2020
|
|
|
Proceedings Paper
|
|
|
STEF92 Technology
|
|
|
International Multidisciplinary Scientific GeoConference-SGEM
|
|
|
SWS Scholarly Society; Acad Sci Czech Republ; Latvian Acad Sci; Polish Acad Sci; Russian Acad Sci; Serbian Acad Sci & Arts; Natl Acad Sci Ukraine; Natl Acad Sci Armenia; Sci Council Japan; European Acad Sci, Arts & Letters; Acad Fine Arts Zagreb Croatia; C
|
|
|
59-66
|
|
|
18 - 24 August, 2020
|
|
|
website
|
|
|
cdrom
|
|
|
6969
|
|
|
Production rules;UAV;Automatic control systems;Systems learning
|
|