Fuzzy Logic Control Systems

One great barrier that has stood in front of computer programmers is that of finally

realizing a dream of building a computer system that realistically models human thinking.

The ethics of realizing such a dream are widely debated. Many believe it would be an

extremely dangerous thing to accomplish, but that hasn’t stopped many from trying. The

two main systems that have been developed so far that come closest to accomplishing this

goal are neural networks and fuzzy logic control systems. This paper will only concern

itself with the latter.

Fuzzy logic control systems are designed to mimic the approximate reasoning of

human thinking and decision making. Instead of standard computer logic, which is based

on only 1’s and 0’s or true and false, fuzzy logic is based on a more loose set of linguistic rules that are called the knowledge base. The fuzzy control system is designed to mimic the effects of a person controlling the machine. In fact, the knowledge base for the control system is often compiled from the knowledge of a real human expert.

The fuzzy control system has four main parts. The first is a fuzzification unit

which takes regular computer data as input and converts it into fuzzy language. Secondly

there is the expert knowledge base, which contains all the rules that are to be used to

achieve the desired control action from the input. Third is a fuzzy reasoning mechanism,

that uses the fuzzified input and the knowledge base, and performs various fuzzy logic

operations to arrive at a correct control action. This control action is then sent to the last part of the control system, which is a defuzzification unit, which converts the fuzzy control action, back into regular computer data which the machine it is controlling can understand.

Unlike neural networks, fuzzy control systems can not learn from the data they

process. So the initial expert knowledge base has to be pre-programmed into the system.

As stated before, this is normally done by trying to convert the knowledge of a real human expert directly into the fuzzy rules. This can be a tedious and unreliable process of trial and error.

Several other methods have been proposed to more reliably build the needed

knowledge base for the fuzzy control systems. One such method is using Neuro-Fuzzy

systems, which combine the learning capacity of neural networks with the knowledge

representation of fuzzy logic.