Re: Tha('c ma('c: Fuzzy logic

[Thanh Tung Truong <tung.truong@e-technik.tu-chemnitz.de>]

Hi cac anh VHLam, PTuan, AiViet and others,

I would like to clearify a few things exchanged in the most recent 
messages of this thread.

Anh Lam wrote
> >toi lai co' tha('c ma('c nu+~a ve^` fuzzy logic:
> >
> >fuzzy control kha'c vo+'i analogous control o+? die^?m na`o?
> >Bo+?i analogous control cu~ng kho^ng die^`u khie^?n theo
> >nguye^n ta('c digital 0/1.

I find the question is misleading.
In technical terminology of control engineering one has to distinguish 
beween design methods and implementation aspects. The terms "analog" and 
"digital" are used to indicate only the implementation techniques. While 
an analog controller is implemented by means of analog electronic 
circuits, a digital controller is realized by programming a (digital) 
microprocessor system, e.g. microcontroller or digital signal processor, 
which has analog interfaces to handle analog input and output signals. A 
digital controller does not neccessarily be a binary logic controller and 
the other way round. 

Examples of binary logic controllers are relais-type controllers, 
time-optimal controller and sliding-mode controllers. All of these 
controller types can be implemented either in analog or digital form.

Fuzzy control is a new control design method based on fuzzy logic. A 
fuzzy controller can be either analog or digital, too. 

Anh Tuan wrote:
> Su+. thu+.c (IMO) fuzzy control ra^'t tho^ so+ vi` no' co`n chu+a co' 
> ddu+o+.c nhu+~ng features cu?a analog (PID) control nhu+ integral 
> action (nhi`n ve^` qua' khu+') va` derivative action (nhi`n ve^` 
> tu+o+ng lai), ma` chi? bie^'t co' hie^.n ta.i.

This applies only to the so-called fuzzy inference engine which infers 
its output from its inputs using the linguistic rules and fuzzy 
operators. A fuzzy controller may yet very well contain integrators and 
differentiators in the input or output channels. That way a fuzzy PID 
controller can easily be constructed. To my experience, it is in many 
cases relatively easy to design a fuzzy PID controller which outperforms 
a conventional PID controller. This is due to the fact that a fuzzy PID 
controller is inherently nonlinear. Hence, it may be able to adapt well 
to a nonlinear plant. Also, performance constraints applied to linear 
controllers may be broken up by utilizing nonlinear controllers.

My opinion is fuzzy control can be very powerful/helpful when being used 
in combination of conventional techniques. For example fuzzy rules can be 
used to on-line or off-line tune the parameters of conventional 
controllers to get more optimal results.

BTW, the conventional P(roportional)I(integral)D(erivative) controller is 
in its origin also a heuristic controller, more specifically a linear 
one. Its abilities to look into the future and the past are just a rough 
technical implemetation of human decision making process. In fact, 
patterns of proportional, integeral and derivative actions can easily be 
watched in social life. Just consider a scene where a man is hostily 
treating his opponent. Firt of all, he may act after the P(roportional) 
principle, and slams his opponent every time he received an assault from 
him. He may also chose to act after the I(ntegral) principle, thus just 
waits until annoyances have been accumulated up to a certain amount then 
heavily slams. The third option is of course the D(erivative) strategy, 
i.e. he watches his poor opponent very carefully, as soon as the mimic of 
this seems to change, he slams immediately.

Cheers,
ThanhTung Truong