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

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

Hi Anh AiViet:

Very nice to get feedback from you.

>   My concern was that whether we have a new "fuzzy" device to realize
> the fuzzy logic instead of the known devices. I have designed few fuzzy
> devices that accept as logic values TRUE,FALSE, I DON'T KNOW 1, I DON'T
> KNOW 2... But that was really a composition of the original logical
> gates.
> So, my point was that we can stay with the binary logic as well.
> The new "invention" is just an implementation achivement not a new
> technology.
>   Am I correct?

Unfortunately,  I don’t know how to understand your last statement. Did 
you refer to fuzzy logic or fuzzy hardware when saying that? As I wrote 
in my previous message, one has to distinguish between design methodology 
and implementation techniques. It would be wrong to judge fuzzy logic 
from the viewpoint of implementation only. 

Fuzzy set theory and fuzzy logic build an independent mathematical 
disciplin and provide a very effective design tool for translating vague 
human knowledge into formal algorithms. This tool can be utilized in 
numerous automated decision making applications in several engineering as 
well as socio-economic areas. We have been talking about fuzzy control. 
Fuzzy control as such is a new DESIGN methodology in control engineering 
and is based on fuzzy logic. What fuzzy control offers is a better user 
interface to the process of translating system insights into controller 
nonlinearities. This is useful since performance improving nonlinearities 
are often underutilized features in conventional control systems. 

And all of the abovementioned have nothing to do with implementation yet.

The implementation aspects jump into discussion as soon as you start 
talking about hardware, chip, device, software et al. Like any other 
algorithm, a fuzzy algorithm is implementable on a standard (analog or 
digital) signal processing platform since it can be broken down to 
classic mathematical operators such as Max, Min, Sum, Difference, 
Multiplication, Division etc. Hence, it is by no means correct to say 
that a fuzzy system MUST be implemented on a fuzzy hardware platform in 
order to make decisions (to infer actions). However, a fuzzy hardware 
platform with a high grade of parallelism would be useful especially when 
it comes to real-time features. I think this is the main motivation for 
special fuzzy hardware, all the more so since fuzzy algorithms require 
much computing performance in general. I doubt that a good fuzzy hardware 
can be build just from binary logic gates as anh AiViet tried to do. It 
rather consists of standard analog circuits carrying out the operations 
Max, Min, Sum, Difference Multiplication, Division etc. I think the most 
important thing which makes the fuzzy hardware distinct from conventional 
ones is the architecture of the chip (or board). I do not know much about 
fuzzy hardware actually, but at least it must contain dedicated hardware 
moduls for fuzzification, inference and defuzzification in order to 
achieve parallel processing capablities. Then, the implementation would 
be more straightforward since a lot of program code, i.e. software, could 
be saved. A problem of fuzzy hardware would be the nummerical accuracy, 
since it depends on analog circuits. 

I can remember some commercial things by name such as fuzzy processor of 
OMRON and PC plug-in card of Togai Infralogic, USA, which are said to be 
based on new hardware architecture. Siemens has been marketing a 
"fuzzified" version of its microcontroller 80C166. This is no new 
hardware architecture, but a general-purpose microcontroller with a 
ROM-based software library containing fuzzy routines.

Sorry if my posting was boring.

Thanh-Tung Truong