Converting Matlab .fis file to C code

[Unknown]

I would like to convert a matlab file with extension .fis, can you explain me how to do this using Fuzylite?

Thanks

Mash

[Unknown]

Hi, I’m new from here and frist thank to Juan for fuzzylite it is great job.

I’m working in a example that could help you but I have a problem to defuzzify the output variable.

1) I’m ussing fuzzylite 5.0 (C++).
2) I have a fis file (created with octave), called fuzzy.fis.

[System]
Name='Direccion-Controlador'
Type='mamdani'
Version=1.0
NumInputs=2
NumOutputs=1
NumRules=6
AndMethod='min'
OrMethod='max'
ImpMethod='min'
AggMethod='max'
DefuzzMethod='centroid'

[Input1]
Name='Error'
Range=[-64 64]
NumMFs=3
MF1='Izquierda':'gaussmf',[35 -64]
MF2='Centro':'gaussmf',[25 0]
MF3='Derecha':'gaussmf',[35 64]

[Input2]
Name='D-Error'
Range=[-1 1]
NumMFs=2
MF1='Negativo':'gaussmf',[1 -1]
MF2='Positivo':'gaussmf',[1 1]

[Output1]
Name='Angulo'
Range=[-45 45]
NumMFs=5
MF1='M-izq':'trapmf',[-46 -45 -30 -15]
MF2='Izq':'trimf',[-30 -15 0]
MF3='Centro':'trimf',[-15 0 15]
MF4='Der':'trimf',[0 15 30]
MF5='M-der':'trapmf',[15 30 45 46]

[Rules]
1 1, 2 (1) : 1
1 2, 1 (1) : 1
2 1, 3 (1) : 1
2 2, 3 (1) : 1
3 1, 5 (1) : 1
3 2, 4 (1) : 1

3) I want to evaluate some input values.

#include <fl/Headers.h>
#include <sstream>
#include <stdio.h>

int main(int argc, char **argv)
{
    fl::Engine* engine = fl::FisImporter().fromFile("/home/ruben/fuzzy.fis");
	
    engine->getInputVariable(0)->setInputValue(15);
    engine->getInputVariable(1)->setInputValue(0.32);

    engine->process();

    //engine->defuzzify();

    fl::OutputVariable* output = engine->getOutputVariable("Angulo");
    
    FL_LOG(output->toString());

    FL_LOG(output->fuzzyOutputValue());
    
    FL_LOG(output->fuzzify(output->getOutputValue()));

}

I have been seaching and trying with different ways but I can’t get the crisp output’s value in float format.

ejemplo.cpp [19]:0.078/Mizq + 0.078/Izq + 0.794/Centro + 0.375/Der + 0.375/Mder
ejemplo.cpp [21]:0.000/Mizq + 0.000/Izq + 0.357/Centro + 0.643/Der + 0.000/Mder
ejemplo.cpp [24]:OutputVariable: Angulo
enabled: true
range: -45.000 45.000
accumulation: Maximum
defuzzifier: Centroid 200
default: nan
lock-previous: false
lock-range: false
term: Mizq Trapezoid -46.000 -45.000 -30.000 -15.000
term: Izq Triangle -30.000 -15.000 0.000
term: Centro Triangle -15.000 0.000 15.000
term: Der Triangle 0.000 15.000 30.000
term: Mder Trapezoid 15.000 30.000 45.000 46.000

If I include the line “engine->defuzzify();” I get error’s compiler message:

error: ‘class fl::Engine’ has no member named ‘defuzzify’
engine->defuzzify();

when I run the fuzzy system in octave with 15 and 0.32, ( evalfis([15,0.32],fis) ) I get 9.6498.

Juan, could you help me?

Thank.

[Unknown]

i find the solution, it was too easy:

int main(int argc, char **argv)
{
    fl::Engine* engine = fl::FisImporter().fromFile("/home/ruben/fuzzy.fis");
	
    engine->getInputVariable(0)->setInputValue(15);
    engine->getInputVariable(1)->setInputValue(0.32);

    engine->process();

    fl::OutputVariable* output = engine->getOutputVariable("Angulo");

    printf("%f\n",output->getOutputValue());
}

I get 9.649305 vs 9.6498 (by octave) very good I think.

[Juan Rada-Vilela (admin)]

Hi,

thank you for your posts and kind words.

Mash to convert fis to C++ you can do:

//Import from FIS
Engine* engine = fl::FisImporter().fromFile("/path/to/yourfile.fis");
//Export to C++ file
fl::CppExporter().toFile(engine, "/path/to/yourfile.cpp");

[Juan Rada-Vilela (admin)]

Ruben,

about the difference in results, fuzzylite is more accurate than octave because the resolution of fuzzylite is 200 to compute the centroid, whereas octave uses 100. If you change the centroid defuzzification method in fuzzylite to use 100, the results should be even closer.

Cheers.

[Author]

Posts