Fuzzy Logic in C

A fuzzy logic Mamdani-like implementation in C. This code implements the classification of crisp values to fuzzy sets and provides a basic inference engine based off of pre-defined rules. Sets can then be de-fuzzified back to a crisp value.

Currently, Triangles, Trapezoids and Rectangular membership functions are supported. Deffuzification is done using centroids.

features

• Easy to get started
• extendable Membership Function States
• No dependencies (except for stdlib)
• Semantic natural language-like syntax for rule definitions

// define membership functions
#define InputMembershipFunctions(X)                                            \
    X(INPUT_LOW, 0.0, 0.0, 15.0, 40.0, TRAPEZOIDAL)                            \
    X(INPUT_MEDIUM, 15.0, 40.0, 60.0, 80.0, TRAPEZOIDAL)                       \
    X(INPUT_HIGH, 60.0, 80.0, 100.0, 100.0, TRAPEZOIDAL)
DEFINE_FUZZY_MEMBERSHIP(InputMembershipFunctions)

// define the system rules
FuzzyRule rules[] = {
    // if input is low then output is high
    PROPOSITION(WHEN(ALL_OF(VAR(Input, INPUT_LOW))), THEN(Output, OUTPUT_HIGH)),

    // if input is not low then output is low
    PROPOSITION(WHEN(ALL_OF(NOT(Input, INPUT_LOW))), THEN(Output, OUTPUT_LOW)),
};

example

Find working examples in the ./example directory:

cd example
make all
#then run
./out/minimal
#and
./out/TecFanController

Usage: ./out/minmal <value>
Usage: ./out/TecFanControl <currentTemperature> <currentTemperatureChange> <currentTECPower> <currentFan>

You can then plot a simple surface of the fan controller example using the provided python script:

python -m venv ./venv
source ./venv/bin/activate
pip install -r requirements.txt
python ./plot.py

Note

This visualization does not represent the entire controller due to its non-linear behavior. To fully represent this model you need a five-dimensional vector space.

legal

Licensed under the Apache License, Version 2.0 (the "License");
See LICENSE.txt file for details.