What should I do if I get "Error: Invalid FIS. PARAMS field of constant output MF must contain a scalar"

I'm a newbie to this forum. I'm facing the following problem in using the fuzzy logic function evalfis.

Below is the code snippet:

a = addvar(Sen_Val,'output','Confidence',[0 20]);
a = addmf(Sen_Val,'output',1,'Very_Low','constant',[0   0.3]);
a = addmf(Sen_Val,'output',1,'Low','constant',[0.30001 0.7]);
a = addmf(Sen_Val,'output',1,'Very_High','constant',[0.70001 1]);

output=evalfis(Fuzz_in,a,5)
% Here Fuzz_in is a 5X3 matrix

Below is the error message displayed:

 ??? Error using ==> evalfismex
Invalid FIS. PARAMS field of constant output MF must contain a scalar.

 Error in ==> test at 163
    output=evalfismex(Fuzz_in,a,5)

Requesting for your help to overcome this error. Thanks in advance Nithya

[Iman Ansari]

What is constant membership function? FIS doesn't have this type of membership Function.

Answer 1

[Sam Chak]

Hi @Nithya Kancherla and @Iman Ansari

The conclusion part of the Sugeno If–Then inference does not have the typical output membership functions (MFs) {triMF, trapMF, gaussMF} defined as the Mamdani fuzzy system does. If x is the input vector and y is the output vector, the output of each rule is defined as a function of the inputs, .

However, the MATLAB Fuzzy Logic Toolbox, unfortunately, still only allows for 'straight line' functions. These can be either linear combinations of the inputs, , or a constant function, . A constant function can be thought of as a scalar-valued function (outputting the same single number for every input value). Thus, the parameter field of the so-called 'constant output MF' must be a scalar, as shown in the following syntax:

fis = addMF(fis, 'outvar', 'constant', 7, 'Name', 'Cfun');

The error message is generated because you have entered a vector instead of a scalar. If you must use a vector to represent a constant function, then you should use the following syntax:

fis = addMF(fis, 'outvar', 'linear', [0, 0, 7], 'Name', 'Cfun');

Strictly speaking, it is incorrect to refer to the outputs of a Sugeno FIS as 'Constant membership function' or 'Linear membership function' because the units of the output values (on the y-axis) of these mathematical functions do not correspond to the membership grades between 0 and 1.

However, the constant function can be reimagined as a Singleton membership function, which is entirely permissible and logically valid.

%% Mamdani Fuzzy System
fis = mamfis;

% Fuzzy Output
fis = addOutput(fis,  [6 8], 'Name', 'out');
% fis = addMF(fis, 'out', 'trimf', [7 7 7], 'Name', '7mf');     % transform triMF into a singleton
fis = addMF(fis, 'out', 'gaussmf', [eps 7], 'Name', '7mf');     % approximation, but good enough

plotmf(fis, 'output', 1, 100001), grid on
title('Singleton MF equivalent to a Constant function')

●