How to generate 2-D Lookup Table data for the Fuzzy PID Controller

Can you please help me? I have a problem when I am making 2D look up table, from .fis file. I use evalfis function, but it only works when I have 1 output in .fis file, but my .fis file has 3 outputs. And then MATLAB shows me this:

Error using evalfis (line 70)

Input data must have as many columns as input variables and as many rows as independent sets of input values.

Error in Lookup (line 21)

LookUpTableData_s = evalfis([S DS], fis_skr);

Please help if you know. Thanks in advance!

[Sam Chak]

Let us create a scenario that causes the error message to appear. A typical Fuzzy PID controller has two inputs: a state (S) and the time derivative of the state (DS). Although the syntax for the older version of evalfis([S, DS], fis_skr) appears to be correct, the input values are arranged in row data form, which triggers the message reminding the user that 'Input data must have as many columns as input variables.' Resolving the issue is as simple as rearranging the input values into column data form.

fis_skr = readfis("tipper");

%% inputs as row data
S       = (0:10)    % state

S = 1×11

0 1 2 3 4 5 6 7 8 9 10

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

DS      = (0:10)    % time derivative of the state

DS = 1×11

0 1 2 3 4 5 6 7 8 9 10

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

%% evaluation
LookUpTableData_s = evalfis(fis_skr, [S DS])

Error using fuzzy.internal.utility.evalfis (line 84)
Input data must have as many columns as input variables and as many rows as independent sets of input values.

Error in evalfis (line 98)
[varargout{1:nargout}] = fuzzy.internal.utility.evalfis(varargin{:});

Answer 1

[Sam Chak]

Hi @Zoran

For simplicity, you can create three individual 2-D lookup tables for the three fuzzy outputs generated from a single MIMO fuzzy system .fis file. In the following demo, I used the Fuzzy Tipper, which has two inputs and one output.

Generally, 2-D lookup table data are created using two for-loops, as there are two dimensions of inputs. However, since the evalfis() command can handle vectorized data, we can reduce the code to use only a single for-loop.

fis = readfis("tipper");

%% inputs (arranged in column data form)
service = (0:10)';
food    = (0:10)';

%% generate 2D Lookup Data
for i = 1:numel(service)
    Tip(:,i) = evalfis(fis, [repmat(service(i), numel(food), 1), food]);
end

%% plot lookup data as a surface
[Service, Food] = meshgrid(service);
figure
surf(Service, Food, Tip)
xlabel("Service"), ylabel("Food"), zlabel("Tip")
title("Visualize 2D Lookup Table Data as a Surface")

%% compare result with gensurf
gensurf(fis)
title("Surface using gensurf")