Error: "Cannot determine number of iterations for a loop..."

Hi Everyone!

So, I having a problem with a MATLAB Function that I am using in SIMULINK. When I try to run my code I get the following error, "Cannot determine the exact number of iterations for a loop with range 0.111:0.001:0.131" , and the model goes into Debug mode.

The general structure of my code is:

function [y] = fcn(x,y_prev)
%#codegen

ctr=1;
for A = (y_prev-0.01):0.001:(y_prev+0.01)
% Some Calculations
B(ctr) = %some algebraic expression based on the variable "x"
ctr=ctr+1;
end

temp = (y_prev-0.01):0.001:(y_prev+0.01);
y = max(temp(B == max(B)));

Where "y_prev" is the previous iteration's value for y (determined by a memory block outside of the MATLAB Function.

I feel like I have tried just about every possible solution. I have added "assert" commands as I first thought that MATLAB couldn't determine bounds for the variables. And I also bounded the output variable range from the Ports & Data Manager. However, if I change "temp" to a constant range (such as 0.2:0.001:0.22) instead of calculating my range from y_prev, it works. I have also noticed that if I restrict the range of the input variable "x" that the code works. I have checked to make sure that B is a real number, and there are no issues there.

I should also mention that I had written this code on a previous computer where it worked beautifully. But, alas that computer died and I am now using a new computer. I have tried to run my code on several other computers just to make sure that it is not my new computer, but I haven't had any luck.

Does anybody have any suggestions or ideas for this?

Answer 1

[Walter Roberson]

Stepsizes that are not integers or powers of 2 (e.g., 1/2) lead to round-off errors that can make it difficult to predict the exact number of iterations. See http://matlab.wikia.com/wiki/FAQ#Why_is_0.3_-_0.2_-_0.1_.28or_similar.29_not_equal_to_zero.3F

You could, for example, use

for At = -10:10
  A = y_prev - At/1000;

[John]

Thanks Walter!

I had tried something similar using "round" but I must have made a mistake. Sometimes it is just helpful to get a fresh eye on the problem.