Matlab Coder and ode45 with extra parameters - What implementation strategy?

Hi all,

I'd like to apply Matlab Coder to an 'ode45' function used with extra parameters. My problem is that Matlab coder does not support anonymous function so I tried to use persistent variable instead. But I obtain an error message.

Do you know where is the mistake in my code or what is the best implementation strategy for this case?

1. As a benchmark, from the Matlab help, it is possible to modify the ‘rigid’ example and to use anonymous function (although there is no meaning here). It works. See section a)
2. But I obtain an error message with the following code when I use persistent variable instead. See section b)

a) Code with anonymous function: no problem

VarAdd = 0.1;

options = odeset('RelTol',1e-4,'AbsTol',[1e-4 1e-4 1e-5]);
[T,Y] = ode45(@(t,y) rigid2(t,y,VarAdd),[0 12],[0 1 1],options);

plot(T,Y(:,1),'-',T,Y(:,2),'-.',T,Y(:,3),'.')

function dy = rigid2(t,y,VarAdd)
    dy = zeros(3,1);    % a column vector
    dy(1) = y(2) * y(3)         + VarAdd;
    dy(2) = -y(1) * y(3)        + VarAdd;
    dy(3) = -0.51 * y(1) * y(2) + VarAdd;

b) Code with persistent variable: problem

Script:

VarAdd = 0.1;

t = [0 12];
y = [0 1 1];
[ T,Y ] = rigidp(t, y, VarAdd);   

plot(T,Y(:,1),'-',T,Y(:,2),'-.',T,Y(:,3),'.')

Function #1

function [ T,Y ] = rigidp(t, y, VarAdd)
%#codegen

    persistent VarAdd2;

    VarAdd2 = VarAdd;    
    rigidp2(t,y,VarAdd2);

    options = odeset('RelTol',1e-4,'AbsTol',[1e-4 1e-4 1e-5]);
    [T,Y] = ode45(@rigidp2, t, y, options);
end

Function #2

function dy = rigidp2(t, y, VarAdd2)
%#codegen

    dy = zeros(3,1);    % a column vector
    dy(1) = y(2) * y(3)         + VarAdd2;
    dy(2) = -y(1) * y(3)        + VarAdd2;
    dy(3) = -0.51 * y(1) * y(2) + VarAdd2;

Here is the error message:

Error using rigidp2 (line 5)
Not enough input arguments.

Thanks.

Nicolas

Answer 1

[Titus Edelhofer]

Hi Nicolas,

one way to do this is the "classical" way, i.e., the way it was done before anonymous functions existed. If you add parameters after the options, then ode45 knows that these are parameters for the ode function:

function Y = compute(VarAdd)
options = odeset('RelTol',1e-4,'AbsTol',[1e-4 1e-4 1e-5]);
[T,Y] = ode45(@rigid2,[0 12],[0 1 1],options,VarAdd);

function dy = rigid2(t,y,VarAdd)
dy = zeros(3,1);    % a column vector
dy(1) = y(2) * y(3)         + VarAdd;
dy(2) = -y(1) * y(3)        + VarAdd;
dy(3) = -0.51 * y(1) * y(2) + VarAdd;

Titus

[Nicolas]

Hi Titus,

thank you for your prompt reply. The 'classical' way works.

Thanks a lot,

Nicolas

[Mike Hosea]

For future reference, the persistent approach needs to have a persistent variable in rigidp2. You use the saved value when 2 arguments are passed, and when 3 arguments are passed, you ignore the first two arguments and just store the 3rd input in the saved value for VarAdd.

Answer 2

[Nicolas]

Two additionnal informations:

1. The program 'b)' works if we substitute 'persistent VarAdd2' by 'global VarAdd2' and if we add this latter in the function 'rigidp2'. But it doesn't work with 'persistent'. I don't know why.
2. Here are two interesting webpages:

http://fr.mathworks.com/help/optim/ug/passing-extra-parameters.html#brhntus-1

http://fr.mathworks.com/help/matlab/math/parameterizing-functions.html#zmw57dd0e23359

Nicolas