There is also possibility to use global variables. You have access to current calculated values for example in OutputFcn. See the example:

options = odeset('OutputFcn',@myfun);

[t,y]=ode45(@vdp1,[0 20],[2 0],options);

plot(t,y(:,1));

and myfun.m:

function status = myfun(t,y,flag)

global tt yy

if nargin < 3 || isempty(flag)

tt = [tt t];

yy = [yy y];

else

switch(flag)

case 'init'

fprintf('start\n');

tt = t(1);

yy = y;

case 'done'

fprintf('done\n');

end

end

status = 0;

In each integration step new calculated values are added to yy and tt and you can use it in each function by means of global variables.

Maybe it is better to use solvers for delay differential equations?

http://www.mathworks.com/help/matlab/delay-differential-equations.html

There is also possibility to use global variables. You have access to current calculated values for example in OutputFcn. See the example:

options = odeset('OutputFcn',@myfun);

[t,y]=ode45(@vdp1,[0 20],[2 0],options);

plot(t,y(:,1));

and myfun.m:

function status = myfun(t,y,flag)

global tt yy

if nargin < 3 || isempty(flag)

tt = [tt t];

yy = [yy y];

else

switch(flag)

case 'init'

fprintf('start\n');

tt = t(1);

yy = y;

case 'done'

fprintf('done\n');

end

end

status = 0;

In each integration step new calculated values are added to yy and tt and you can use it in each function by means of global variables.

I found out a way to do this via the function "persistent"...

I was about to say that. Persistent is more useful in this situation that global. Moreover, if you are using embedded Matlab in Simulink, you cannot use globals and have to resort to presistent variables.

I understand that you use persistent variable in your ode function.

But what if an integration failed and solver will call your function for previous time step?

Didn't quite get your question Grzegorz... I was just using the persistent function to store data that is one step behind..

ode45 is a solver with adaptive time step, and sometimes integration can fail, so solver automatically decreases time step, and "has to go back".

See example:

function describing a problem:

function dydt = test(t,y)

persistent tt yy tmp

if isempty(tt) || isempty(yy)

tt = t;

yy = y;

tmp = 0;

else

tt = [tt t];

yy = [yy y];

end

dydt = (1-y^2)*y-y/max(abs(t),1e-10)*sign(t);

if tt(end)>0.9 && tmp==0

figure(1)

clf

plot(tt,yy,'o-')

figure(2)

clf

plot(tt(1:end-1),diff(tt))

tmp = 1;

end

And solution of this problem:

[t,y] = ode45(@test,[-1 1],1);

figure(1)

hold on

plot(t,y,'r')

ylim([-1 25])

xlim([-0.2 0.02])

figure(2)

hold on

plot(t(1:end-1),diff(t),'r')

Two solutions aren't the same. This is why I suggest solution from my previous answer.

Grzegorz you are right.. I checked my code and apparently the persistent function gives all the values of the variables which may not be the solution but are just attempts of the solver to integrate the problem. This is not what I want.. Any other solution that you may suggest?

I think I will just use the OutputFcn... Can't figure out any other simpler methods. Last time when I did my programming using S-function in Simulink, I had no issue collecting the past variables because the S-function acts like a state-space formulation. But now for coding using the conventional scripting, I found out that there are restrictions to how I can use the ode solvers. I could write my own code for the solvers but that would not be a very efficient use of time..

@Bilal Siddiqui: we have used global variables with embedded matlab function in simulink. It works fine. The only problem is accessing that data from a different model.

i suggest easy method, at end of any calculation, time interval is Zero, therefore you can add this code in main function

Dear Yong, I wish to do a similar thing. Did you find out how to do it?

@Grzegorz Knor Can you provide a further elaboration in the global variable method? Thanks.

@Tamiru Alemu Lemma Did you know how to do it now?

Consider looking into outputfcn in MATLAB documentation