Re: Non homogeneous differential equations

[François Poulain]

I know the "lsode" function, but I didn't find how work with it in the
case of non homogeneous equations (witch mean that I have some input
time varying signals in my equations).

I am working about non linear control theory, and I works with models
like the following (in LaTeX format) :

\begin{align}
\dot x_1 &= - a x_4 sin x_3 + u1(t)                     \\
\dot x_2 &= - a x_4 cos x_3 + u2(t)                     \\
\dot x_3 &= - x_4 + u3(t)                               \\
\dot x_4 &= b ( x_1 sin x_3 + x_2 cos x_3) - \tau(t)
\end{align}
with measured output $y = (x_1,  x_2)$, parameters  $a,b>0$,
inputs $(u1, u2, u3)$.

Under matlab, when I create any function dot_x = f(x,t,u1,u2,...), I can
integrate it, because matlab use the only 2 firsts arguments to
integrate, but the others argument are used to give input signals (and
we need to interpolate them at the current time t in the function).

I didn't find any way of doing it under GNU/Octave, and it's a problem
for me, because it's the only one privative software that I am using
currently.

Thank you for your help.

François.

Le lundi 21 mai 2007 à 10:38 +1000, Geordie McBain a écrit :
> I believe lsode is a fair bit more sophisticated than ode45, using
> variable step size and variable order up to quite high order (12 or
> 13?).  It should be more efficient for problems with reasonably smooth
> solutions.
> 
> Hope this helps.  (I wasn't too sure what u (t) was in your question.
> Did I miss something?)
> 
> Geordie McBain 
-- 
François Poulain <address@hidden>