Question: Further details then

Well,

Basically I know that.

I just hoped there is some walkaround this problem.

It seems now there is not.

The problem I am trying to solve is how I can construct an iterative procedure where function is computed from numerical solution of the ODE system (done) and then this very function is used to recalculate the solution of this ODE system.

This is my attempt to impement some sort of  gradient projection method to the optimal control problem. However at this point I am not sure this is possible with Maple at all.

If function is defined only through procedure calls, then every next step contains more and more functions of time and thus all the process stops after 3-4 steps.

Namely: Given u(x(t),t) initally const

I solve ODE like diff(x(t),t)=f(x(t))+g(u(x(t),t)) (controlled equation) for u(x(t),t)=const, say, u_0.

Then u_1 is given by u_1=u_0+h(x(t),t). To calculate this I use solution of the system above, say, x_1(t) given by a procedure from dsolve/numeric.

Now at the second step I need to solve the ODE diff(x(t),t)=f(x(t))+g(u(x(t),t)) with u(x(t),t)=u_1(u_0,x_1(t),t)

which is also solved, but then the process repeats and then <i get stuck sinc u_3 is a function of x_1;x_2;x_3 ß 3 functions of time and dsolve cannot handle it (in sufficiently complicated system).

May be someone has any ideas how I can handle this situation?

May be some sort of forget function may help....

Thanks for help