I think that sum(x[i](r[i,t]-u[i]))*sum(x[j](r[j,t]-u[j])) = w[t]^2 is OK, since i,j are independent, so you can rewrite it (use i instead of j) as sum(x[i](r[i,t]-u[i]))*sum(x[i](r[i,t]-u[i])) = w[t]^2.

x[1]:=1.5;

for n from 1 to 20 while (Error>DesiredError) do
x[n+1]:=x[n]-(p(x[n])/eval(diff(p(x),x),x=x[n])):
Error:=abs(x[n+1]-x[n]):
end do;

Or you can compute the derivative before the loop (to avoid computing it in each cycle):

dp:=diff(p(x),x);

x[1]:=1.5;

for n from 1 to 20 while (Error>DesiredError) do
x[n+1]:=x[n]-(p(x[n])/eval(dp,x=x[n])):
Error:=abs(x[n+1]-x[n]):
end do;

It is not undefined for x=y, but I don't know if it's a problem.

Is there a command for diagonalizing of matrices?

I have managed to compute it numerically by

dsolve([diff(N(t), t) = N(t)*(1-(1/1000)*N(t))-piecewise(200 < N(t), 300, N(t) <= 200, (3/2)*N(t)), N(0) = 500], numeric)

for particular values. I guess it is not possible to obtain a "general analytical solution".

But as I had written earlier this didn't help me a lot. I would like to see a concrete example.

The most important thing is probably the dsolve call (= dsolve({sys,ic},numeric,method=rosenbrock,jacobian=??)).

I have tried to use an example from Help section but haven't managed to run it so even a simple (but concrete) example

would be great.

Regards,

Jarekkk

Thanks for the comment. I know there are several solvers for ODEs in Maple. What more I have used all of them. The goal

I've been searching for is to compare all the solvers and find one which suits best my problem. I don't know if specifying

jacobian matrix would help rosenbrock solver to compute faster but I have to try it. And don't know how.

Jacobian is a constant matrix in my case and I can't find out how to tell Maple's rosenbrock solver that the matrix is jacobian of the system.