Implementation of time varing function in CasADi

[oaldaify]

Implementation of time varing function in CasADi

[jgillis]

If you introduce parameters, you can keep those pure-numeric dot_m(t) , T_ram(t) helper functions as-is:

params = struct;
params.dot_m = MX.sym('dot_m')
params.T_ram = MX.sym(...)
...


p = [params.dot_m;params.T_ram;...]

f = Function('f', {x, u, p}, {dynamics(x,u,d,params)}, {'x','u', 'p'},{'x_dot'});
% In the dynamics, write params.dot_m instead of dot_m(t), and similar for others..


%Integration options
G = struct;
G.tf = T/N;

%Construct DAE
dae = struct;
dae.x = x; %define the states
dae.u = u; %define the control parameter
dae.p = p; %define paramaeters
% dae.t = t; %define the time parameter 
dae.ode = f(x,u,p); %define the RH side of ode
intg = integrator('intg','cvodes',dae, G); %Integration command

res = intg('x0',x,'p',p,'u',u); %integrate symbolically 
x_next = res.xf; %Assign the final state of x to x_next

%Returns x_next
F = Function('F',{x, u, p},{x_next},{'x','u','p'},{'x_next'});

% This loop is now devoid of any symbolic operations -> faster
for i = 1:length(t0)
	u_input = g(i, :);
	pp_num = [dot_m(i);T_ram(i)]
	x_next = F(x_current, u_input, p_num);

	x_current = x_next;
	X(:, i) = full(x_current); % Store the state at the current time step     

end

If you are willing to rewrite your helpers, you could consider using interpolant to CasADi Functions out of dot_m , T_ram, and use a time-dependant ode formulation

f = Function('f', {x, u, t}, {dynamics(t,x,u,d)}, {'x','u', 't'},{'x_dot'});
...
dae.t = t; %define the time parameter 
dae.ode = f(x,u,t); %define the RH side of ode
...

[jgillis]

Do you mean a lot of time to implement, or to solve?

[oaldaify]

to solve.

[oaldaify]

Thank you so much @jgillis

Out of curiostiy, Is there an example on how to use (interpolant) to CasADi functions out of helper functions I can look at?