cvx_begin variable k(n_0,1) minimize ( norm(D_2*k,1) + 0.2*norm(k,1) ) subject to norm ( B*(cos(B*inv_radius) - sin(B*inv_radius).*(B*(k-inv_radius))) - x_LLF,2) <= 0.5 norm ( B*(sin(B*inv_radius) + cos(B*inv_radius).*(B*(k-inv_radius))) - y_LLF,2) <= 0.5 cvx_end
Well, apparently your problem is infeasible. You haven’t provided the input data, therefore we can’t diagnose it for you.
You can try solving a problem where, instead of the current objective function, you minimize the LHS of one or both of the constraints (subject, or not, to the other constraint) to determine how close the problem is to feasibility.
Hier is my Problem. I have estimatet k_hat. It is a 751x1 Vector. Why is my Problem infeasible?
Thanks a lot
a_x = 0; gamma=0;
I have k_hat. It is a 751x1 Vector. The hadamard Product i implemented with .*
You still haven’t provided data to make this a reproducible problem, so I can;t further diagnose it.
IIt appears, however, that you did not properly implement the 2-norm constraint. (5.27) shows a single 2-norm constraint applied to a stacked vector of x^i and y^i for each i. That is not the same as implementing separate 2-norm constraint for x^i and y^i, as you have. Obviously the 2-norm constraint(s) would be feasible with sufficiently large value of \epsilon^i.
Is your N = 1? You’re also missing the constraint on |\kappa|.
Thank you for your answer.
My N is 751. ´
The B is a 751x751 Matrix and k_hat is 751x1 Vector.
Do i need a for loop ? Which Data do you need?
Thanks a lot
You an implement the constraints in a for loop.
But really, it is your problem, so you should diagnose why is is not feasible. Obviously, feasibility depends on the input data., certainly to include the values of \epsilon^i. Of course, it appears that you haven’'t gotten the model implemented correctly yet. My comments only highlight some of your mistakes, and don;t necessarily capture all of them, which I I have no way of knowing.
Also note that inearizing constraints, as your extract says was done to produce the expression inside the 2-norm, can be dangerous, and depending on how it is done, could in some cases result in a feasible nonlinear constraint becoming an infeasible linearized constraint. I don’t know whether that can happen in this model.
subject to for i=1:n_0 norm( [B(i,i)*(cos(B(i,i)*inv_radius(i,:)) - sin(B(i,i)*inv_radius(i,:)).*(B(i,i)*(k(i,:)-k_hat(i,:)))) ; B(i,i)*(sin(B(i,i)*inv_radius(i,:)) + cos(B(i,i)*inv_radius(i,:)).*(B(i,i)*(k(i,:)-k_hat(i,:))))] - x_ref(i,:) , 2) <= 2; end sum(abs((k))) <= 0.0015;
So Sir, my x_LLF is a 751x1 Vector and y_LLF also. I have combined them to a 1502x1 Vector.
B is a 751x751 Matrix.
I get the same problem? Where is my mistake? Sorry for asking so much. it is for my thesis . .
I don;t know where your mistake is. Note that it is your thesis, not mine.
How did you get the right-hand side values? As I’ve said, if those are high enough, the problem will become feasible.
Perhaps you can start with a small-scale version of the problem. Diagnose and make that work. Then scale up to the full-size problem of interest.