Solver Not Applicable (mosek)

[Aras Selvi]

Dear Professor Löfberg and Yalmip users,

I am trying to solve a quadratic constraint in Yalmip. When I don't put it in Conic Representable form, I can only solve it via Knitro. I want to use Mosek, so I convert the model to a Conic Representable one. I get the following error:

Warning: Solver not applicable (mosek)

Shall I write with sqrt(.) <= something? Because now I am using norm([f(x);g(x)],2)<=d(x) form. 

Just in case, I am pasting the relevant part of the constraint and the code:

tau = sdpvar(1,1);

u = sdpvar(2,1);

V = sdpvar(2,2);

r = sdpvar(2,1);
Objective = tau;

Constraints = [norm([(1+r(1)+r(2)+u(1)+u(2)-tau)/(2+2*r(1)+2*r(2)); (1.4142*(V(1,1)+V(1,2))-1.7678*(V(2,1)+V(2,2)))/(1+r(1)+r(2));(0.9354*V(2,1)+V(2,2))/(1+r(1)+r(2))],2) <= (1+r(1)+r(2)-u(1)-u(2)+tau)/(2+2*r(1)+2*r(2))];

optimize(Constraints,Objective,sdpsettings('solver','mosek'));

Note that u,V, r are variables

[Johan Löfberg]

Well, it is obviously not an SOCP as it is nastily nonlinear in r

However, as far as I can tell, you can eliminate the denominator 1 + 1^Tr

[Aras Selvi]

Yes, I have noticed the 1+1^Tr thing. I eliminated it, but I still have 

Constraints = [Constraints,norm([(1+(u(1)+u(2)-tau)*(1+r(1)+r(2)))/2; 1.4142*(V(1,2))-1.7678*(V(2,1)+V(2,2)); 0.9354*(V(2,1)+V(2,2))],2) <= (1-(u(1)+u(2)-tau)*(1+r(1)+r(2)))/2]; 

Which is still not applicable. Is is because of the term:    (1+(u(1)+u(2)-tau)*(1+r(1)+r(2)))/2

Thank you very much!

[Johan Löfberg]

You still have products between r and u and tau, so not SOCP. An SOCP is in the form norm(linear)<= linear

However, as I said, 1+1^Tr can be eliminated completely. I don't understand how you managed to get a product between r and tau and u when eliminating the common denominator, unless the picture is incorrect

[Aras Selvi]

I will try more. Thank you very much.