Here is my code for the problem formulation
cvx_begin
variable f(J,1) complexminimize( (w * Hm * f)’ * (w’ * Hm * f))
subject to
(1) a.’ * f == 1;
(2) f’ * f <= 1/(J/gamma);
(3a) for i = 1:K
(3b) (Hm(i , : ) * f)’ * (Hm(i , : ) * f) <= tau. * (f’ * f);
(3c) end
cvx_end
where \mathbf{H}_\mathrm{m} is known complex square matrix, \mathbf{w} and \mathbf{a} are known complex vectors, and tau, eps, gamma, J, K are all known real scalars. And the error invalid constraint {convex} <= {convex} will occur in constraint (3b).
Please let me first explain why I wrote the constraint (3), although I knew it would be an invalid constraint ({convex} <= {convex}). The actual meaning of constraint (3) is that, for the i th receiving antenna of all K receiving antennas, the received power from the transmit signal (unit power transmit signal) after the transmitting weight \mathbf{f} and the channel \mathbf{H}_\mathrm{m}(i, : ) (\mathbf{H}_\mathrm{m} is the channel matrix between transmitting and receiving antennas) is less a certain level, which is proportional to the transmitting power. The transmit power level is calculated by \mathbf{f}^{\prime}\mathbf{f} and the relative loss is tau.
Is that possible that I further formulate this constraint to get rid of the invalid constraint error and make it be accepted by CVX?