Hi. I am currently reforming a research paper related to SINR and power distribution problems.

The researchers said that this problem can be implemented in CVX. I want to reform it using CVXPY, but I am a beginner at CVX or CVXPY. Can somebody tell me whether CVX can solve this so I can implement it in CVXPY?

Article title: Multi-UAV Deployment for Throughput Maximization in the Presence of Co-Channel Interference

Article link: Multi-UAV Deployment for Throughput Maximization in the Presence of Co-Channel Interference | IEEE Journals & Magazine | IEEE Xplore

it looks straightforward in CVX or CVXPY, if you use `log_sum_exp`

in the first constraint.

Mosek would be the preferred solver for this in CVX due to the exponential term.

Thank you so much! It helps me a lot.

Hello MatthewSin, sorry to bother but I have a question which is similar to your work. I am now doing a cell-communication based proragmme which also requires the SCA/first-order Taylor expansion to convert my nonconvex formula into approximately convex form. (And it is based on this Paper) .

However, from the comman window, my feasible programme only gets “feasible” results, as shown below. It means that no matter how big the max bound for the feasible problem codes set, the programme will only give me the maxmim value based on what I set. I wonder if you met the same problem and how do you solve it? I am really looking forward to your reply.

Hi everyone. I have encountered problem {log-convex} >= {log-concave} when dealing with constraints (24b). I have tried to reformulate the problem several times but still cannot solve it. Do there have any tricks for solving this type of problem?

My CVX code:

cvx_begin gp

variable z(1,n)

variable v(1,n)

variable y(1,n)

variable psi_z

maximize( psi_z )

subject to

for i = 1:size(Ku_cover)

for j = 1:size(UAV)

if j ~= Ku_cover(i,3) % exclude mth UAV

v(j) + power(norm(UAV(j,1:2)-Ku_cover(i,1:2)),2) + sigma >= exp(-y(j)); % (24b)

v(j)+power(z_appr(j),2) <= 2*z_appr(j)*z(j); % (24c)

end

end

end

cvx_end

The power() term and sigma are constants.

Why are you using gp mode? Are there any multiplications or divisions of CVX variables? is `z_appr`

input data? I don’t see why that error message would correspond to this program (maaybe it has something to do with gp mode). Perhaps you’re not showing us everything? The program as it is doesn’t make any sense, because the variable `psi_z`

is unconstrained, hence the problem is unbounded presuming it is feasible.

Whatever your problem is actually supposed to be, have you proven it is a convex optimization problem?