I don’t know what problem you’re trying to solve. Your CVX “code” isn’t even CVX code.

If you are trying to solve non-convex SDPs, maybe you should be using YALMIP instead of CVX. You might be able to solve them without relaxing non-convexities.

What is your difficulty entering the SDR into CVX? It looks straightforward if you use for loop for the last constraint. If A is n by n by K, you can use sum(A,3)for the sum of the A_k.

As to whether that SDR is any good for your purpose, is a different matter, which is out of scope of this forum.

Infeasibility of the model suggests that the model formulation and/or it’s input data may not be adequate, or your real world problem has no solution.

If the problem you enter in an optimization modeling tool is infeasible, it may also be that any approximation made between that problem and the problem you really want to solve is inadequate.

If the problem involving the sum of logs in the objective function is a convex optimization problem, it might be enterable in CVX.