Cvx was used to solve the two variables through the alternate feasibility optimization method. In the first iteration, both variables had feasible solutions, but in the second iteration, the second variable did not have feasible solutions. Could you please give me some advice

I don’t know the exact algorithm you are using. But it sounds like a crude unsafeguarded implementation, at least similar in spirit to Successive Convex Approximation for purposes of the applicability of the following:

Your algorithm might not converge at all. Whether or not your algorithm converges, or what it converges to, might depend on the starting value for the optimization variables. However, if you are only trying to find a feasible solution, rather than an optimal solution to an optimization problem having (non-constant) objective function, you can at least easily verify whether the “solution” is optimal, meaning (primal) feasible…

OK,thank you Mark! thank you very much~