Algorithm 1 Our algorithm based on SDR and CCP

Input: ${\sigma }_{\mathrm{A}}{}^2,{\sigma }_{{\mathrm{E}}_k}{}^2,{\sigma }_{h_{\mathrm{ab}}}{}^2,{\eta }_1,{\eta }_2,{\eta }_3,\mathsf{\beta },\Psi ,M,\epsilon ,Q_{\max }$

Output: $V$, $Q$

1: Initialize $Q_{\min }=0$, and Set $Q^{\left(t\right)}=\left(Q_{\max }+Q_{\min }\right)/2$.

2: repeat (Bisection Search for $Q$)

3:  repeat (Our algorithm based on SDR and CCP)

4:    For given $Q^{\left(t\right)}$, when $q=1$, initialize a positive semi-definite matrix $V^{\left(1\right)}$ randomly; when

$q>1$, given $V^{\left(q\right)}$, solve the above optimization problem (28) for $V^{q+1}$.

5:    Update $q=q+1$.

6:  until the optimization variable $V$ reaches convergence.

7:    Record $V_{opt}=V^{\left(q\right)}$ of the current iteration if the above problem can be solved, then update

$Q_{\min }=Q^{\left(t\right)}$; if $Q^{\left(t\right)}$ is unreachable, then update $Q_{\max }=Q^{\left(t\right)}$

8: until $Q_{\max }-Q_{\min }<\epsilon$.

9: Recover $v$ from $V_{opt}$ by Gaussian randomization.