It can be easily verified that the optimal solution of (10) is
also optimal to (9). As discussed in [34], an iterative dual-layer
distributed algorithm can be directly derived to solve (10), and
is outlined as Algorithm P. In Step 1, the dual decomposition
Algorithm P Primitive Dual-Layer Algorithm for Nr = 1
Initialize the auxiliary variables arbitrarily and the dual variables
to zero.
For outer iteration n,
Step 1. Fix the auxiliary variables in (10), and solve it with
respect to the original variables, using the subgradient method
[32]. The subgradient method, which is an iterative method by
itself, carries out the inner iterations of this algorithm.
Step 2. Update the auxiliary variables by assigning the values
of the corresponding original variables in Step 1 to them.
Stop until the total transmit power of (9) satisfies a predefined
convergence criterion.
[32] is applied to deal with the coupling constraints. The
corresponding partial Lagrangian of (10) subject to constraint
(10d) is given by