Network sum-rate maximizing power allocation over time-varying multiple-access interference channels

In this paper, power allocation over time-varying multiple-access interference channels is studied. Particularly, a stochastic network sum-rate maximizing power allocation problem is formulated, capturing the random nature of communication channels. Typically, a centralized controller must have perfect knowledge of global instantaneous channel state information (CSI) for dynamic optimal power allocation; however, this may not be possible, due to the computational complexity and communication overheads/delays involved. Based on the second-order statistics of the CSI, the stochastic problem formulation is transformed into an optimal deterministic representation in terms of ergodic capacity, while ensuring satisfactory quality of service via target outage probability. However, such deterministic reformulation happens to be non-convex and thus is computationally expensive. In turn, a sub-optimal reformulation is derived and solved via an iterative low-complexity algorithm. Simulation results demonstrate that the proposed deterministic sub-optimal power allocation reformulation coincides with its optimal deterministic counterpart, with the proposed algorithm converging in a finite number of iterations.