Latency-Aware Sum-Rate Maximization for 5G Software-Defined Radio Access Networks

In this paper, we consider the downlink power allocation problem where the objective is to maximize the sum-rate in software-defined radio access network (SDRAN). Using SDRAN, the power allocation is performed at a central controller with the channel state information (CSIs) received from the base stations (BSs). However, due to time variable nature of the fading channel and the control path delay between the controller and BSs, the central decisions at the controller might become outdated by the time each BS wants to transmit. Modeling channel fading by using a time-correlated channel model, the power allocation problem would consist of parameters which are random variables. In order to solve this problem, we use the stochastic optimization framework and propose two practical schemes. The first solution works in a completely centralized manner while the second one is performed collaboratively in the controller and BSs. We analyze these schemes for cellular networks with arbitrary number of BSs and present procedures to find the optimum power allocations. Finally, using different numerical experiments, we evaluate the performance of our proposed schemes. Significant improvement in comparison to the scenario without considering CSI variation is observed.