Proactive Online Power Allocation for Uplink NOMA-IoT Networks with Delayed Gradient Feedback

In this letter, we propose a proactive online power allocation algorithm aiming to maximize the ergodic sum rate for an uplink multi-carrier non-orthogonal multiple access enabled Internet of Things (IoT) network, in which IoT devices (IoTDs) are subject to both instantaneous and ergodic transmit power constraints. The proposed algorithm enables a natural distributed implementation where each IoTD chooses its own transmit power proactively for future time slots without requiring instant channel power gain (CPG) information but only upon a delayed feedback of the sum rate gradient from the base station and a self-maintaining virtual queue. We show that the optimal power allocation for each IoTD can be easily obtained by a low-complexity bisection method. Moreover, the proposed algorithm achieves an O(V),O(1/V)-tradeoff between the virtual queue length and the ergodic sum rate optimality, where V is a positive parameter. Simulation results show that our algorithm has a comparable convergence speed and insignificant performance loss compared to a centralized drift-plus-penalty based algorithm upon instant CPG information.