Dynamic user pairing and power allocation for throughput maximization in NOMA systems

Non-orthogonal multiple access (NOMA) can enhance the spectral efficiency of the fifth generation (5G) wireless networks. System level optimization on power allocation and user pairing for improving the throughput of NOMA systems has been well studied. However, most of the existing works have not taken into account dynamic traffic arrival and possible packet transmission failure. In this paper, we consider downlink transmission of a power-domain NOMA system with dynamic packet arrival, where the base station supports both NOMA and orthogonal multiple access (OMA). We propose a packet-level scheduling scheme for the base station to decide using either NOMA or OMA, and to determine user pairing and power allocation, with an objective to maximize the aggregate throughput. To tackle the challenges introduced by dynamic packet arrivals and possible transmission failure, we first derive the probability of successful packet transmission with limited feedback on channel state information (CSI). We then formulate a throughput maximization problem as a stochastic network optimization problem taking into account the backlog stabilities. We decompose the problem into two subproblems for NOMA and OMA, respectively, and obtain the optimal user pairing and power allocation. Packet-level simulations show that the proposed scheme obtains a higher throughput compared with the distance-based user pairing scheme and the CSI-based power allocation scheme.