Pair Outage Probability Minimization for NOMA Among Resource-Constrained IoT Users

Non-orthogonal multiple access (NOMA) is a favourable technique for future wireless networks to meet massive connectivity and high throughput. The primary concept of NOMA is to share the same radio resource block among multiple users. However, each user has a predefined data rate demand. Under the quality of service (QoS) agreement, guaranteeing all users' diversified data rate requirements is necessary. Therefore, this article mainly focuses on designing a QoS-aware NOMA system with two users from a decoding order and power allocation viewpoint. We evaluate the network's performance regarding the system's pair outage probability (POP). To characterize the realistic network's performance, we assume imperfect successive interference cancellation (SIC) at receivers. Focusing on the possibilities of different decoding orders for a NOMA system, we derive POP expressions in closed-form for each possible decoding order with the linear model of imperfect SIC. To achieve the best pair outage performance of the system, we formulate the problem of minimizing the POP by jointly optimizing the decoding order and the power allocation. The formulated optimization problem is combinatorial. Therefore, we solve it by optimizing pair outages analytically over power allocation for each possible decoding order. Thus, the optimal power allocation solution to the formulated joint optimization problem is provided in closed-form. Lastly, numerical results are presented, which validate the analytical claims and highlight the effect of various parameters on achieved pair outage performance. The results show that following the conventional decoding order and higher power allocation to the weaker user than the strong one is not always necessary. We benchmark the jointly optimal solution against different benchmarks FDFP (fixed decoding order and fixed power allocation), FDOP (fixed decoding order and optimal power allocation), and ODFP (optimal decoding order and fixed power allocation), and obtain average gain of 25%, 41% and 15%.