When Asymmetry Helps: Joint Power and Blocklength Optimization for Non-Orthogonal Multiple Access in Downlink Communications

This paper investigates downlink non-orthogonal multiple access (NOMA) in short-packet communications, specifically focusing on a scenario where an Access Point (AP) serves two users with varying channel conditions. The study explores the finite blocklength regime in order to support low latency scenarios. In this context, the conventional Shannon’s capacity theorem becomes inadequate due to the non-negligible decoding error probability, resulting from finite blocklengths that do not approach infinity, which is the case of the asymptotic regime of the capacity theorem. NOMA has great potential to improve spectral efficiency compared to orthogonal multiple access by adapting power allocation to take benefit from the asymmetrical channel conditions between served users. In this study, we take the concept further by suggesting that, in the finite blocklength regime, the user with the worst communication channel can benefit from a longer blocklength. The approach considers both power and blocklength partitioning as the degrees of freedom to be tuned to enhance overall system performance. In particular, we derive the maximum achievable rates of both users when different blocklengths are assumed. We are evaluating the impact of promoting the user with the worst channel condition not only through higher power but also through a longer block on its maximum achievable rate while ensuring a minimum achievable rate for the other user. Numerical results of the exact resolution are provided. The comparative analysis of the proposed scheme with the conventional one where both users have the same blocklength shows that the asymmetry of blocklengths is as beneficial as the asymmetry in power. This observation has been verified by the achievable rate region analysis, using Pareto Frontier and genetic algorithm tools.