Energy Efficient Hybrid NOMA-OMA IoT Systems: A Hybrid PPO Approach

초록

a hybrid orthogonal multiple acInternet-of-things (IoT) systems, where a group of interested IoT devices is assigned to subchannels, some using NOMA and the others using OMA. Since multicasting is heavily constrained by the weakest channel gain of involved IoT devices, an advanced resource allocation strategy is required to enhance energy efficiency. This paper investigates a simultaneous optimization of power allocation and subchannel assignment to maximize energy efficiency in HMA, not resorting to alternating optimization (AO). Specifically, an optimization problem is formulated as a mixedinteger nonlinear programming (MINLP) problem, posing a significant challenge in finding a globally optimal solution. To address this, we propose two novel algorithms: the first is designed for small search spaces, using the Big-M method for linearization of a bilinear term followed by an iterative and thresholdbased rounding, and the second is tailored for large search spaces, utilizing a hybrid proximal policy optimization (HPPO) technique to train hybrid action space. The simulation results confirm that both the proposed algorithms surpass other AObased benchmarks by 7.1% and 5.6%, respectively, in terms of energy efficiency.

키워드