Energy Minimization for NOMA-Based Data Collection and Computation in UAV-Assisted Marine IoT

Unmanned aerial vehicles (UAVs) have been emerging as promising tools for data collection and processing due to their mobility and line-of-sight conditions, especially in maritime operations. In this paper, we deploy a UAV-assisted marine Internet of Things system, where a UAV equipped with a mobile edge computing server acts as an aerial base station for data collection and computation. The UAV first collects data from sensing devices (SDs) using the non-orthogonal multiple access (NOMA) technology. It then performs real-time processing on the collected data and feeds the computation results to the corresponding SDs to optimize their sensing behaviors. Our goal is to minimize the total energy consumption by jointly optimizing the transmit power of the SDs, the trajectory of the UAV, and the allocation of computational resources for data processing, under the constraints of the maximum system latency and the required collected data volume. Due to the non-convexity of the proposed problem, we first analyze the correlation among constraints to reformulate an equivalent one and then employ the twin delayed deep deterministic policy gradient algorithm to solve it. Numerical results demonstrate the efficiency of our proposed scheme in UAV trajectory optimization and energy efficiency maximization. Notably, our NOMA-based scheme reduces the total energy consumption by 20.21% and 32.34% compared to the frequency-division multiple access-based and time-division multiple access-based schemes, respectively.