Energy-Efficient UAV-Enabled MEC Systems: NOMA, FDMA, or TDMA Offloading?

TL;DR

This paper tackles energy-efficient offloading in UAV-enabled MEC by comparing NOMA, FDMA, and TDMA under both infinite and finite blocklength regimes. It develops a comprehensive energy-minimization framework that accounts for local and remote computing energies as well as offloading power, with a two-UE UAV system and DVFS-based CPU control. A novel alternating optimization algorithm (BCD/SCA) jointly optimizes task-offload portions, offloading times, and UAV location for the NOMA-F problem, with parallel formulations for the other schemes. Results show TDMA outperforms FDMA in both blocklength regimes, while NOMA's advantage over FDMA depends on blocklength and channel symmetry, and SIC-related considerations under finite blocklength are crucial. The proposed algorithm significantly reduces MEC energy consumption compared to benchmarks, highlighting the practical impact of joint MA selection, offloading allocation, and UAV positioning in energy-constrained environments.

Abstract

Unmanned aerial vehicle (UAV)-enabled mobile edge computing (MEC) systems can use different multiple access schemes to coordinate multi-user task offloading. However, it is still unknown which scheme is the most energy-efficient, especially when the offloading blocklength is finite. To answer this question, this paper minimizes and compares the MEC-related energy consumption of non-orthogonal multiple access (NOMA), frequency division multiple access (FDMA), and time division multiple access (TDMA)-based offloading schemes within UAV-enabled MEC systems, considering both infinite and finite blocklength scenarios. Through theoretically analysis of the minimum energy consumption required by these three schemes, two novel findings are presented. First, TDMA consistently achieves lower energy consumption than FDMA in both infinite and finite blocklength cases, due to the degrees of freedom afforded by sequential task offloading. Second, NOMA does not necessarily achieve lower energy consumption than FDMA when the offloading blocklength is finite, especially when the channel conditions and the offloaded task data sizes of two user equipments (UEs) are relatively symmetric. Furthermore, an alternating optimization algorithm that jointly optimizes the portions of task offloaded, the offloading times of all UEs, and the UAV location is proposed to solve the formulated energy consumption minimization problems. Simulation results verify the correctness of our analytical findings and demonstrate that the proposed algorithm effectively reduces MEC-related energy consumption compared to benchmark schemes that do not optimize task offloading portions and/or offloading times.

Figures (10)

• Figure 1: A two-UE UAV-enabled MEC system.
• Figure 2: Illustrations of the time-frequency transmit power allocations and the remote computing CPU frequency allocations in different offloading schemes.
• Figure 3: Isolines of $\mathcal{A}(\rho_1,\rho_2,N)$ and $\mathcal{B}(\rho_1,\rho_2,N)$ over mesh $(\rho_1L_1,\rho_2L_2)$ for $N = 300$.
• Figure 4: MEC-related energy consumption required by different multiple access offloading schemes versus the portion of task offloaded by UE 2 $\rho_2$ for different values of $d$ when $\eta = \delta = 0.5$, $D = 100$ m, $\rho_1 = 0.5$, $B = 0.5$ MHz, $t = 0.3$ ms, $L_1 = L_2 = 1.2$ kbit, and $\epsilon_1 = \epsilon_2 = 10^{-5}$.
• Figure 5: MEC-related energy consumption required by different multiple access offloading schemes versus the UAV location $d$ for different cases of $\bm{\rho}$ when $\eta = \delta = 0.5$, $D = 100$ m, $B = 0.5$ MHz, $t = 0.3$ ms, $L_1 = L_2 = 1.2$ kbit, and $\epsilon_1 = \epsilon_2 = 10^{-5}$.