Utility Maximization Integrating Secrecy, Energy Consumption, and Latency for UAV-Assisted MEC Systems via Dual-Replay TD3
TL;DR
A utility maximization problem to jointly optimize UAV trajectory and task-offloading decisions in UAV-assisted MEC systems against multiple eavesdroppers is formulated and an enhanced twin-delayed deep deterministic policy gradient (TD3) framework integrating Hindsight Experience Replay (HER) and Prioritized Experience Replay (PER) is proposed to improve convergence efficiency and learning stability.
Abstract
Unmanned aerial vehicle (UAV)-assisted mobile edge computing (MEC) has emerged as a practical means of providing computation and communication support for geographically dispersed Internet of Things (IoT) terminals. However, the broadcast nature of wireless links makes offloading data vulnerable to cooperative eavesdropping. Moreover, the limited onboard energy of UAVs and the latency-sensitive characteristics of MEC services lead to a challenging trade-off between the secrecy rate, energy consumption, and latency. To address this issue, we formulate a utility maximization problem to jointly optimize UAV trajectory and task-offloading decisions in UAV-assisted MEC systems against multiple eavesdroppers. Due to the strong coupling among optimization variables and the non-convexity of the problem, an enhanced twin-delayed deep deterministic policy gradient (TD3) framework integrating Hindsight Experience Replay (HER) and Prioritized Experience Replay (PER) is proposed to improve convergence efficiency and learning stability. Furthermore, a system utility-driven reward function is designed to balance the secrecy rate, energy consumption, and processing latency under different application requirements. The simulation results demonstrate that the proposed approach consistently outperforms DDQN, DDPG, and conventional TD3 in terms of system utility, secrecy performance, convergence speed, and adaptability to different scenarios.
