UAV-Assisted Security-Aware Vehicular Edge Computing: A TD3-Enhanced Scheme

Abstract

Unmanned aerial vehicles (UAVs) equipped with edge computing capabilities offer a promising solution for the coverage and flexibility of terrestrial networks, but they also face challenges in low-latency, security-aware data transmission. To address this, a UAV-assisted, security-aware vehicular edge computing system is established and supported by comprehensive channel, communication, and computation models. To balance computing latency and security-aware offloading in the presence of an eavesdropper, we formulate a problem that minimizes the maximum computation latency by optimizing offloading, UAV movement, and vehicle association. Considering the dynamic nature of vehicular networks and the need for real-time decision-making, a twin delayed deep deterministic policy gradient (TD3)-based UAV-assisted security-aware vehicular edge computing system is proposed to dynamically adjust movement and offloading policies, thereby satisfying system constraints. Extensive simulations demonstrate that the TD3 scheme exhibits outstanding convergence stability, outperforming other benchmark schemes in cumulative reward by at least 20%, reducing overall system latency by at least 1.7%, and showcasing its robustness and adaptability through an algorithmic detail comparison.