Secrecy-Aware UAV Path Planning and Offloading Strategy Optimization Using Deep Reinforcement Learning and Particle Swarm Optimization

초록

Efficient management of limited resources, reduced data transmission latency, and improved overall network performance are critical in ground networks. This paper proposes a hierarchical sensing and offloading framework for intelligent transportation systems (ITS), where unmanned aerial vehicles (UAVs) act as mobile data aggregators to support ITS monitoring in areas beyond roadside unit (RSU) coverage, during RSU outages, or under severe ground network congestion. By enabling localized offloading, the framework ensures the timely delivery of critical ITS data. The proposed secrecy-aware approach enhances the resilience, efficiency, and integrity of ITS sensing and communications. Although UAVs offer advantages such as high mobility, on-demand deployment, and reliance on line-of-sight (LoS) communication channels, they remain vulnerable to security threats. To counter eavesdropping and jamming threats, the proposed method combines policy-gradient reinforcement learning with protective jamming and secrecy-aware transmission scheduling, enabling UAVs to adaptively adjust flight paths, transmit power, and time slot assignments. A multi-objective reward function is designed to jointly optimize secrecy rate, communication delay, and energy consumption under adversarial conditions. Extensive simulations confirm the proposed model's effectiveness in enhancing communication security and operational efficiency across varying threat scenarios.

키워드