Multi-Type Disaster Scenario Task Offloading in Air–Ground Integrated Search and Rescue Networks: A Blockchain-Assisted MFL Approach
Peng Da Qin, Wang Yi-fei, Jing Zhang, Peilin Li, Yang Fu · IEEE Transactions on Vehicular Technology · 2025
Inrecent years, autonomous aerial vehicle (UAV) has demonstrated significant potential in Search and Rescue (SAR) operations by providing real-time high-definition image and video data. However, their limited computational power and battery capacity may hinder performance in executing computationally intensive tasks. To address this challenge, this paper proposes an innovative edge computing model that leverages flexibly deployed autonomous ground vehicle (UGV) to assist UAVs in performing SAR missions. Our approach optimizes UAV task splitting, computing resource allocation, offloading decisions, and flight trajectories to minimize system energy consumption as well as network coverage overlap areas. To tackle the coupling issue between short-term offloading decisions and long-term queue constraints, Lyapunov optimization is utilized. Subsequently, a blockchain-assisted algorithm based on Meta-Federated Learning (MFL) is designed, which can significantly enhance system reliability and security. Simulation experiments conducted in various SAR scenarios demonstrate the superior performance of our approach in terms of convergence speed, energy consumption control, queue backlog suppression, and security compared to baseline methods.