A Task-Aware Federated MEC Architecture Assisted by UAVs and Robotic Dogs for Post-Disaster Emergency Medical Rescue

Existing unmanned aerial vehicle (UAV)-assisted edge computing (EC) architectures in post-disaster rescue operations lack integration of artificial intelligence (AI)-powered tasks, and fail to systematically consider the differences in computing resource demands and delay sensitivity among various types of computing tasks. Moreover, the computing resources of multiple EC server nodes are not fully utilized, which significantly restricts the responsiveness and intelligence level of rescue systems. To address these challenges, we proposed a task-aware federated mobile edge computing (MEC) architecture, employing UAVs and robotic dogs as carriers for environmental perception, scenario analysis, communication and computing. At the same time, we established a system model that better approximates real-world rescue scenarios and proposed a federated offline-online transformer-based resource management (FO2T-RM) method to minimize system energy consumption while satisfying heterogeneous task latency requirements. Simulation results validate that the proposed method can achieve superior system performance compared to the other benchmark schemes.