Semantic task offloading and resource allocation in UAV-assisted mobile edge computing network for disaster rescue

In disaster scenario, the damage of ground base stations leads to the interruption of traditional communication. Unmanned Aerial Vehicle (UAV) equipped with edge servers becomes the key infrastructure for emergency rescue. However, the UAV communication face strict bandwidth constraints, while aerial image transmission and ground search-rescue tasks both have high data redundancy. Simultaneously processing bidirectional UAV-ground tasks faces the dual pressure of delay and energy consumption. To address these challenges, this paper constructs a UAV-assisted semantic edge computing network. We introduce semantic compression technology to reduce the amount of redundant information in aerial images and search-rescue task data. We minimize the system energy consumption by collaboratively optimizing the task offloading decision, semantic compression ratio, edge computing resource allocation, transmission power, bandwidth allocation, and UAV three-dimensional flight trajectory. We decompose the complex non-convex joint optimization problem into five subproblems and propose an alternating optimization strategy to solve them iteratively. Specifically, we propose a branch and bound algorithm based on continuous relaxation and Lagrangian duality to solve the offloading decision subproblem. To solve the optimal semantic compression ratio, we develop an optimization algorithm based on block coordinate descent. We propose a sequential convex approximation algorithm to solve the power and bandwidth allocation subproblem. Then we design an improved artificial lemming algorithm to plan the UAV 3D flight trajectory. Simulation results show that our proposed scheme can effectively overcome the bandwidth bottleneck, significantly reduce the system energy consumption and guarantee the delay requirement of tasks.