Joint Optimization of AAV Trajectory, Task Offloading, and Resource Allocation in AAV-Aided Emergency Response Operations

In emergency response operations, the role of autonomous aerial vehicles (AAVs) has become vital for performing tasks such as human detection. However, these tasks are computationally intensive, and AAVs have limited resources. Therefore, fixed-wing AAV mother-ships with edge servers (UMECs), which usually have more speed and payload capacity than rotary-wing AAVs, can be deployed near AAV scouts (generate the tasks) to provide processing services. However, the design of fixed-wing AAVs requires adherence to a minimum turning radius, which limits maneuverability and impacts both energy consumption and task execution time, thus necessitating careful trajectory planning. Furthermore, UMECs may lack the capacity to process all types of tasks due to memory restrictions and require both central processing units and graphics processing units for faster execution of tasks. Therefore, this article aims to minimize the energy consumption in the network and task execution time by jointly optimizing the trajectory of UMECs, task offloading, and resource allocation (TTR) problem while taking into account UMECs movement constraints, task performance limitations, and computational resource budgets. In this regard, we first propose a particle swarm optimization (PSO)-based TTR scheme (P-TTR), where a PSO algorithm finds the trajectory solution, and a block coordinate descent-based method is developed to solve the task offloading and resource allocation problem. Then, for real-time decision-making, a deep-reinforcement-learning-based algorithm, namely soft-actor–critic-based TTR (S-TTR), is introduced to address the problem, which can adapt to dynamic environments and have lower computational overhead. Simulation results confirm that the proposed approaches can outperform the baseline method.