Computation-Efficient Grouping, Trajectory, and Resource Allocation for UAV Swarm-Assisted Aerial–Ground Collaborative Computing Networks

Unmanned aerial vehicle (UAV) swarms have found widespread applications in executing high-complexity and remote-risk missions. However, the limited onboard resources and energy of UAV swarms may hinder their support for computation-intensive yet delay-sensitive applications, especially when faced with exponentially growing big data. This article focuses on investigating a UAV swarm-assisted aerial–ground collaborative computing system, where one UAV swarm is divided into different groups and collaborates with a remote ground base station (BS) to provide computation services for ground smart mobile devices (SMDs). A comprehensive optimization framework is presented to maximize the system’s computation efficiency by jointly designing group formation, UAV trajectories, and resource allocation. The formulated problem involves a fractional structure with nonlinear coupling of different variables, rendering it highly nonconvex. To address this challenge, we propose a Dinkelbach-based looped iterative optimization (DLIO) algorithm. Specifically, Dinkelbach’s method is initially adopted to reformulate the original problem into a parametric structure, which is then decomposed into subproblems for group formation, resource allocation, and UAVs’ trajectory scheduling. Subsequently, these subproblems are addressed by a looped iterative optimization (LIO) algorithm. The outer loop determines group forming and resource allocation, while the inner loop utilizes the method of successive convex approximation (SCA) to solve trajectory scheduling for UAVs. Simulation results validate the effectiveness of our proposed DLIO algorithm, ensuring rapid convergence and significant improvements in the system’s computation efficiency compared to other benchmarks.