Service Capacity Maximization in Multi-UAV MEC: Deployment, Offloading, and Resource Allocation

To address the increasing computation-intensive tasks, we consider a multi-unmanned aerial vehicle (UAV)-enabled computing offloading system and aim to explore its maximum service capacity. The existing solutions aim to optimize user service experience in terms of delay or energy efficiency, but this fundamentally assumes that all tasks are completable. We highlight that all computational tasks in practical application environments have specific delay constraints. As such, it is unrealistic to consider that all tasks can always be completed on schedule when resources fall short of demand. To this end, we propose a service-maximized bilevel (SMax-bilevel) optimization scheme, and formulate a task completion rate maximization problem. Due to its non-convex complexity, we decompose the original problem into a tractable bilevel structure. At the upper level, an encoding mechanism-enhanced adaptive differential evolution algorithm (ADEEM) is developed to determine UAV deployment locations that maximize service coverage. At the lower level, a theorem is presented to guide user selection, task offloading, and resource allocation, aiming to maximize the number of completed tasks within their delay constraints. The two levels iteratively interact to produce an optimal solution. Simulation results demonstrate that the proposed SMax-bilevel scheme significantly enhances system service capacity while reducing algorithmic complexity, outperforming existing benchmark solutions.