Integrated UAV-Enabled MEC Network and STAR-RISs for IoT Communications: A Computation Rate Maximization Approach

Abstract

Due to limited computing resources and severe blockage in dense urban environments, uncrewed aerial vehicle (UAV)-enabled mobile edge computing (MEC) faces significant challenges in serving the Internet of Things (IoT) devices. In this article, by deploying multiple simultaneously transmitting and reflecting reconfigurable intelligent surfaces (STAR-RISs), we design a multi-UAV-enabled MEC network for computation offloading enhancement. A sum computation rate maximization problem is formulated under UAV mobility, energy, and reconfigurable intelligent surface (RIS) constraints. By decomposing the mixed-integer nonconvex problem into tractable subproblems, we develop an iterative algorithm that alternately optimizes resource allocation, network association, STAR-RIS phase shifts, and 3-D UAV trajectories. Specifically, closed-form solutions are first derived for power and frequency allocation via dual decomposition under the TDMA framework with maximum ratio combining (MRC) reception. Then, a greedy algorithm with local search is proposed for device–UAV–RIS association, followed by successive convex approximation (SCA) for trajectory optimization. Convergence analyses show that the surrogate objective sequence generated by the proposed block coordinate descent (BCD) algorithm is monotonically nondecreasing for the baseline formulation, and is empirically preserved under the proposed extensions. Robustness is further assessed under low UAV-to-RIS Rician-factor fading, external co-channel interference with an interference-aware trajectory, and rotary-wing propulsion-energy feasibility. Generality is validated on a representative mixed UAV fleet configuration, and performance is benchmarked against demonstration-initialized deep reinforcement learning baselines (TD3 and SAC). Extensive numerical results validate significant performance improvements over baseline schemes, confirming the effectiveness of joint STAR-RIS assistance and 3-D trajectory design.