Trustworthy aerial edge computing with robust, priority-aware, and learning-based optimization for 6G networks

Abstract

Aerial edge computing serves remote and disaster-affected regions where terrestrial coverage is sparse or unavailable. Trust in these deployments reflects the consistency with which an uncrewed aerial vehicle (UAV) fulfills the tasks it accepts. The existing literature generally develops trust measures without explicitly connecting them to the admission of Internet of things (IoT) devices. These measures typically associate trust with regions, data contributions, or device populations. IoT device admission meanwhile focuses on UAV resources such as coverage, residual energy, and computing capacity. This resource-oriented view describes what a UAV can offer rather than what it has reliably delivered. Offloading commits each device to the service the selected UAV provides, which makes trust an important consideration in aerial edge computing. This thesis proposes a trust-aware aerial edge computing framework that delivers trustworthy mobile edge computing (MEC) from UAVs to IoT devices. The framework evaluates each UAV through multiple trust dimensions and admits an IoT device only when the selected UAV meets the trust requirement of its task. We formulate joint UAV deployment, device association, and resource allocation as a multi-criteria mixed-integer nonlinear program (MINLP) balancing connectivity, trust, cost, and task latency. A penalty-guided optimization (PGO) algorithm based on sequential quadratic programming (SQP) solves the resulting problem and achieves near-optimal connectivity relative to branch-and-bound at substantially lower computational cost. We extend the framework to uncertain IoT device states, differentiated task requirements, and device mobility. Uncertainty in residual energy and device location enters the IoT-UAV association decision as an explicit input. Task priority accompanies trust within the same association decision and captures the deadline-sensitive requirements of individual tasks. IoT mobility follows a Gauss-Markov process within the extended problem formulation. UAV deployment and the associated decision-making proceed under centralized, hybrid, and fully distributed control schemes. Connectivity remains near the optimal benchmark under both device-state uncertainty and task heterogeneity. Mobile settings reveal a trade-off between accumulated reward and trust assurance across the three control strategies. Trust therefore functions as a decision criterion that complements resource availability, task requirements, and network dynamics in aerial edge computing.