Joint communication, computation, and trajectory optimization for UAV-enabled mobile edge computing in smart city IoT networks

Smart-city Internet of Things (IoT) devices generate latency-sensitive computation tasks but have limited processing and energy resources. UAVs Computing (MEC) servers can serve them, yet probabilistic air-to-ground propagation, nonlinear rotary-wing propulsion, strict per-task deadlines, backhaul availability and finite onboard energy make the joint coordination of communication, computation and mobility difficult. We formulate a mixed-integer nonconvex problem that jointly determines task admission, binary edge/cloud execution, onboard Central Processing Unit (CPU) allocation and horizontal trajectory, minimizing a normalized combination of device and UAV mission energy under single-user Time-Division Multiple Access (TDMA) access, same-slot end-to-end deadlines, computation-capacity, backhaul-availability, mobility, minimum-completion and battery-reserve constraints. The optimal frequency is shown to admit a closed form, reducing the integer block to a binary linear program solved to proven optimality; the trajectory is updated by a constrained trust-region nonlinear program using the original expressions. Every accepted state satisfies the original constraints, and the best-so-far objective sequence converges. Over 50 realizations with 12 devices and 120 slots, the method attains a mean normalized objective of 0.334, reducing it by $$11.10%$$ , $$7.16%$$ , $$7.02%$$ and $$0.98%$$ against local-only execution, fixed CPU allocation, an estimation-of-distribution algorithm and a straight-line trajectory (Holm-adjusted p=3.02× 10⁻⁹ ); a cloud-only design is feasible in only 5 of 50 cases. Code and data are openly available.