Joint Computation Offloading and Trajectory Planning for UAV-Assisted Edge Computing

With excellent flexibility, unmanned aerial vehicles (UAVs) can act as airborne computing servers to assist smart terminals (STs) with their computationally-intense and delay-sensitive tasks. This paper presents a new UAV-assisted edge computing framework, which jointly optimizes the trajectory and CPU frequency of a fixed-wing UAV, and the offloading schedule to minimize the energy consumption of the UAV. The key idea is that we reveal the condition for the convexity of the optimization, when the UAV flies a linear trajectory. Under the condition, alternating optimization-and successive convex approximation (SCA)-based algorithms are developed to efficiently achieve the globally optimal linear trajectory, CPU configuration, and offloading schedule. Another important aspect is that we prove the SCA-based algorithm can achieve a local optimum satisfying the Karush-Kuhn-Tucker (KKT) conditions, when the revealed condition is unmet or the UAV flies horizontally in two dimensions. By analyzing the KKT conditions, we also unveil the underlying patterns for the optimal CPU frequency and offloading schedule. Extensive simulations validate the patterns and corroborate the merits of our schemes