In this dissertation, energy-efficient scheduling and optimization algorithms are proposed for unmanned aerial vehicle (UAV) mobile networks. The design and implementation of scheduling and optimization are challenging especially for the energy-limited UAV networks. Therefore, this dissertation proposes novel algorithms for solving the following three challenging problems in energy-efficient scheduling. First of all, a multi-UAV charging scheduling algorithm is proposed in terms of joint optimization between scheduling and energy delivery. This algorithm computes message-passing based scheduling by deactivating selected surveillance Closed-Circuit Television (CCTV) cameras located in overlapping areas. After this computation, optimal matching between UAVs and charging towers is performed. Second, an algorithm for joint UAV scheduling and deep learning-based active energy sharing among charging facilities is proposed. The cooperative energy sharing among towers is designed via multi-agent deep reinforcement learning, and thus intelligent sharing can be realized. Lastly, an adaptive learning computation outsourcing algorithm is designed in distributed big-data outsourcing systems. In order to make the outsourcing decision when a single UAV cannot conduct deep learning computation alone, the edges which can compute the learning computation instead of the UAV can be scheduled under the concept of max-weight. After the scheduling, Lyapunov-based transmit power allocation is considered for stabilized time-average UAV energy consumption minimization in order to deliver the learning data to scheduled edges. Based on the proposed three energy-efficient scheduling and learning algorithms, UAV mobile networks can extend their lifetime for scalable, flexible, and robust operations.
