Autonomous aerial vehicles (AAVs) with wireless power transfer (WPT) technology offer a promising solution to extend the lifetime of wireless rechargeable sensor networks (WRSNs) by swiftly recharging multiple sensor nodes. Existing strategies often prioritize optimizing the AAV trajectory and charging schedule. While overlooking diverse energy needs and spatial distribution of sensor nodes, many approaches rely on positioning the AAV at the center of the region above a fixed height. This also leads to potential energy wastage by unnecessarily recharging already adequately powered nodes near the AAV. A key challenge lies in optimizing the 3-D position of the AAV to maximize energy harvested by sensor nodes, considering their diverse energy demands and spatial distribution. To tackle this challenge, we propose a novel approach that jointly optimizes the 3-D deployment of the AAV and energy harvested by sensor nodes. We formulate this complex, nonconvex problem as a Markov decision process (MDP) and leverage deep reinforcement learning based on the soft actor-critic (SAC) algorithm to address it. The AAV learns to navigate continuous state and action spaces, enabling it to discover optimal hovering locations. Our novel reward function allows the AAV to prioritize energy-critical sensor nodes and regions with higher sensor node density, maximizing energy harvesting within limited hovering times. Simulation results demonstrate that our proposed method significantly outperforms conventional AAV deployment strategies, substantially improving sensor node energy levels and network sustainability.

Vehicular fog computing (VFC) has emerged as a promising research paradigm to address the demands of vehicular application requests. The dynamic nature of VFC poses challenges in distributing fog computing resources to the vehicular tasks. In VFC task scheduling problems, the optimization of the success rate of tasks and minimization of the energy consumption of vehicles ensure efficient completion of critical operations. Moreover, failing to consider the dependencies among the tasks leads to increased delays. Therefore, in this paper, we propose a dependency and mobility-aware intelligent task scheduling (DMITS) mechanism to optimize the success rate of vehicular tasks and energy consumption of vehicles by considering the moving paths of the vehicles, the social relationships among the vehicles and dependencies of tasks. In this study, we model the social relationships among the vehicles based on their communication patterns and social characteristics to improve the success rate of the tasks. We consider the mobility of vehicles using the Markov renewal process (MRP) technique. Initially, we formulate a mixed integer nonlinear programming (MINLP) problem for the proposed problem. Further, we employ a federated deep reinforcement learning mechanism that uses the improved soft actor-critic (iSAC) technique to allocate tasks. We compare our proposed algorithm with existing task scheduling approaches. The simulation results demonstrate that the proposed algorithm performs efficiently in addressing task scheduling problems, achieving higher task success rates and lower energy consumption for vehicles.