Deep Reinforcement Learning Based Energy Consumption Optimization in UAV-Aided MEC Networks

In recent years, the deployment of unmanned aerial vehicles (UAVs) equipped with mobile edge computing (MEC) servers to provide computing resources to ground users has become an emerging technology. For the UAV-assisted multi-user mobile edge computing system, a model is formulated to minimize users' average energy consumption. It involves optimizing both the UAV's flight trajectory and the scheduling of user computing strategies. Deep reinforcement learning (DRL) was used to solve the energy consumption optimization problem, and an optimization algorithm based on Soft Actor-Critic (SAC) was proposed. The algorithm employs maximum entropy to explore optimal strategies and utilizes efficient iterative updates to obtain them, enhancing exploration ability and accelerating convergence in the training process by retaining high-return strategies. Simulation results demonstrate that compared to existing algorithms, it effectively reduces average user energy consumption while maintaining good stability and convergence.