Deep Reinforcement Learning Based Energy Efficient Multi-UAV Data Collection for IoT Networks

Unmanned aerial vehicles (UAVs) are regarded as an emerging technology, which can be effectively utilized to perform the data collection tasks in Internet of Things (IoT) networks. However, both UAVs and sensors these networks energy-limited devices, necessitates energy-efficient procedure ensure network lifetime. In this paper, we propose a multi-UAV-assisted network, where fly ground control sensor's transmit power during time. Our goal is minimize total energy consumption sensors, needed accomplish mission. We formulate problem into three sub-problems single UAV navigation, sensor well multi-UAV scheduling model each part finite-horizon Markov Decision Process (MDP). deploy deep reinforcement learning (DRL)-based frameworks solve part. Specifically, use deterministic policy gradient (DDPG) method generate best trajectory for obstacle-constraint environment, given its starting position target sensor. also DDPG collection. To schedule activity plans visit multi-agent Q-learning (DQL) approach by taking on path account. simulations show that find safe optimal their trips. Continuous achieves better performance over fixed approaches terms addition, compared two commonly used baselines, our framework near-optimal results.