DRED: A DRL-Based Energy-Efficient Data Collection Scheme for UAV-Assisted WSNs

Abstract

In Wireless Sensor Networks (WSNs), sensors collect and transmit information to the sink node through single-hop or multi-hop wireless communication links. However, the traditional static sink node solution will cause the hotspot problem due to the energy limitation of sensor nodes. To alleviate the above problem, the Unmanned Aerial Vehicle (UAV)-assisted WSNs, which employs a UAV as the sink node, is proposed to flexibly adjust the routing scheme and prolong the lifetime of sensor nodes. However, the movement of the UAV needs to adapt to the sensor nodes' energy consumption during the transmission in the WSNs, which is a challenging task. Therefore, we propose DRED, an energy-efficient data collection scheme for UAV-assisted WSNs, to control the dynamic routing and the movement of the UAV based on Deep Reinforcement Learning (DRL). The simulation results show that DRED can achieve high network performance in terms of network lifetime.