Deep Reinforcement Learning Based Data Collection in IoT Networks

Abstract

Unmanned aerial vehicles (UAVs) are an emerging technology that can be effectively utilized to perform data collection tasks in the Internet of Things (IoT) networks. However, both the UAV and the sensors in these networks are energy-limited devices, necessitating an energy-efficient data collection procedure to ensure network lifetime. In this paper, we consider a UAV-assisted network, where a UAV flies to the ground sensors according to a predetermined schedule and controls the sensor’s transmit power when hovering above the sensor. Our goal is to minimize the total energy consumption of the UAV and the sensors, which is needed to accomplish the data collection mission. We formulate this problem into two sub-problems of UAV navigation and sensor power control and model each part as a finite-horizon Markov Decision Process (MDP). We deploy the deep deterministic policy gradient (DDPG) method to generate the best trajectory for the UAV in an obstacle-constrained environment and to control the sensor’s transmit power during data collection. Our simulations show that the UAV can find a safe and energy-efficient path for each trip. In addition, continuous sensor power control achieves better performance against the fixed-power and fixed-rate approaches in terms of the total energy consumption during data collection.