Path planning of multi-UAVs based on deep Q-network for energy-efficient data collection in UAVs-assisted IoT

In recent years, Unmanned Aerial Vehicles (UAVs) can effectively alleviate the problems of unstable links and low transmission efficiency, which have been applied for Wireless Sensor Networks (WSNs) to speed up data collection and transmission. However, when the multiple UAVs collect data onto the same area, there is a problem of overlapping coverage areas, which will result in low energy efficiency. Therefore, this paper studies the energy-efficient collaborative path planning problem to maximize data collection of UAVs from distributed sensors. Based on built multi-UAVs assisted system for collecting sensors data, we formulate the optimization objective to maximize the data collected by the UAV group within the limits of energy and the total covered area. To solve the problem of UAVs' collaborative path planning, we propose a Hexagonal Area Search (HAS) algorithm, which is combined with multi-agents Deep Q-Network(DQN), called HAS-DQN. By limiting the total coverage of UAVs, HAS-DQN can effectively avoid collision problems with UAVs. Experiments show that HAS-DQN can effectively solve the path overlap problem of multiple UAVs moving at the same cost in an unknown environment.