Priority-oriented trajectory planning for UAV-aided time-sensitive IoT networks

Abstract

Unmanned Aerial Vehicles (UAVs) have been widely employed in the Internet of Things (IoT) networks due to their high mobility and high probability of line-of-sight (LoS) propagation. Equipped with certain payloads, UAVs are able to gather data from sensors located in a particular area where no ground base station is available for transmitting data, such as oceans and mountains. However, for a time-sensitive network, the latency has to be minimized, especially in heterogeneous scenarios where each sensor has its own latency tolerance, which emphasizes the importance of trajectory design of UAVs. In this paper, we propose a priority-oriented trajectory planning problem for a UAV-aided time-sensitive heterogeneous IoT network, based on which we provide a solution for satisfying the latency tolerance of the network within a given period of time. Aiming at optimizing trajectories, we employ continuous Deep Q-Learning Network (DQN) which is proven to be capable of identifying a relatively optimal trajectory compared to the benchmarks through a large number of experiments. Simulation results are provided for demonstrating that the proposed DQN-based algorithm outperforms the benchmarks. More particularly, the proposed DQN-based algorithm is capable of achieving in excess of 49% and 10% improvements in system costs over the greedy algorithm and Q-Learning algorithm, respectively.