UAV 3D Trajectory Design for IoRT Data Collection against Malicious Jamming

As an emerging field beyond the Fifth-generation (5G) mobile communication technology, the space-air-ground (SAG) integrated networks provide a promising data collection paradigm for Internet of remote things (IoRT) networks where terrestrial infrastructures are usually unavailable. In this paper, we explicitly consider the presence of malicious ground jammers (GJs) in the IoRT networks, where a fixed-wing UAV is dispatched to collect data stored in ground sensors (GSs) with the aid of LEO satellites. And we exploit a more practical aerodynamic model of UAV and extend the data collection scenario to a 3D environment. To solve this continuous control and long-term sequential decision-making problem, we propose a learning-based IoRT data collection against jamming (LIDCAJ) framework, which contains a deep neural network (DNN) model based on DRL for policy learning, and a temporal sequence model based on LSTM for historical trajectory modeling. Furthermore, we conduct extensive simulations to verify the effectiveness of the LIDCAJ. Compared with the state-of-the-art methods, our proposed approach is superior in the aspects of energy efficiency, data uploaded rate, and geographic fairness.