Spectrum-Energy Efficient Sink-AAV-LEO Backbone Networks to Ensure Ubiquitous Data Collection Timeliness

The rise of low Earth orbit (LEO) satellites and autonomous aerial vehicles (AAVs) has facilitated the integration of space-air–terrestrial networks, enabling ubiquitous data collection in Internet of Remote Things (IoRT) networks. However, limited battery capacity of sinks/AAVs and scarce spectrum in these networks make it challenging to build spectrum-energy efficient backbone networks for timely data collection. In this article, we explore the resource allocation problem in Sink-AAV-LEO-based data collection scenarios, where sink nodes gather data from sensors, and AAVs relay it to LEO satellites. To ensure data collection timeliness while optimizing energy and spectrum use, we focus on optimizing the three-hop network structure, power control for sink nodes, path selection and power control for AAVs, and bandwidth allocation for AAV-LEO links. Although these tasks should theoretically be modeled as a global optimization problem, solving them within tolerable time complexity is difficult. Therefore, we first establish a three-hop network using our proposed optimization mechanism. Then, we describe the Sink-AAV and AAV-LEO resource allocation problems, and propose the corresponding solutions via utilizing our improved deep reinforcement learning algorithm, namely ISS-proximal policy optimization. Finally, we solve the global problem based on the solutions to the above two local problems. Besides, our simulation results are compared with various benchmark schemes, validating the effectiveness and superiority of our proposed solutions across different perspectives.