Joint Scheduling and Adaptive Compression for Minimizing the Age of Distorted Information in Satellite-Terrestrial Networks
Abstract
Recent advances in Low Earth Orbit (LEO) satellite constellations enable global connectivity, particularly for remote, infrastructure-scarce regions. For real-time monitoring in such environments, effective visual monitoring requires both high information freshness and visual fidelity. Moreover, practical deployment relies on terrestrial-satellite terminals (TSTs) that are powered by finite batteries and stochastic ambient energy harvesting (EH). Transmitting under these constraints necessitates balancing information freshness against compression-induced visual distortion. To address this, we propose a Deep Q-Network (DQN) framework that jointly optimizes transmission scheduling and compression bit rate to minimize the Age of Distorted Information (AoDI), a metric that jointly considers timeliness and fidelity. We develop both distributed and centralized algorithms, which can flexibly accommodate diverse network settings and LEO satellites with varying computational capabilities. Numerical simulations demonstrate that the proposed methods reduce time-averaged AoDI compared to representative baselines, and the centralized algorithm outperforms the distributed one with additional signaling overhead.