Intelligent AoI-energy optimization for IoT over LEO satellite networks.

Abstract

Low Earth orbit (LEO) non-terrestrial networks (NTNs) provide direct satellite access for remote energy harvesting Internet of Things (EH-IoT) monitoring, but status update scheduling is constrained by energy-causality at the nodes and intermittent satellite coverage. We study a contact aware weighted AoI–energy cost minimization problem, where scheduling decisions determine whether each node defers or transmits by selecting the feasible LEO link. We cast this sequential scheduling problem as a Markov Decision Process (MDP) and solve it using an intelligent lightweight digital-twin assisted proximal policy optimization (DT-PPO) framework. The digital twin augments the MDP state with look-ahead contact and link information, enabling the PPO agent to minimize long-term AoI–energy cost. A safety layer then ensures that each selected action satisfies update availability, battery feasibility, visibility, contact-time, and satellite-admission constraints. Simulation results based on LEO contacts and real energy-harvesting traces show that the proposed DT-PPO policy achieves a better freshness–energy tradeoff than myopic link based scheduling and reinforcement-learning baselines.