In-Space Computing for IoT Data Processing via Low Earth Orbit Satellites

The Internet of Things (IoT) technology has become widespread in numerous domains, especially in remote regions where resources are scarce. The deployment of IoT technology has been facilitated by distributed edge computing (EC) systems spanning terrestrial to aerospace environments. This paper focuses on optimizing data routing, selection of computing nodes, and buffering policies for processing data in space by exploiting Low Earth Orbit (LEO) satellites. We introduce a constrained optimization problem to reduce the combined energy and latency costs associated with space data processing. This problem is structured as a hierarchical decision-making model using the Markov Decision Process (MDP) methodology and is addressed through Reinforcement Learning (RL) techniques. The proposed HRL framework employs two RL agents leveraging Deep Q Networks (DQN) to enhance data processing by optimizing node selection and improving buffering strategies for task execution. Simulation results show that the proposed framework outperforms existing methods in latency and energy efficiency.