Fuzzy Logic-Based Orchestration of Multi-Access Edge Computing in LEO Satellite Constellations

In the fast-evolving domain of satellite communications, massive constellations of LEO satellites are increasingly gaining notice in both academic and industrial sectors. Operating these growing constellations has become a complex challenge, and new obstacles have emerged with integrating cellular networks in the space sector. Some techniques like Multi-access Edge Computing (MEC) were initially designed for static terrestrial networks and should be adapted to the dynamic nature of Non-Terrestrial Networks (NTN). This study introduces an innovative Fuzzy Logic-based controller tailored explicitly for determining the operational status of satellite MEC servers based on historical task loads and task processing failures. Our approach is grounded in a distributed control framework, where each satellite node operates autonomously relying on its own data and limited information from neighboring nodes. We evaluated our fuzzy controller against a baseline strategy, demonstrating significant improvements in reducing the overall active time of MEC nodes. The orchestration effectiveness of these strategies was tested by using our custom-built continuous time emulation framework for satellite constellations. This framework integrates 1) the Simplified General Perturbations Model 4 (SGP4) for orbit propagation, 2) Mininet for network virtualization, 3) Docker for operating system virtualization, and 4) MEC task offloading and orchestration.