Satellite-Edge Computing for Monitoring Remote Natural Areas of Biological Relevance

This paper addresses the key issue of fostering biodiversity conservation by designing a framework for the long-term management and monitoring of remote natural areas of biological relevance. The approach is based on a distributed network of smart sensors that collect relevant environmental signals using high-sensitivity microphones to monitor ambient sounds through ecoacoustic analysis. Then, the gathered data are processed by means of deep learning techniques to automatically detect and analyze relevant information on the fly. Since protected areas are, in most cases, remote and likely not served by terrestrial telecommunication infrastructure, our solution exploits non-terrestrial networks (NTN) to provide both data delivery and elaboration in such a scenario. This paper analyzes the feasibility of relying on a NTN based on Low Earth Orbit (LEO) satellites for environmental data transmission and elaboration in support of biodiversity conservation. Moreover, the performance improvement that can be achieved by means of the envisioned satellite-edge offloading solution is evaluated with respect to local computing at the sensor nodes.