Resumen
Environmental sensing is based on long-term measurements of ecological, meteorological, and hydrological variables. Its
aim is to provide the necessary information to understand trends, establish benchmarks, and inform policy. To move from
"environmental sensing" to "environmental intelligence", disciplines such as artificial intelligence (AI), and the Internet of
Things (IoT) must work together to improve the services that can be offered. The main goal of this project is to propose
solutions to seamlessly integrate AI with IoT, overcoming the many challenges, with novel approaches and the rethinking of
the overall architecture for the communication and processing to allow environmental intelligence applications to meet the
requirements in latency, reliability, and sensible use of resources.
The primary objective of MORELLINO is to devise solutions for the seamless integration of AI with IoT, leveraging the
potential of TinyML and edge computing. The size and computational constraints of IoT devices pose a difficulty in
accommodating traditional AI models, necessitating the optimization of these models without compromising their accuracy.
Furthermore, integrating TinyML, designed to run machine learning models on ultra-low power devices, presents a challenge
in ensuring consistent power management and model efficiency. Additionally, as we shift towards edge computing, we deal
with issues related to data synchronization, storage solutions that cater to high-volume, real-time data, and the development
of robust, low-latency and long-distance communication protocols. Ensuring the security and privacy of data processed at
the edge is also a crucial concern. MORELLINO will propose solutions for these technical aspects by re-envisioning the
overarching communication and processing architecture such that environmental intelligence applications will meet
expectations in terms of latency, reliability, and efficient resource allocation. By identifying main gaps in connectivity and
affordable data analytics and through interleaved research, development, and validation in a real-world setting, MORELLINO
will address the challenge of bringing IoT and data analytics systems a step closer to seamless, energy efficient and secure
deployment.
