MIT Portugal SMART project is the winner of AI Moonshot Challenge

SMART will use a physics-guided deep learning approach to combine large amounts of satellite data with high-resolution ocean models, in-situ validation, and citizen science to create probability of plastic occurrence maps. This will allow authorities to create strategies for ocean remediation while making decisions under uncertain conditions. 

The Project PI, Leonardo Azevedo, from CERENA – Técnico, highlights the innovative and creative approach of the proposal by using advanced methods of  Machine Learning (ML) and Artificial Intelligence (AI) with particular focus on “Physics Informed GANs and Deep Neural Networks.” Leonardo Azevedo explains that the integration of satellite image data from the Copernicus program (ESA) with advanced models that simulate the behavior of the ocean, ML / AI methods, and direct observations from autonomous vehicles marine species will result in probability of plastic occurrence maps: “It’s our desire that the resulting maps will be a tool to help local, regional and global  authorities to implement strategies to mitigate the problem of accumulation of debris, and in particular plastics, in the ocean.” 

Over the summer, the AI Moonshot Challenge was calling for proposals on fighting Earth’s problems with Artificial Intelligence and Space tech. MIT Portugal assembled a team of Portuguese and MIT researchers to tackle this global problem. “SMART is a result of MPP’s effort in the strategic domain of Earth Systems and Data Science,” says the National Co-Director of MIT Portugal, Zita Martins, who is also a member of the project’s research team. 

The SMART consortium gathers 12 Portuguese researchers from Portuguese research and academic institutions and 6 MIT researchers with diverse but complementary expertise to tackle the exciting challenge of identifying and tracking floating plastic debris from satellite images. Leonardo Azevedo, project’s PI, highlights the multidisciplinary and scientific quality of the team promoted by “a generation of young researchers from different areas: space sciences, ocean sciences and oceanography.”

The project leader is the Center for Natural Resources and Environment (CERENA) – Técnico) and will contribute with the experience acquired by Centro de Química Estrutural (CQE -Técnico), the Associated Laboratory Center for Environmental and Sea Studies(CESAM) from Aveiro University  (UA), the Underwater Systems and Technology Laboratory, (LSTS) from the Engineering Faculty of Porto University (FEUPorto), and the Portuguese Hydrographic Office (IH).
To complement these resources, the SMART consortium also includes students and faculty from the MIT Portugal Program.These partners will contribute with invaluable experience and skills in fields related to deep-learning and physics-informed machine learning and will comprise regular visits from MIT to the different Portuguese partners.
This proposal has also the support of MACC – Minho Advanced Computing Center, from IBM, EPhysLab from the University of Vigo and the Portuguese Institute of the Sea and the Atmosphere.

The SMART project will be conducted entirely from Portugal with the rest of the academic and research institutions, for a period of up to two years. 

AI Moonshot Challenge is a global competition, promoted  by The Foundation for Science and Technology (FCT), Portugal Space Agency (PT Space), European Space Agency (ESA) and Unbabel Labs through the Portuguese Innovation Agency (ANI) to advance monitoring of plastic litter on a planetary scale.  The AI ​​Moonshot Challenge will return in 2021. 

Watch the SMART proposal pitch at the Web Summit here.