SAI4EO – Symbiotic Artificial Intelligence for Earth Observation

SAI4EO (Symbiotic Artificial Intelligence for Earth Observation) is a research and innovation project exploring the use of generative and multimodal Artificial Intelligence to make the analysis of satellite and geospatial data simpler, more intuitive and more accessible.

The project was developed within FAIR – Future Artificial Intelligence Research, a programme funded by Italy's National Recovery and Resilience Plan (NRRP), specifically under Spoke 6, dedicated to Symbiotic AI, coordinated by the University of Bari Aldo Moro.

Developed through a collaboration between Planetek Italia and the SWAP research group at the University of Bari, SAI4EO applies the concept of Symbiotic Artificial Intelligence to Earth Observation: an approach in which AI works alongside people, enhances their analytical capabilities and facilitates the interpretation of complex information, while keeping humans at the centre of the decision-making process.

The main objective is to develop a conversational AI assistant specialised in Earth Observation, capable of understanding natural language requests, querying geospatial data, analysing satellite imagery and providing answers, maps and insights that are easy to understand, even for non-specialist users.

The research focuses on two application scenarios: detecting and analysing objects in high-resolution satellite imagery, and monitoring land use and land cover changes across European coastal areas, also using data from the Copernicus programme.

Planetek Italia contributes to the project by providing its expertise in Earth Observation data analysis, dataset selection and preparation, use case definition and AI model evaluation. The company is also responsible for developing and validating the conversational AI assistant prototype, which builds on language and multimodal models developed by the SWAP research group.

The research outcomes have paved the way for the evolution of SAI4EO towards Agentic AI, an approach in which intelligent agents work together to perform complex geospatial analyses. Starting from a simple natural language request, the system can coordinate specialised tools to identify and prepare satellite data, perform processing tasks and generate maps, statistics and reports. This evolution aims to make geospatial analysis increasingly accessible, allowing users to focus on their objectives and decisions rather than the complexity of the underlying technologies.

Read more