Sustainable Artificial Intelligence for Sustainable 6G

Our goal is to address the sustainability of developing and using AI systems and to promote the use
of AI towards the sustainable development goals. The priority is therefore on the design of AI systems for sustainable development while targeting the sustainability of AI training and usage. Our research is multi-disciplinary in nature and spans several research areas including data science, computer science, network science, information engineering, wireless communications, energy engineering, environmental engineering and remote sensing.
The core focus is the definition, design and implementation of energy-aware, high-accuracy and interpretable ML methods for networked cyber-physical systems. The is to provide new methodological approaches for highly efficient learning systems from the theoretical definition
up to the implementation into computing platforms, being here the main objectives: decentralization (e.g., federated learning, continual learning, transfer learning), interpretability, security and energy-efficiency.
The common ground of our research is ’sustainable design’: our methods are intended for supreme
efficiency and not only for high accuracy, as normally done in state-of-the-art ML. For this reason, we usually include complexity and energy assessment of the computational models proposed as our research output, and advocate highly efficient computing hardware, such as neuromorphic and reservoir computing. Our methodology is general, flexible and scalable to be applied to a vast range of sectors and verticals, e.g., artificial intelligence of things (AIoT), smart cities, automotive, energy, health, factories of the future, civil protection. More than this, we devise our research to support the deployment of digital twins, facilitate the simulation of real conditions of the underlying physical world, what-if scenarios and visualize the outcomes instantly, having in mind the UN Sustainable Development Goals as targets.