PRIN PNRR 2022 Digit4Circle
DIGItal Twins for CIRCuLar Economy (DIGIT4Circle)
DIGItal Twins for CIRCuLar Economy (DIGIT4Circle)
Digit4Circle advocates for a middleware architecture to federate domain-specific application islands and a DT model that can be applied to circular economy settings. In specific, the architecture itself combines different levels of applications of different domains of information technologies, aiming to: (i) develop a virtualized platform and mechanisms allowing the on-demand federation of resources (e.g., production units so to accomplish the process goals) and convergence of digitized models to represent the distributed production process. Dynamicity is an essential feature, allowing for potential adaptations of the processes, relying on localized intelligent feedback; (ii) design an integrated data pipeline and MLOPs practices used to establish possible product recycling and/or reuse, capable of providing predictive and prescriptive receipts, triggering potential process adaptations at runtime. To this end, we will also explore distributed machine learning techniques for the classification of functional behaviours, and detection of anomalies inside the production cycle, also allowing for portability and scalability; and (iii) design a solution for the exchange of information across the different domains involved, which may be structural entities of different kinds (users/customers and producers) as well as multiple ownerships. This requires the creation of a secure dataspace with specific privacy-ensuring techniques to enable a seamless exchange of information and incentivize the adoption of the proposed techniques.
Given this context, the teams have proactively addressed the primary innovation pillars from the outset, focusing on key building blocks that holistically encompass all the identified innovation objectives. Their work has ensured a coherent and comprehensive approach to tackling these challenges. Specifically, the group adopted an iterative approach, refining and expanding upon earlier design decisions, enhancing the functionality and robustness of the developed framework and mechanisms. During the project the emphasis was on the following areas: (i) Architecture Design and Resource Cooperation Mechanisms: efforts were dedicated to optimizing system architecture and enabling efficient, seamless collaboration across diverse resources; (ii) Secure Data Ecosystems and Federated Unlearning: the team prioritized developing secure ecosystems for data sharing, coupled with advanced techniques for federated unlearning to enhance privacy and adaptability; and (iii) Data Acquisition Procedures: concerning the efficient, decentralized & cooperative data gathering procedures enacted in order to acquire visibility on the (production) process, and more generally a monitored phenomenon. Additional details on the activities and (planned) scientific output during this Q3 are provided throughout this document.
Virtualized environments align distributed production units to process goals.
Integration of MLOps for predictive analytics, reuse potential, and process adaptation.
Privacy-preserving dataspaces for seamless, incentivized collaboration across stakeholders.