Practical Learning of Artificial intelligence on the Edge for IndusTry 4.0

Unipi Team Leader: Daniele Mazzei, Department of Computer Science 

The project aims at filling the gap between scientific research on Artificial Intelligence (AI) and Machine Learning (ML) and its industrial application as enabling technology for the I4.0 paradigm. AI and ML improve the data acquisition and analysis typical of the I4.0, leading to the optimization of the industrial processes through fast and well-performing algorithms. The academic research efforts on AI have followed a trend of development of complex algorithms that require cloud-centric architectures while the industrial architectures for data acquisition are in most cases fragmented and resource-constrained. Recent researches have demonstrated the need of a decentralized use of AI and ML where algorithms for data acquisition and analysis are executed directly on the machine side. It is evident that a new generation of AI and ML experts, able to adapt these technologies to the industrial needs and to foster their role as the key players of the 4th industrial revolution, is needed. PLANET4 enables a knowledge transfer between academia and industry by achieving the following objectives: a) design of a b-learning course for the porting and integration of AI techniques in I4.0 applications;b) evaluation of a novel method for the description of industrial digitalization needs and pains aimed at enabling fast identification of the most appropriate AI methodologies; c) formalization of a framework of soft skills and related training materials for 4.0 Innovation and Change Management training workshops;d) development of a portal for the collection and sharing of best practices in the applications.The project approach is cross-disciplinary and focuses on both hard skills in AI and ML technologies and soft competencies needed to manage the changes introduced in the industrial ecosystem. Academics will have the possibility to gather needs and requirements from the industrial world, allowing the adaptation of ML teaching to better fit the real-world industrial pains.