Machine Learning techniques to shape Intelligent Human-Centered Cyber-Physical Systems

During the last years of technological development, we are seeing exponential growth in the number of Internet of Things (IoT) devices we interact with daily, which is leading to an increasing pervasiveness of sensing and computation, resulting in the generation of massive amounts of data, which requires scalable and efficient decentralized approaches to storage and processing. Thus, systems and control researchers and computer scientists have been pioneers in developing powerful engineering methods and tools for Cyber-Physical Systems (CPS). CPSs represent the new generation of systems with integrated computational and physical capabilities that can interact with humans lives in many ways and can expand the powers of the world of physical objects through computation, communication and control by making a significant contribution across the most critical domains such as healthcare, energy, finance, industry, transportation, agriculture, waste management, home security, and many others. Despite being considered a nascent technology with significant room for growth, CPS can add more intelligence to social life, making it better. The uncertainty and variability of the environment in which humans have surrounded means that these CPS need to integrate intelligence within them from which they can not only continuously learn over time from historical data but so that humans can leverage them to make correct decisions. In such a context, Artificial Intelligence (AI) technology and, specifically, Machine Learning (ML) and Deep Learning (DL) techniques play an essential role in this area and represent an important step towards continuous improvement of CPS and human’s Quality of Life (QoL) by transforming, effectively, CPS, into “Human-Centered Cyber-Physical Systems” (HCCPS), and engaging humans as part of the system instead of placing them outside by leading, at the same time, their lives towards better well-being. Such HCCPSs with the integration of ML techniques may be called smart HCCPSs, which are expected to be safer, more reliable and efficient, enabling the implementation of a new type of intelligent systems that can perform autonomous actions based on the operating environment and human activities. This doctoral thesis aims to study and design ML and DL algorithms aimed at multi-risk and multi-objective analysis to manage and investigate a new computing architecture to develop the next generation of intelligent HCCPSs. The experimental results, tested on real application contexts like healthcare, building, industry, and economy, all key components of a smart city, prove the promising benefits of the proposed techniques in terms of efficiency and effectiveness.