Towards High-Fidelity Digital Twins in Internet of Things

With the proliferation of devices in every aspect of social activities and production, Internet of Things (IoT) has developed rapidly, interconnecting physical devices embedded with sensors, software, and other technologies. It enables devices to collect and exchange data over the Internet and facilitates automation and intelligent decision-making. With the recent wave of digital transformation, more and more IoT entities can be characterized by their digital copies. In this context, Digital Twin (DT) technology has emerged as a promising bridge between physical and virtual realms that enables the synergy of various things. Despite ongoing exploration of DT in extensive domains, current implementations often suffer from low fidelity due to a lack of standards in DT construction, utilization, and management. In this dissertation, we explore the development of high-fidelity DT in IoT scenarios leveraging multiple advanced technologies. First, we propose a Mobile Edge Computing (MEC)-based Human Activity Recognition (HAR) DT deployment scheme in the smart home domain. In particular, our proposed scheme tracks the essential data with Deep Reinforcement Learning (DRL) to maintain sensor update consistency between physical and digital twins while being energy-efficient in different residential environments. The second study further extends the previous HAR scenario and explores the potential for IoT startups to overcome cold-start challenges with limited business knowledge and seed budget. By utilizing every resource of digital assets, especially the trust DT, IoT startup initiates services in a trustworthy blockchain environment with the assistance of expertise-diverse resource providers and trustworthy advice from expert IoT companies. Thirdly, we propose a sustainable data collection and management approach for data acquisition of DTs in the balance between enduring data collection and the information loss associated with stale data. Our proposed method optimizes the metrics of data fidelity and reveal delay to ensure both sustainable energy and sustainable information, which could be further organized in a decentralized and distributed manner. Furthermore, as the pioneering effort in establishing a comprehensive DT deployment pipeline within the IoT domain, this study illuminates the promising prospects of DT implementation across sectors and lays the groundwork for the future Internet paradigm of Metaverse.