Consensus-enabled and Value-oriented Collaboration in Distributed IoT Systems: Mechanisms, Design, and Implementation
Abstract
The ongoing convergence of Internet of Things (IoT), artificial intelligence and big data analytics has inspired many innovative IoT applications. Enabling these new applications requires accurate and reliable capabilities in data sensing, exchange and processing, which can be best fulfilled by collaborative IoT systems. Nevertheless, the dynamic condition of IoT networks may lead to ever-changing demand and objectives among devices, making it difficult for reliable and efficient collaboration. To overcome these challenges, this thesis develops a new framework on consensus-enabled and value-oriented collaboration, which resolves two critical technical challenges, i.e., low latency consensus creation and value-oriented decision-making, to enable collective mindset, promote collaborative behavior, and eventually enhance situation-aware resource sharing in distributed IoT systems.
First, consensus creates a foundation of collaboration among distributed devices. However, reaching consensus usually involves a time-consuming negotiation process, which may significantly degrades the system real-time performance. To resolve this issue, a smart futures based resource trading scheme is proposed, which implements resource trading in advance by predicting onsite resource supply and demand and signing futures contracts, so as to avoid the latency for conventional onsite negotiation. Apart from consensus creation, collaboration participants also need to make specific decisions, e.g., resource allocation and task scheduling schemes, based on the time-changing situation of their needs and interests. Conventional decision-making mechanisms focus on the optimization of specific system performance, while overlooking how users actually benefit from the improved performance. We address this issue by a proposed concept of value of service (VoS), which characterizes user-perceived value by a value function and enables value-oriented decision-making to optimize comprehensive functional benefits brought to users under fast-changing system situations.
Finally, the consensus enabled collaboration is implemented in two realistic applications, i.e., 1) a collaborative rendering scheme which opportunistically leverages dynamic IoT resource to offer real-time and high-quality rendering, and 2) a collaborative multi-camera system which offers real-time 3D reconstruction of dynamic scene via optimal viewpoints planning.