Off-campus UNL users: To download campus access dissertations, please use the following link to log into our proxy server with your NU ID and password. When you are done browsing please remember to return to this page and log out.
Non-UNL users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
Distributed Artificial Intelligence over Edge-Assisted Internet-Of-Things
Abstract
Serving as the bridge between physical and cyber world, Internet-of-Things (IoT) connects a sheer volume of objects and humans with the functions of communication, storage, learning, and control. IoT collects big data from physical environment and processes data with Artificial Intelligence (AI) on cloud server, the edge of network and user devices to make complex decisions for interaction with users or controlling devices. By distributing AI over its architecture, IoT gains stronger learning ability than using centralized AI thanks to the offloaded computation, reduced network traffic and quicker responses. However, the adoption of distributed AI over IoT still encounters serious issues from the perspective of resource constraints. Most edge nodes and IoT devices have limited hardware capabilities, network bandwidth and power supply. Their constrained resources in computation and communication raises challenges to deploy distributed AI over IoT, since they can hardly support as sophisticated AI as on cloud server. In this dissertation, we improve the efficiency of resource utilization for distributed AI over IoT to guarantee sustainable operation of IoT systems. We firstly discuss the background of AI-driven techniques that empower IoT systems in perceiving, learning and behaving abilities. Then, we identify three key questions to be addressed for improving efficiency of distributed AI over IoT applications. After that, we investigate the effectiveness and efficiency of utilizing AI on IoT by proposing a fast yet accurate AI-driven modeling method for energy status estimation of IoT devices. In addition, we discuss tradeoff balancing between AI model accuracy and modeling overhead over IoT architecture with the proposed adaptive AI framework. Furthermore, we discuss the user diversity issue in collaborative learning on the edge of IoT that requires both efficiency and personalization from AI model training. Finally, we draw conclusions of this dissertation and identify some open directions for the future research in AI-empowered IoT.
Subject Area
Engineering|Artificial intelligence|Computer Engineering
Recommended Citation
Shen, Shuaiqi, "Distributed Artificial Intelligence over Edge-Assisted Internet-Of-Things" (2022). ETD collection for University of Nebraska-Lincoln. AAI29319224.
https://digitalcommons.unl.edu/dissertations/AAI29319224