A High-Level Data Decentralized Processing Platform for AIoT Applications

Tools

Tong, Kit-Lun (2025) A High-Level Data Decentralized Processing Platform for AIoT Applications. Doctoral thesis, University of East Anglia.

[thumbnail of KL_Tong_Thesis_redacted.pdf]
Preview
 PDF
Download (10MB) | Preview

Abstract

Artificial Intelligence of Things (AIoT), the fusion of Internet of Things (IoT) and Artificial Intelligence (AI), is changing manufacturing and navigation alongside many other industries. However, complexities in device scaling, data management, and lack of skilled personnel hinder the wide adoption of AIoT. A high-level IoT platform integrates communication protocols, databases, and application program interfaces (APIs). These centralize the management of an IoT solution to make it simpler to develop and deploy applications while also addressing device communication, data processing, and security factors.

On the other hand, traditional IoT platforms are often cloud-based or data-centralized, suffering inefficiencies in routing and process scaling limitations. To address these problems, we aim to develop DAIoTtalk, an AIoT platform with a data-decentralized architecture that builds upon IoTtalk. Additionally, DAIoTtalk supports flexible networking and low-code configuration by enabling gRPC-based Pub-Sub communications.

To showcase its applicability, we created case studies defined by different industries: SewingTalk and GNSS-EStalk. SewingTalk improves the productivity of textile manufacturing by analyzing logs of smart sewing machines using unsupervised learning to estimate daily completion. Meanwhile, GNSS-EStalk identifies sources of GNSS errors using an AI-driven temporal-spatial approach.

Through a series of experiments and case studies, we demonstrate the effectiveness of DAIoTtalk across multiple domains, addressing challenges related to communication efficiency, deployment versatility, and resource scalability.

Item Type: Thesis (Doctoral)
Faculty \ School: Faculty of Science > School of Computing Sciences
Depositing User: Chris White
Date Deposited: 10 Nov 2025 14:05
Last Modified: 10 Nov 2025 14:05
URI: https://ueaeprints.uea.ac.uk/id/eprint/100939
DOI:

Downloads

Downloads per month over past year

Actions (login required)

View Item View Item