Information Fusion for Autonomous Internet of Things

Abstract

The Internet of Things (IoT) extends the electronic connectivity into millions of IoT nodes around the world. In order to achieve the autonomy to make sound decisions based on the collected and analysed information, artificial intelligence algorithms are for networks to leverage to turn into autonomous IoT (AIoT). The core value of these AIoT applications is from analysing the information provided by these smart entities and performing autonomous decision-making. However, general issues in networks, such as the outlier raging, communication overhead, memory shortage, etc., render the information fusion process ineffective, which leads to the unreliable AIoT applications. Thus, the advanced information fusion methods must be proposed to address such issues for the sake of fully utilizing these smart applications and services. To solve such real-life problems, various information fusion approaches are developed for the AIoT systems. For the outlier raging problems in AIoT applications, the distance-based method inspired by the KNN rule has been used. In addition, artificial intelligence algorithms, such as the deep learning and reinforcement learning, are integrated with consensus theories to handle the distributed learning problems in AIoT applications for the sake of solving the communication overhead and memory shortage in networks. Also, the comprehensive evaluation works show the proposed approaches emerge the powerful capability, and they provide important operational advantages over traditional centralized methods and typical distributed approaches. The main contributions of this thesis are proposing two outlier detection methods with compelling features, such as low computational complexity and memory usage, for the outlier raging in networks, and one novel distributed learning framework that can save the average training time and memory storage compared with the typical methods.

Date of Award	1 Aug 2021
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Zhirun Hu (Supervisor) & Zhengtao Ding (Supervisor)