Wireless intelligent distributed consensus system for autonomous driving

Li, Zongyao (2025) Wireless intelligent distributed consensus system for autonomous driving. PhD thesis, University of Glasgow.

Abstract

The rapid advancements in embedded processing, sensing, artificial intelligence (AI), and communication technologies have accelerated the adoption of connected and autonomous systems (CAS). However, as devices in CAS become more intelligent and autonomous, their application scenarios—such as autonomous driving—are growing increasingly complex and dynamic. To enable intelligent, connected, and autonomous (ICA) nodes to engage in deeper deliberation and mutual understanding, joint decision-making has emerged as an effective solution. Joint decision-making is a process where multiple autonomous agents collectively analyze information, deliberate, and reach consensus to make unified decisions that align with common goals. However, traditional joint decision-making approaches face significant challenges when applied to the stringent demands of modern CAS. For instance, centralized decision-making (CDM) offers streamlined processes and high consistency but suffers from limitations like single points of failure (SPOF), scalability issues, and dependence on centralized infrastructure. By contrast, the decentralized nature of distributed decision-making (DDM) enhances scalability and system reliability, leveraging the intelligence of individual nodes to achieve collective intelligence, making it a promising alternative. In this context, distributed consensus (DC) protocols as a key element in distributed systems are essential to enabling DDM, with features such as data consistency and fault tolerance drawing significant research attention in recent years. This thesis focuses on the application, optimization, and development of wireless DC protocols to enable ICA nodes in CAS, with a particular focus on autonomous driving, to achieve robust and expressive joint decision-making.

First, the study proposes Intelligent Distributed Consensus (IDC) and introduces the first IDC protocol, Intelligent-Raft, which builds upon the traditional Raft algorithm. Additionally, to facilitate the deployment of IDC in practical CAS environments, the study introduces Wireless Intelligent Distributed Consensus System (WIDCS) which leverages distributed wireless communication combined with the Intelligent-Raft algorithm to enable ICA nodes to make collective joint-decisions and ensure fault tolerance. A practical hardware module of WIDCS is implemented using microcontroller-based systems, which is named ‘AIR-RAFT’. To validate the feasibility and effectiveness of WIDCS, we undertake research and evaluations within an autonomous driving scenario, specifically at uncontrolled intersections, utilizing both mathematical modeling and practical scenario testing. Numerical and experimental results, in good alignment, demonstrate that WIDCS substantially improves autonomous driving safety

Second, this study enhances WIDCS by incorporating the functions of ad hoc network formation, management, and dismissal, improving its ability to provide better data consistency and joint decision-making services for ICA nodes. Additionally, we have developed the second-generation WIDCS module, RaBee, which enables distributed nodes to achieve Intelligent-Raft consensus via a ZigBee-based ad hoc network. In addition, we develope mathematical probability models to evaluate and compare the reliability of centralized decision-making systems and WIDCS. Employing autonomous driving in onramp merging as a case study, we further formulated a mathematical model to assess the safety of Autonomous Vehicles (AVs) under different decision-making frameworks. By integrating the RaBee module with AV, we conduct safety tests in practical on-ramp scenarios, and the results demonstrate that WIDCS notably enhances AV safety, indicating substantial potential for future CAS.

Third, this study proposes a novel IDC protocol, Converging-Raft, which leverages the collective intelligence of all nodes to make globally optimal joint decisions—a capability not present in Intelligent-Raft. To enhance the adaptability, we propose the Heterogeneous Intelligent Joint Decision System (HIntS), an architecture which integrates CDM, Intelligent-Raft, and Converging-Raft within a hybrid network combining ad hoc and cellular networks. Our self-developed hardware module at the core of HIntS, ‘5G-MInd’, is designed to verify the system’s feasibility and performance in practical experiments. We develop a mathematical model to analyze and compare the reliability and latency of HIntS under different working modes and validate these findings through joint-decision experiments using 5G-MInd modules. Our quantitative and qualitative results demonstrate the advantages and characteristics of different combinations of joint-decision mechanisms and network structures. These findings highlight HIntS’s adaptability to complex, dynamic environments and provide critical guidance for the practical deployment of future wireless joint-decision mechanisms.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Subjects: T Technology > T Technology (General); T Technology > TL Motor vehicles. Aeronautics. Astronautics
• Colleges/Schools: College of Science and Engineering > School of Engineering
• Supervisor's Name: Zhang, Professor Lei
• Date of Award: 2025
• Depositing User: Theses Team
• Unique ID: glathesis:2025-85204
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 17 Jun 2025 12:58
• Last Modified: 17 Jun 2025 13:00
• URI: https://theses.gla.ac.uk/id/eprint/85204

Related URLs

• Conference proceeding
• Conference proceeding
• Conference proceeding
• Conference proceeding