[Luận án Tiến sĩ] - Serverless And Reinforcement Learning-Based Resource Management For Quantum Computing
3 tháng 7, 2026
[Giới thiệu luận án tiến sĩ của tác giả Nguyễn Thanh Hòa- Giảng viên Khoa Mạng máy tính và Truyền thông]Tên đề tài: SERVERLESS AND REINFORCEMENT LEARNING-BASED RESOURCE MANAGEMENT FOR QUANTUM COMPUTINGTác giả: Nguyễn Thanh HòaPrincipal Supervisor: Prof. Rajkumar BuyyaCo-Supervisor: A/Prof. Muhammad Usman‐ Năm: 2025‐ Orcid: 0000-0001-6904-6312School of Computing and Information Systems - The University Of Melbourne, AustraliaTừ khóa: Quantum computing, quantum cloud computing,quantum resource management, deep reinforcement learning, quantum serverless, quantum cloud simulation.Tóm tắt:Quantum computing promises to address computationally intractable problems beyond the capabilities of classical computers across a variety of industrial sectors, including drug discovery, finance, machine learning, and cybersecurity. Cloud-based quantum computing has emerged as a solution to the challenges of operating and maintaining physical quantum systems, as well as the high costs and specialised expertise required to access them. This paradigm democratises access to remote quantum computers through cloud services, enabling the practical deployment of quantum applications. However, current quantum cloud computing environments are characterised by heterogeneous quantum backends with varying capabilities, noise levels, and availability patterns. The limitations of Noisy Intermediate-Scale Quantum (NISQ) devices create distinct scheduling and orchestration complexities that are compounded by the hybrid nature of quantum-classical workflows, where quantum tasks must be seamlessly integrated with classical processing steps. Additionally, current quantum computing resources are limited in capacity and costly, requiring efficient resource utilisation. These challenges are further exacerbated by the dynamic nature of quantum computing environments and the need to balance execution fidelity with time constraints in practical quantum applications. Hence, quantum cloud providers require adaptive quantum application deployment and sophisticated resource management techniques to harness the full potential of quantum computing tailored for application-specific scenarios while accommodating the inherent limitations of current quantum hardware and software. This thesis focuses on adaptive resource management solutions for quantum cloud computing environments by developing serverless architectures for seamless quantum application deployment and reinforcement learning-based techniques for efficient task orchestration with time-aware and fidelity-aware optimisation. Because NISQ devices are inherently noisy and time-sensitive, we propose serverless architectures to manage the complexities of quantum task scheduling and hybrid quantum-classical workflow integration. Furthermore, because the dynamic nature of NISQ-era quantum systems challenges efficient resource utilisation, we introduce reinforcement learning-based techniques to address quantum hardware heterogeneity and optimise dynamic backend selection. The thesis advances the state-of-the-art by aligning these solutions with the following key contributions: 1. A comprehensive systematic mapping study and taxonomy of quantum cloud computing from different aspects, including service models, platforms, applications, resource management approaches, security, and privacy. 2. A holistic Quantum Function-as-a-Service framework that enables seamless integration of quantum computation within classical cloud environments through serverless architecture with adaptive backend selection and cold start mitigation strategies. 3. A comprehensive modelling and discrete-event simulation framework for quantum computing environments that facilitates systematic evaluation of quantum resource management algorithms, incorporating realistic quantum system models and multi-use case support. 4. A novel deep reinforcement learning-based approach for time-aware quantum task placement using the Deep Q-Network technique to adapt to dynamic quantum cloud environments and optimise task completion efficiency. 5. A fidelity-aware quantum task orchestration framework using deep reinforcement learning that effectively balances execution fidelity and time constraints in NISQ-era quantum systems through noise-aware performance modelling and Proximal Policy Optimisation approaches. 6. A detailed study outlining challenges and research directions for quantum cloud resource management, establishing foundational approaches for future research to advance quantum computing paradigms.Nguồn: Library - The University Of Melbourne, Australia.https://minerva-access.unimelb.edu.au/items/1b01ee58-7d1d-4d4a-8c87-4dea2fef8edbBiểu ghi luận án tại Thư viện UIT: đang cập nhậtToàn văn: đang cập nhật
