This thesis work examines the application of state-of-the-art reinforcement learning algorithms to the quantum state stabilization problem, specifically focusing on the robustness w.r.t. different types of noise. In the first part, this work aims to provide the basics of both the quantum mechanics formalism and the reinforcement learning framework. The second part of the thesis focuses on the development of three different reinforcement learning methods with the purpose of stabilizing a target state for a quantum system via feedback control. The algorithms that follows from these setups are trained without any noise acting on the system, but their robusteness is then tested by adding different types of quantum noise to the system dynamics. An in-depth numerical analysis of the performance of the presented methods for a significant case study is developed, and the results are compared in order to identify the approach that is least sensitive to the noise addition. Our results provide some valuable insights about the robustness of the reinforcement learning framework applied to quantum state stabilization and clearly indicate that model-free approaches as the best candidates for a controlling quantum systems in uncertain environments. This work also suggests interesting new research directions, including the scalability of the performances when dealing with quantum systems of increasing size.

This thesis work examines the application of state-of-the-art reinforcement learning algorithms to the quantum state stabilization problem, specifically focusing on the robustness w.r.t. different types of noise. In the first part, this work aims to provide the basics of both the quantum mechanics formalism and the reinforcement learning framework. The second part of the thesis focuses on the development of three different reinforcement learning methods with the purpose of stabilizing a target state for a quantum system via feedback control. The algorithms that follows from these setups are trained without any noise acting on the system, but their robusteness is then tested by adding different types of quantum noise to the system dynamics. An in-depth numerical analysis of the performance of the presented methods for a significant case study is developed, and the results are compared in order to identify the approach that is least sensitive to the noise addition. Our results provide some valuable insights about the robustness of the reinforcement learning framework applied to quantum state stabilization and clearly indicate that model-free approaches as the best candidates for a controlling quantum systems in uncertain environments. This work also suggests interesting new research directions, including the scalability of the performances when dealing with quantum systems of increasing size.

Reinforcement Learning for robust quantum state stabilization

GUATTO, MANUEL
2022/2023

Abstract

This thesis work examines the application of state-of-the-art reinforcement learning algorithms to the quantum state stabilization problem, specifically focusing on the robustness w.r.t. different types of noise. In the first part, this work aims to provide the basics of both the quantum mechanics formalism and the reinforcement learning framework. The second part of the thesis focuses on the development of three different reinforcement learning methods with the purpose of stabilizing a target state for a quantum system via feedback control. The algorithms that follows from these setups are trained without any noise acting on the system, but their robusteness is then tested by adding different types of quantum noise to the system dynamics. An in-depth numerical analysis of the performance of the presented methods for a significant case study is developed, and the results are compared in order to identify the approach that is least sensitive to the noise addition. Our results provide some valuable insights about the robustness of the reinforcement learning framework applied to quantum state stabilization and clearly indicate that model-free approaches as the best candidates for a controlling quantum systems in uncertain environments. This work also suggests interesting new research directions, including the scalability of the performances when dealing with quantum systems of increasing size.
2022
Reinforcement Learning for robust quantum state stabilization
This thesis work examines the application of state-of-the-art reinforcement learning algorithms to the quantum state stabilization problem, specifically focusing on the robustness w.r.t. different types of noise. In the first part, this work aims to provide the basics of both the quantum mechanics formalism and the reinforcement learning framework. The second part of the thesis focuses on the development of three different reinforcement learning methods with the purpose of stabilizing a target state for a quantum system via feedback control. The algorithms that follows from these setups are trained without any noise acting on the system, but their robusteness is then tested by adding different types of quantum noise to the system dynamics. An in-depth numerical analysis of the performance of the presented methods for a significant case study is developed, and the results are compared in order to identify the approach that is least sensitive to the noise addition. Our results provide some valuable insights about the robustness of the reinforcement learning framework applied to quantum state stabilization and clearly indicate that model-free approaches as the best candidates for a controlling quantum systems in uncertain environments. This work also suggests interesting new research directions, including the scalability of the performances when dealing with quantum systems of increasing size.
Quantum Control
Reinforcement
Learning
Robustness
PPO
File in questo prodotto:
File Dimensione Formato  
Guatto_Manuel.pdf

accesso aperto

Dimensione 7.77 MB
Formato Adobe PDF
7.77 MB Adobe PDF Visualizza/Apri

The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.12608/46148