Model-Predictive Quantum Control via Hamiltonian Learning
Model-Predictive Quantum Control via Hamiltonian Learning
复制标题
DOI：
10.1109/tqe.2022.3176870
复制
发表时间：
2022
影响因子：
--
通讯作者：
Maison Clouâtré;M. J. Khojasteh;M. Win
中科院分区：
文献类型：
--
作者：
Maison Clouâtré;M. J. Khojasteh;M. Win
本文提出了一种端到端的闭合量子系统学习控制框架。所提出的学习技术是第一种利用分层设计的此类技术，该设计将探测控制、量子状态层析成像、量子过程层析成像和哈密顿学习分层来识别内部和控制哈密顿量。在此背景下，提出了一种新的量子过程层析算法，它涉及到对酉群，即酉算符空间的优化，以确保物理上有意义的预测。我们的可伸缩哈密顿学习算法具有较低的存储需求和可调的计算复杂度。一旦学习了哈密顿量，我们就将数据驱动的模型-预测量子控制(MPQC)形式化。该技术利用所学习的模型在闭环模拟中计算量子控制参数。然后，将最优控制输入以开环方式提供给物理量子系统。仿真结果表明，当使用序列二次规划(SQP)来解决每个控制问题时，模型预测量子控制比目前最先进的量子最优控制更有效。
This article proposes an end-to-end framework for the learning-enabled control of closed quantum systems. The proposed learning technique is the first of its kind to utilize a hierarchical design, which layers probing control, quantum state tomography, quantum process tomography, and Hamiltonian learning to identify both the internal and control Hamiltonians. Within this context, a novel quantum process tomography algorithm is presented, which involves optimization on the unitary group, i.e., the space of unitary operators, to ensure physically meaningful predictions. Our scalable Hamiltonian learning algorithms have low memory requirements and tunable computational complexity. Once the Hamiltonians are learned, we formalize data-driven model-predictive quantum control (MPQC). This technique utilizes the learned model to compute quantum control parameters in a closed-loop simulation. Then, the optimized control input is given to a physical quantum system in an open-loop fashion. Simulations show model-predictive quantum control to be more efficient than the current state-of-the-art, quantum optimal control, when sequential quadratic programming (SQP) is used to solve each control problem.