We generalize a standard benchmark of reinforcement learning, the classical cartpole balancing problem, to the quantum regime by stabilizing a particle in an unstable potential through measurement and feedback. We use state-of-the-art deep reinforcement learning to stabilize a quantum cartpole and find that our deep learning approach performs comparably to or better than other strategies in standard control theory. Our approach also applies to measurement-feedback cooling of quantum oscillators, showing the applicability of deep learning to general continuous-space quantum control.
@article{arxiv.1910.09200,
title = {Deep Reinforcement Learning Control of Quantum Cartpoles},
author = {Zhikang T. Wang and Yuto Ashida and Masahito Ueda},
journal= {arXiv preprint arXiv:1910.09200},
year = {2020}
}
Comments
5+4 pages, 2+2 figures, 2+2 tables, 5 videos at an external link