Reinforcement Learning for Quantum Optimal Control

Artificial Intelligence agents trained through reinforcement learning algorithms have demonstrated a superior ability in maximizing long term rewards in complicated and adversarial environments, such as those found in game theory like Chess or Go. We aim to explore how such algorithms can be of use to the quantum control community.

The evolution of quantum systems is often complicated, sensitive to changes in control parameters, and subject to various decoherence effects; this makes optimizing their control a difficult problem. Reinforcement learning offers a way to tackle the complexity of these problems, while remaining adaptable to changes in the environment with relatively low computational complexity. Applications of this include quantum sensing, Hamiltonian engineering, quantum gate design, and more.