Speaker
Description
Quantum sensors achieve measurement precisions well beyond classical bounds, though in order to reach such sensitivities methods for quantum optimal control are often employed, particularly in systems where decoherence is important; however, optimizing the control of quantum sensors is made difficult by the fact that not all Hamiltonian parameters are known. Here we investigate the use of reinforcement learning (RL) for optimal control of a spin magnetometer, wherein a single spin evolves according to an unknown background field. We apply the soft actor-critic RL algorithm to learn a policy, from which a set of transverse control fields may be sampled to improve the sensitivity of the magnetometer in time, in the presence of decoherence. The agent is trained so as to maximize the quantum Fisher information of the spin with respect to the unknown field. We train the agent on numerical simulations of the system, and then apply the resulting agent to various conditions not seen in training. We find that the agent is sensitive to certain changes in the system Hamiltonian, but overall is able to generalize well, supporting the use of such algorithms in quantum optimal control problems.
| Keyword-1 | Quantum Sensing |
|---|---|
| Keyword-2 | Quantum Optimal Control |
| Keyword-3 | Reinforcement Learning |