Deep Reinforcement Learning Control of Quantum Cartpoles

Area of Science:

• Quantum Control
• Quantum Computing
• Machine Learning

Background:

• The classical cartpole problem is a standard benchmark in reinforcement learning.
• Controlling quantum systems often requires sophisticated measurement and feedback strategies.
• Deep reinforcement learning (DRL) has shown promise in complex control tasks.

Purpose of the Study:

• To generalize the cartpole problem to the quantum regime.
• To apply state-of-the-art deep reinforcement learning for stabilizing a quantum system.
• To demonstrate the applicability of DRL to continuous-space quantum control problems.

Main Methods:

• Developed a quantum version of the cartpole balancing problem.
• Utilized deep reinforcement learning algorithms to learn control policies.
• Employed measurement and feedback for stabilizing a particle in an unstable potential.

Main Results:

• Successfully stabilized the quantum cartpole using DRL.
• DRL approach demonstrated performance comparable to or exceeding traditional control theory strategies.
• The method was also applied to measurement-feedback cooling of quantum oscillators.

Conclusions:

• Deep reinforcement learning is effective for stabilizing quantum systems.
• This work extends DRL applications to continuous-space quantum control.
• The approach shows potential for quantum feedback cooling and other quantum technologies.