Domino-cooling Oscillator Networks with Deep Reinforcement Learning

TL;DR

This paper demonstrates how deep reinforcement learning can be used to control coupled harmonic oscillators in a network, actively cooling them to their ground states and showing potential in quantum applications.

Contribution

It introduces a novel deep RL-based control scheme for oscillator networks, enabling active cooling and potential quantum regime applications.

Findings

Successful thermal cooling of oscillators in various network configurations

Effective control of oscillator states using deep reinforcement learning

Potential applicability in quantum regime control

Abstract

The exploration of deep neural networks for optimal control has gathered a considerable amount of interest in recent years. Here, we utilize deep reinforcement learning to control individual evolutions of coupled harmonic oscillators in an oscillator network. Our work showcases a numerical approach to actively cool internal oscillators to their thermal ground states through modulated forces imparted to the external oscillators in the network. We present our results for thermal cooling of all oscillators in multiple network configurations and introduce the utility of our scheme in the quantum regime.