Reinforcement Learning Approach to Shortcuts between Thermodynamic States with Extra Constraints

TL;DR

This paper introduces a reinforcement learning method to identify optimal paths minimizing entropy production between equilibrium states in classical and quantum systems, accommodating continuous control parameters.

Contribution

It presents a novel RL-based approach for optimizing thermodynamic paths with constraints, applicable to both classical and quantum systems.

Findings

Successfully applied to classical two-level systems

Effective in quantum two-level systems

Handles continuous control parameters

Abstract

We propose a systematic method based on reinforcement learning (RL) techniques to find the optimal path that can minimize the total entropy production between two equilibrium states of open systems at the same temperature in a given fixed time period. Benefited from the generalization of the deep RL techniques, our method can provide a powerful tool to address this problem in quantum systems even with two-dimensional continuous controllable parameters. We successfully apply our method on the classical and quantum two-level systems.