Reinforcement Learning for Quantum Network Control

APPLICATIONS OF TECHNOLOGY:

Development of a quantum network or a quantum data center for distributed quantum computing

BENEFITS:

Improves the performance of quantum information throughput

BACKGROUND:

Quantum network components can be enhanced using optimized control frameworks, with the engineered pulses enhancing the fidelity of quantum operations. However, in quantum physics, model-based optimization techniques are insufficient due to the difficulty of modeling quantum systems coupled to decoherence channels. Particularly, the components for quantum networks are subject to noise that can degrade the quantum information throughput.

TECHNOLOGY OVERVIEW:

Researchers at Berkeley Lab have developed a Deep Reinforcement Learning (DRL) control methodology to enhance the efficiency of quantum information throughput in quantum networks.

The agent interacts with the quantum network component and performs pulses on the amplitudes of electromagnetic control signals. It is connected to an automatic wave generator to implement the desired control signals. The final product can be connected to nodes of a future quantum internet or quantum data center.

This Artificial Intelligence (AI) solution lends an algorithm to learn the best pulses to maximize quantum network performance. Currently, there is no competing technology in the quantum networking domain, as quantum network control using AI has never been done before. Researchers developed the control theory for optomechanical quantum transducers as well as a digital twin of an experiment to generate training data using their control framework. They have also created a proof-of-concept demonstration where the pulses determined by the RL agent can be implemented on a real device.

DEVELOPMENT STAGE: Proven principle

FOR MORE INFORMATION:

https://crd.lbl.gov/news-and-publications/news/2021/deep-learning-tactics-speed-quantum-simulations/

https://arxiv.org/pdf/2009.07876.pdf

PRINCIPAL INVESTIGATORS: