Deep reinforcement learning for key distribution based on quantum repeaters
Simon D. Reiß, Peter van Loock
Abstract
This work examines secret key rates of key distribution based on quantum repeaters in a broad parameter space of the communication distance and coherence time of the quantum memories. As the first step in this task, a Markov decision process modeling the distribution of entangled quantum states via quantum repeaters is developed. Based on this model, a simulation is implemented, which is employed to determine secret key rates under naively controlled, limited memory storage times for a wide range of parameters. The complexity of the quantum state evolution in a multiple-segment quantum repeater chain motivates the use of deep reinforcement learning to search for optimal solutions for the memory storage time limits, typically referred to as memory cutoffs. The main contribution in this work is to explore very general cutoff strategies which dynamically adapt to the state of the quantum repeater. An implementation of this approach is presented, with our focus on four-segment quantum repeaters, achieving a proof of concept of its validity by finding exemplary solutions that outperform the naive strategies.