Learning-based Protocol for Routing in Quantum Networks

Abstract

With the imminent advent of quantum communications head-on and its landing in what is going to be called the Quantum Internet, it is essential to research different ways of how it could be done in its most efficient way. In this work, we propose a game theory based self-adaptive system that, by accumulating reward, manages to learn how to minimize the latency of packets that are sent in a congested quantum network. To begin with, we study under which circumstances the quantum protocol outperforms its classical equivalent. First, analyzing the minimum number of packets moving through the network as a function of the size of the network necessary for guaranteeing an improvement in the quantum protocol. Then, studying the minimum size of the network based on the number of nodes needed to ensure the aforementioned advantage. Furthermore, taking advantage of the concepts of non-cooperative games and reinforcement learning, a protocol is designed and the behavior of a network with quantum nodes is evaluated. In order to analyze this protocol, different algorithms and scenarios are tested by measuring their efficiency and adaptability.