AI-Powered Persistent Entanglement Distribution in Quantum Networks

Entanglement distribution is a fundamental problem in quantum networks. Most existing entanglement distribution works only focused on provisioning quantum memory and quantum channels, but ignored Entangled Photon Sources (EPSes) which are also precious resources. In addition, motivated by the presumed fast entanglement decoherence (in the order of seconds), they destroy all created Entanglement Links (ELs) at the end of each time slot, which could result in serious quantum resource waste and low network throughput. In this paper, we propose PED to solve the Persistent Entanglement Distribution problem. Given the network topology and a set of demands, PED first selects an Entanglement Path (EP) for each demand and then leverages neural networks to online determine EPS provisioning. The salient features of PED include (i) a created EL can be maintained for multiple time slots before it has to be eventually destroyed due to decoherence, and (ii) every EPS agent hosts a specific neural network to calculate its EPS provisioning scheme based on its local information. Extensive simulations show that the EP selection and EPS provisioning approaches in PED can help improve the network throughput by up to 10.25% and 611.14%, respectively.