AI-Assisted Polarization Basis Alignment for Quantum Key Distribution System Receivers

In this work, we propose and experimentally validate a polarization basis alignment method based on a hybrid approach, using the Particle Swarm Optimization (PSO) and a gradient search algorithm. This method uses the Quantum Bit Error Ratio (QBER) as metric to be minimized. At first, the PSO runs one iteration to perform a coarse basis alignment, followed by the gradient search algorithm which performs a fine basis alignment, until the QBER is below the threshold. The polarization basis alignment method here proposed is implemented and tested at the receiver of a Quantum Key Distribution (QKD) system testbed. With this proposed hybrid basis alignment approach, a QBER below 1% can be achieved after, on average, nineteen QBER estimations, when using five particles for the PSO. These nineteen QBER estimations represent a 24% improvement regarding the method mostly used in the state of the art which.