A White-Box Digital Twin for Real-Time Polarization Tracking

Tracking polarization in real-time is a long-term challenge. The conventional heuristic and gradient-descent-based algorithms for polarization tracking lack efficiency and interpretability. To resolve the problem, a white-box digital twin modeling the entire polarization tracking system is derived by calculating with Stokes vectors and Mueller matrices. Moreover, the real-time polarization tracking enabled by the white-box digital twin is experimentally demonstrated, which is over 7 times faster than the commonly used stochastic parallel gradient descent (SPGD) on average. The adoption of digital twin allows the algorithm to bypass the loop of perturbation, sampling, and adjusting over the real system, thereby significantly reducing the sample times and recovery time. The proposed white-box digital-twin-based algorithm has strong interpretability and high efficiency, which has substantial potential to become a standard approach to achieve real-time polarization tracking. A white-box digital twin is proposed to realize real-time polarization tracking. The digital twin derived by rigorous polarization optics calculation accurately models the real system. Digital twin bypasses the loop of perturbation, sampling, and adjusting over the real system, and requires merely 9 samples to complete one-step tracking, which is averagely over 7 times faster than conventional SPGD. image