ESPGD Algorithm to Improve the Convergence Speed for Adaptive Single-Mode Fiber Coupling

Abstract

The adaptive single-mode fiber (SMF) coupling technique is normally adopted since the coupling efficiency (CE) significantly determines the performance of the free-space optical communication (FSOC) systems. The stochastic parallel gradient descent (SPGD) algorithm is the most commonly used control algorithm in adaptive SMF fiber coupling system. This paper proposes an improved SPGD algorithm named estimation-based stochastic parallel gradient descent (ESPGD) algorithm to accelerate the sytem convergence when applied to a practical adaptive SMF coupling system based on fast steering mirror (FSM). Applying the perturbed voltages, FSM dynamic response and then recording the performance metrics is the basic and most time-consuming process in actual adaptive SMF coupling system control. The ESPGD algorithm uses a different gradient estimation method based on adaptive parameter estimation method. The algorithm only needs to perform this process once in one iteration while the original SPGD algorithm needs to perform it twice to obtain the estimated gradient. This greatly reduce the time consumed by one iteration of the algorithm, thereby reducing the convergence time. The simulation and experimental results show that the ESPGD algorithm reduces the system convergence time by nearly half when correcting static angular errors and more than doubles the control bandwidth when correcting sinusoidal angular jitters.