A Novel Efficient Coupling Algorithm for Adaptive Fiber Coupling

The high-speed free space optical communication systems based on the fiber techniques play more and more important roles nowadays, the laser beam propagating in the free space should be coupled into the single-mode fiber (SMF) at the input of the receiver. However, propagation through the atmospheric turbulence degrades the laser beam's spatial coherence and the jitters of the launch platform also decrease the convergent efficiency. Based on the existing adaptive coupling system, this paper studies the coupling algorithm based on the gradient descent (GD). We propose a novel algorithm that estimating the gradient through the simultaneous perturbation stochastic approximation (SPSA) and combining it with some traditional deep learning algorithms. Besides, we compare the coupling efficiency and speed of these algorithms. The simulation results demonstrate that the new algorithm combining the SPSA and the iteration point update strategy of Adagrad (SP-Adagrad) algorithm has a great advantage in coupling performance and is robust compared with other traditional algorithms, which improves the performance of the adaptive fiber coupling system in coupling speed and stability.