Deep learning piston aberration control of fiber laser phased array by spiral phase modulation

Abstract

The stochastic parallel gradient descent (SPGD) algorithm is usually employed as the control strategy for phase-locking in fiber laser phased array systems. However, the convergence speed of the SPGD algorithm will slow down as the number of array elements increases. To improve the control bandwidth, the convolutional neural network is introduced to quickly calculate the initial piston aberration in a single step. In addition, the irrationality of the commonly used Mean Square Error (MSE) evaluation function in existing convolutional neural networks is analyzed. A new evaluation function NPCD (Normalized Phase Cosine Distance) is proposed to improve the accuracy of the neural networks. The results show that the piston aberration residual is 0.005 and the power in the bucket (PIB) is 0.993 after accurate preliminary compensation, which means that the system directly enters the co-phase state. We also demonstrate the robustness and scalability by adding additional disturbance and expanding the scale of the array.

References in corpus (3)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Experimental phase control of a 100 laser beam array with quasi-reinforcement learning of a neural network in an error reduction loop
• Deep learning-based phase control method for coherent beam combining and its application in generating orbital angular momentum beams