Laser communication system distortion wavefront prediction method based on LSTM network

Abstract

The invention discloses a laser communication system distortion wavefront prediction method based on an LSTM network, and the prediction process comprises the steps: simulating and generating a dynamic phase screen which accords with the statistical distribution of atmospheric turbulence in a laser communication system in MATLAB through a phase covariance function, and simulating distortion wavefront; dividing the dynamic phase screen into two groups as a training set and a test set of the LSTM network prediction model, inputting historical continuous distorted wavefront in the training process, outputting the next frame of wavefront to be corrected, generating a plurality of groups of prediction models after iteration, performing effect verification by using the test set, taking a residual average RMS value of a truth value and a prediction value as an evaluation index of the prediction model, selecting a model with the minimum residual average RMS value as a final prediction model, and deploying the model to an actual laser communication system to realize forward prediction of the distorted wavefront. According to the invention, the nonlinear fitting capability of the neural network is exerted, the long and short term memory capability of the LSTM network is fully utilized to carry out distortion wavefront prediction, and the energy concentration and communication capability of transmission laser are improved.

Description

technical field [0001] The invention belongs to the technical field of wavefront prediction control, and relates to a method for predicting a distorted wavefront of a laser communication system based on an LSTM network, which is suitable for wavefront prediction and correction in an adaptive optical system. Background technique [0002] Adaptive Optics (AO) technology is an effective method that can compensate atmospheric turbulence in real time, and has been widely used in wavefront correction of laser communication systems. However, the actual AO system usually has a delay of 2-3 sampling periods due to the delay in reading data from the wavefront sensor and the delay in control calculation. In the case of correcting the distorted wavefront of atmospheric turbulence with a high temporal frequency, this delay error will cause the compensation wavefront on the deformable mirror to obviously lag behind the change of the distorted wavefront, which severely limits the correctio...

AI Technical Summary

Problems solved by technology

[0006] The technical problem to be solved by the present invention is: in the case of correcting the distorted wavefront of atmospheric turbulence with high time frequency, the inherent time delay error of the AO system will cause the compensation wavefront on the deformable mirror to obviously lag behind the change of the distorted wavefront, Severely limit the correction performance of AO technology, which can easily lead to a sharp drop in energy and an increase in the bit error rate in the laser communication system, resulting in communication failure

Images