Investigation On Anti-Turbulence Method For Underwater Vortex Optical Communication

With the increasing of underwater activities,the demand for the improvement of information transmission rates of underwater communication systems has become stronger and stronger.Due to its high dimensional and orthogonality,orbital angular momentum(OAM)mode,a new degree of the optical signal,has also been applied in the underwater wireless optical communication systems for improving the capacity and the bandwidth utilization efficiency.Here,the beam carrying OAM mode is also called vortex beam.However,the vortex beam is easily disturbed by oceanic turbulence(OT)when it propagates through the underwater environment,resulting in its wavefront is distorted and the crosstalk between different OAM modes is occured,that heavily reduces the quality of the underwater communication system.Adaptive optics(AO)is a technique to effectively suppress the distortion effect by compensating the distorted beam so that it can improve the communication quality.This paper studies the method of anti-oceanic turbulence in underwater vortex optical communication.The main research contents and achievements are as follows:(1)We experimentally compare the compensation performance of three typical AO algorithms for beam distortion of OAM mode caused by OT,including Shack-Hartmann(SH)wavefront correction algorithm,Gerchberg-Saxton(GS)wavefront phase recovery algorithm and stochastic parallel gradient descent(SPGD)recovers algorithm,where we setup an experimental platform of the underwater vortex optical communication system consisting of the AO realization,and the random phase screen is used to simulate the OT by a spatial light modulator(SLM)in the platform.The results show that the three AO algorithms can compensate the distorted OAM beams under weak OT,reduce the crosstalk between modes,and improve the mode purity of the beams.Under the strong OT,the GS algorithm has the best compensation effect for the distorted OAM beam,followed by the SPGD algorithm,and the SH algorithm has the worst compensation effect.(2)We propose an AO compensation method based on generative adversarial network(GAN),where the GAN model is used to train the mapping relationship between the intensity distribution of distorted Gaussian beams and the phase screen representing OT.In the experiment,the compensation performance of the AO method based on GAN for the distorted OAM beam was verified,and the intensity distribution of the distorted beam before and after compensation was recorded by the Charge Coupled Device(CCD)camera.The predictive phase screen is then compensated by the reverse phase processing,and the distorted beam is compensated and repaired by the compensated phase screen.The results show that the AO compensation method based on the GAN can correct the beam distortion caused by OT,reduce the cross-talk between modes,improve the mode purity of the beam,and achieve a good compensation effect.(3)We futher propose an AO compensation method based on diffraction deep neural network(DDNN),where each diffractive layer of a DDNN model is used to train the mapping relationship between the intensity distribution of distorted OAM beams and the phase screen representing OT.In the experiment,each diffraction layer of the DDNN model was recorded,solidified and loaded on the SLM.A CCD camera was used to record the intensity distribution of the beam.The results show that the AO method based on the DDNN can compensate the beam distortion,restore the intensity distribution of the beam,improve the mode purity of the beam,and achieve a good compensation effect.In addition,a mode recognition method of OAM based on DDNN is presented.The DDNN model is used to train and intensity the azimuthal index and radial index of different OAM modes.The results show that the recognition accuracy of this method can reach 1 without OT and with weak OT,while the recognition accuracy will decrease under the strong OT.It shows that the OAM mode recognition method based on DDNN has certain anti-turbulence characteristics.(4)Moreover,we propose an OAM mode recognition based on photoelectric hybrid deep neural network(PHDNN),where the PHDNN integrates the advantages of the optical neural network(diffraction deep neural network,DNN)and the electronic neural network(convolution neural network,CNN).The DDNN model is used to train the mapping relationship between the intensity distribution of the distorted OAM mode and that of the non-distorted OAM mode.The input of the CNN model is the output of the DDNN model,and the azimuthal index and radial index of the OAM mode are identified.In the experiment,the diffraction layer of the DDNN model was recorded,solidified and loaded onto a SLM,and the intensity distribution of the beam was recorded by a CCD camera.The intensity distribution of the beam is processed and input into the CNN for feature extraction and mode recognition.The results show that this method can overcome the disturbance of OT to the OAM mode and recognize the azimuthal index and radial index of the OAM mode accurately.