Application Research Of Underwater Orbital Angular Momentum Wireless Communication Based On Deep Learning

Underwater wireless optical communication has become a research hotspot for underwater wireless communication because of its faster transmission rate and lower propagation loss.Since the vortex beams with different orbital angular momentum（OAM）modes are orthogonal to each other,the OAM modes can be taken as arbitrary values,and the OAM is orthogonal to the amplitude and phase dimensions,so OAM can be used for digital coding and combined with the existing techniques to significantly improve the channel capacity and the coding efficiency.Since the intensity distributions of the vortex beams with different OAM modes are different,and the intensity distributions of the different composite vortex beams are also vastly different,image recognition of the received light intensity distribution maps can be performed to obtain the OAM modes of the vortex beams.In recent years,with the great breakthrough of deep learning technology and the improvement of computer hardware performance,convolutional neural networks（CNN）have been widely used in underwater wireless optical communication system based on orbital angular momentum.In this thesis,the existing technologies have been investigated and two underwater wireless optical communication systems based on orbital angular momentum have been proposed,which have the advantages of fast modulation rate and strong anti-interference capability.To better simulate the real water environment,the ocean turbulence model with a wider range of applicability and CNN for classification and identification to output digital information have been used in this thesis.To reduce the interference of ocean turbulence,a reed-solomon（RS）coding and decoding technique is introduced to provide a better demodulation effect when the classification capability of CNN is limited.The feasibility and effectiveness of the proposed system are demonstrated by theoretical analysis and simulation.The main works of this thesis are as follows:1.A deep learning based underwater orbital angular momentum keyed wireless optical communication system is proposed,using a power spectrum model with a large range of Plante/Schmidt number,which is more widely applicable and closer to the realistic water environment.Using CNN for demodulation not only has high recognition accuracy but also does not require too much manual intervention,which not only reduces the cost but also improves the system performance.In this thesis,the image to be transmitted is converted into a binary bit stream and divided into four channels by serial-parallel conversion,thus modulating the Gaussian beam incident on the spatial light modulator and a time-varying composite vortex beam would be exported.The output composite vortex beams are passed through the ocean turbulence channel,and is classified and identified at the receiver side with the help of CNN to demodulate the corresponding digital information and recover the original image.The effects of factors such as temperature,salinity and turbulence intensity on the system performance are discussed in detail through numerical simulations.The comparative analysis proves that the Res Net101 model used in this paper has better recognition accuracy under certain conditions.2.An underwater orbital angular momentum keyed wireless optical communication system combined with RS coding is proposed to effectively solve the problem that the CNN-based demodulation end is less effective under strong ocean turbulence conditions.Based on the deep learning-based underwater orbital angular momentum keyed wireless optical communication system proposed in this thesis,RS coding is introduced to reduce the impact of CNN classification errors on information transmission due to ocean turbulence.The simulation results show that the BER of the system can be reduced from 2×10 ^2-to 1×10^3-by using RS coding technique combined with CNN under certain conditions.The performance comparison with the deep learning-based underwater orbital angular momentum keyed wireless optical communication system proposed in this thesis illustrates that when the classification capability of convolutional neural network is limited,the improved system can effectively reduce the BER and has a better ability to fight against ocean turbulence interference.