Turbulence-resistant free space optical communication method and system based on joint optimization of transmission and reception

Abstract

The application discloses an anti-turbulence free space optical communication method based on transceiving joint optimization, through a Polar coding mode of an atmospheric turbulence channel, balance code length and code rate, ensure the overhead of coding and decoding; at the same time, a convolution enhanced Transformer model is designed at the receiving end, the learning of local features is strengthened through convolution, and the connection between phase screen pixels is established, and the conjugate phase screen is given by the model to compensate the turbulence. The application aims at the problem that the atmospheric turbulence effect disturbs the OAM light beam in the free space optical transmission system, and utilizes the joint optimization mode of the transmitting end and the receiving end to relieve the influence of the atmospheric turbulence disturbance on the free space optical communication. After the Polar coding of the transmitting end corrects the error code caused by the atmospheric turbulence, and the phase compensation of the receiving end, the coupling efficiency of the received light spot is improved, so that the mode purity of the corrected light spot image is improved, and the system error rate is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of optical communication technology, more particularly, to an anti-turbulence free space optical communication method and system based on joint optimization of transmission and reception BACKGROUND

[0002] Free space optical communication (FSO) is a wireless optical communication technology that uses the atmosphere as a signal transmission medium. Compared with traditional optical fiber communication, free space optical communication system does not need to dig the ground or erect cables, and has the advantages of high bandwidth, anti-electromagnetic interference and flexible deployment. In some difficult-to-wire situations, such as historical buildings, protected areas or complex terrain areas, FSO provides an effective communication means. Therefore, FSO is widely used in metropolitan area network expansion, optical fiber backup, wireless cellular network backhaul and post-disaster emergency recovery, etc. It has important significance for "last mile" communication and is considered as one of the most potential wireless connection technologies in the next generation. The research of free space optical communication system is of great significance for building high-speed optical communication network.

[0003] However, the free space optical communication link is greatly affected by the complex atmospheric channel. Atmospheric molecules and other molecules can cause absorption, scattering and turbulence of signal light, causing power attenuation, pulse broadening, beam expansion and angle of arrival fluctuation, etc. Ultimately, it causes the average signal-to-noise ratio of the communication system to decrease and causes the random fluctuation of the probe signal. Among them, the most serious one is atmospheric turbulence. Atmospheric turbulence and bad weather can cause random fluctuations in the amplitude or phase of the signal, thereby degrading the quality of the optical signal in atmospheric transmission, increasing the communication bit error rate and reducing the transmission quality. Therefore, it is crucial to design a free space optical transmission scheme that can alleviate the degradation of optical signal transmission caused by atmospheric channel and realize anti-atmospheric turbulence interference. SUMMARY

[0004] The purpose of the present application is to provide an anti-turbulence free space optical communication scheme based on joint optimization of transmission and reception to alleviate the interference of atmospheric turbulence in free space on the signal.

[0005] In order to achieve the above purpose, the technical scheme adopted by the present application is as follows: an anti-turbulence free space optical communication method based on joint optimization of transmission and reception, comprising:

[0006] S1, at the transmitting end, the construction of the code in Polar coding is determined by the output of the neural network model, the reliability of each subchannel is analyzed, and it is given whether the subchannel is a frozen bit or an information bit under the current channel according to the turbulence channel;

[0007] S2, modulate the encoded information onto the optical carrier, after collimating and expanding, the size of the Gaussian beam is controlled by the diaphragm, and a spiral phase plate is placed at the back end to convert the Gaussian beam into an OAM beam with better anti-turbulence performance;

[0008] S3, simulate the atmospheric turbulence channel by rotating the turbulence phase plate, and the OAM beam is disturbed by the atmospheric turbulence;

[0009] S4, at the receiving end, the received light is divided into two beams by a beam splitter, one beam of light is irradiated into an infrared camera, and the infrared camera is used to collect images and input them into a neural network, the neural network analyzes the phase of the turbulence channel from the light intensity image and gives the conjugate phase, and the compensated phase image is displayed by the SLM;

[0010] S5, the other light divided by the beam splitter is compensated by the conjugate phase on the SLM, and then coupled into the optical fiber by using the displacement table, and then enters the photodetector to complete the photoelectric conversion;

[0011] S6, finally, the data input by the PD is collected by the oscilloscope, saved and analyzed by MATLAB for off-line processing of the bit error rate.

[0012] Preferably, the code construction process in the Polar encoding is to sort the reliability of N sub-channels generated after the channel polarization operation from large to small, and then select the first K sub-channels as information bits, and the remaining N-K sub-channels as frozen bits.

[0013] Preferably, the Polar encoding process includes:

[0014] Collect the state information of the sub-channels by simulation, including the bit error rate of the received signal under different turbulence intensities as the channel transmission characteristics, and select the Bussgang parameter of each sub-channel after channel polarization and the sub-channel index as the features of the network model for training;

[0015] Construct a data set, use the known Polar encoding construction method as a benchmark, mark whether each sub-channel is an information bit or a frozen bit, and use it as a training label;

[0016] Construct a network model to accept turbulence intensity and encoding parameters as input, and output the reliability prediction of each sub-channel, and express the polarization weight of each sub-channel as a probability that can be learned by a neural network to represent the sub-channel as an information bit.

[0017] Define the loss function, the error between the feedback result and the true value.

[0018] Preferably, the network model comprises an input layer for receiving feature vectors of the subchannels, such as bit error rate, coding parameters; a plurality of fully connected layers for extracting features and learning nonlinear mapping; an LSTM layer for processing channel state variation; and an output layer for outputting probability or reliability score of each subchannel as information bit.

[0019] Preferably, when defining the loss function, a binary cross-entropy loss function is used as the basic loss to determine whether the current subchannel is a frozen bit or an information bit:

[0020]

[0021] where N is the number of samples, y i is the actual reliability label of the subchannel, is the probability predicted by the model. In order to consider the constraint of code rate, penalize the deviation of model prediction from the target code rate, and add the code rate loss shown in the following formula:

[0022]

[0023] where R target is the target code rate, usually set to 0.5, λ R is the weight for balancing the code rate loss. In order to constrain the code length, we add the following formula to the loss function, and the code length is set to 2048:

[0024] L Length = λ L N

[0025] In order to quantify the change of channel condition, the signal-to-noise ratio is used as the quantitative indicator of channel quality, and the bit error rate is used as the measure of performance difference before and after the change of channel condition, then the loss term under the turbulent channel is shown in the following formula:

[0026] L channel = λ ch · f(ΔBER)

[0027] where f is a non-negative penalty factor proportional to the size of the channel condition change, ΔBER is the change amount of bit error rate after the change of channel condition, λ ch is a hyperparameter for balancing the loss weight;

[0028] Combining the above, the final loss function is shown in the following formula,

[0029] L total = L BCE + L RATE + L Length + L channel .

[0030] Preferably, the receiving end neural network determines the turbulence intensity of the current channel according to the light intensity image and feeds back to the network model of the transmitting end. After experiencing the identification of the turbulence intensity of the received light intensity image, the trained model is used to evaluate the new sub-channel set to obtain the reliability prediction of each sub-channel. The output of the model is used as the input of the Polar coding construction to determine the final information bit and frozen bit.

[0031] Preferably, the receiving end uses a convolution-enhanced Transformer model to learn the mapping relationship between the received light intensity image and the turbulence phase screen.

[0032] Preferably, the convolution-enhanced Transformer model uses convolution layers in the encoder to gradually reduce the spatial dimensions of the feature map, then uses upsampling and convolution layers in the decoder to gradually restore the spatial dimensions of the feature map, and uses Transformer modules in the encoder and the decoder to capture the dependency between long-distance pixels in the feature map. The model combines the encoder, the decoder, and the final convolution layer to generate a prediction result.

[0033] Meanwhile, the turbulence intensity of the current channel is also fed back to the neural network of the transmitting end in the last output layer. The output of the network includes the turbulence intensity fed back to the transmitting end and the current turbulence phase screen. Based on the convolution-enhanced Transformer model architecture, a convolution module is combined in the architecture.

[0034] The anti-turbulence free-space optical communication system based on transceiver joint optimization includes a receiving end neural network model, a transmitting end network model, a Polar encoder, a waveform generator, an intensity modulator, a laser, a collimating beam expander, a spiral phase plate, a turbulence phase plate, a beam splitter, a photodetector, an oscilloscope, an infrared camera, and a spatial light modulator.

[0035] The transmitting end designs a neural network model output coded for the turbulence channel. A training sequence is transmitted first to allow the receiving end neural network model to receive the light intensity image and determine the turbulence intensity of the current atmospheric channel and feed back to the transmitting end network model. The Polar encoder gives the distribution of the Polar coded information bits under the current channel condition. The coded information is input to the arbitrary waveform generator according to the output of the model. The radio frequency signal output by the waveform generator is loaded to the intensity modulator. The laser outputs a 1550nm wavelength laser input to the intensity modulator. The bias controller is used to make the intensity modulator work in the linear region. The coded information is modulated to the optical carrier. The output of the intensity modulator is collimated and expanded by the collimating beam expander, and then converted to an OAM beam with stronger anti-turbulence ability by the spiral phase plate.

[0036] In the channel part, the turbulence phase plate rotation is used to simulate the interference of the OAM light beam in the atmospheric turbulence channel, and at the receiving end, the light beam enters a 10:90 beam splitter, wherein 10% of the light beam is displayed as a spot image by an infrared camera and input into a network; the other part passes through a spatial light modulator for phase compensation, and the phase pattern displayed on the spatial light modulator is given by the output of the receiving end neural network model; finally, the compensated spatial light is coupled into an optical fiber by using a displacement table and connected to a photoelectric detector and an oscilloscope, and after recording data, the error rate is analyzed offline.

[0037] Preferably, the receiving end neural network model is a CNN_Transformer model.

[0038] Compared with the prior art, the advantages of the present application are that:

[0039] The present application is aimed at the problem of atmospheric turbulence effect on OAM light beam interference in free space optical transmission system, and the present application uses the joint optimization of the transmitting and receiving ends to alleviate the influence of atmospheric turbulence interference on free space optical communication. After the Polar encoding of the transmitting end corrects the bit error caused by atmospheric turbulence, and the phase compensation of the receiving end improves the coupling efficiency of the received spot, the mode purity of the corrected spot image is improved, and the system bit error rate is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0041] Figure 1 is a flow chart of the anti-turbulence free space optical communication method of the present application based on joint optimization of transmitting and receiving;

[0042] Figure 2 is a Polar encoding flow chart of the present application;

[0043] Figure 3 is a convolution enhanced Transformer model architecture diagram of the present application;

[0044] Figure 4 is an architecture diagram of the MBConv module of the present application;

[0045] Figure 5 is an architecture diagram of the anti-turbulence free space optical communication system of the present application based on joint optimization of transmitting and receiving. DETAILED DESCRIPTION

[0046] The preferred embodiments of the present application are described in detail below with reference to the accompanying drawings, so that the advantages and features of the present application can be more easily understood by those skilled in the art, and the scope of protection of the present application can be more clearly defined.

[0047] The present application provides a turbulence-resistant free-space optical communication method based on transceiver joint optimization, which adopts encoding and compensation methods at the transmitting end and the receiving end respectively to alleviate the interference of atmospheric turbulence on signals. The present application designs a Polar encoding method for atmospheric turbulence channels at the transmitting end, balances the code length and code rate, and ensures the overhead of encoding and decoding. Meanwhile, a convolution-enhanced Transformer model is designed at the receiving end, which learns local features through convolution enhancement and establishes the connection between phase screen pixels. The conjugate phase screen is given by the model to compensate for turbulence.

[0048] As shown in Figure 1 , the specific steps include:

[0049] S1, at the transmitting end, the construction of the code in the Polar encoding is determined by the output of the neural network model, the reliability of each subchannel is analyzed, and it is given whether the subchannel is a frozen bit or an information bit under the current channel according to the turbulence channel;

[0050] S2, the encoded information is modulated onto an optical carrier, transmitted after collimation and beam expansion, and the size of the Gaussian beam is controlled by an aperture, a spiral phase plate is placed at the back end to convert the Gaussian beam into an OAM beam with better anti-turbulence performance;

[0051] S3, the rotating turbulence phase plate is used to simulate the atmospheric turbulence channel, and the OAM beam is disturbed by the atmospheric turbulence;

[0052] S4, at the receiving end, the received light is divided into two beams by a beam splitter, one beam of light is irradiated into an infrared camera, and the image is collected by the infrared camera and input into the neural network, the neural network analyzes the phase of the turbulence channel from the light intensity image and gives the conjugate phase, and the compensated phase image is displayed by the SLM;

[0053] S5, the other light divided by the beam splitter is compensated by the conjugate phase on the SLM, then coupled into an optical fiber using a displacement table, and then enters a photodetector to complete the photoelectric conversion;

[0054] S6, finally, the oscilloscope collects the data input by the PD, saves it and then processes it offline by MATLAB to analyze the bit error rate.

[0055] Preferably, the construction process of the code in the Polar encoding is to sort the reliability of the N subchannels generated after the channel polarization operation from large to small, then select the first K subchannels as information bits, and the remaining N-K subchannels as frozen bits.

[0056] The following describes the Polar coding-based transmitting end process in detail:

[0057] As a new channel coding scheme, the Polar code not only has a structured construction method and superior asymptotic performance, but also has lower complexity, but its application in practical scenarios such as FSO systems is relatively small.

[0058] The construction process of the Polar code is to sort the reliabilities of N sub-channels generated after channel polarization from large to small, and then select the first K sub-channels as information bits, and the remaining N-K sub-channels as frozen bits. The core of the Polar code is channel polarization. When designing the Polar code in the atmospheric turbulence channel, evaluating the reliability of the sub-channel is a key step, which not only affects the selection of information bits and frozen bits, but also relates to the control of computational complexity and time delay.

[0059] At the same time, the atmospheric turbulence channel is a time-varying channel model, and the state information of the channel needs to be predicted to adjust the parameters of the Polar code, such as code length, code rate and frozen bit distribution. After the information at the transmitting end is encoded and modulated and transmitted, the received signal is calculated at the receiving end by using the Successive Cancellation (SC) decoding algorithm to calculate the Bhattacharyya parameter of each sub-channel.

[0060] Through continuous circulation and large-scale simulation, the Bhattacharyya parameter obtained by each sub-channel is accumulated, and the smaller the value of the Bhattacharyya parameter, the higher the reliability of the sub-channel. Then, the cumulative sum of the Bhattacharyya parameter of each sub-channel is sorted, and the K sub-channels with the smallest cumulative sum are selected as information bits.

[0061] Therefore, in order to balance the error correction capability and the coding and decoding efficiency, the Polar weight combined with deep learning method is used in the present application, the Polar weight can be used as the basis for selecting information bits and frozen bits, and the sub-channel with higher weight is considered to be more reliable, and therefore is more suitable for transmitting information bits. The model is trained in a data-driven manner to learn and evaluate the reliability of each sub-channel after polarization, and to predict whether the sub-channel is an information bit or a frozen bit.

[0062] As shown in Figure 2 , in the present embodiment, the Polar coding process includes:

[0063] The state information of the sub-channel is collected by simulation, including the bit error rate of the received signal under different turbulence intensities as the channel transmission characteristics, and the Bhattacharyya parameter of each sub-channel after channel polarization and the sub-channel index are selected as the features of the network model for training.

[0064] Construct the dataset, use the known Polar encoding construction method as the benchmark, mark whether each subchannel is an information bit or a frozen bit, and use it as a training label;

[0065] Construct the network model to accept the turbulence intensity and the encoding parameters (code length, code rate) as inputs, and output the reliability prediction of each subchannel, and represent the polarization weight of each subchannel as a probability that can be learned by a neural network to represent the subchannel as an information bit.

[0066] Then, define the loss function, the error between the feedback result and the true value.

[0067] Preferably, the network model includes an input layer for receiving a feature vector of a subchannel, such as a bit error rate, an encoding parameter; a plurality of fully connected layers for extracting features and learning nonlinear mapping; an LSTM layer for processing channel state changes; and an output layer for outputting a probability or reliability score of each subchannel as an information bit.

[0068] Preferably, when defining the loss function, a binary cross-entropy loss function is used as the basic loss to determine whether the current subchannel is a frozen bit or an information bit:

[0069]

[0070] where N is the number of samples, y i is the actual reliability label of the subchannel, is the probability predicted by the model. In order to consider the constraint of the code rate, the deviation of the model prediction from the target code rate is punished, and the code rate loss shown in the following formula is added:

[0071]

[0072] where R target is the target code rate, usually set to 0.5, λ R is the weight for balancing the code rate loss. In order to constrain the code length, we add the following formula to the loss function, and the code length is set to 2048:

[0073] L Length = λ L N

[0074] In order to quantify the change of channel conditions, the signal-to-noise ratio is used as a quantitative indicator of channel quality, and the bit error rate is used as a measure of the performance difference before and after the change of channel conditions. The loss term under the turbulence channel is shown in the following formula:

[0075] L channel = λ ch · f(ΔBER)

[0076] where f is a non-negative penalty factor proportional to the size of the channel condition variation, ΔBER is the change in the bit error rate after the channel condition variation, and λ ch to balance the loss weight hyperparameters;

[0077] Combining the above, the final loss function is shown in the following formula,

[0078] L total = L BCE + L RATE + L Length + L channel .

[0079] Further, in actual use, since the network of the transmitting end is unaware of the channel condition, it is necessary to use the neural network of the receiving end together, which judges the turbulence intensity of the current channel according to the light intensity image and feeds back to the network model of the transmitting end. After experiencing the identification of the turbulence intensity of the received light intensity image, the trained model is used to evaluate the new sub-channel set to obtain the reliability prediction of each sub-channel. According to the output of the model as the input of the Polar coding construction, the final information bit and frozen bit are determined.

[0080] As Figure 3 shown, in the embodiment, a convolution-enhanced Transformer model is designed at the receiving end, which is used to learn the mapping relationship between the received light intensity image and the turbulence phase screen.

[0081] Preferably, the convolution-enhanced Transformer model gradually reduces the spatial dimension of the feature map in the encoder using the convolution layer, then gradually recovers the spatial dimension of the feature map in the decoder using the up-sampling and convolution layer, and uses the Transformer module in the encoder and the decoder to capture the dependency between the long-distance pixels in the feature map, respectively. The model generates a prediction result in combination with the encoder, the decoder and the final convolution layer;

[0082] At the same time, the turbulence intensity of the current channel is also fed back to the neural network of the transmitting end in the last output layer. The output of the network includes two parts: the turbulence intensity fed back to the transmitting end and the current turbulence phase screen. On the basis of the convolution-enhanced Transformer model architecture, in view of the problem that the traditional Transformer has poor high-bit information extraction capability for the feature map, a convolution module is combined in the architecture to increase the learning of local image features by the network.

[0083] As Figure 4As shown, the Encoder part contains multiple MBConv modules, each followed by an SE (Squeeze-and-Excitation) module. In detail, the SE module learns the relationships between channels in the feature map through global average pooling and fully connected layers, and uses the sigmoid function to output weights to adjust the channel responses of the feature map. The network also includes MaxPool2d layers for downsampling. Similarly, the Decoder part contains multiple ConvTranspose2d convolutional transpose layers for upsampling. Each upsampling layer is followed by a Transformer module, which includes a multi-head self-attention mechanism and a feedforward network for interactions between different regions of the feature map, enhancing feature representation. The Decoder part progressively restores the spatial resolution of the features while maintaining rich feature representations through Skip Connections and the Transformer module.

[0084] Furthermore, the MBConv module is one of the most frequently used modules in the encoder section. Essentially, it is an improved convolutional module, with the following specific structure: Figure five As shown, it mainly includes input convolutional layers, normalization, depthwise convolutional layers, SE modules, pointwise convolutional layers, and downsampling convolutions, using depthwise separable convolutions and SE modules to capture more complex features.

[0085] like Figure 5 As shown, the anti-turbulence free-space optical communication system based on transmit-receive joint optimization includes a receiver neural network model CNN_Transformer, a transmitter network model, a Polar encoder, a waveform generator AWG, an intensity modulator, a laser, a collimating beam expander Col, a spiral phase plate SPP, a turbulence phase plate, a beam splitter BS, a photodetector PD, an oscilloscope, an infrared camera, and a spatial light modulator SLM.

[0086] The transmitter, designed for the turbulent channel, first sends a training sequence after the neural network model outputs the encoded signal. This allows the receiver's neural network model, CNN_Transformer, to receive the light intensity image and determine the turbulence intensity of the current atmospheric channel, which is then fed back to the transmitter's network model. The Polar encoder provides the distribution of Polar encoded information bits under the current channel conditions. According to the model's output, the encoded information is input into an arbitrary waveform generator (AWG). The radio frequency signal output by the AWG is loaded onto the intensity modulator. The laser outputs a 1550nm wavelength laser, which is input into the intensity modulator. A bias controller is used to make the intensity modulator work in the linear region. The encoded information is modulated onto the optical carrier. The output of the intensity modulator passes through a collimating beam expander (Col) and then through a spiral phase plate (SPP) to convert it into an OAM beam with stronger anti-turbulence capability.

[0087] In the channel part, the rotation of the turbulent phase plate is used to simulate the interference of the OAM beam in the atmospheric turbulence channel. At the receiving end, the beam enters a 10:90 beam splitter, where 10% of the beam is displayed as a spot image by an infrared camera and input into the network; the other part passes through a spatial light modulator for phase compensation. The phase pattern displayed on the spatial light modulator SLM is given by the output of the receiving end neural network model. Finally, we use a displacement table to couple the compensated spatial light into an optical fiber and connect it to a photodetector PD after connecting to an oscilloscope. After recording the data, offline processing is performed to analyze the bit error rate.

[0088] Although the embodiments of the present application are described in conjunction with the drawings, various modifications or changes can be made by the patent owner within the scope of the appended claims, as long as they do not exceed the protection scope described in the claims of the present application, and should be within the protection scope of the present application.

Claims

1. A method for turbulence-resistant free-space optical communication based on joint optimization of transmission and reception, characterized in that, The application relates to a Polar coding method based on neural network model. At the transmitting end, the output of the neural network model determines the construction of the Polar code, the reliability of each subchannel is analyzed, and it is given whether the subchannel is a frozen bit or an information bit under the current channel; The coded information is modulated onto an optical carrier, transmitted after collimation and beam expansion, and the size of the Gaussian light beam is controlled by an aperture, and a spiral phase plate is placed at the rear end to convert the Gaussian light beam into an OAM light beam with better anti-turbulence performance; The rotating turbulence phase plate simulates the atmospheric turbulence channel, and the OAM light beam is disturbed by the atmospheric turbulence; At the receiving end, the received light is divided into two beams by a beam splitter, one beam of light irradiates an infrared camera, and the infrared camera is used to collect images and input into the receiving end neural network, the receiving end neural network analyzes the phase of the turbulence channel from the light intensity image and gives the conjugate phase, and the compensated phase image is displayed by an SLM; The other light divided by the beam splitter is compensated by the conjugate phase on the SLM, and then coupled into an optical fiber by using a displacement table, and then enters a photodetector to complete photoelectric conversion; Finally, the data input by the PD is collected by an oscilloscope, saved and analyzed by MATLAB for off-line processing of the bit error rate; The neural network of the receiving end judges the turbulence intensity of the current channel according to the light intensity image and feeds back to the neural network model of the transmitting end, after the light intensity image is identified, the trained model is used to evaluate the new subchannel set, the reliability prediction of each subchannel is obtained, and the output of the model is used as the input of the Polar coding construction to determine the final information bit and frozen bit. 2.The transceiving joint optimization based anti-turbulence free space optical communication method according to claim 1, characterized in that: The Polar coding construction process is to sort the reliability of N subchannels generated after the channel polarization operation from large to small, then select the first K subchannels as information bits, and the remaining N-K subchannels as frozen bits. 3.The transceiving joint optimization based anti-turbulence free space optical communication method according to claim 2, characterized in that, The Polar coding process includes: Collecting the state information of the subchannels by simulation, including the bit error rate of the received signal under different turbulence intensities as the channel transmission characteristics, and selecting the Bhattacharyya parameter of each subchannel after channel polarization and the subchannel index as the features of the neural network model for training; Building a data set, using the known Polar coding construction method as a benchmark, marking whether each subchannel is an information bit or a frozen bit, and using it as a training label; Constructing a neural network model to accept the turbulence intensity and coding parameters as input and output the reliability prediction of each subchannel, and expressing the polarization weight of each subchannel as a probability representing the probability of the subchannel as an information bit learned by the neural network; Defining a loss function to feedback the error between the result and the true value. 4.The transceiving joint optimization based anti-turbulence free space optical communication method of claim 2, wherein: The neural network model includes an input layer for receiving the feature vector of the subchannel; a plurality of fully connected layers for extracting features and learning nonlinear mapping; an LSTM layer for processing channel state changes; and an output layer for outputting the probability or reliability score of each subchannel as an information bit.

5. The transceiving jointly optimized anti-turbulence free-space optical communication method of claim 3, wherein: When defining the loss function, a binary cross-entropy loss function is used as the basic loss to judge whether the current subchannel is a frozen bit or an information bit: where N is the number of samples, y i is the actual reliability label of the subchannel, is the predicted probability by the model, to consider the constraint of code rate, the deviation of the prediction of the model from the target code rate is punished by adding the code rate loss shown in the following formula: where R is the target code rate, usually set to 0.5, λ is the weight for balancing the code rate loss, and L is the code length, usually set to 2048. target where R is the target code rate, usually set to 0.5, λ is the weight for balancing the code rate loss, and L is the code length, usually set to 2048. R where R is the target code rate, usually set to 0.5, λ is the weight for balancing the code rate loss, and L is the code length, L Length = λ L N In order to quantify the change of channel condition, the signal-to-noise ratio is used as the quantitative indicator of channel quality, and the bit error rate is used as the measure of performance difference before and after the change of channel condition, then the loss term under the turbulence channel is as follows: L channel = λ ch · f(ΔBER) In the formula, f is a non-negative penalty factor proportional to the size of the channel condition change, ΔBER is the change in the error rate after the channel condition change, and λ ch a loss balancing hyperparameter; The above combinations form the final loss function as follows, L total = L BCE + L RATE + L Length + L channel .

6. The transceiving jointly optimized anti-turbulence free space optical communication method of claim 1, wherein: The receiving end uses a convolution-enhanced Transformer model to learn the mapping relationship between the received light intensity image and the turbulence phase screen.

7. The transceiving jointly optimized anti-turbulence free-space optical communication method of claim 6, wherein: The convolution-enhanced Transformer model uses convolution layers in the encoder to gradually reduce the spatial dimensions of the feature map, then uses upsampling and convolution layers in the decoder to gradually restore the spatial dimensions of the feature map, and uses Transformer modules in the encoder and the decoder to capture the dependence between pixels with long distances in the feature map, and the model combines the encoder, the decoder and the final convolution layer to generate a prediction result; At the same time, the turbulence intensity of the current channel is also fed back to the neural network of the transmitting end in the last output layer, and the output of the network includes the turbulence intensity fed back to the transmitting end and the current turbulence phase screen, and the convolution module is combined in the architecture based on the convolution-enhanced Transformer model architecture.

8. The turbulence-immune free-space optical communication system based on joint optimization of transmission and reception, characterized by: The receiving end neural network model, the transmitting end network model, the Polar encoder, the waveform generator, the intensity modulator, the laser, the collimating beam expander, the spiral phase plate, the turbulence phase plate, the beam splitter, the photodetector, the oscilloscope, the infrared camera and the spatial light modulator are included. After the transmitting end designs the neural network model output for the turbulence channel, a training sequence is first sent, which is used to let the receiving end neural network model receive the light intensity image to judge the turbulence intensity of the current atmospheric channel and feed back to the transmitting end network model, and the Polar encoder gives the distribution of the Polar encoded information bits under the current channel condition, and the encoded information is input to the arbitrary waveform generator according to the output of the transmitting end network model, the radio frequency signal output by the waveform generator is loaded to the intensity modulator, the laser outputting 1550nm wavelength laser is input to the intensity modulator, the bias controller is used to make the intensity modulator work in the linear region, the information after encoding is modulated to the optical carrier, and the output of the intensity modulator is collimated and expanded by the collimating beam expander, and then converted to an OAM beam with stronger anti-turbulence ability by the spiral phase plate. In the channel part, the turbulence phase plate rotation is used to simulate the interference of the OAM beam in the atmospheric turbulence channel, and in the receiving end, the beam enters the 10:90 beam splitter, of which 10% of the beam is displayed by the infrared camera and input to the receiving end neural network; the other part is phase compensated by the spatial light modulator, and the phase pattern displayed on the spatial light modulator is given by the output of the receiving end neural network model, finally the compensated spatial light is coupled into the optical fiber by using the displacement table and connected to the photodetector, and the oscilloscope is connected after the photodetector, the data is recorded, and the bit error rate is analyzed offline. The neural network of the receiving end judges the turbulence intensity of the current channel according to the light intensity image and feeds back to the network model of the transmitting end. After experiencing the identification of the turbulence intensity of the received light intensity image, the trained model is continuously used to evaluate the new sub-channel set to obtain the reliability prediction of each sub-channel. The output of the model is used as the input of the Polar coding construction to determine the final information bit and frozen bit.

9. The transceiving jointly optimized anti-turbulence free space optical communication system based on claim 8, characterized in that: The neural network model of the receiving end is a CNN_Transformer model.

Images