Tilt aberration pre-correction method based on free space optical communication receiving end

Abstract

The tilt aberration pre-correction method based on the free space optical communication receiving end comprises the following steps: S1, capturing continuous six frames of speckle images through a CCD (Charge Coupled Device) camera, and converting the continuous six frames of speckle images into a centroid coordinate sequence by utilizing a Gray Scale Centroid module; s2, inputting the centroid coordinate sequence into an LSTM module, and obtaining speckle centroid coordinate sequence change characteristics on a continuous time domain; s3, centroid coordinate feature information predicted by the LSTM module is mapped back to a physical coordinate system through a full connection layer, the exact position of the speckle centroid at the future moment is determined, a control signal is generated to drive the optical fiber end face to deviate, and the coupling efficiency is improved; and S4, building a loop simulation system based on the spatial light modulator for testing, and verifying the effective performance under the simulated atmospheric turbulence condition. According to the invention, accurate prediction and pre-correction of the speckle centroid position at the future moment are realized, and the coupling efficiency and the signal stability of the free space optical communication receiving end under the complex atmospheric condition are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of free space optical communication, and in particular to a tilt aberration pre-correction method based on a free space optical communication receiving end. Background Art

[0002] As a high-bandwidth, low-latency wireless communication method, free-space optical communication has been widely used in point-to-point data transmission in recent years. However, the aberration caused by atmospheric turbulence seriously limits the performance and reliability of FSO systems, especially over long distances or in bad weather conditions. In order to improve communication quality and ensure the stability and efficiency of data transmission, it is necessary to develop a technical solution that can correct the aberration of the receiving end in real time.

[0003] Traditional technical solutions usually use static or semi-static optical compensation methods to counter the effects of atmospheric turbulence. These methods include using deformable mirrors or liquid crystal spatial light modulators in adaptive optical systems to correct wavefront errors. Although this method can improve signal quality to a certain extent, it often requires a long response time and is difficult to respond to rapidly changing atmospheric conditions. In addition, traditional solutions mostly rely on post-processing mechanisms and cannot achieve predictive and preventive correction of aberrations, which makes it have certain limitations in practical applications.

[0004] In the existing technology, some advanced adaptive optical systems have begun to try to introduce machine learning algorithms to predict wavefront errors and correct them in advance. However, most of these systems focus on optimizing the transmitter, but pay less attention to dynamic aberration compensation at the receiver. Moreover, when processing speckle images, the existing prediction methods lack sufficient exploration of time series characteristics and fail to effectively capture the changing laws of the speckle centroid coordinates, resulting in insufficient prediction accuracy and failure to meet the needs of high-speed, high-precision real-time correction.

[0005] Therefore, the present invention provides a tilt aberration pre-correction method based on a free-space optical communication receiving end, which solves the above-mentioned technical problems. Summary of the invention

[0006] The present invention is based on a tilt aberration pre-correction method for a free-space optical communication receiving end, comprising:

[0007] S1: 6 consecutive speckle images captured by a CCD camera, which are converted into a centroid coordinate sequence using the Gray Scale Centroid module;

[0008] S2: Input the centroid coordinate sequence into the LSTM module to obtain the change characteristics of the speckle centroid coordinate sequence in the continuous time domain;

[0009] S3: The centroid coordinate feature information predicted by the LSTM module is mapped back to the physical coordinate system through the fully connected layer to determine the exact position of the speckle centroid at the future moment, and a control signal is generated to drive the fiber end face to deviate, thereby improving the coupling efficiency.

[0010] S4: Build a loop simulation system based on the spatial light modulator for testing and verify the effective performance under simulated atmospheric turbulence conditions.

[0011] Preferably, in S1, a CCD camera captures a speckle image, and 6 consecutive frames of the captured speckle images are analyzed to obtain a speckle coordinate sequence in a continuous time domain.

[0012] Preferably, the Gray Scale Centroid module in S1 converts 6 consecutive speckle images into a centroid coordinate sequence, specifically comprising: the Gray Scale Centroid module reads the gray value of each pixel in each speckle image frame, calculates the centroid position of the spot in the speckle image according to the gray value, calculates the weighted coordinate average of the pixel points in the x-dimension and the y-dimension, accumulates the gray values ​​of the pixels in each row and column of each frame of the image, and divides them by the total gray value respectively, and the formula is: Among them I ij Represents the gray value of the position (i, j), N and M are the number of row and column pixels of the two-dimensional image of the CCD camera, and the centroid coordinates (x, y) corresponding to the image are obtained.

[0013] Preferably, in S2, the centroid coordinate sequence is input into the LSTM module to obtain the variation characteristics of the speckle centroid coordinate sequence in the continuous time domain, comprising the following steps:

[0014] S2.1: The centroid coordinates (x, y) corresponding to the six speckle images generated by the Gray Scale Centroid module are organized into a two-dimensional time series data set;

[0015] S2.2: The time series data set is input into the LSTM module and processed through a multi-layer neuron structure. Each time step in the LSTM module receives the state from the previous time series and the center of mass coordinate input at the current moment.

[0016] S2.3: Use the LSTM module to process the centroid coordinates time step by time step, calculate the new state value based on the memory state of the previous moment and the current input, and pass it to the next time step;

[0017] S2.4: After iterative processing of all time steps, the LSTM module outputs the future centroid coordinate feature information.

[0018] Preferably, each pair of centroid coordinates in the two-dimensional time series data set in S2.1 represents position information of a frame of speckle image at a specific time point, and are arranged in chronological order to form an input sequence.

[0019] Preferably, the future centroid coordinate position in S2.4 is predicted by acquiring the speckle centroid coordinate sequence variation characteristics in the continuous time domain through the LSTM module; the speckle centroid coordinate sequence variation characteristics include but are not limited to the displacement, velocity and acceleration dynamic properties of the centroid coordinates in the time series; the speckle centroid coordinate sequence variation characteristics cover how the centroid position moves between consecutive frames, the rate of change of the centroid position per unit time, and the change of the centroid movement rate; the speckle centroid coordinate sequence variation characteristics also include the time correlation, periodicity or random volatility of the centroid coordinates; the dynamic behavior of the speckle in the receiving plane is captured by analyzing the change of the centroid coordinates over time, and the future centroid coordinate position is predicted.

[0020] Preferably, the future center of mass coordinate feature information output by the LSTM module in S3 is passed as input to the fully connected layer, and the fully connected layer receives the future center of mass coordinate prediction value and transforms it into a higher-level feature through nonlinear transformation; the nonlinear transformation includes weighted summation combined with an activation function, and the future center of mass coordinate prediction value is mapped to the physical coordinate system through the activation function sTanh of the last layer of the fully connected layer, and the formula is: sTanh(l)=20×tanh(0.3×l), where l is the input center of mass coordinate value.

[0021] Preferably, the fully connected layer maps the future center of mass coordinate feature information predicted by the LSTM module into a specific numerical value in the physical coordinate system, converts the specific numerical value into a control signal, calculates the required offset according to the mapped physical coordinate value, and converts the offset into an electrical signal or a digital instruction; the electrical signal or digital instruction is transmitted to the actuator to adjust the position of the optical fiber end face.

[0022] Preferably, the construction of the loop simulation system in S4 utilizes Python's AOtools toolkit to generate a random phase screen to simulate atmospheric turbulence, and realizes continuous change of the dynamic phase screen by extrapolation. The generated phase screen is loaded onto a spatial light modulator to modulate the incident laser beam; the laser beam modulated by the spatial light modulator is focused by a lens, and the beam is guided to a CCD camera to capture a speckle image and calculate the center of mass coordinates.

[0023] Preferably, the loop simulation system includes Optisystem's free-space optical communication simulation platform, which includes signal encoding and modulation at the transmitting end, atmospheric turbulence simulation of the transmission channel, and optoelectronic conversion and signal analysis modules at the receiving end, forming a complete communication link simulation environment.

[0024] Compared with the prior art, the technical solution of this application has the following technical effects:

[0025] The present invention uses a CCD camera to capture 6 consecutive speckle images and uses the Gray Scale Centroid module to convert them into a centroid coordinate sequence, which provides a solid foundation for subsequent aberration prediction and is conducive to the subsequent modules to quickly complete the time domain feature extraction of tilt aberration, so that the system can maintain stable and efficient performance under complex and changeable atmospheric conditions.

[0026] The present invention solves the compensation delay problem caused by delays in calculation and response equipment by inputting the centroid coordinate sequence into the LSTM module to obtain the time domain variation characteristics of the speckle centroid coordinate sequence in the continuous time domain. The powerful time series processing capability of the LSTM module is utilized to process the centroid coordinates time step by time step, and a new state value is calculated according to the memory state of the previous moment and the current input, thereby improving the accuracy of the prediction, enabling it to quickly respond to changes in environmental factors such as atmospheric turbulence, achieving high-precision prediction of the speckle centroid position, and improving the coupling efficiency and signal quality of the receiving end of free-space optical communication.

[0027] The present invention maps the future centroid coordinate feature information output by the LSTM module back to the specific value in the physical coordinate system, generates a control signal to drive the fiber end face offset, solves the problem of being unable to achieve predictive correction, uses a fully connected layer to perform nonlinear transformation on the received future centroid feature information, accurately maps it to the physical coordinate system, and then calculates the required offset and converts it into an electrical signal or digital instruction to pass it to the actuator. It achieves a seamless connection from prediction to actual correction, ensures that the fiber end face can quickly make corresponding adjustments, corrects the upcoming tilt aberration in advance, and improves the response speed and correction accuracy of the device.

[0028] The present invention solves the problem of difficulty in realistically reproducing complex atmospheric conditions in an experimental environment by building a loop simulation system based on a spatial light modulator, including using Python's AOtools toolkit to generate a random phase screen to simulate atmospheric turbulence. The continuous change of the dynamic phase screen is achieved through extrapolation, and the generated phase screen is loaded onto the spatial light modulator to modulate the incident laser beam to simulate a communication scenario close to the real thing. This not only tests and verifies the effectiveness of the algorithm, but also evaluates the performance of the system in different environments to ensure its reliability in practical applications. Combined with Optisystem's free-space optical communication simulation platform, a complete communication link simulation environment is formed.

[0029] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application so that it can be implemented in accordance with the contents of the specification, and to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following is a detailed description of the preferred embodiments of the present application in conjunction with the accompanying drawings as follows.

[0030] Based on the detailed description of the specific embodiments of the present application in combination with the accompanying drawings below, those skilled in the art will become more aware of the above and other objects, advantages and features of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can also be obtained based on these drawings without creative work. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, each element or part is not necessarily drawn according to the actual scale.

[0032] Figure 1 It is a flow chart of the tilt aberration pre-correction method based on the free space optical communication receiving end of the present invention;

[0033] Figure 2 It is a flow chart of the future centroid coordinate position of the tilt aberration pre-correction method based on the free space optical communication receiving end of the present invention;

[0034] Figure 3 The structural principle diagram of the tilt aberration pre-correction method based on the free space optical communication receiving end of the present invention;

[0035] Figure 4 The hardware system diagram of the tilt aberration pre-correction method based on the free space optical communication receiving end of the present invention;

[0036] Figure 5 A loss value curve diagram of the tilt aberration pre-correction method based on a free-space optical communication receiving end of the present invention;

[0037] Figure 6 It is a simulation comparison diagram of the tilt aberration pre-correction method based on the free space optical communication receiving end of the present invention;

[0038] Figure 7 This is a comparison diagram of the convergence speed of the tilt aberration pre-correction method based on the free space optical communication receiving end of the present invention;

[0039] Figure 8The probability distribution diagram of the coupling efficiency of the open loop and the closed loop under different turbulences of the tilt aberration pre-correction method based on the free space optical communication receiving end of the present invention;

[0040] Fig. 9 The network structure framework diagram of the tilt aberration pre-correction method based on the free space optical communication receiving end of the present invention;

[0041] Fig.10 A communication eye diagram of the tilt aberration pre-correction method based on a free-space optical communication receiving end of the present invention;

[0042] Fig.11 It is a comparison diagram of bit error rates of the tilt aberration pre-correction method based on the free space optical communication receiving end of the present invention. DETAILED DESCRIPTION

[0043] To make the purpose, technical scheme and advantages of the embodiment of the present application clearer, the technical scheme in the embodiment of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiment of the present application. Obviously, the described embodiment is a part of the embodiment of the present application, rather than all of the embodiments. In the following description, specific details such as specific configuration and components are provided only to help fully understand the embodiments of the present application. Therefore, it should be clear to those skilled in the art that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. In addition, for clarity and brevity, the description of known functions and structures is omitted in the embodiment.

[0044] It should be understood that the references to "one embodiment" or "this embodiment" throughout the specification mean that the specific features, structures, or characteristics associated with the embodiment are included in at least one embodiment of the present application. Therefore, the references to "one embodiment" or "this embodiment" appearing throughout the specification do not necessarily refer to the same embodiment. In addition, these specific features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0045] In addition, the present application may repeat reference numerals and / or letters in different examples. This repetition is for the purpose of simplicity and clarity, and does not in itself indicate the relationship between the various embodiments and / or settings discussed.

[0046] The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist at the same time. The term " / and" in this article describes another type of association object relationship, indicating that there can be two relationships. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the previous and next associated objects are in an "or" relationship.

[0047] The term "at least one" in this article is merely a description of the association relationship of associated objects, indicating that there may be three relationships. For example, at least one of A and B can mean: A exists alone, A and B exist at the same time, and B exists alone.

[0048] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusions.

[0049] Example 1

[0050] This embodiment mainly describes a tilt aberration pre-correction method based on a free-space optical communication receiving end. Figure 1 As shown, the following steps are included:

[0051] S1: 6 consecutive speckle images captured by a CCD camera, which are converted into a centroid coordinate sequence using the Gray Scale Centroid module;

[0052] Furthermore, the Gray Scale Centroid module converts the six consecutive speckle images into a centroid coordinate sequence, specifically including: the Gray Scale Centroid module reads the gray value of each pixel in each speckle image frame, calculates the centroid position of the spot in the speckle image according to the gray value, calculates the weighted coordinate average of the pixel points in the x-dimension and y-dimension, accumulates the gray values ​​of each row and column of each frame of the image, and divides them by the total gray value respectively. The formula is: Among them I ij Represents the gray value of the position (i, j), N and M are the number of row and column pixels of the two-dimensional image of the CCD camera, and the centroid coordinates (x, y) corresponding to the image are obtained.

[0053] S2: Input the centroid coordinate sequence into the LSTM module to obtain the change characteristics of the speckle centroid coordinate sequence in the continuous time domain;

[0054] Furthermore, the centroid coordinate sequence is input into the LSTM module to obtain the change characteristics of the speckle centroid coordinate sequence in the continuous time domain, such as Figure 2 As shown, the following steps are included:

[0055] S2.1: The centroid coordinates (x, y) corresponding to the six speckle images generated by the Gray Scale Centroid module are organized into a two-dimensional time series data set;

[0056] S2.2: The time series data set is input into the LSTM module and processed through a multi-layer neuron structure. Each time step in the LSTM module receives the state from the previous time series and the center of mass coordinate input at the current moment.

[0057] S2.3: Use the LSTM module to process the centroid coordinates time step by time step, calculate the new state value based on the memory state of the previous moment and the current input, and pass it to the next time step;

[0058] S2.4: After iterative processing of all time steps, the LSTM module outputs the predicted future centroid coordinate feature details.

[0059] Furthermore, each pair of centroid coordinates in the two-dimensional time series data set in S2.1 represents the position information of a frame of speckle image at a specific time point, and is arranged in chronological order to form an input sequence.

[0060] Furthermore, the future centroid coordinate position in S2.4 is predicted by obtaining the change characteristics of the speckle centroid coordinate sequence in the continuous time domain through the LSTM module; the change characteristics of the speckle centroid coordinate sequence include but are not limited to the displacement, velocity and acceleration dynamic properties of the centroid coordinates in the time series; the change characteristics of the speckle centroid coordinate sequence cover how the centroid position moves between consecutive frames, the rate of change of the centroid position per unit time, and the change of the centroid movement rate; the change characteristics of the speckle centroid coordinate sequence also include the time correlation, periodicity or random volatility of the centroid coordinates; by analyzing the change of the centroid coordinates over time, the dynamic behavior of the speckle in the receiving plane is captured, and the future centroid coordinate position is predicted.

[0061] S3: The centroid coordinate feature information predicted by the LSTM module is mapped back to the physical coordinate system through the fully connected layer to determine the exact position of the speckle centroid at the future moment, and a control signal is generated to drive the fiber end face to deviate, thereby improving the coupling efficiency.

[0062] Furthermore, the future center of mass coordinate feature information output by the LSTM module is passed as input to the fully connected layer, and the fully connected layer receives the future center of mass coordinate feature information and transforms it into a higher-level feature through nonlinear transformation; the nonlinear transformation includes weighted summation combined with activation function, and the future center of mass coordinate prediction value is mapped to the physical coordinate system through the activation function sTanh of the last layer of the fully connected layer. The formula is: sTanh(l)=20×tanh(0.3×l), where l is the input center of mass coordinate value.

[0063] Furthermore, the fully connected layer maps the future centroid coordinate feature information predicted by the LSTM module into specific values ​​in the physical coordinate system, converts the specific values ​​into control signals, calculates the required offset according to the mapped physical coordinate values, and converts the offset into electrical signals or digital instructions; the electrical signals or digital instructions are transmitted to the actuator to adjust the position of the optical fiber end face.

[0064] S4: Build a loop simulation system based on the spatial light modulator for testing and verify the effective performance under simulated atmospheric turbulence conditions.

[0065] Furthermore, the construction of the loop simulation system uses Python's AOtools toolkit to generate a random phase screen to simulate atmospheric turbulence, and realizes the continuous change of the dynamic phase screen through extrapolation. The generated phase screen is loaded onto the spatial light modulator to modulate the incident laser beam; the laser beam modulated by the spatial light modulator is focused through a lens, and the beam is guided to the CCD camera to capture the speckle image and calculate the center of mass coordinates.

[0066] Furthermore, the loop simulation system includes Optisystem's free-space optical communication simulation platform, which includes signal coding and modulation at the transmitter, atmospheric turbulence simulation of the transmission channel, and optoelectronic conversion and signal analysis modules at the receiver, forming a complete communication link simulation environment.

[0067] This embodiment combines the continuous speckle images captured by the CCD camera with the time series analysis of the LSTM module to achieve accurate prediction of the speckle centroid position at future moments, maps the prediction results back to the physical coordinate system through the fully connected layer, generates a control signal to drive the optical fiber end face offset, thereby improving the coupling efficiency, and improves the accuracy and timeliness of the prediction by introducing the time series alignment mechanism and deep learning algorithm, ultimately bringing more stable and reliable communication performance.

[0068] Example 2

[0069] This embodiment describes in detail the structural model of the tilt aberration pre-correction method based on the free space optical communication receiving end. Figure 3 As shown, specifically including:

[0070] The structural model includes: a beam splitter, a coupling lens, a CCD camera, and a PC. The CCD camera is used to detect the received light spot information, and a control system with predictive capability is used to output a control signal to drive the optical fiber end face to deviate, thereby improving the coupling efficiency.

[0071] For the overall offset of the speckle centroid coordinates in the structural model, rmse is used as the evaluation index, and the formula is: Where (x p ,y p) is the predicted value of the center of mass coordinate of the light spot at time t, (x t ,y t ) is the true value of the light spot centroid coordinate at time t, m is the number of light spot centroid coordinates, and i is the i-th light spot centroid coordinate.

[0072] This embodiment describes in detail the structural model of the tilt aberration pre-correction method based on the free-space optical communication receiving end, and provides an evaluation index formula, which can more objectively determine the overall offset of the speckle centroid coordinates and achieve accurate measurement.

[0073] Example 3

[0074] This embodiment is based on Embodiment 1, and describes in detail the data generation and performance test of the tilt aberration pre-correction method based on the free space optical communication receiving end, which specifically includes:

[0075] Based on the atmospheric freezing flow hypothesis, the von Karman turbulence power spectrum inversion method built into the AOtools toolkit of Python is used to generate a random phase screen, and the wavefront distortion of atmospheric turbulence is measured; the atmospheric turbulence phase screen of fixed size is slid and intercepted, and the phase covariance function is used to repeatedly calculate the newly added phase information by using the extrapolation method, thereby realizing the dynamic simulation of the atmospheric turbulence phase screen; the numerical simulation of 1 km laser atmospheric horizontal transmission is set up, in which the receiving aperture is 0.1m, the spatial sampling is 1920×1200, the beam wavelength is 1550nm, and the sampling frequency is 500Hz; the detailed parameters of the atmospheric turbulence simulation are shown in Table 1, D / r0 is used to describe the influence of turbulence on the beam, where D is the diameter of the telescope and r0 is the Freid parameter;

[0076] Table 1 Atmospheric turbulence simulation parameter settings

[0077]

[0078] A hardware loop system for laser atmospheric transmission based on a spatial light modulator is built on the optical platform. The phase screen is loaded into the spatial light modulator to modulate the incident laser to achieve the wavefront distortion effect of atmospheric turbulence. The spatial light modulator has a resolution of 1920×1200 and a frame rate of 60 Hz. Finally, it is focused by a lens with a focal length of 300 mm and the spot data is collected by a CCD camera. Figure 4 As shown in the figure, a is the system diagram, b is the phase screen loaded by the spatial light modulator, and c is the speckle received by the CCD; the speckle data set and verification set used in this article are both generated by the hardware loop system of the laser atmospheric transmission part.

[0079] Five sets of continuous-time coordinate data simulated with different D / r0 are simultaneously input into the coordinate pre-correction model to improve the robustness of the model. In order to adapt to the time delay in most systems and avoid excessive calculation and overfitting problems, 6 consecutive frames of data are selected as the input of the model, that is, the center of mass position coordinates from time t to time t+5 are used as the model input, and the center of mass position at time t+8 is set as the data label, which effectively avoids the problem of the model processing too much data and solves the compensation delay problem caused by the equipment operation delay. 12,000 sets of data are generated under each D / r0 condition, the first 10,800 sets of data are used as training sets, and the last 1,200 sets of data are used as verification data. A total of 54,000 sets of data are used to constitute the training set, and 6,000 sets of verification data are used as the verification set.

[0080] The model training process is carried out on a desktop workstation, which includes AMD Ryzen 97950X, NVIDIA GeForce RTX 4090 and other devices. The code is implemented using the keras framework and uses the Adam optimizer mechanism to adaptively adjust the learning rate based on momentum and second-order matrix. The initial learning rate is 0.001. The Dropout mechanism is introduced in the training process to randomly discard 20% of the neurons to prevent overfitting of the model training. Figure 5 The figure shows the loss value curves of the model training set and test set. In the figure, the training set basically completed convergence fitting in 20 rounds. At the same time, the loss value curve of the model in the test set basically coincides with that of the training set, indicating that the model has completed convergence and there is no overfitting problem.

[0081] We tested the prediction ability of the trained model under different circumstances. We obtained the centroid position at time t+8 from the six consecutive frames of data from time t to t+5 in the simulation, and compared it with the ideal value prepared in advance. The test process used the RMS of the overall coordinate offset of the ideal and predicted results to compare and measure the prediction effect. The formula is: Where m is the number of light spot centroid coordinates, and i is the coordinate of the i-th light spot centroid;

[0082] like Figure 6As shown in the figure, (a) is the RME value of the predicted value and the true value of the test data set, and (b) is the residual RMS of the predicted value and the true value of the test data set; the RMS value of the original beam center of mass position offset (Truth RMS), the RMS value of the predicted center of mass position offset (Prediction RMS) and the RMS value of the residual (Residual RMS) of the two are plotted, and the change curve of each turbulence in 600 consecutive frames. From the curve in the figure, it can be seen that the larger the atmospheric turbulence D / r0, the larger the RMS and the ResidualRMS, but the value of Residual RMS is stable below 0.04λ, which shows that the pre-correction method has a good prediction ability for tilt aberrations of different turbulence intensities, and good generalization performance in different environments.

[0083] In order to test the closed-loop rate advantage of the pre-correction model under different atmospheric turbulence, the performance is compared with that of the classic SPGD algorithm to verify the technical advancement of the model. The model and the traditional SPGD algorithm are used to correct and compensate for the centroid offset of the receiving end speckle at different D / r0. The performance of the two algorithm models in the closed loop is as follows: Figure 7 As shown, the result is the average of the iteration time of 20 sets of independent data. In the face of different turbulence, SPGD requires 5 to 7 iterations to complete the closed loop, but the prediction model proposed in this application only requires one prediction to enable the system to complete the closed-loop control, and the prediction time is 0.04ms. The closed-loop efficiency is two orders of magnitude higher than the SPGD algorithm.

[0084] For the test of coupling efficiency, the coupling efficiency of spatial light to single-mode fiber is a widely used indicator in free-space optical communication. However, the bit error rate is nonlinearly related to the coupling efficiency. Even with a higher coupling efficiency, the signal will still have a large jitter. Therefore, the probability distribution of the coupling efficiency at the open-loop and closed-loop moments of the pre-correction network, as well as the corresponding variance, are tested and analyzed. The results are as follows: Figure 8 As shown in the figure, at the closed-loop moment, the coupling efficiency under different turbulence conditions is greatly improved, and the probability distribution of the coupling efficiency of weaker turbulence being greater than 0.7 is more than 90%, and even when D / r0=3, the coupling efficiency can be increased to above 0.6; in the field of laser transmission, D / r0 is usually used to describe the effect of turbulence on the light beam, where D is the diameter of the telescope and r0 is the Freid parameter.

[0085] In the simulation test, we calculated and compared the variance of the coupling efficiency at the open and closed loop moments. As the turbulence increases from small to large, the variances at the open loop moment are 0.03359, 0.047829, 0.044962, 0.040021, and 0.036353, respectively; after the closed loop is implemented, these values ​​drop to 0.000553, 0.001817, 0.002181, 0.002538, and 0.002552, respectively; after closed-loop control, the higher coupling efficiency and lower variance show a lower bit error rate.

[0086] For the test of communication performance, a single-aperture transmission and single-aperture receiving system of free-space optical communication was built in the OptiSystem simulation system to verify the performance of the test pre-correction model in the communication system; the transmitter uses non-return-to-zero coding (NRZ), and the Mach-Zehnder modulator (MZ Modulator) loads the pseudo-random sequence electrical signal onto the 1550nm laser carrier. The laser transmission power is -4dBm, and the system code rate is 10G bit / s. After propagation on the free-space optical link (FSOChannel), the coupling efficiency is loaded as multiplicative noise onto the receiving end beam signal, and then the photoelectric conversion is used to receive the signal, and the signal bit error rate (BER) is analyzed after passing through the low-pass Gaussian filter (LPG Filter); Fig. 9 This is the network framework diagram of the OptiSystem simulation system.

[0087] like Fig.10 As shown, (a) is an open loop, and (b) is a closed loop; Fig.10 The eye diagrams of the communication system are compared during open-loop and closed-loop periods. After the loop is closed, the system eye diagram signal line is narrowed, the eye height is increased, and the line clarity is improved, which indicates that the signal voltage noise is reduced, the time domain jitter is weakened, the inter-code crosstalk is improved, and the signal quality is significantly improved.

[0088] In order to explore the influence of turbulence intensity on the communication quality of the pre-correction model, the bit error rate of the pre-correction model at the open-loop and closed-loop moments was tested under different D / r0 environments. 15 sets of comparative data were tested in each turbulence environment. The results are as follows: Fig.11 As shown in the figure, at the open loop moment, the bit error rate under each turbulence is higher than 0.001, and the average value reaches 0.02191. After the pre-correction model completes the closed loop, the link bit error rate is significantly reduced to 1.1e -9 Below this, the telecommunication level of the communication system is reached; when D / r0=1, the system bit error rate is basically reduced to 0, and the effect is most significant.

[0089] This embodiment describes in detail the data generation and performance test of the tilt aberration pre-correction method at the receiving end of free-space optical communication, and simulates under five turbulence intensities. It can be seen from the above content that the method has good generalization under different turbulences, and can complete convergence more quickly than the traditional SPGD algorithm. At the same time, the method significantly improves the coupling efficiency and greatly reduces the variance of the coupling efficiency. The optical communication system is built by using Optisystem simulation, which has a significant improvement in bit error rate and communication stability, verifying its effectiveness. At the same time, without adding equipment, the closed-loop bandwidth is effectively improved, which is a key advantage for the deployment of free-space optical communication network equipment, and can achieve autonomous operation and avoid parameter optimization process.

[0090] The above are only preferred embodiments of the present invention, which do not limit the scope of protection of the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any changes, modifications, replacements, integrations and parameter changes to these embodiments within the spirit and principles of the present invention through conventional substitutions or without departing from the principles and spirit of the present invention fall within the scope of protection of the present invention.

Claims

1. A tilt aberration pre-correction method based on a free space optical communication receiving end, characterized in that: include: S1: 6 consecutive speckle images captured by a CCD camera, which are converted into a centroid coordinate sequence using the Gray Scale Centroid module; S2: Input the centroid coordinate sequence into the LSTM module to obtain the change characteristics of the speckle centroid coordinate sequence in the continuous time domain; S3: The centroid coordinate feature information predicted by the LSTM module is mapped back to the physical coordinate system through the fully connected layer to determine the exact position of the speckle centroid at the future moment, and a control signal is generated to drive the fiber end face to deviate, thereby improving the coupling efficiency. S4: Build a loop simulation system based on the spatial light modulator for testing and verify the effective performance under simulated atmospheric turbulence conditions.

2. The tilt aberration pre-correction method based on a free space optical communication receiving end according to claim 1, characterized in that: In S1, a CCD camera is used to capture 6 consecutive frames of speckle images, and the centroid coordinates of the captured 6 consecutive frames of speckle images are analyzed to obtain a centroid coordinate sequence in a continuous time domain.

3. The tilt aberration pre-correction method based on a free space optical communication receiving end according to claim 1 or 2, characterized in that: The Gray Scale Centroid module in S1 converts 6 consecutive speckle images into a centroid coordinate sequence, which specifically includes: the Gray Scale Centroid module reads the gray value of each pixel in each speckle image frame, calculates the centroid position of the spot in the speckle image according to the gray value, calculates the weighted coordinate average of the pixel points in the x-dimension and the y-dimension, accumulates the gray values ​​of each row and column of each frame of the image, and divides them by the total gray value respectively. The formula is: Among them I ij Represents the gray value of the position (i, j), N and M are the number of row and column pixels of the two-dimensional image of the CCD camera, and the centroid coordinates (x, y) corresponding to the image are obtained.

4. The tilt aberration pre-correction method based on a free space optical communication receiving end according to claim 1, characterized in that: In S2, the centroid coordinate sequence is input into the LSTM module to obtain the variation characteristics of the speckle centroid coordinate sequence in the continuous time domain, including the following steps: S2.1: The centroid coordinates (x, y) corresponding to the six speckle images generated by the Gray Scale Centroid module are organized into a two-dimensional time series data set; S2.2: The time series data set is fed into the LSTM module and processed through a multi-layer neuron structure, with each time step in the LSTM module receiving the state from the previous time series and the center of mass coordinate input at the current moment; S2.3: Use the LSTM module to process the centroid coordinates time step by time step, calculate the new state value based on the memory state of the previous moment and the current input, and pass it to the next time step; S2.4: After iterative processing of all time steps, the LSTM module outputs the predicted future centroid coordinate feature information.

5. The tilt aberration pre-correction method based on a free space optical communication receiving end according to claim 1, characterized in that: Each pair of centroid coordinates in the two-dimensional time series data set in S2.1 represents the position information of a frame of speckle image at a specific time point, and is arranged in time sequence to form an input sequence.

6. The tilt aberration pre-correction method based on a free space optical communication receiving end according to claim 1, characterized in that: The future centroid coordinate position in S2.4 is predicted by obtaining the change characteristics of the speckle centroid coordinate sequence in the continuous time domain through the LSTM module; the change characteristics of the speckle centroid coordinate sequence include but are not limited to the displacement, velocity and acceleration dynamic properties of the centroid coordinate in the time series; the change characteristics of the speckle centroid coordinate sequence cover how the centroid position moves between consecutive frames, the rate of change of the centroid position per unit time, and the change of the centroid movement rate; the change characteristics of the speckle centroid coordinate sequence also include the time correlation, periodicity or random volatility of the centroid coordinate; the dynamic behavior of the speckle in the receiving plane is captured by analyzing the change of the centroid coordinate over time, and the future centroid coordinate position is predicted.

7. The tilt aberration pre-correction method based on a free space optical communication receiving end according to claim 1, characterized in that: The future centroid coordinate feature information output by the LSTM module in S3 is passed as input to the fully connected layer, and the fully connected layer receives the future centroid coordinate feature information and converts it into physical information, i.e., coordinate prediction value, through nonlinear transformation. The nonlinear transformation includes weighted summation combined with activation function. Through the activation function sTanh of the last layer of the fully connected layer, the future centroid coordinate feature information is mapped to the physical coordinate system. The formula is: sTanh(l)=20×tanh(0.3×l), where l is the input centroid coordinate feature information.

8. The tilt aberration pre-correction method based on a free space optical communication receiving end according to claim 1 or 7, characterized in that: The fully connected layer maps the future center of mass coordinates predicted by the LSTM module into specific values ​​in the physical coordinate system, converts the specific values ​​into control signals, calculates the required offset according to the mapped physical coordinate values, and converts the offset into electrical signals or digital instructions; the electrical signals or digital instructions are transmitted to the actuator to adjust the position of the optical fiber end face.

9. The tilt aberration pre-correction method based on a free space optical communication receiving end according to claim 1, characterized in that: The construction of the loop simulation system in S4 uses Python's AOtools toolkit to generate a random phase screen to simulate atmospheric turbulence, and realizes the continuous change of the dynamic phase screen through extrapolation. The generated phase screen is loaded onto the spatial light modulator to modulate the incident laser beam; the laser beam modulated by the spatial light modulator is focused by a lens, and the beam is guided to a CCD camera to capture the speckle image and calculate the center of mass coordinates.

10. The tilt aberration pre-correction method based on a free space optical communication receiving end according to claim 1 or 9, characterized in that: The loop simulation system includes Optisystem's free-space optical communication simulation platform, which includes signal coding and modulation at the transmitting end, atmospheric turbulence simulation of the transmission channel, and optoelectronic conversion and signal analysis modules at the receiving end, forming a complete communication link simulation environment.

Images