Coherent detection spatial light cooperative transmission method based on local oscillator elastic light splitting

Abstract

The invention relates to a coherent detection spatial light cooperative transmission method based on local oscillator elastic light splitting, and belongs to the technical field of spatial light communication. The method comprises the following steps: at an optical transmitting end, modulating a bit sequence into an optical signal by adopting a quadrature amplitude modulation method, and then dividing the optical signal into a plurality of channel-irrelevant sub-signals through a fixed optical splitter for broadcast transmission; at a light receiving end, the signals transmitted by the even-numbered links are subjected to phase shifting through a 90-degree phase shifter, and phase orthogonality of the signals of the even-numbered links and the odd-numbered links is achieved; meanwhile, local oscillation optical power is distributed to the photoelectric detectors corresponding to the links in a self-adaptive mode through an elastic optical splitter; and an output signal of the photoelectric detector is demodulated by the coherent receiver to restore a bit sequence. The power distribution of the elastic optical splitter is controlled by a voltage controller, and the output voltage of the voltage controller is dynamically determined according to the atmospheric channel state through a DDPG algorithm. The method can flexibly cope with the influence of different atmospheric channel conditions on the transmission performance.

Description

Technical Field

[0001] The invention belongs to the field of space optical communication counting and relates to a coherent detection space optical cooperative transmission method based on local oscillator elastic splitting. Background Art

[0002] Free Space Optical (FSO) communication has become an effective supplementary solution for large-capacity wireless communication due to its advantages such as no need for spectrum licensing, large channel capacity, and flexible deployment. At present, FSO communication technology is mainly divided into two communication systems: intensity modulation / direct detection (IM / DD) and multi-order modulation / coherent detection. Among them, the multi-order modulation / coherent detection system is widely used in existing optical communication terminals due to its higher detection sensitivity, more flexible modulation method and better wavelength selectivity. It is also an important research direction for future space optical communication technology. Although the FSO system has many advantages, factors such as atmospheric turbulence and path loss greatly restrict the communication quality of the FSO link.

[0003] As an effective technology for FSO communication to combat the effects of atmospheric turbulence and path loss, cooperative communication has been widely studied. However, since FSO is directly exposed to the atmospheric environment, this may lead to differences in the received optical power of different transmission paths in the cooperative transmission system. The traditional proportional allocation scheme of local oscillator optical power leads to unreasonable allocation of optical power resources, making it difficult to maximize the mixing gain of the local oscillator. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a coherent detection spatial optical cooperative transmission method based on local oscillator elastic splitting. In view of the difference in received optical power under different transmission paths caused by the time-varying nature of the atmospheric channel state, an elastic splitter that can adaptively allocate local oscillator optical power according to channel conditions is used, and a deep deterministic policy gradient algorithm is combined to determine the output scheme of the voltage controller, thereby determining the power allocation scheme of the elastic splitter, thereby improving the flexibility of the system and the transmission efficiency.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] A method for coherent detection spatial light cooperative transmission based on local oscillator elastic splitting, the method comprising:

[0007] At the optical transmitter, the bit sequence is modulated into an optical signal using orthogonal amplitude modulation, and then divided into multiple channels of unrelated sub-signals for broadcast transmission through a fixed optical splitter;

[0008] At the optical receiving end, the signal transmitted by the even-numbered link is phase-shifted by a 90° phase shifter to achieve phase orthogonality between the even-numbered link and the odd-numbered link signals; at the same time, the local oscillator optical power is adaptively allocated to the photodetectors corresponding to each link through an elastic beam splitter; the output signal of the photodetector is demodulated by a coherent receiver to restore the bit sequence.

[0009] The power distribution of the elastic splitter is controlled by a voltage controller, wherein the output voltage of the voltage controller is dynamically determined according to the atmospheric channel state through the DDPG algorithm.

[0010] Furthermore, the DDPG algorithm calculates the output voltage of the voltage controller with the optimization objective of minimizing the bit error rate; in the DDPG algorithm, action a L (t) is the power allocation ratio of the local oscillator light, the state s(t) is the channel state of each link, and the reward function is expressed as:

[0011] r L (t)=|log 10 BER(t)|-r th

[0012] In the formula, r th Indicates the initial threshold of the system, which refers to the |log 10 BER|, BER(t) represents the bit error rate; if action a L (t) Reward function r after being applied to the environment L When (t)>0, it means that the local oscillator optical power ratio in this state improves the system performance.

[0013] The steps of the DDPG algorithm to solve the output voltage of the voltage controller include:

[0014] 1) Initialize the experience replay pool D and the Actor network parameters θ a and Critic network parameters θ c , set the target network parameters θ c′ ←θ c ,θ a′ ←θ a ; Set the learning rate and system initial threshold r th and discount factor γ;

[0015] 2) At each moment, the Actor network follows its learned strategy Select a group of local oscillator optical power allocation ratio actions a L (t), the Critic network uses the value function Q(s(t),a L (t)|θ c ) to evaluate the Actor network μ(s t ,at |θ a ) and receives a reward r L (t) and the next state s(t+1);

[0016] 3) The experience tuple (s(t), a L (t),r L (t), s(t+1)) are stored in the experience replay pool D; when the agent accumulates enough experience tuples, it starts training and randomly selects M experience tuples from the experience replay pool D for training; the parameters θ are updated according to the loss function and the policy gradient function respectively c and θ a ; Update the objective function parameters according to the soft update;

[0017] 4) After the training is completed, the atmospheric channel state is sensed in real time through the channel state sensor, and the DDPG algorithm is input to calculate the output voltage of the voltage controller to adjust the power allocation ratio of the elastic splitter in real time.

[0018] Wherein, the loss function is expressed as:

[0019]

[0020] R t =Q(s(t),a L (t)|θ c )

[0021] R′ t =r L (t)+γ·Q(s(t+1),a L (t+1)|θ c )

[0022] In the formula, R t Represents the long-term reward of the Critic network, R′ t represents the estimated long-term reward;

[0023] The policy gradient function is expressed as:

[0024] The beneficial effect of the present invention is that the present invention uses the DDPG algorithm to dynamically estimate the local oscillator splitting ratio according to the channel conditions for different atmospheric channel conditions, and sends the decision result to the voltage controller in real time, and the voltage controller further controls the dynamic allocation of the local oscillator optical power. The present invention uses the local oscillator elastic splitting structure to adaptively determine the optical power allocation scheme of different links, which can flexibly respond to the impact of different atmospheric channel conditions on transmission performance, and can effectively improve the stability and efficiency of the system under dynamic channel conditions, showing a strong ability to resist channel attenuation and environmental changes.

[0025] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0027] Figure 1 The block diagram of the IQ modulation / coherent detection spatial optical cooperative transmission system with local oscillator elastic splitting;

[0028] Figure 2 This is the schematic diagram of the DDPG algorithm;

[0029] Figure 3 is the bit error rate performance of the system based on a fixed optical splitter;

[0030] Figure 4 The bit error rate performance of the system based on the elastic splitter;

[0031] Figure 5 is the system performance under different refractive index constants;

[0032] Figure 6 To compare the performance under different channel conditions;

[0033] Figure 7 The performance comparison is performed under similar IQ channel conditions. DETAILED DESCRIPTION

[0034] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0035] In order to ensure the rationality of local oscillator optical power distribution and reduce the bit error rate of system transmission, an embodiment of the present invention provides a coherent detection spatial optical cooperative transmission method based on local oscillator elastic splitting. In view of the difference in received optical power under different transmission paths caused by the time-varying state of the atmospheric channel, this method designs an elastic splitter that can adaptively distribute local oscillator optical power according to channel conditions, determines the output scheme of the voltage controller through the DDPG algorithm, and the voltage controller further determines the power distribution scheme of the elastic splitter.

[0036] Figure 1 The figure shows the IQ modulation / coherent detection spatial optical cooperative transmission system with local oscillator elastic splitting, which mainly consists of three parts: optical transmitter, atmospheric channel and low-complexity coherent receiver. The transmitter adopts quadrature amplitude modulation (QAM), and then divides the modulated signal into multiple channel-independent sub-signals for broadcast transmission. At the receiving end, the second and fourth (even) links are passed through a 90° phase shifter to intersect with the first and third (odd) links. At this time, the first and third (odd) links are I-channel signals, and the second and fourth (even) links are Q-channel signals orthogonal to the I-channel signal. Then, a simplified coherent structure is used to demodulate the I-channel and Q-channel signals respectively to achieve symbol recovery.

[0037] At the receiving end of the coherent receiver, an elastic optical splitter is used to distribute the local oscillator optical power to the photodetectors of each channel. The distribution ratio of the local oscillator optical power is determined by the output scheme of the voltage controller determined by the DDPG algorithm. The specific process of determining the output scheme of the voltage controller by the DDPG algorithm is as follows:

[0038] 1. First, in the DDPG algorithm, action a L (t) is set to the local oscillator optical power ratio, and the state is set to the channel state of each link s(t) = [h1(t-1),h2(t-1),…,h n (t-1)],h n (t-1) represents the channel state of link n. L (t) After being applied to the environment, an immediate reward r can be obtained L (t) to evaluate action a L (t) performance in state s(t); In order to increase the utilization efficiency of training data, an experience replay mechanism is introduced, that is, saving the historical state, action and reward sequence, and randomly extracting a batch of samples from it for training. Among them, the reward function is defined as:

[0039] r L (t)=|log 10 BER(t)|-r th

[0040] In the formula, r th is the initial threshold of the system, which refers to the |log 10 BER|. When action a L (t) After application to the environment L When (t)>0, it means that the local oscillator optical power ratio in this state improves the system performance. Therefore, the goal of the DDPG algorithm is to maximize the long-term reward.

[0041] 2. Initialization: Initialize the experience replay pool D and initialize the Actor network parameters θ a and Critic network parameters θ c , set the target network parameters θ c′ ←θ c ,θ a′ ←θ a , set the learning rate, system initial threshold r th , discount factor γ and other parameters.

[0042] 3. Interaction process: At each moment, the Actor network acts according to its learned strategy Select a group of local oscillator optical power ratio actions a L (t), the Critic network uses the value function Q(s(t),a L (t)|θ c ) to evaluate the Actor neural network μ(s t ,a t |θ a ) to optimize the Actor neural network during online training.

[0043] 4. When the agent has accumulated enough experience tuples, training begins and the Critic network replays these tuples to obtain the long-term reward of the state s(t):

[0044] R t =Q(s(t),a L (t)|θ c )

[0045] At the same time, the instant reward r L (t) is stored in the corresponding tuple, and the estimated long-term reward is obtained using the Bellman equation:

[0046] R′ t =r L (t)+γ·Q(s(t+1),a L (t+1)|θ c )

[0047] Since the training of the Critic network is to learn how to more accurately evaluate the actions of the Actor network, its loss function is defined as:

[0048]

[0049] In the formula, M represents the number of samples sampled during training. During the training process, the parameters θ of the Critic network are updated by minimizing the loss function c .

[0050] Actor network parameters θ a The policy gradient function is used for update, and its formula is as follows:

[0051]

[0052] 5. By repeating the interaction process, the Actor network will gradually improve its action selection in the environment and enhance the performance of the strategy under a given state. After a certain number of training steps, the DDPG algorithm converges, which is reflected in the fact that the actions generated by the Actor network tend to be stable in the state space and the prediction error of the Critic network decreases. Based on the DDPG algorithm, the output of the voltage controller can be adjusted in real time under dynamically changing channel conditions, thereby adjusting the splitting ratio of the elastic splitter in real time.

[0053] The steps of the DDPG algorithm are as follows:

[0054] 1: Initialize the experience replay pool D;

[0055] 2: Initialize Actor network θ a and the parameters θ of the Critic network c ;

[0056] 3: Set the target network parameters θ c′ ←-θ c ,θ a′ ←θ a ;

[0057] 4: Initialize the environment and execute the local oscillator light average distribution optical power action to obtain the initial state s0 and initial

[0058] Threshold r th ;

[0059] 5: for episodes=0,1,…,L do

[0060] 6: Initialize random process for action exploration;

[0061] 7: for t=0,1,…,T do

[0062] 8: Select the local oscillator optical power ratio a based on the strategy function and exploration noiseL (t)

[0063] 9: Execute action a L (t) Get reward r L (t) and the next state s(t+1);

[0064] 10: The experience tuple (s(t), a L (t),r L (t), s(t+1)) are stored in the experience replay pool D;

[0065] 11: Randomly select M samples from D (s(t), a L (t),r L (t),s(t+1));

[0066] 12: Update θ according to the minimized loss function and policy gradient function a and θ c ;

[0067] 13: Update the objective function parameters according to the soft update;

[0068] 14: end for

[0069] 15: end for

[0070] In order to analyze the relative advantages of the DDPG algorithm in optimizing optical power allocation, this embodiment uses the power allocation results of the Lagrangian method to compare with the DDPG algorithm. In order to analyze the performance advantages of the elastic splitter (ES) over the fixed splitter (FS), a comparative analysis is performed from the two aspects of atmospheric attenuation and atmospheric turbulence. Figure 3 and Figure 4 The performance comparison of the coherent detection space optical cooperative transmission system using FS and ES without adding atmospheric turbulence is demonstrated respectively.

[0071] from Figure 3 and Figure 4 It can be seen that the system bit error rate increases with the link L I,1 / L Q,2 and L I,3 / L Q,4 At the same time, when the received optical power of the I / Q component link is lower than -34dBm, the transmission performance of the system will decrease sharply. I,1 / L Q,2 and L I,3 / L Q,4When the received optical powers are not equal, the BER of the coherent detection space optical cooperative transmission system using ES is lower than that of the coherent detection space optical cooperative transmission system using FS. I,1 / L Q,2 and L I,3 / L Q,4 The greater the difference in received optical power, the more the performance of the FSO cooperative communication system using ES low-complexity coherent detection is improved (e.g. Figure 3 and Figure 4 The system performance can be improved by 1 to 2 orders of magnitude at most. This is because the local oscillator plays the role of optical amplification in the coherent detection process, and the flexible optical splitter can flexibly adjust the local oscillator power splitting ratio according to the received optical power of each link to achieve the maximum mixing gain.

[0072] In order to further understand the effectiveness of ES in improving the anti-turbulence capability of the coherent detection space optical cooperative transmission system, this embodiment compares the effects of the coherent detection space optical cooperative transmission system of ES and FS under different turbulence. I,1 / L Q,2 =6dB / km and L I,3 / L Q,4 =10dB / km, the refractive index structure constant is 1×10 -18 m -2 / 3 ~1×10 -14 m -2 / 3 Changes between. Figure 5 As shown in the figure, the average BER of the system increases with the increase of the refractive index structure constant of the link. It will affect the optical signal strength fluctuation in a small range, so the average BER can better reflect the impact of the refractive index structure constant on system performance. The average BER is the value of each refractive index structure constant of the link. Results of 10 experiments. Figure 5 The error bars in the figure represent the statistical distribution of the BER values ​​of the 10 simulations. When , the ES coherent detection space light cooperative transmission system no longer has a system gain relative to the FS coherent detection space light cooperative transmission system. Figure 5 It can be seen that the ES coherent detection space optical cooperative transmission system is 1×10 -18 m -2 / 3 ~1×10 -15 m -2 / 3 There is a large gain in the weak to moderate turbulence range, with the BER improving by 1 to 1.5 orders of magnitude at most.

[0073] Figure 6The performance of ES and FS in the coherent detection space optical cooperative transmission system under different channel conditions is compared. The channel parameters under different channel conditions are shown in Table 1. Figure 6 It can be seen that the BER of the ES supported by DDPG applied to the coherent detection FSO cooperative system gradually converges with the increase of the number of iterations. The average BER in the figure is the average BER of 10 trainings per iteration. The algorithm reaches convergence after 44 iterations, and the converged BER is close to 10 -5 At the same time, the ES supported by the Lagrangian optimization algorithm applied to the coherent FSO cooperative system cannot optimize the system performance, and may even make it worse. This is because the local oscillator optical power ratio obtained by the Lagrangian optimization method cannot be applied to different channel conditions. Figure 6 It can be found that the BER of FS applied to the coherent detection space optical cooperative transmission system is higher than 10 -3 , and the ES supported by DDPG applied to the coherent detection space optical cooperative transmission system can improve the system performance by more than 2 orders of magnitude compared with FS.

[0074] Table 1

[0075]

[0076]

[0077] In order to compare whether the DDPG algorithm can still maintain its advantage when the overall channel states of the I component and the Q component are similar, this embodiment compares the performance of the Lagrangian optimization algorithm and the DDPG algorithm. The system channel parameter settings are shown in Table 2. Figure 7 The performance comparison between the ES supported by DDPG and the ES supported by Lagrangian algorithm in the coherent detection space optical cooperative transmission system under the condition of similar overall channel state of IQ components is shown. The average BER is the average BER of 10 trainings per iteration. The ES supported by DDPG in the coherent detection space optical cooperative transmission system began to converge at the 47th iteration, and the optimal BER converged to 4.4×10 -5 The ES supported by the Lagrangian optimization algorithm is applied to the coherent detection space optical cooperative transmission system. It only needs to calculate the optimal solution according to the expression, without the need for a convergence process, and the BER can be optimized to 2.25×10 -5 Compared with the DDPG algorithm, the Lagrangian algorithm can achieve optimization more simply and efficiently under this channel condition.

[0078] Table 2

[0079]

[0080] pass Figure 6 and Figure 7The performance comparison of the Lagrangian optimization algorithm and the DDPG algorithm applied to the ES-based low-complexity coherent FSO cooperative communication system under multiple channel conditions shows that both algorithms have their own advantages and disadvantages. The DDPG algorithm is more advantageous in dealing with complex channel conditions, while the Lagrangian optimization algorithm works more efficiently when the channel conditions are not very different.

[0081] In summary, the present invention proposes a method for spatial optical cooperative transmission of coherent detection based on local oscillator elastic splitting. In view of the difference in received optical power under different transmission paths caused by the time-varying nature of the atmospheric channel state, the output scheme of the voltage controller is dynamically determined based on the Lagrangian and DDPG algorithms according to the dynamically changing channel conditions. Furthermore, the voltage controller determines the power allocation scheme of the elastic splitter, thereby realizing the on-demand allocation of local oscillator optical power. According to different channel conditions, the elastic splitting structure designed by the present invention can flexibly adjust the output local oscillator optical power. It is no longer limited to the equal output of the local oscillator optical power, but is flexibly adjusted according to actual conditions. The simulation results show that the present invention is effective in improving the stability and efficiency of the system under dynamic channel conditions, as well as its powerful ability to combat channel attenuation and environmental changes.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A method for spatial light cooperative transmission of coherent detection based on local oscillator elastic spectrometry, characterized in that: The method includes: At the optical transmitter, the bit sequence is modulated into an optical signal using orthogonal amplitude modulation, and then divided into multiple channels of unrelated sub-signals for broadcast transmission through a fixed optical splitter; At the optical receiving end, the signal transmitted by the even-numbered link is phase-shifted by a 90° phase shifter to achieve phase orthogonality between the even-numbered link and the odd-numbered link signal; at the same time, the local oscillator optical power is adaptively allocated to the photodetector corresponding to each link through an elastic optical splitter; the output signal of the photodetector is demodulated by a coherent receiver to restore the bit sequence; The power distribution of the elastic beam splitter is controlled by a voltage controller, wherein the output voltage of the voltage controller is dynamically determined according to the atmospheric channel state by a DDPG algorithm.

2. The method according to claim 1, characterized in that The DDPG algorithm calculates the output voltage of the voltage controller with the optimization goal of minimizing the bit error rate; in the DDPG algorithm, action a L (t) is the power allocation ratio of the local oscillator light, the state s(t) is the channel state of each link, and the reward function is expressed as: r L (t)=|log 10 BER(t)|-r th In the formula, r th Indicates the initial threshold of the system, which refers to the |log 10 BER|, BER(t) represents the bit error rate; if action a L (t) Reward function r after being applied to the environment L When (t)>0, it means that the local oscillator optical power ratio in this state improves the system performance.

3. The method according to claim 1 or 2, characterized in that: The steps of solving the output voltage of the voltage controller by the DDPG algorithm include: Initialize the experience replay pool D and initialize the Actor network parameters θ a and Critic network parameters θ c , set the target network parameter θ c′ ←θ c ,θ a′ ←θ a ; Set the learning rate and system initial threshold r th and discount factor γ; At each moment, the Actor network follows its learned strategy Select a group of local oscillator optical power allocation ratio actions a L (t), the Critic network uses the value function Q(s(t), a L (t)|θ c ) to evaluate the Actor network μ(s t , a t |θ a ) and receives a reward r L (t) and the next state s(t+1); The experience tuple (s(t), a L (t), r L (t), s(t+1)) are stored in the experience replay pool D; when the agent accumulates enough experience tuples, it starts training and randomly selects M experience tuples from the experience replay pool D for training; the parameters θ are updated according to the loss function and the policy gradient function respectively c and θ a ; Update the objective function parameters according to the soft update; After the training is completed, the atmospheric channel state is sensed in real time through the channel state sensor, and the DDPG algorithm is input to calculate the output voltage of the voltage controller to adjust the power allocation ratio of the elastic splitter in real time.

4. The method according to claim 3, characterized in that The loss function is expressed as: R t =Q(s(t),a L (t)|θ c ) R′ t =r L (t)+γ·Q(s(t+1),a L (t+1)|θ c ) In the formula, R t Represents the long-term reward of the Critic network, R′ t represents the estimated long-term reward; The policy gradient function is expressed as:

Images