Active polarization control method and system based on RMS-Prop algorithm

Abstract

The invention provides an active polarization control method and system based on an RMS-Prop algorithm, and relates to the technical field of polarization control, and the method comprises the steps: receiving a voltage signal fed back by a photoelectric detector, carrying out the optimization processing of the voltage signal through the RMS-Prop algorithm, and generating a control voltage needed by the implementation of polarization control; a control voltage is loaded to piezoelectric ceramics in the polarization controller, and the piezoelectric ceramics are controlled to extrude optical fibers to generate phase delay, so that the purpose of active polarization control is achieved. According to the invention, active polarization control can be effectively implemented in the non-polarization-maintaining fiber laser, and linear polarization laser output with high extinction ratio is realized.

Description

Technical Field

[0001] This invention relates to the field of polarization control technology, and in particular to an active polarization control method and system based on the RMS-Prop algorithm. Background Technology

[0002] With the continuous development of high-power laser technology, high-power linearly polarized lasers, as an important high-brightness laser source, have been widely used in fields such as coherent combining, spectral combining, and nonlinear frequency conversion, and have broad application prospects. Currently, linearly polarized lasers are mainly obtained based on the following two schemes: one is to build a fiber laser using fully polarization-maintaining devices to achieve linearly polarized laser output; the other is to achieve linearly polarized laser output by implementing active polarization control in a non-polarization-maintaining fiber amplifier.

[0003] Fully polarization-maintaining fiber lasers, with polarization-maintaining fiber as their core, possess birefringence characteristics and can effectively suppress birefringence depolarization effects, thus obtaining stable linearly polarized laser output. However, as power increases, the fiber core temperature rises, and the thermo-optical effect of the fiber causes changes in birefringence characteristics, leading to birefringence depolarization and a decrease in the polarization degree of the output laser. Furthermore, the combined effects of physical factors such as mode instability and stimulated Brillouin scattering pose significant technical challenges to increasing the power of fully polarization-maintaining fiber lasers. Compared to fully polarization-maintaining fiber lasers, non-polarization-maintaining fiber amplifiers have higher power thresholds for mode instability and nonlinear effects. Therefore, employing active polarization control technology in non-polarization-maintaining fiber lasers can serve as a solution to obtain high-power linearly polarized fiber laser output. Active polarization control technology, by real-time detection and adjustment of the laser's polarization state, ensures that the output laser remains linearly polarized, and is a key technology for implementing active polarization control in non-polarization-maintaining fiber lasers.

[0004] However, existing active polarization control methods cannot achieve linearly polarized laser output with high extinction ratios, resulting in poor flexibility and adaptability. Therefore, there is an urgent need for an active polarization control method that can effectively implement active polarization control in non-polarization-maintaining fiber lasers to achieve linearly polarized laser output with high extinction ratios. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides an active polarization control method and system based on the RMS-Prop algorithm, used to implement active polarization control in non-polarization-maintaining fiber lasers and improve the polarization extinction ratio of polarization control.

[0006] In a first aspect, the present invention provides an active polarization control method based on the RMS-Prop algorithm, the method comprising: The voltage signal fed back from the photodetector is received, and the voltage signal is optimized using the RMS-prop algorithm to generate the control voltage required for polarization control. The control voltage is applied to the piezoelectric ceramic in the polarization controller, and the active polarization control is achieved by controlling the piezoelectric ceramic to squeeze the optical fiber to generate a phase delay.

[0007] Furthermore, the voltage signal is optimized using the RMS-prop algorithm, including: the RMS-Prop algorithm applies the change in the evaluation function caused by random disturbance to the gradient estimation to update the control voltage; and searches for the extremum of the evaluation function along the gradient descent direction by iteratively iterating the control parameters until the evaluation function meets the convergence condition.

[0008] Furthermore, the control parameters include: an evaluation function, a disturbance voltage of random disturbances, a learning rate, a small constant, and a decay rate.

[0009] Furthermore, the iterative formula is:

[0010] in, and They represent the current time. and the next moment The control voltage applied to the polarization controller; Indicates the learning rate; Represents a small constant; express The cumulative squared gradient at time step; This represents the gradient estimate of the evaluation function; This represents the disturbance voltage caused by random disturbances.

[0011] Furthermore, the RMS-Prop algorithm applies the changes in the evaluation function caused by random perturbations to the gradient estimation to update the control voltage, including: Set the initial control voltage, initial cumulative squared gradient, small constants greater than zero, learning rate, and decay rate; The RMS-Prop algorithm generates random perturbations based on the Bernoulli distribution. During the application of voltage to the random perturbations, the algorithm employs a bidirectional perturbation method, generating positive and negative perturbation voltages based on the perturbation voltages of the random perturbations, and then generating positive and negative evaluation functions respectively. Gradient estimates are generated based on the positive and negative evaluation functions, and the cumulative squared gradient is calculated to update the control voltage.

[0012] Furthermore, the cumulative square ladder is calculated as follows: ; in, Indicates the attenuation rate. and They represent the current time. and the next moment The cumulative squared gradient; This represents the gradient estimate.

[0013] Furthermore, the gradient estimate is calculated as follows: ; in, , representing a positive evaluation function, This represents the positive disturbance voltage of the entire polarization controller; , represents the negative evaluation function. This represents the negative disturbance voltage of the entire polarization controller.

[0014] Furthermore, the optimal value of the perturbation voltage is adjusted as follows: first, the perturbation voltage is set to the maximum value within the perturbation voltage range, and then gradually reduced until the RMS-Prop algorithm loses lock in the closed loop. The value before the loss of lock is then selected as the optimal value of the perturbation voltage.

[0015] Furthermore, when using the Jones matrix to describe the polarization characteristics of light, the Jones matrix expression for the output linearly polarized laser is: ; in, , These represent the output linearly polarized laser at... and The amplitude of the electric field intensity in the direction; The Jones matrix is ​​a function of phase delay; , which represents the control voltage applied to the piezoelectric ceramic in the polarization controller, and is proportional to the phase delay; , These represent the input beam at... and The amplitude of the electric field intensity in the direction.

[0016] In a second aspect, the present invention provides an active polarization control system based on the RMS-Prop algorithm, comprising a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of any of the methods described above.

[0017] In summary, this invention provides an active polarization control method and system based on the RMS-Prop algorithm. Compared with existing technologies, the technical solution conceived in this invention can achieve the following beneficial effects: This invention utilizes the RMS-prop algorithm to optimize the voltage signal, generating the control voltage required for polarization control. This control voltage is then applied to a piezoelectric ceramic in the polarization controller. Active polarization control is achieved by controlling the piezoelectric ceramic to compress the optical fiber, creating a phase delay. This control method significantly improves the convergence speed of polarization control, especially in scenarios where fiber nonlinearity and mode instability lead to drastic polarization changes, reducing the number of adjustment iterations and lowering system adjustment delay. Furthermore, it suppresses gradient jitter and drift, maintaining high polarization degree in complex nonlinear environments, reducing polarization state degradation, and improving the stability of output laser power. In addition, it allows for smoother polarization state adjustment, reducing polarization controller adjustment errors and improving adaptability to complex modes in non-polarization-maintaining fibers. In summary, this invention effectively implements active polarization control in non-polarization-maintaining fiber lasers, achieving linearly polarized laser output with a high extinction ratio. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the active polarization control method and system based on the RMS-Prop algorithm provided by the present invention, which implements active polarization control in a non-polarization-maintaining fiber amplifier. Figure 2 This is a schematic diagram of the polarization control structure of an active polarization control method and system based on the RMS-Prop algorithm provided by the present invention. Figure 3 This is a schematic diagram illustrating the influence of the evaluation function of an active polarization control method and system based on the RMS-Prop algorithm provided by this invention on the polarization control effect. Figure 4 This is a schematic diagram illustrating the influence of disturbance voltage on the polarization control effect of an active polarization control method and system based on the RMS-Prop algorithm provided by this invention. Figure 5 This is a schematic diagram illustrating the influence of the learning rate on the polarization control effect of an active polarization control method and system based on the RMS-Prop algorithm provided by this invention. Figure 6 This is a schematic diagram illustrating the influence of attenuation rate on the polarization control effect of an active polarization control method and system based on the RMS-Prop algorithm provided by this invention. Figure 7This is a schematic diagram illustrating the influence of a small constant on the polarization control effect of an active polarization control method and system based on the RMS-Prop algorithm provided by this invention. Figure 8 This is a schematic diagram of the polarization control simulation structure of an active polarization control method and system based on the RMS-Prop algorithm provided by the present invention. Figure 9 This is a schematic diagram of the polarization extinction ratio of the output laser of an active polarization control method and system based on the RMS-Prop algorithm provided by this invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings and embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0021] It should be noted that, in the description of the embodiments of the present invention, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a method, step, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to the method, step, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the method, step, or apparatus that includes that element.

[0022] To implement active polarization control in non-polarization-maintaining fiber lasers and improve the polarization extinction ratio, this invention proposes an active polarization control method and system based on the RMS-Prop algorithm, starting from the working principle of the polarization controller. Specifically, the method includes: receiving a voltage signal from a photodetector; optimizing the voltage signal using the RMS-Prop algorithm to generate the control voltage required for polarization control; applying the control voltage to a piezoelectric ceramic in the polarization controller; and achieving active polarization control by controlling the piezoelectric ceramic to compress the fiber and generate a phase delay.

[0023] It should be noted that the root mean square propagation (RMS-Prop) algorithm is an optimization algorithm in deep learning. Its principle is based on gradient descent to search for the extreme value of the evaluation function. It has the characteristics of fast convergence speed and high control precision. It can dynamically adjust the learning rate according to historical gradient information, making the algorithm more stable and efficient in the optimization process.

[0024] like Figure 1As shown, based on the theoretical model and simulation results of the RMS-Prop algorithm, a structure for implementing active polarization control in a non-polarization-maintaining fiber laser is designed, including a linearly polarized single-frequency seed laser (Seed), a polarization controller (PC), non-polarization-maintaining amplifiers (AMPs), a collimator (CO), a polarization beam splitter (PBS), a high-reflection mirror (Mirror), a photodetector (PD), and an RMS-prop algorithm module.

[0025] Specifically, the output pigtail of the linearly polarized single-frequency seed laser is connected to the non-polarization-maintaining single-mode input pigtail of the polarization controller via a flange. After passing through the polarization controller, the seed laser enters the subsequent non-polarization-maintaining amplifier via the non-polarization-maintaining single-mode output pigtail. The amplified signal laser then enters the polarization beam splitter via a collimator. The polarization beam splitter splits the output laser into horizontally polarized light (P-beam) transmitted through the beam splitter and vertically polarized light (S-beam) reflected by the beam splitter. The P-beam directly enters power meter 1 (PM1), where the output power of the P-beam is measured and recorded as P1. The S-beam is incident on a high-reflection mirror, and the reflected light power of the S-beam is measured using power meter 2 (PM2) and recorded as P2. The weak transmitted light of the S-beam through the high-reflection mirror enters the photodetector, which is used to extract the feedback signal required for active polarization control.

[0026] For example, this invention employs a commercial polarization controller based on the electro-optic effect of lithium niobate crystal (LiNbO3), the structural principle of which is as follows: Figure 2 As shown, the input and output optical fibers of this polarization controller are both single-mode non-polarization-maintaining fibers, which contain four piezoelectric ceramics. The piezoelectric ceramics are driven by an external control circuit. This polarization controller has the characteristic of short response time and can be well applied to active polarization control.

[0027] The polarization controller works as follows: When an external control circuit applies a voltage signal to the piezoelectric ceramic, the piezoelectric ceramic compresses the optical fiber, causing it to produce an elastic-optical effect. This alters the birefringence within the fiber, resulting in different phase delays for the o-ray and e-ray components transmitted in the fiber. Therefore, when the polarization controller is running, each pair of piezoelectric ceramics can be considered equivalent to a birefringent delay plate, and the magnitude of the phase delay depends on the voltage signal applied to the piezoelectric ceramic by the external control circuit.

[0028] It should be noted that, in mathematical models, the Poincaré sphere, elliptic equations, and Jones matrices are typically used to describe the polarization characteristics of light. However, this invention, which focuses on the active polarization control of non-linearly polarized lasers, prioritizes the use of Jones matrices to describe the polarization characteristics of light and the working principle of the polarization controller, considering the convenience of establishing theoretical models and formula derivations.

[0029] will along A beam of light propagating in a specific direction is divided into mutually perpendicular beams. and The directional component, then the input beam in and The amplitudes of the electric field intensity in the directions are as follows: ; ; in, , They represent and Amplitude in direction; Indicates the frequency of light; Represents the wave vector; , They represent and The phase of the light field in the direction.

[0030] If the beam is the input beam of the polarization controller, then after normalization, the polarization characteristics of the light can be described using the Jones matrix as follows: ; in, This indicates that the input beam is in and The ratio of the amplitude of the electric field intensity in the direction; ,express and The phase difference of the electric field in the direction.

[0031] like Figure 2 As shown, the piezoelectric ceramics P1, P2, P3, and P4 inside the polarization controller are all equivalent to a birefringent delay waveplate, which changes the ratio of the electric field intensity amplitude. and the phase difference of the electric field in the x and y directions The value of is changed, thereby altering the polarization state of the output laser.

[0032] Using the Jones matrix, an ideal linear birefringent waveplate can be represented as: ; in, This represents the rotation angle of polarized light, which is also the angle between the fast axis and the horizontal direction; , which represents the phase delay generated in two polarization directions by the piezoelectric ceramic extruding the optical fiber.

[0033] The phase delay in the two polarization directions is controlled by changing the magnitude of the control voltage applied to the polarization controller. This process outputs the desired linearly polarized laser. The Jones matrix expression for the linearly polarized laser output after passing through the polarization controller is: The polarization extinction ratio of the output laser is: .

[0034] in, , These represent the output linearly polarized laser at... and The amplitude of the electric field intensity in the direction; , which represents the phase delay generated in two polarization directions due to the piezoelectric ceramic extruding the optical fiber; The Jones matrix is ​​a function of phase delay; , These represent the input beam at... and The amplitude of the electric field intensity in the direction; , They represent and Output optical power in two polarization directions.

[0035] Due to the Jones matrix of the polarization controller It is a phase delay The function, and the phase delay The polarization is generated by piezoelectric ceramics extruding optical fibers in two polarization directions. In practice, the control voltage applied to the piezoelectric ceramics in the polarization controller... With phase delay Proportional to the polarization characteristics of light, therefore, as an example, when using the Jones matrix to describe the polarization characteristics of light, the Jones matrix expression for the output linearly polarized laser can be: ; in, , These represent the output linearly polarized laser at... and The amplitude of the electric field intensity in the direction; The Jones matrix is ​​a function of phase delay; , which represents the control voltage applied to the piezoelectric ceramic in the polarization controller, and is proportional to the phase delay; , These represent the input beam at... and The amplitude of the electric field intensity in the direction.

[0036] By optimizing the evaluation function And search for the evaluation function based on the RMS-Prop algorithm. The extreme values ​​are used to output the desired linearly polarized laser.

[0037] As an example, the RMS-prop algorithm is used to optimize the voltage signal, including: the RMS-Prop algorithm applies the change in the evaluation function caused by random disturbance to the gradient estimation to update the control voltage; and searches for the extremum of the evaluation function along the gradient descent direction by iterating the control parameters until the evaluation function meets the convergence condition.

[0038] like Figure 2 As shown, the polarization controller changes the polarization state of the output laser by squeezing the optical fiber with four internal piezoelectric ceramics. This invention optimizes the RMS-Prop algorithm to enable its application to active polarization control. The RMS-Prop algorithm applies the evaluation function change caused by random perturbation to the gradient estimation to update the control voltage. The control voltage is applied to the piezoelectric ceramics in the polarization controller. By controlling the piezoelectric ceramics to squeeze the optical fiber, a phase delay is generated, and the desired linearly polarized laser is output.

[0039] Specifically, the control parameters include: evaluation function, perturbation voltage of random perturbation, learning rate, small constant, and decay rate.

[0040] It should be noted that the RMS-prop algorithm updates the control voltage based on gradient estimation. First, the evaluation function is set. Evaluation function It is the control voltage applied to the piezoelectric ceramic in the polarization controller. The function. When the RMS-prop algorithm runs, it provides a control voltage. Apply a small random perturbation voltage Evaluation function This will also result in a corresponding change. The RMS-Prop algorithm will process randomly perturbed voltages. The resulting change in the evaluation function Apply to gradient estimation, then based on the amount of change. Update control voltage The evaluation function is searched for its extreme value along the gradient descent direction by iteratively adjusting the control parameters until the evaluation function reaches its extreme value. The convergence condition is met.

[0041] Furthermore, the iterative formula is:

[0042] in, and They represent the current time. and the next moment The control voltage applied to the polarization controller; Indicates the learning rate; Represents a small constant; express The cumulative squared gradient at time step; This represents the gradient estimate of the evaluation function; This represents the disturbance voltage caused by random disturbances.

[0043] As an example, the RMS-Prop algorithm applies changes in the evaluation function caused by random perturbations to gradient estimation to update the control voltage, including: S101: Set initial control voltage Initial cumulative squared gradient Small constants greater than zero Learning rate and attenuation rate .

[0044] Based on the operating principle of the RMS-Prop algorithm, in the evaluation function During the extreme value search process, the learning rate is used... Adjust the gradient weights of the search. When the learning rate... When the learning rate is small, the convergence speed will be relatively slow; when the learning rate is large... If the learning rate is too large, it will cause oscillations during the search process. During simulation and experimentation, the learning rate needs to be continuously adjusted based on the algorithm's convergence performance. Adjustments are made. As the RMS-Prop algorithm iterates and approaches the extreme point, the learning rate is adjusted. It should be gradually reduced to more accurately reach the evaluation function. The extreme point. Attenuation rate. It will affect the cumulative squared gradient By setting an appropriate attenuation factor, the oscillation phenomenon during the search process can be effectively mitigated.

[0045] The RMS-Prop algorithm generates random perturbations based on the Bernoulli distribution. During the application of voltage to these random perturbations, it employs a bidirectional perturbation approach, generating both positive and negative perturbation voltages based on the perturbation voltage. These, in turn, generate corresponding positive and negative evaluation functions. Gradient estimates are then generated based on these positive and negative evaluation functions, leading to the calculation of the cumulative squared gradient, which is used to update the control voltage. This iterative optimization of the control parameters continues until the evaluation function meets the convergence condition.

[0046] It should be noted that when the disturbance voltage When the voltage follows a Bernoulli distribution, the algorithm converges quickly, and perturbation voltages following a Bernoulli distribution can be easily generated using hardware circuits. This is because the convergence speed is directly related to the perturbation voltage. It is positively correlated, so the larger the perturbation voltage, the faster the convergence speed, but a larger perturbation voltage will lead to a decrease in the accuracy of the RMS-Prop algorithm.

[0047] To balance convergence speed and control accuracy, this invention further incorporates the disturbance voltage. The parameters are set to be adjustable online. To promote rapid convergence, the optimal value of the perturbation voltage is adjusted as follows: first, the perturbation voltage is set to the maximum value within the perturbation voltage range, and then gradually decreased until the RMS-Prop algorithm loses its closed-loop lock. The value before the lock loss is selected as the optimal value of the perturbation voltage. Preferably, the perturbation voltage is adjusted by changing the light intensity signal detected by the photodetector.

[0048] As an example, the cumulative square ladder is calculated as follows: ; in, Indicates the attenuation rate. and They represent the current time. and the next moment The cumulative squared gradient; This represents the gradient estimate.

[0049] As an example, the gradient estimate is calculated as follows: ; in, , representing a positive evaluation function, This represents the positive disturbance voltage of the entire polarization controller; , represents the negative evaluation function. This represents the negative disturbance voltage of the entire polarization controller.

[0050] Specifically, the positive disturbance voltage of the entire polarization controller and negative disturbance voltage They are respectively: ; ; in, Indicates the first The positive perturbation voltage of a piezoelectric ceramic Indicates the first The negative perturbation voltage of a piezoelectric ceramic; Indicates the first The control voltage of the piezoelectric ceramic Indicates the first The perturbation voltage of a piezoelectric ceramic is randomly perturbed.

[0051] Based on the theoretical model of the RMS-Prop algorithm, the main control parameters include: evaluation function, perturbation voltage of random disturbance, learning rate, small constant, and decay rate. To accurately evaluate the impact of control parameters on the running time and control accuracy of the RMS-Prop algorithm, and to better apply it to polarization control experimental systems, this invention also conducts simulation analysis and experimental verification of the influence of each control parameter on the polarization control effect of the RMS-Prop algorithm.

[0052] Analysis and evaluation functions Impact on the polarization control performance of the RMS-Prop algorithm. Evaluation function. It can be: , or .in, , They represent and The light intensity in a direction depends on the different expressions of the evaluation function. , , Representing degree of polarization, Light intensity and polarization extinction ratio in different directions.

[0053] In the simulation analysis, respectively, , , The RMS-Prop algorithm's performance evaluation function uses a polarization degree of 0.95 as the convergence condition. Furthermore, the program is run 1000 times during simulation, and the average is taken to better reflect statistical significance. The RMS-Prop algorithm employs a bidirectional perturbation method, with the perturbation voltage generated using a Bernoulli distribution. In the simulation, the perturbation voltage... Learning rate attenuation rate and small constants Set them to: 0.05, 0.1, 0.1, and 10 respectively. -4 The convergence curves and polarization control effects of different evaluation functions are as follows: Figure 3 As shown, the horizontal axis represents the number of iteration steps.

[0054] like Figure 3 As shown in (a), the following is adopted: , When used as the evaluation function of the RMS-Prop algorithm, the number of iterations required to achieve a polarization degree of 0.95 is 50 and 45 steps respectively, while using When used as an evaluation function for the RMS-Prop algorithm, it struggles to meet the convergence condition of a polarization degree of 0.95 and exhibits slight oscillations. For example... Figure 3 The polarization extinction ratio curve shown in (b) is compared to... , ,use Using it as an evaluation function for polarization control yields the best polarization control effect, with a converged polarization extinction ratio exceeding 60 dB.

[0055] The results show that, considering both the number of iterations required for algorithm convergence and the control effect, the preferred method is... The light intensity in a given direction is used as the evaluation function.

[0056] Disturbance voltage It is one of the important parameters of the RMS-Prop algorithm; therefore, it is necessary to analyze the disturbance voltage. The impact on the polarization control effect of the RMS-Prop algorithm. The RMS-Prop algorithm uses... As the evaluation function, a two-way perturbation method is adopted, and the perturbation voltage is generated using a Bernoulli distribution. The learning rate in the simulation is... attenuation rate and small constants Set them to: 0.1, 0.1, and 10 respectively. -4 The convergence and polarization extinction of the RMS-Prop algorithm under different perturbation voltages, for example... Figure 4 As shown.

[0057] like Figure 4 The polarization degree convergence curve shown in (a) indicates that when the perturbation voltage... When the voltage values ​​are 0.02, 0.03, 0.04, 0.05, 0.06, 0.07, 0.08, and 0.09 respectively, the number of iterations required to achieve a polarization degree of 0.95 are 60, 52, 46, 46, 43, 43, 44, and 44 respectively. That is to say, with the change in perturbation voltage... As the voltage gradually increases, the number of iterations required for the RMS-Prop algorithm to meet the convergence condition gradually decreases and tends to stabilize. However, when the perturbation voltage... When the value is 1, the convergence curve exhibits some fluctuations and fails to meet the convergence condition. For example... Figure 4 As shown in (b), with the disturbance voltage As the voltage gradually increases, the polarization extinction ratio after algorithm convergence gradually decreases, from the perturbation voltage... The value gradually decreased from over 70 dB at 0.02 to 11 dB at 1.

[0058] The results show that as the disturbance voltage increases, the number of iterations required for the algorithm to reach the convergence condition gradually decreases, and the control accuracy of the algorithm also gradually decreases.

[0059] Analyze the learning rate The impact of the learning rate on the polarization control performance of the RMS-Prop algorithm. Adjusting gradient weights allows the current gradient to guide the descent direction of the next gradient, thereby accelerating the search.

[0060] The RMS-Prop algorithm uses As the evaluation function, a two-way perturbation method is adopted, and the perturbation voltage is generated using a Bernoulli distribution. In the simulation, the perturbation voltage... attenuation rate and small constants Set them to: 0.05, 0.1, and 10 respectively. -4 Different learning rates The convergence and polarization extinction of the RMS-Prop algorithm, for example Figure 5 As shown.

[0061] like Figure 5 The polarization degree convergence curve shown in (a) increases with the learning rate. As the learning rate gradually increases, the number of iterations required for the RMS-Prop algorithm to meet the convergence condition decreases significantly. When the values ​​are 0.05, 0.075, 0.1, 0.125, 0.15, 0.2, 0.25, and 0.3, the number of iterations required to achieve a polarization degree of 0.95 are 85, 58, 47, 39, 33, 25, 22, and 19, respectively. Figure 5 (b) shows the polarization extinction ratio, which changes with the learning rate. As the value increased from 0.05 to 0.3, the polarization extinction ratio after algorithm convergence gradually decreased from greater than 65 dB to 44 dB.

[0062] The results show that the smaller the learning rate, the better the polarization control effect, but the longer the algorithm takes to run.

[0063] The RMS-prop algorithm requires the control parameter decay rate when calculating the cumulative squared gradient during the search process. Analyze the attenuation rate The impact on the polarization control effect of the RMS-Prop algorithm.

[0064] The RMS-Prop algorithm uses As the evaluation function, a two-way perturbation method is adopted, and the perturbation voltage is generated using a Bernoulli distribution. In the simulation, the perturbation voltage... Learning rate and small constants Set them to: 0.05, 0.1, and 10 respectively. -4 Different attenuation rates The convergence and polarization extinction of the RMS-Prop algorithm, for example Figure 6 As shown.

[0065] like Figure 6(a) shows the polarization degree convergence curve, when the attenuation rate When the decay rate is 0.1, 0.3, 0.5, 0.7, and 0.9 respectively, the number of iterations required for the algorithm to converge is 44, 45, 43, 40, and 34. As the decay rate increases, the number of iterations required for convergence decreases. Figure 6 As shown in (b), the polarization extinction ratios under different attenuation rates are all above 60 dB, indicating that there is no significant difference in the polarization control effect of the algorithm under different attenuation rates.

[0066] The RMS-prop algorithm also contains a small constant greater than zero. Analyze small constants The impact on the polarization control effect of the RMS-Prop algorithm.

[0067] The RMS-Prop algorithm uses As the evaluation function, a two-way perturbation method is adopted, and the perturbation voltage is generated using a Bernoulli distribution. In the simulation, the perturbation voltage... Learning rate and attenuation rate The values ​​were set to 0.05, 0.1, and 0.1 respectively. Different small constants. The convergence and polarization extinction of the RMS-Prop algorithm, for example Figure 7 As shown.

[0068] like Figure 7 As shown in (a), the small constant δ Take 10 respectively -3 10 -4 10 -5 10 -6 10 -7 With small constant δ Gradually from 10 -3 Reduce to 10 -7 The number of iterations required for the algorithm to converge to a polarization degree of 0.95 shows a trend of first decreasing and then stabilizing, being 67, 45, 41, 40, and 43 respectively. Figure 7 As shown in (b), with the small constant δ From 10 -3 Reduce to 10 -7 The polarization extinction ratios after algorithm convergence show a decreasing trend, at 63, 65, 41, 30, and 27 dB respectively, with small constants. δ The polarization extinction ratio is significantly affected, and the polarization extinction ratio increases with a small constant. δ The decrease showed a clear downward trend.

[0069] The evaluation function was analyzed through the above simulation. Disturbance voltage Learning rate attenuation rate and small constants The impact of various control parameters on the performance of the RMS-prop algorithm was investigated, providing a basis for the selection and optimization of control parameter settings.

[0070] Taking into account both the number of iterations required for algorithm execution and the polarization control effect, and after multiple simulation tests, the preferred control parameter settings are as follows: The directional light intensity is used as the evaluation function, employing a bidirectional perturbation method. The perturbation voltage is generated using a Bernoulli distribution, with a perturbation voltage of 0.06, a learning rate of 0.125, an attenuation rate of 0.1, and a small constant of 10. -4 This combination method serves as the basic control parameter for the RMS-prop algorithm. For example... Figure 8 As shown, based on the settings of this combination of control parameters, the RMS-prop algorithm requires 37 iterations to converge to a polarization degree of 0.95, and the polarization extinction ratio can exceed 55 dB.

[0071] Based on the theoretical model and simulation analysis results of the RMS-Prop algorithm, an experimental system was designed to experimentally verify the polarization control effect of the algorithm. The experimental structure is as follows: Figure 1 As shown, a linearly polarized single-frequency seed laser with a center wavelength of ~1071.5 nm and an output power of ~42 mW is used as the input beam. The polarization extinction ratio of the output laser is calculated according to the formula... To calculate.

[0072] This invention uses the intensity of S-beams as an evaluation function. A small circular aperture with a diameter of approximately 1.5 mm is added to a photodetector to detect the intensity of the S-beams. Then, the photodetector converts the intensity signal of the output laser into a voltage signal and transmits it to the RMS-Prop algorithm controller.

[0073] An RMS-prop algorithm controller was designed and implemented based on a Field-Programmable Gate Array (FPGA). The RMS-prop algorithm controller mainly consists of one input port, four output ports, and an RMS-prop algorithm module. The input port receives a voltage signal from the photodetector. The RMS-prop algorithm module optimizes this voltage signal to generate the control voltage required for polarization control. This control voltage is then applied to the piezoelectric ceramics in the polarization controller through the four output ports. By controlling the four piezoelectric ceramics to compress the optical fiber and generate a phase delay, active polarization control is achieved. This invention optimizes the voltage signal using the RMS-prop algorithm to minimize the intensity of the S-ray, thereby outputting a linearly polarized laser with the desired high extinction ratio.

[0074] First, a power meter was placed at the collimator output terminal to calibrate the amplifier output power to 500 mW. Then, active polarization control was implemented, setting control parameters such as perturbation voltage, learning rate, attenuation rate, and small constant. Keeping the amplifier output power constant, the control system was kept in open-loop mode (i.e., active polarization control was not running). Multiple sets of output power for P-beams and S-beams were recorded over a 5-minute period. The results are as follows: Figure 9 As shown. When the system is in open-loop mode, the polarization extinction of the output laser fluctuates between 1.2 dB and 6.4 dB, indicating that the polarization state of the output laser is random. Then, an active polarization control system based on the RMS-prop algorithm is run, keeping the control system in closed-loop mode, and multiple sets of output powers for the P-beam and S-beam are recorded over 5 minutes. Figure 9 As shown, when active polarization control is performed, the polarization extinction ratio of the output laser can be stably maintained above 10 dB, that is, the P-light power accounts for more than 90% of the total output power.

[0075] Experimental results show that the active polarization control method based on the RMS-Prop algorithm can effectively implement active polarization control in non-polarization-maintaining fiber lasers, thereby achieving linearly polarized laser output with high extinction ratio.

[0076] On the other hand, the present invention also provides an active polarization control system based on the RMS-Prop algorithm, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of any of the methods described above. The technical features of the system are consistent and will not be repeated here.

[0077] In summary, this invention not only significantly improves the convergence speed of polarization control, especially in scenarios where fiber nonlinearity and mode instability lead to drastic polarization changes, reducing the number of adjustment iterations and lowering system adjustment delay, but also suppresses gradient jitter and drift, maintains high polarization degree in complex nonlinear environments, reduces polarization state degradation, and improves output laser power stability. Furthermore, it enables smoother polarization state adjustment, reduces polarization controller adjustment errors, and enhances adaptability to complex modes in non-polarization-maintaining fibers. In conclusion, this invention effectively implements active polarization control in non-polarization-maintaining fiber lasers, achieving linearly polarized laser output with a high extinction ratio.

[0078] It should be noted that, for the sake of simplicity, the foregoing embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0079] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0080] In the several embodiments provided in this application, it should be understood that the disclosed methods or systems can be implemented in other ways. For example, the embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0081] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of embodiments of this disclosure upon considering the specification and practicing the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described herein. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

[0082] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0083] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An active polarization control method based on the RMS-Prop algorithm, characterized in that, The method includes: The voltage signal fed back from the photodetector is received, and the voltage signal is optimized using the RMS-prop algorithm to generate the control voltage required for polarization control. The control voltage is applied to the piezoelectric ceramic in the polarization controller, and the active polarization control is achieved by controlling the piezoelectric ceramic to squeeze the optical fiber to generate a phase delay.

2. The active polarization control method based on the RMS-Prop algorithm according to claim 1, characterized in that, The voltage signal is optimized using the RMS-prop algorithm, which includes: applying the change in the evaluation function caused by random disturbance to the gradient estimation to update the control voltage; and searching for the extremum of the evaluation function along the gradient descent direction by iteratively iterating the control parameters until the evaluation function meets the convergence condition.

3. The active polarization control method based on the RMS-Prop algorithm according to claim 2, characterized in that, The control parameters include: evaluation function, perturbation voltage of random perturbation, learning rate, small constant, and decay rate.

4. The active polarization control method based on the RMS-Prop algorithm according to claim 2, characterized in that, The iterative formula is: in, and They represent the current time. and the next moment The control voltage applied to the polarization controller; Indicates the learning rate; Represents a small constant; express The cumulative squared gradient at time step; This represents the gradient estimate of the evaluation function; This represents the disturbance voltage caused by random disturbances.

5. The active polarization control method based on the RMS-Prop algorithm according to claim 2, characterized in that, The RMS-Prop algorithm applies changes in the evaluation function caused by random perturbations to gradient estimation to update the control voltage, including: Set the initial control voltage, initial cumulative squared gradient, small constants greater than zero, learning rate, and decay rate; The RMS-Prop algorithm generates random perturbations based on the Bernoulli distribution. During the application of voltage to the random perturbations, the algorithm employs a bidirectional perturbation method, generating positive and negative perturbation voltages based on the perturbation voltages of the random perturbations, and then generating positive and negative evaluation functions respectively. Gradient estimates are generated based on the positive and negative evaluation functions, and the cumulative squared gradient is calculated to update the control voltage.

6. An active polarization control method based on the RMS-Prop algorithm according to any one of claims 4 or 5, characterized in that, The cumulative square ladder is calculated as follows: ； in, Indicates the attenuation rate. and They represent the current time. and the next moment The cumulative squared gradient; This represents the gradient estimate.

7. The active polarization control method based on the RMS-Prop algorithm according to claim 6, characterized in that, The gradient estimate is calculated as follows: ； in, , representing a positive evaluation function, This represents the positive disturbance voltage of the entire polarization controller; , represents the negative evaluation function. This represents the negative disturbance voltage of the entire polarization controller.

8. The active polarization control method based on the RMS-Prop algorithm according to claim 1, characterized in that, The optimal value of the disturbance voltage is adjusted as follows: first, the disturbance voltage is set to the maximum value within the disturbance voltage range, and then gradually reduced until the RMS-Prop algorithm loses its closed-loop lock. The value before the lock loss is then selected as the optimal value of the disturbance voltage.

9. The active polarization control method based on the RMS-Prop algorithm according to claim 1, characterized in that, When using the Jones matrix to describe the polarization characteristics of light, the Jones matrix expression for the output linearly polarized laser is: ； in, , These represent the output linearly polarized laser at... and The amplitude of the electric field intensity in the direction; The Jones matrix is ​​a function of phase delay; , which represents the control voltage applied to the piezoelectric ceramic in the polarization controller, and is proportional to the phase delay; , These represent the input beam at... and The amplitude of the electric field intensity in the direction.

10. An active polarization control system based on the RMS-Prop algorithm, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 9.

Images