Intelligent decision-based method and system for coupling of dfb laser chip with photodetector probe

Abstract

The application discloses a DFB laser chip photoelectric probe coupling method and system based on intelligent judgment, adopts a double-path vision mode to drive the photoelectric probe, that is, uses a top image to control the contact between the photoelectric probe and the laser chip, collects an infrared spot image generated after a driving current is applied to the laser chip after the contact, and predicts the power distribution based on the infrared spot image; then, the optimal displacement instruction is generated by combining historical coupling data and real-time optical power and with the aid of an intelligent decision model; then, the photoelectric probe is moved based on the optimal displacement instruction, and a near-infrared spot image is collected again after the movement, so that the movement path of the photoelectric probe is continuously adjusted; thus, the application constructs a full-closed-loop coupling technology of multi-modal perception-intelligent decision-precise execution, realizes the real-time adjustment of the movement path of the photoelectric probe, and therefore, the coupling precision can be greatly improved, so that the latest industry demand can be met.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of optoelectronic device testing and packaging, and particularly relates to a DFB laser chip photoelectric probe coupling method and system based on intelligent judgment. BACKGROUND

[0002] As a core light source of 5G optical communication, data center interconnection and silicon optical integration, the chip-level testing of the DFB laser needs to achieve high-precision coupling between the photoelectric probe and the sub-micron waveguide (typical width 1.5-3 um); wherein, the early manual operation stage completely depends on the operator's microscope visual adjustment, and the positioning accuracy is limited by the resolution limit of the human eye (>1 um), and the coupling process depends on the operator's experience, and the maximum value point of the coupling power is searched in the optical power detection interface; at the same time, according to the production line report of an internationally renowned optoelectronic device company, the single-chip testing time under the manual operation mode is about 25 minutes, and the labor cost accounts for more than 60% of the total testing cost, and under the manual operation mode, the precision fluctuation between different products of the same production line is significant, and the standard deviation of the coupling loss of different operators may reach 2dB (the ideal value should be <0.2dB), therefore, the discrete nature of such manual testing data forces the design margin of the chip on the driving current parameter to be enlarged, which will greatly shorten the service life of the device.

[0003] Since 2015, some leading enterprises have begun to try to complete the coupling work by using a semi-automatic visual positioning system, that is, the image is first collected by a CCD camera and the chip is positioned by edge detection, once the system obtains the chip positioning information, the stepper motor can be adjusted according to the preset motion path to adjust the probe position, finally, whether the coupling standard is reached is judged by a fixed threshold optical power; in this way, under the application of semi-automatic technology, the coupling efficiency of the photoelectric probe is greatly improved, the single-chip testing time is shortened to 8-12 minutes, and the precision is improved to ±0.5 um; however, the fixed scanning path cannot adapt to the process fluctuation of the chip, and a ±0.2 um deviation of the waveguide width may lead to coupling failure, and the repeated positioning accuracy of the motor is affected by the guide rail return difference, thermal expansion, etc., resulting in a maximum deviation of ±1.5 um in the positioning accuracy; therefore, with the evolution of the semiconductor laser industry to 800G DR8 (8-channel DFB) and CPO (co-packaged optics), the coupling precision of the semi-automatic technology has failed to meet the industry demand; based on this, how to provide a DFB laser chip photoelectric probe coupling method with high coupling precision has become a problem to be solved. SUMMARY

[0004] The purpose of the present application is to provide a DFB laser chip photoelectric probe coupling method and system based on intelligent judgment, to solve the problem of low coupling precision caused by the fixed path used by the existing semi-automatic coupling technology for photoelectric probe coupling.

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

[0006] In a first aspect, a DFB laser chip photoelectric probe coupling method based on intelligent decision is provided, comprising:

[0007] Obtaining a top image of the laser chip;

[0008] Based on the top image, controlling the photoelectric probe on the nanometer positioning platform to contact the electrode plane of the laser chip and judging whether the preset contact condition is met;

[0009] If yes, applying a driving current to the laser chip, and after applying the driving current, collecting a near-infrared light spot image formed by the waveguide end of the laser chip;

[0010] Extracting the light spot spatial intensity distribution characteristics of the near-infrared light spot image, and inputting the light spot spatial intensity distribution characteristics into a power prediction model to obtain light spot power distribution information and an optimal moving direction of the probe;

[0011] Obtaining historical photoelectric probe coupling data and real-time optical power readings of the laser emitted by the laser chip;

[0012] Inputting the light spot power distribution information, the optimal moving direction of the probe, the real-time optical power readings and the historical photoelectric probe coupling data into a decision control model to obtain an optimal displacement instruction of the photoelectric probe;

[0013] Adjusting the position of the photoelectric probe according to the optimal displacement instruction, and re-collecting the near-infrared light spot image after adjusting the position until the real-time optical power readings reach the preset condition, thereby completing the photoelectric probe coupling of the laser chip.

[0014] Based on the above disclosure, this invention first acquires a top image of the laser chip; then, based on this top image, it controls a photoelectric probe on a nano-positioning platform to contact the electrode plane of the laser chip, and when the contact between the two meets a preset contact condition, it applies a driving current to the laser, so as to acquire a near-infrared spot image formed by the laser chip after applying the driving current; then, this invention extracts the spatial intensity distribution characteristics of the near-infrared spot image and inputs it into a power prediction model to obtain the spot power distribution information and the optimal movement direction of the probe; next, it acquires historical photoelectric probe coupling data and real-time optical power readings of the laser emitted by the laser chip, and inputs them together with the output data of the power prediction model into a decision control model to obtain the optimal displacement command of the photoelectric probe; then, according to the optimal displacement command, the position of the photoelectric probe can be adjusted, and after adjustment, the driving current is reapplied to reacquire the near-infrared spot image; finally, the above process is repeated to continuously adjust the movement path of the photoelectric probe until the acquired real-time optical power reading reaches the preset condition, thus completing the photoelectric probe coupling of the laser chip.

[0015] Through the above design, this invention employs a dual-path vision approach to drive the photoelectric probe. Specifically, it uses a top image to control the contact between the photoelectric probe and the laser chip. After contact, it acquires an image of the infrared light spot emitted after applying a driving current to the laser chip, using this image to predict power distribution. Then, combining historical photoelectric probe coupling data and real-time optical power, and utilizing an intelligent decision-making model, it generates the optimal displacement command for the photoelectric probe. Next, based on the optimal displacement command, the photoelectric probe moves, and after movement, a driving current is reapplied to reacquire the near-infrared light spot image. This process is repeated to continuously adjust the movement path of the photoelectric probe until the acquired real-time optical power reading reaches a preset condition, thus completing the photoelectric probe coupling to the laser chip. In this way, this invention constructs a fully closed-loop system of multimodal perception, intelligent decision-making, and precise execution, achieving real-time adjustment of the photoelectric probe's movement path. Therefore, compared to the fixed-path coupling method in semi-automatic technology, this invention can significantly improve coupling accuracy, meeting the latest industry demands and making it highly suitable for large-scale application and promotion.

[0016] In one possible design, based on the top image, the photoelectric probe on the nanopositioning platform is controlled to contact the electrode plane of the laser chip, and it is determined whether preset contact conditions are met, including:

[0017] Based on the top image, the nanopositioning platform is driven to operate so that the photoelectric probe on the nanopositioning platform is parallel to the electrode plane of the laser chip;

[0018] The photoelectric probe is controlled to descend vertically at a first speed, and the contact force on the photoelectric probe is monitored in real time to determine whether the contact force is greater than the contact force threshold.

[0019] If so, the operating mode of the photoelectric probe is switched to fine mode, wherein the fine mode is in which the photoelectric probe descends vertically at a second speed, and the second speed is less than the first speed;

[0020] The contact force on the photoelectric probe in fine mode is collected in real time, and it is determined whether the contact force of the photoelectric probe in fine mode meets the preset contact conditions.

[0021] If so, the driving current is applied to the laser chip.

[0022] In one possible design, driving the nanopositioning platform based on the top image includes:

[0023] Image recognition is performed on the top image to determine the coordinates of the electrode center of the laser chip;

[0024] The coordinates of the electrode center are transformed into the mechanical coordinate system of the nano-positioning platform to obtain the actual coordinates of the electrode center;

[0025] The nano-positioning platform is driven to operate based on the actual coordinates of the electrode center.

[0026] In one possible design, image recognition is performed on the top image to determine the coordinates of the electrode center of the laser chip, including:

[0027] The top image is preprocessed to obtain a preprocessed image;

[0028] A pre-trained deep learning segmentation model is used to perform image recognition and segmentation on the preprocessed image to segment the electrode pixel region image in the top image;

[0029] The centroid of the electrode pixel region image is calculated, and the centroid is used as the coordinates of the electrode center.

[0030] In one possible design, the coordinates of the electrode center are transformed to the mechanical coordinate system of the nano-positioning platform to obtain the actual coordinates of the electrode center, including:

[0031] Obtain the pixel equivalents in the X and Y axes of the pixel coordinate system corresponding to the top image, and the translation offsets in the X and Y axes of the mechanical coordinate system;

[0032] Based on the pixel equivalent and the translation offset, the coordinates of the electrode center are transformed to the mechanical coordinate system to obtain the actual coordinates of the electrode center.

[0033] In one possible design, the drive current is calculated in the following manner;

[0034] Obtain a drive current mapping table, wherein the drive current mapping table stores the minimum threshold current corresponding to different types of sample laser chips;

[0035] In the drive current mapping table, find the minimum threshold current corresponding to the target model, wherein the target model is the model corresponding to the laser chip;

[0036] The driving current is calculated based on the found minimum threshold current.

[0037] In one possible design, the spatial intensity distribution features of the near-infrared spot image are extracted, including:

[0038] The near-infrared spot image is preprocessed to obtain a preprocessed spot image;

[0039] The centroid coordinates and ellipticity of the preprocessed spot image are calculated.

[0040] The spatial intensity distribution characteristics of the light spot are formed by using the centroid coordinates and the ellipticity.

[0041] In one possible design, the decision control model is a trained reinforcement learning agent, wherein the historical photoelectric probe coupling data includes multiple coupling tuples, and each coupling tuple includes a historical state, a historical action, a historical reward, and a next state.

[0042] Correspondingly, the historical state includes: the spatial intensity distribution characteristics of the light spot at a historical moment, the optical power reading at a historical moment, and the position of the nano-positioning platform at a historical moment, wherein the position of the nano-positioning platform is the position of the photoelectric probe on the nano-positioning platform;

[0043] The historical actions include: the optimal displacement command executed at a historical moment;

[0044] Historical rewards include: the increased coupling efficiency resulting from performing the historical action, where coupling efficiency is the increase in optical power;

[0045] The next state includes: after executing the historical action, the new spatial intensity distribution characteristics of the light spot, the new position of the nano-positioning platform, and the optical power reading at the new position of the nano-positioning platform.

[0046] In one possible design, a visible light camera is used to acquire the top image, a near-infrared camera is used to acquire the near-infrared spot image, and an optical power meter is used to acquire the real-time optical power reading;

[0047] Before acquiring the top image of the laser chip, the method further includes:

[0048] The nano-positioning platform is subjected to five-axis zero-point calibration, and the calibration parameters of the visible light camera, the camera gain of the near-infrared camera, and the sampling frequency of the optical power meter are loaded so that the top image can be acquired by the visible light camera after loading is completed.

[0049] Secondly, a DFB laser chip optoelectronic probe coupling system based on intelligent decision-making is provided, comprising:

[0050] Image acquisition unit, used to acquire the top image of the laser chip;

[0051] The position adjustment unit is used to control the photoelectric probe on the nanopositioning platform to contact the electrode plane of the laser chip based on the top image, and to determine whether the preset contact conditions are met.

[0052] The driving and near-infrared image acquisition unit is used to apply a driving current to the laser chip when it is determined that the preset contact conditions are met, and to acquire the near-infrared spot image formed by the laser chip waveguide end after the driving current is applied.

[0053] The prediction unit is used to extract the spatial intensity distribution features of the near-infrared spot image and input the spatial intensity distribution features of the spot into the power prediction model to obtain the spot power distribution information and the optimal movement direction of the probe.

[0054] The acquisition unit is used to acquire historical photoelectric probe coupling data and real-time optical power readings of the laser emitted by the laser chip;

[0055] The decision control unit is used to input the spot power distribution information, the optimal movement direction of the probe, the real-time optical power reading and historical photoelectric probe coupling data into the decision control model to obtain the optimal displacement command of the photoelectric probe.

[0056] The photoelectric probe coupling unit is used to adjust the position of the photoelectric probe according to the optimal displacement command, and after adjusting the position, re-acquire the near-infrared spot image until the acquired real-time optical power reading reaches the preset condition, thus completing the photoelectric probe coupling of the laser chip.

[0057] Thirdly, a DFB laser chip optoelectronic probe coupling device based on intelligent decision-making is provided. Taking the device as an electronic device as an example, it includes a memory, a processor, and a transceiver that are connected in sequence. The memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the DFB laser chip optoelectronic probe coupling method based on intelligent decision-making as described in the first aspect or any possible design of the first aspect.

[0058] Fourthly, a storage medium is provided, on which instructions are stored, which, when executed on a computer, perform the DFB laser chip photoelectric probe coupling method based on intelligent decision-making as described in the first aspect or any possible design of the first aspect.

[0059] Fifthly, a computer program product containing instructions is provided, which, when executed on a computer, causes the computer to perform the DFB laser chip optoelectronic probe coupling method based on intelligent decision-making as described in the first aspect or any possible design of the first aspect.

[0060] Beneficial effects:

[0061] (1) The present invention constructs a multimodal perception-intelligent decision-precise execution closed-loop system, which realizes real-time adjustment of the photoelectric probe moving path. Therefore, compared with the fixed path coupling method in semi-automatic technology, the present invention can greatly improve the coupling accuracy, thereby meeting the latest industry needs. Therefore, it is very suitable for large-scale application and promotion. Attached Figure Description

[0062] Figure 1 A flowchart illustrating the steps of the DFB laser chip optoelectronic probe coupling method based on intelligent decision-making provided in this embodiment of the invention;

[0063] Figure 2 A structural diagram of the DFB laser chip optoelectronic probe coupling system based on intelligent decision-making provided in an embodiment of the present invention;

[0064] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0065] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in conjunction with the accompanying drawings and descriptions of the embodiments or the prior art. Obviously, the following description of the structure of the accompanying drawings is only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0066] It should be understood that although the terms first, second, etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit, without departing from the scope of the exemplary embodiments of the invention.

[0067] It should be understood that the term "and / or" that may appear in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" that may appear in this document describes another relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " that may appear in this document generally indicates that the related objects before and after it are in an "or" relationship.

[0068] Example:

[0069] See Figure 1As shown in the figure, the DFB laser chip photoelectric probe coupling method based on intelligent decision-making provided in this embodiment adopts a dual-path vision approach to drive the photoelectric probe. Specifically, it uses a top image to control the contact between the photoelectric probe and the laser chip. After contact, it acquires an image of the infrared light spot emitted after applying a driving current to the laser chip, using this image to predict power distribution. Then, it combines historical photoelectric probe coupling data and real-time optical power, and uses an intelligent decision-making model to generate the optimal displacement command for the photoelectric probe. Next, based on the optimal displacement command, it moves the photoelectric probe, and after movement, it re-acquires the near-infrared light spot image, repeating the aforementioned process to continuously adjust the movement path of the photoelectric probe. Until the optimal coupling state is reached; thus, this method constructs a fully closed-loop system of multimodal perception-intelligent decision-making-precise execution, which realizes real-time adjustment of the photoelectric probe's moving path. Therefore, compared with the fixed-path coupling method in semi-automatic technology, this method can significantly improve coupling accuracy, thereby meeting the latest industry requirements. Therefore, it is very suitable for large-scale application and promotion. For example, this method can be run on the control end side of the laser chip photoelectric probe coupling system. Optionally, the control end can be, but is not limited to, an industrial computer. It is understood that the aforementioned execution subject does not constitute a limitation on the embodiments of this application. Accordingly, the operation steps of this method can be, but are not limited to, as shown in steps S1 to S7 below.

[0070] S1. Acquire the top image of the laser chip; In this embodiment, a dual-path vision system is provided to drive the photoelectric probe on the nano-positioning platform; Specifically, the dual-path vision system includes a visible light camera for positioning the electrodes and a near-infrared camera for capturing the light spot, that is, the visible light camera is used to acquire the aforementioned top image; At the same time, before acquiring the top image, the nano-positioning platform needs to be started, the platform calibrated, and the system parameters of the aforementioned dual-path vision system need to be loaded; In addition, the aforementioned laser chip photoelectric probe coupling system is also equipped with an optical power meter to measure the real-time optical power of the laser emitted by the laser chip, thereby providing a data basis for the subsequent movement control of the photoelectric probe.

[0071] Furthermore, the nano-positioning platform is subjected to five-axis zero-point calibration to ensure repeatability accuracy verification <±5nm; then, the dual-path vision system is activated to load the calibration parameters of the visible light camera, the camera gain of the near-infrared camera, and the sampling frequency of the optical power meter, so that after loading is completed, the top image is acquired through the visible light camera.

[0072] Yes, for example, you can set the near-infrared camera gain to 12dB, the optical power meter sampling frequency to 10kHz, and the measurement range to -70 to +10dBm.

[0073] Simultaneously, after completing the zero-point calibration of the nano-positioning platform and loading the parameters of the dual-path vision system, a training model can also be loaded. The training model includes a power prediction model and a decision control model. First, a Physical Information Neural Network (PINN) model is loaded as the power prediction model. Its input layer receives a 128-dimensional light field feature vector (i.e., the subsequent light spot spatial intensity distribution feature), and the output layer generates a 512×512 resolution power distribution heatmap. The network weights are obtained from 100,000 sets of simulation data through transfer learning. The decision control model belongs to the instantiated reinforcement learning agent (TD3Agent class) and loads policy network weights optimized for 25G DFB chips. This network includes an Actor network (3 fully connected layers, 256 nodes / layer).

[0074] Thus, after the model loading is completed, the photoelectric probe can be controlled to contact the electrode plane of the laser chip based on the top image of the laser chip acquired by the visible light camera (i.e., the top view image), so as to provide a positional basis for subsequent photoelectric coupling; wherein, the photoelectric probe contact control is as shown in step S2 below.

[0075] S2. Based on the top image, control the photoelectric probe on the nano-positioning platform to contact the electrode plane of the laser chip, and determine whether the preset contact conditions are met. In specific applications, the photoelectric probe is first controlled to reach a parallel state with the electrode plane to prepare for subsequent vertical contact. Then, the photoelectric probe is controlled to descend vertically until the contact with the electrode plane reaches the preset contact conditions. Optionally, the above process can be, but is not limited to, as shown in steps S21 to S25 below.

[0076] S21. Based on the top image, drive the nano-positioning platform to operate so that the photoelectric probe on the nano-positioning platform is parallel to the electrode plane of the laser chip; in this embodiment, image recognition technology is used to identify the center coordinates of the electrodes of the laser chip; then, based on the center coordinates of the electrodes, drive the nano-positioning platform to adjust its position; wherein, the aforementioned position adjustment process of the nano-positioning platform is as shown in the following steps S21a to S21c.

[0077] S21a. Perform image recognition on the top image to determine the center coordinates of the electrodes of the laser chip; in specific implementations, for example, but not limited to, the following steps S21a1 to S21a3 can be used to identify the center coordinates of the electrodes.

[0078] S21a1. Perform image preprocessing on the top image to obtain a preprocessed image; in specific implementations, image preprocessing may include, but is not limited to, grayscale conversion, noise reduction, and contrast enhancement.

[0079] Thus, after image preprocessing is completed, image recognition and segmentation can be performed, as shown in step S21a2 below.

[0080] S21a2. A pre-trained deep learning segmentation model is used to perform image recognition and segmentation on the preprocessed image to segment the electrode pixel region image in the top image. In this embodiment, the pre-trained deep learning segmentation model can be, but is not limited to, a trained U-Net model or a Mask R-CNN model. During training, the input of the aforementioned models is the sample top image of each sample laser chip, and the output of the models is the electrode pixel region image in the sample top image corresponding to each sample laser chip. Thus, after the model is trained in the aforementioned manner, it can be used for image recognition and segmentation of the top image of the laser chip.

[0081] Based on this, after completing the image recognition and segmentation of the preprocessed image, the coordinates of the electrode center can be determined from the electrode pixel region image, as shown in step S21a3 below.

[0082] S21a3. Calculate the centroid of the electrode pixel region image and use the centroid as the coordinates of the electrode center.

[0083] After determining the electrode center coordinates of the laser chip through the aforementioned steps S21a1 to S21a3, coordinate transformation can be performed. That is, the electrode center coordinates belong to the pixel coordinates, but to control the nano-positioning platform, it is necessary to transform them to the mechanical coordinate system of the nano-positioning platform. The coordinate transformation process is shown in step S21b below.

[0084] S21b. Transform the electrode center coordinates to the mechanical coordinate system of the nano-positioning platform to obtain the actual coordinates of the electrode center. In this embodiment, for example, but not limited to, first obtain the pixel equivalents in the X and Y axes of the pixel coordinate system corresponding to the top image, and the translational offsets in the X and Y axes of the mechanical coordinate system; then, based on the pixel equivalents and the translational offsets, transform the electrode center coordinates to the mechanical coordinate system to obtain the actual coordinates of the electrode center.

[0085] Optionally, for example, but not limited to, using the following formula, coordinate transformation can be performed.

[0086] ;

[0087] In the formula, The pixel coordinates representing the center of the electrode in the pixel coordinate system (i.e., the image coordinate system) are... These represent the pixel's horizontal and vertical coordinates in the electrode center coordinate system, respectively. This represents the actual coordinates of the electrode center in the mechanical coordinate system (in micrometers), i.e. These are the x and y coordinates of the actual coordinate system of the motor center, respectively. These represent the pixel equivalent along the X-axis and the pixel equivalent along the Y-axis in the pixel coordinate system, respectively (obtained through camera calibration). These represent the translational offset along the X-axis and the translational offset along the Y-axis in the mechanical coordinate system, respectively (determined by the initial system calibration); simultaneously, Let be the image center pixel coordinates of the electrode pixel region image, where These represent the horizontal coordinate and vertical coordinate of the image center, respectively.

[0088] Thus, after completing the coordinate transformation using the aforementioned formula, the position of the nano-positioning platform can be adjusted, as shown in step S21c below.

[0089] S21c. Drive the nano-positioning platform to operate according to the actual coordinates of the electrode center.

[0090] Based on the aforementioned steps S21a to S21c, the driving platform is positioned to make the photoelectric probe parallel to the electrode plane, and then vertical contact can be made. The process is shown in steps S22 to S25 below.

[0091] S22. Control the photoelectric probe to descend vertically at a first speed, and monitor the contact force on the photoelectric probe in real time to determine whether the contact force is greater than the contact force threshold. In this embodiment, the first speed can be set to, but is not limited to, 10 μm / s, and the contact force threshold can be set to 0.3 mN. When the contact force Fz on the photoelectric probe is greater than 0.3 mN, the operating mode can be switched, and the process is shown in step S23 below.

[0092] S23. If so, the operating mode of the photoelectric probe is switched to fine mode, wherein the fine mode is when the photoelectric probe descends vertically at a second speed, and the second speed is less than the first speed; in specific implementation, for example, the second speed is 1μm / s, that is, when the contact force is greater than 0.3mN, the fine mode is switched (the speed is reduced to 1μm / s), at which time the contact force can continue to be monitored until it meets the preset contact conditions, then the driving current can be applied; wherein, the contact force judgment process in fine mode is as shown in step S24 below.

[0093] S24. Real-time acquisition of the contact force on the photoelectric probe in fine mode, and determination of whether the contact force of the photoelectric probe in fine mode meets the preset contact condition; in specific applications, for example, the preset contact condition is: the contact force of the photoelectric probe in fine mode is equal to 5.0±0.2mN and lasts for 200ms; in this way, when the aforementioned preset contact condition is met, the driving current can be applied, and the process is as shown in step S25 below.

[0094] S25. If so, apply the driving current to the laser chip.

[0095] After the photoelectric probe and the electrode plane of the laser chip are made into perpendicular contact through the aforementioned steps S21 to S25, the movement control of the photoelectric probe on the motor plane can be performed. In this embodiment, a near-infrared camera is used to acquire images of the near-infrared light spot emitted by the laser chip. Then, the power distribution is predicted using the aforementioned loaded power prediction model. Subsequently, by combining historical photoelectric probe coupling data and the real-time optical power of the laser emitted by the laser chip, and using a decision control model, displacement commands for the photoelectric probe are generated, thereby enabling the movement control of the photoelectric probe.

[0096] The aforementioned motion control process is described in steps S3 to S7 below.

[0097] S3. If so, apply a driving current to the laser chip, and after applying the driving current, acquire an image of the near-infrared spot formed by the emission from the waveguide end of the laser chip; in specific implementation, the driving current is calculated as follows: first, obtain a driving current mapping table (wherein, the driving current mapping table stores the minimum threshold current corresponding to different models of sample laser chips); then, find the minimum threshold current corresponding to the target model in the driving current mapping table, where the target model is the model corresponding to the laser chip; finally, the driving current can be calculated based on the found minimum threshold current.

[0098] In this embodiment, the system applies an initial driving current to the laser chip according to the pre-stored chip current-optical power (IL) characteristic lookup table (i.e. the aforementioned driving current mapping table). This process aims to ensure that the light spot can be clearly captured by the near-infrared camera, while strictly avoiding chip damage or spectral distortion due to excessive current.

[0099] Regarding the selection of the initial drive current, the system first retrieves the minimum threshold current corresponding to the DFB laser chip model from the pre-stored data table, and sets the initial drive current to 1.1 times the minimum threshold current. In this way, the current value is slightly higher than the threshold, which is sufficient to excite the chip to generate a stable and easily detectable weak laser, while being far lower than the rated operating current that may cause chip aging or damage.

[0100] After applying a driving current to the laser chip, a near-infrared camera can be used to acquire an image of the near-infrared spot formed by the emission from the end of the chip's waveguide. Simultaneously, after acquiring the near-infrared spot image, the real-time optical power reading of the laser emitted by the current laser chip is also acquired. It is then determined whether the real-time optical power reading meets the preset conditions. If the preset conditions are not met, the spot power distribution needs to be predicted based on the near-infrared spot image, as shown in step S4 below.

[0101] S4. Extract the spatial intensity distribution features of the near-infrared spot image and input the spatial intensity distribution features of the spot into the power prediction model to obtain the spot power distribution information and the optimal movement direction of the probe; in specific implementation, for example, but not limited to, the following steps S41 to S43 can be used to extract the aforementioned spatial intensity distribution features of the spot.

[0102] S41. The near-infrared spot image is preprocessed to obtain a preprocessed spot image. In this embodiment, the preprocessing of the near-infrared spot image may include, but is not limited to, grayscale conversion, noise reduction, and contrast enhancement. Thus, after the image preprocessing is completed, feature extraction can be performed, as shown in step S42 below.

[0103] S42. Calculate the centroid coordinates and ellipticity of the preprocessed spot image; in specific implementations, for example, but not limited to, the centroid coordinates can be obtained by calculating the intensity weighted center of the gray values ​​of the spot image.

[0104] Optionally, the formula for calculating the centroid coordinates is:

[0105] ;

[0106] In the formula, Indicates the centroid coordinates. For pixels The intensity value, i.e., the pixel value. The grayscale value.

[0107] After calculating the centroid coordinates based on the aforementioned formula, the ellipticity can be calculated. For example, but not limited to, it can be obtained by calculating the second moment of the spot image and then fitting the ratio of its major and minor axes of an approximate ellipse. This parameter characterizes the shape asymmetry of the spot.

[0108] After feature extraction is completed based on the aforementioned step S42, the two features can be used to form the spatial intensity distribution feature of the light spot, as shown in step S43 below.

[0109] S43. Using the centroid coordinates and the ellipticity, the spatial intensity distribution characteristics of the light spot are formed.

[0110] Thus, based on the aforementioned steps S41 to S43, after extracting the spatial intensity distribution features of the light spot, the power distribution of the light spot can be predicted using a pre-trained power prediction model. That is, these light spot feature parameters are used as inputs to the PINN model, and the PINN model predicts the possible power distribution and optimal movement direction under the current light spot state.

[0111] After predicting the possible power distribution and optimal movement direction under the current spot state, displacement commands can be generated by combining historical photoelectric probe coupling data and the optical power of the laser chip. The process is shown in steps S5 and S6 below.

[0112] S5. Obtain historical photoelectric probe coupling data and real-time optical power readings of the laser emitted by the laser chip; in specific implementation, the real-time optical power readings are obtained using the aforementioned optical power meter.

[0113] Meanwhile, as explained above, the decision control model is a trained reinforcement learning agent. Therefore, the historical photoelectric probe coupling data mentioned in the example is the coupling experience explored and stored by the agent before this coupling process (including the online learning stage and the pre-training stage). It includes multiple coupling tuples, and each coupling tuple includes [historical state, historical action, historical reward, next state].

[0114] Furthermore, historical states can include, but are not limited to, the spatial intensity distribution characteristics of the light spot at a historical moment, the optical power reading at a historical moment, and the position of the nano-positioning platform at a historical moment. The position of the nano-positioning platform is the position of the photoelectric probe on the nano-positioning platform, and the historical moment is the coupling moment corresponding to the coupling element. For example, if the coupling data obtained is within 2 seconds before the current moment (3 seconds), assuming a sampling interval of 1 second, then the historical moments corresponding to the two coupling elements are the 1st second and the 2nd second. Of course, the above examples are just examples, and this embodiment is not limited to them.

[0115] Meanwhile, the historical actions mentioned above may include, but are not limited to: the optimal displacement command executed at a historical moment; the historical rewards may include, but are not limited to: the improved coupling efficiency after executing the historical actions (wherein, the coupling efficiency is the increase in optical power); and the next state may include, but is not limited to: the new spatial intensity distribution characteristics of the light spot obtained after executing the historical actions, the new position of the nano-positioning platform, and the optical power reading at the new position of the nano-positioning platform.

[0116] Thus, after obtaining historical photoelectric probe coupling data and the real-time optical power reading at the current moment, the output of the aforementioned power prediction model can be combined to make decisions and controls, as shown in step S6 below.

[0117] S6. Input the spot power distribution information, the optimal movement direction of the probe, the real-time optical power reading, and the historical photoelectric probe coupling data into the decision control model to obtain the optimal displacement command of the photoelectric probe. In this embodiment, the decision control model (TD3Agent) integrates the output of the PINN model, the real-time optical power meter reading, and the historical coupling data to calculate the final optimal displacement command (i.e., the optimal movement distance and movement direction) to drive the nano-positioning platform to adjust the probe position. The process is shown in step S7 below.

[0118] S7. According to the optimal displacement command, adjust the position of the photoelectric probe, and after adjusting the position, re-acquire the near-infrared spot image until the acquired real-time optical power reading reaches the preset condition, thus completing the photoelectric probe coupling of the laser chip; In this embodiment, after the driving nano-positioning platform adjusts the probe position, the driving current can be reapplied to the laser chip (the current magnitude remains unchanged), and then the near-infrared spot image is re-acquired until the acquired real-time optical power reading remains stable, which indicates that the optimal coupling state has been reached; Specifically, the criterion for judging that the optical power reading remains stable is: the difference between the real-time optical power readings obtained by several consecutive couplings is less than the preset value, such as the difference between the real-time optical power readings of three or five consecutive couplings being less than the preset value.

[0119] Therefore, through the DFB laser chip photoelectric probe coupling method based on intelligent decision-making described in detail in steps S1 to S7 above, this invention constructs a fully closed-loop system of multimodal sensing-intelligent decision-making-precise execution. It realizes real-time adjustment of the photoelectric probe's moving path. Therefore, compared with the fixed-path coupling method in semi-automatic technology, this invention can achieve sub-micron level coupling accuracy, thus meeting the latest industry requirements. At the same time, the closed-loop automatic control of the entire process can reduce the coupling time of a single chip and improve coupling efficiency. Therefore, this invention is very suitable for large-scale application and promotion.

[0120] like Figure 2As shown, the second aspect of this embodiment provides a hardware system for implementing the DFB laser chip photoelectric probe coupling method based on intelligent decision-making described in the first aspect of the embodiment, comprising:

[0121] The image acquisition unit is used to acquire an image of the top of the laser chip.

[0122] The position adjustment unit is used to control the photoelectric probe on the nanopositioning platform to contact the electrode plane of the laser chip based on the top image, and to determine whether the preset contact conditions are met.

[0123] The driving and near-infrared image acquisition unit is used to apply a driving current to the laser chip when it is determined that the preset contact conditions are met, and to acquire the near-infrared spot image formed by the emission from the waveguide end of the laser chip after the driving current is applied.

[0124] The prediction unit is used to extract the spatial intensity distribution features of the near-infrared spot image and input the spatial intensity distribution features of the spot into the power prediction model to obtain the spot power distribution information and the optimal movement direction of the probe.

[0125] The acquisition unit is used to acquire historical photoelectric probe coupling data and real-time optical power readings of the laser emitted by the laser chip.

[0126] The decision control unit is used to input the spot power distribution information, the optimal movement direction of the probe, the real-time optical power reading and historical photoelectric probe coupling data into the decision control model to obtain the optimal displacement command of the photoelectric probe.

[0127] The photoelectric probe coupling unit is used to adjust the position of the photoelectric probe according to the optimal displacement command, and after adjusting the position, re-acquire the near-infrared spot image until the acquired real-time optical power reading reaches the preset condition, thus completing the photoelectric probe coupling of the laser chip.

[0128] The working process, working details and technical effects of the system provided in this embodiment can be found in the first aspect of the embodiment, and will not be repeated here.

[0129] like Figure 3 As shown, the third aspect of this embodiment provides a DFB laser chip photoelectric probe coupling device based on intelligent decision-making. Taking the device as an electronic device as an example, it includes: a memory, a processor, and a transceiver connected in sequence. The memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the DFB laser chip photoelectric probe coupling method based on intelligent decision-making as described in the first aspect of the embodiment.

[0130] For specific examples, the memory may include, but is not limited to, random access memory (RAM), read-only memory (ROM), flash memory, first-in-first-out (FIFO) memory, and / or first-in-last-out (FILO) memory, etc.; specifically, the processor may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor may be implemented using at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), PLA (Programmable Logic Array). The processor may also include a main processor and a coprocessor. The main processor, also known as the CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state.

[0131] In some embodiments, the processor may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. For example, the processor may not be limited to microprocessors of the STM32F105 series, reduced instruction set computer (RISC) microprocessors, x86 architecture processors, or processors with integrated neural network processing units (NPUs). The transceiver may be, but is not limited to, a Wi-Fi transceiver, a Bluetooth transceiver, a General Packet Radio Service (GPRS) transceiver, a ZigBee (a low-power LAN protocol based on the IEEE 802.15.4 standard) transceiver, a 3G transceiver, a 4G transceiver, and / or a 5G transceiver. Furthermore, the device may also include, but is not limited to, a power module, a display screen, and other necessary components.

[0132] The working process, working details and technical effects of the electronic device provided in this embodiment can be found in the first aspect of the embodiment, and will not be repeated here.

[0133] The fourth aspect of this embodiment provides a storage medium that stores instructions containing the DFB laser chip photoelectric probe coupling method based on intelligent decision-making as described in the first aspect of the embodiment. That is, the storage medium stores instructions that, when executed on a computer, perform the DFB laser chip photoelectric probe coupling method based on intelligent decision-making as described in the first aspect of the embodiment.

[0134] The storage medium refers to a carrier for storing data, which may include, but is not limited to, floppy disks, optical disks, hard disks, flash memory, USB flash drives, and / or memory sticks. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0135] The working process, working details, and technical effects of the storage medium provided in this embodiment can be found in the first aspect of the embodiment, and will not be repeated here.

[0136] The fifth aspect of this embodiment provides a computer program product containing instructions that, when executed on a computer, cause the computer to perform the DFB laser chip photoelectric probe coupling method based on intelligent decision as described in the first aspect of the embodiment, wherein the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device.

[0137] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A smart decision based DFB laser chip opto-probe coupling method, characterized in that, The method comprises the following steps: acquiring a top image of a laser chip; based on the top image, controlling a photoelectric probe on a nano-positioning platform to contact an electrode plane of the laser chip and determining whether a preset contact condition is met; if yes, applying a driving current to the laser chip, and after applying the driving current, collecting a near-infrared light spot image formed by light emitted from the end of a waveguide of the laser chip; extracting a light spot spatial intensity distribution feature of the near-infrared light spot image, and inputting the light spot spatial intensity distribution feature into a power prediction model to obtain light spot power distribution information and an optimal moving direction of the probe, wherein the power prediction model adopts a physical information neural network model; acquiring historical photoelectric probe coupling data and a real-time optical power reading of laser emitted by the laser chip; inputting the light spot power distribution information, the optimal moving direction of the probe, the real-time optical power reading, and the historical photoelectric probe coupling data into a decision control model to obtain an optimal displacement instruction of the photoelectric probe, wherein the decision control model adopts an instantiated reinforcement learning intelligent agent; adjusting the position of the photoelectric probe according to the optimal displacement instruction, and after adjusting the position, re-collecting the near-infrared light spot image until the real-time optical power reading reaches a preset condition, thereby completing the photoelectric probe coupling of the laser chip.

2. The method of claim 1, wherein, based on the top image, controlling a photoelectric probe on a nano-positioning platform to contact an electrode plane of the laser chip and determining whether a preset contact condition is met, comprising: based on the top image, driving the nano-positioning platform to operate so that the photoelectric probe on the nano-positioning platform is parallel to the electrode plane of the laser chip; controlling the photoelectric probe to vertically descend at a first speed, and monitoring the contact force on the photoelectric probe in real time to determine whether the contact force is greater than a contact force threshold; if yes, switching the operation mode of the photoelectric probe to a fine mode, wherein the fine mode is that the photoelectric probe vertically descends at a second speed, and the second speed is less than the first speed; collecting the contact force on the photoelectric probe in the fine mode in real time, and determining whether the contact force of the photoelectric probe in the fine mode reaches the preset contact condition; if yes, applying the driving current to the laser chip.

3. The method of claim 2, wherein, based on the top image, driving the nano-positioning platform to operate, comprising: performing image recognition on the top image to determine the electrode center coordinates of the laser chip; converting the electrode center coordinates to actual electrode center coordinates in a mechanical coordinate system of the nano-positioning platform; and driving the nano-positioning platform to operate according to the actual electrode center coordinates.

4. The method of claim 3, wherein, performing image recognition on the top image to determine the electrode center coordinates, comprising: performing image preprocessing on the top image to obtain a preprocessed image; performing image recognition and segmentation on the preprocessed image by using a pre-trained deep learning segmentation model to segment out an electrode pixel area image in the top image; calculating the centroid of the electrode pixel area image, and taking the centroid as the electrode center coordinates.

5. The method of claim 3, wherein, Converting the electrode center coordinates into a mechanical coordinate system of the nano-positioning platform to obtain actual electrode center coordinates, including: Obtaining pixel equivalents in X-axis and Y-axis directions of a pixel coordinate system corresponding to the top image, and translation offsets in X-axis and Y-axis directions of a mechanical coordinate system; Converting the electrode center coordinates into the mechanical coordinate system according to the pixel equivalents and the translation offsets to obtain the actual electrode center coordinates.

6. The method of claim 1, wherein, The driving current is calculated in the following manner: Obtaining a driving current mapping table, wherein the driving current mapping table stores minimum threshold currents corresponding to different types of sample laser chip; In the driving current mapping table, finding out the minimum threshold current corresponding to the target type, wherein the target type is the type corresponding to the laser chip; According to the found minimum threshold current, the driving current is calculated.

7. The method of claim 1, wherein, The spot spatial intensity distribution features of the near-infrared spot image are extracted, including: Preprocessing the near-infrared spot image to obtain a preprocessed spot image; Calculating the centroid coordinates and ellipticity of the preprocessed spot image; Using the centroid coordinates and the ellipticity to form the spot spatial intensity distribution features.

8. The method of claim 1, wherein, The decision control model is a trained reinforcement learning agent, wherein the historical opto-probe coupling data includes a plurality of coupling tuples, and each coupling tuple includes a historical state, a historical action, a historical reward, and a next state; Correspondingly, the historical state includes the spot spatial intensity distribution features at the historical time, the optical power reading at the historical time, and the position of the nano-positioning platform at the historical time, wherein the position of the nano-positioning platform is the position of the opto-probe on the nano-positioning platform; The historical action includes the optimal displacement instruction executed at the historical time; The historical reward includes the coupling efficiency improved after the historical action is executed, wherein the coupling efficiency is the optical power increase; The next state includes the new spot spatial intensity distribution features obtained after the historical action is executed, the new position of the nano-positioning platform, and the optical power reading at the new position of the nano-positioning platform.

9. The method of claim 1, wherein, The top image is collected by a visible light camera, the near-infrared spot image is collected by a near-infrared camera, and the real-time optical power reading is collected by an optical power meter; Before the top image of the laser chip is obtained, the method further includes: Five-axis zero-point calibration is performed on the nano-positioning platform, and the calibration parameters of the visible light camera, the camera gain of the near-infrared camera, and the sampling frequency of the optical power meter are loaded, so that after the loading is completed, the top image is collected by the visible light camera.

10. A smart decision based DFB laser chip opto-probe coupling system, characterized in that, Including: An image acquisition unit configured to obtain a top image of a laser chip; A position adjustment unit configured to control an opto-probe on a nano-positioning platform to contact an electrode plane of the laser chip based on the top image, and determine whether a preset contact condition is met. The driving and near-infrared image acquisition unit is configured to apply a driving current to the laser chip when it is determined that the preset contact condition is met, and to acquire a near-infrared spot image formed by the laser chip waveguide end after the driving current is applied. The prediction unit is configured to extract a spot spatial intensity distribution feature of the near-infrared spot image, and input the spot spatial intensity distribution feature into a power prediction model to obtain spot power distribution information and an optimal probe movement direction, wherein the power prediction model adopts a physical information neural network model. The acquisition unit is configured to acquire historical photoelectric probe coupling data and real-time optical power readings of the laser emitted by the laser chip. The decision control unit is configured to input the spot power distribution information, the optimal probe movement direction, the real-time optical power readings and the historical photoelectric probe coupling data into a decision control model to obtain an optimal displacement instruction of the photoelectric probe, wherein the decision control model adopts an instantiated reinforcement learning intelligent agent. The photoelectric probe coupling unit is configured to adjust the position of the photoelectric probe according to the optimal displacement instruction, and to re-acquire the near-infrared spot image after the position is adjusted, until the real-time optical power readings reach a preset condition, thereby completing the photoelectric probe coupling of the laser chip.

Images