A new application of light assembly in unmanned vehicle

Abstract

This invention relates to the field of autonomous driving and discloses a novel optical component for autonomous driving applications, comprising: an optical functional module for transmitting, modulating, and receiving optical signals in autonomous driving; a parameter acquisition module for acquiring parameters of the optical functional module; a data preprocessing module for time synchronization, noise reduction, and standardization; a drift prediction module for predicting parameter drift trends; a deviation tracing module for differentiating deviation sources and obtaining tracing results; a difference decision module for customizing differentiated compensation strategies and generating compensation control commands; and a compensation execution module for adjusting the optical and electrical parameters of the optical functional module. Through the organic integration of parameter drift prediction using time-series modeling and multi-source data fusion, deviation source differentiation using causal inference and feature matching, differentiated compensation strategies based on deviation causes, and closed-loop control, active and high-precision dynamic calibration of the parameters of the autonomous driving optical component is achieved.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, specifically a novel optical component for autonomous driving applications. Background Technology

[0002] With the rapid development of artificial intelligence and autonomous driving technologies, driverless vehicles have become an important development direction in the transportation sector. They achieve accurate identification of the surrounding environment, distance calculation, and obstacle judgment through various sensing devices. As a component of the autonomous driving environment perception system, optical components are responsible for the functions of light signal transmission, modulation, and reception. Their operational stability and parameter accuracy directly determine the reliability of environmental perception data, thereby affecting the decision-making safety and driving stability of autonomous vehicles.

[0003] Currently, optical components used in autonomous driving applications generally employ a passive compensation scheme based on parameter acquisition threshold comparison. This involves collecting optical, operational, and electrical parameters of the optical component during operation via sensors and comparing them with preset fixed thresholds. When the threshold range is exceeded, a preset compensation command is triggered. However, these compensation strategies are mostly based on preset fixed logic and do not differentiate the causes of parameter deviations. The entire process relies on a passive response mode triggered by exceeding the limit, making it impossible to predict parameter drift trends in advance. This causes the compensation action to lag behind parameter changes, easily resulting in distortion of perceived data in a short period of time. The compensation strategy lacks specificity and is difficult to adapt to the complex operating conditions required in autonomous driving scenarios, affecting the long-term operational stability and environmental perception accuracy of the optical component. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a novel optical component for autonomous driving, solving the problems that existing optical components for autonomous driving applications are prone to causing distortion of perception data in a short period of time, making it difficult to adapt to the usage requirements of complex working conditions in autonomous driving scenarios, and affecting the long-term operational stability and environmental perception accuracy of the optical component.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a novel optical component for autonomous driving, comprising: Optical functional modules are used for the transmission, modulation, and reception of optical signals in autonomous driving. The parameter acquisition module is used to acquire the optical parameters, operating parameters, and electrical parameters of the optical functional module to obtain the raw acquisition data. The data preprocessing module is used to perform time synchronization, noise reduction, and standardization on the raw collected data to obtain preprocessed data. The drift prediction module is used to predict the parameter drift trend based on the preprocessed data by time series modeling and multi-source data fusion, and obtain the drift prediction result. The deviation tracing module is used to distinguish the sources of deviation based on the drift prediction results through causal inference and feature matching, and obtain the tracing results. The differential decision-making module is used to customize differential compensation strategies and generate compensation control instructions based on the source tracing results. The compensation execution module is used to adjust the optical and electrical parameters of the optical function module based on the compensation control command to obtain the calibrated parameter state.

[0006] By adopting the above technical solutions, through the organic integration of parameter drift prediction by time-series modeling and multi-source data fusion, deviation source differentiation by causal inference and feature matching, and differential compensation strategies based on deviation causes and closed-loop control, the system can capture parameter drift dynamics in advance, locate the root causes of deviations such as environmental interference and component aging, and specifically adapt to parameter calibration requirements under different working conditions. This enables active and high-precision dynamic calibration of optical component parameters for autonomous driving, solving the problem that existing optical components for autonomous driving applications are prone to short-term perception data distortion, making it difficult to adapt to the usage requirements of complex working conditions in autonomous driving scenarios, and affecting the long-term operational stability and environmental perception accuracy of optical components.

[0007] Preferably, the optical parameters include laser wavelength, output power, beam offset angle, and carrier rejection ratio; the operating parameters include operating temperature and vibration acceleration; and the electrical parameters include MZM bias voltage and laser drive current. Obtaining the preprocessed data includes the following steps: Time synchronization of different types of data in the original collected data is performed based on time stamps to ensure data time sequence consistency and obtain synchronized data. The Kalman-Particle fusion filtering algorithm is used to denoise the synchronization data, removing high-frequency noise and low-frequency drift to obtain denoised data; The noise reduction data is standardized, and after unifying the data units, it is classified and organized according to the parameter type to form preprocessed data.

[0008] Preferably, obtaining the drift prediction result includes the following steps: Extract the time-series features of the target parameters from the preprocessed data and construct a three-dimensional feature vector, which includes the current value, historical period average, and rate of change. The Holt-Winters exponential smoothing method is used to model the three-dimensional feature vector, and the horizontal component, trend component and seasonal component are calculated to obtain the initial predicted value. By fusing preprocessed data through Bayesian filtering, the initial prediction values ​​are corrected to obtain the drift prediction results; The drift prediction result is compared with the dynamic threshold. If the drift prediction result exceeds the dynamic threshold, the drift prediction result is sent to the deviation tracing module.

[0009] Preferably, setting the dynamic threshold includes the following steps: An initial threshold coefficient is set based on an ideal reference baseline of the optical functional module. Each preset duration is based on the cumulative usage time of the optical components and historical calibration data. The initial threshold coefficient is finely adjusted to form a threshold range that dynamically adapts to the aging of the components, thus obtaining a dynamic threshold.

[0010] Preferably, obtaining the tracing result includes the following steps: Construct a directed acyclic causal graph with the correlation between operating conditions, parameters, and deviations; By fixing the confusion variables in the directed acyclic causal graph using the do-calculus operator, the correlation between the deviation and the operating parameters and the cumulative usage time of the optical components was observed. Based on the difference between the drift prediction result and the ideal reference baseline, the real-time deviation characteristics are calculated. Calculate the Mahalanobis distance between the real-time deviation characteristics and the preset environmental interference characteristic library and component aging characteristic library; By combining the causal inference results obtained through observations using the do-calculus operator with the Mahalanobis distance matching degree, the source of the deviation is determined to be environmental interference or component aging, and the source tracing results are output.

[0011] Preferably, the generation of compensation control instructions includes the following steps: Based on the source tracing results, distinguish between environmental interference and component aging types; When there is environmental interference, the adjustment amounts of beam angle, optical power, bias voltage and drive current are calculated based on the correlation between temperature, vibration and parameter deviation. When components are aging, a step-by-step adjustment strategy is adopted to calculate the safe adjustment amount of each parameter; The upper limit of the adjustment constraint is adjusted by combining the preset parameters, and the adjustment amount is integrated to form a standardized compensation control command, which is then output to the compensation execution module.

[0012] Preferably, the step-by-step adjustment strategy for calculating the safe adjustment amount of each parameter includes the following steps: Based on the cumulative usage time of optical components and historical degradation data of various parameters, aging stages are divided. For each aging stage, calculate the safe adjustment range of the corresponding parameter to ensure that the single adjustment does not exceed the preset proportion of the parameter's rated value. The adjustment amount is allocated in stages based on the current aging stage and the magnitude of the real-time deviation. Summarize the adjustments allocated at each stage to form the safety adjustment amount for each parameter.

[0013] Preferably, obtaining the calibrated parameter status includes the following steps: Based on the compensation control command, analyze the adjustment amount of each parameter and adjust the optical and electrical parameters of the autonomous driving system. The system collects and adjusts real-time parameters, generates a calibrated parameter status, and feeds it back to the drift prediction module.

[0014] This invention provides a novel optical component for use in autonomous driving. It offers the following advantages: 1. This invention achieves proactive and high-precision dynamic calibration of unmanned driving optical component parameters by combining time-series modeling and multi-source data fusion for parameter drift prediction, causal inference and feature matching for deviation source differentiation, differentiated compensation strategies based on deviation causes, and closed-loop control. This allows for the early detection of parameter drift dynamics, positioning environmental interference, and component aging as the root causes of deviations, and targeted adaptation to parameter calibration requirements under different working conditions.

[0015] 2. This invention constructs a causal graph of working condition-parameter-deviation correlation by relying on causal inference and feature matching. After fixing the confusion variables, the correlation is observed. Then, the similarity between the real-time deviation features and the preset feature library is calculated by combining Mahalanobis distance. This enables the tracing of the source of environmental interference and component aging deviation, thereby distinguishing the source of deviation and strengthening the pertinence of the compensation strategy.

[0016] 3. When facing environmental interference, this invention calculates the adjustment amount based on the correlation between operating conditions and parameters; when dealing with component aging, it adopts a step-by-step adjustment strategy to control the single adjustment ratio, and combines parameter rating constraints to ensure adjustment safety, thereby flexibly adapting to complex operating conditions such as extreme temperature and vibration in autonomous driving scenarios and improving the parameter calibration accuracy of optical components. Attached Figure Description

[0017] Figure 1 This is an architecture diagram of a novel optical component proposed in this invention for use in autonomous driving. Figure 2 This is a flowchart illustrating a novel application method of optical components in autonomous driving, as proposed in an embodiment of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: In a first embodiment of the present invention, the present invention provides a novel optical component for use in autonomous driving, such as... Figure 1 As shown, it includes: Optical functional modules are used for the transmission, modulation, and reception of optical signals in autonomous driving. Specifically, the laser emitting unit in the optical functional module employs a specific wavelength distributed feedback laser, which outputs a stable optical signal through a driving circuit. The optical transmission path is adjusted via a silicon-based micromirror array to adapt to different sensing angle requirements. The modulation unit uses a Mach-Zehnder modulator to modulate the amplitude or phase of the optical signal, and the receiving unit converts the reflected optical signal into an electrical signal using an indium gallium arsenide photodiode. Auxiliary optical components include collimating lenses, fiber optic gratings, and optical couplers, which are fixed by pre-set slots on a low-temperature co-fired ceramic substrate to ensure optical path coaxiality. The optical path adopts a sealed design with a hydrophobic coating to reduce environmental interference, ensure the stability of optical signal transmission, and provide a reliable foundation for subsequent parameter acquisition and calibration.

[0020] The parameter acquisition module is used to acquire the optical parameters, operating parameters, and electrical parameters of the optical functional module to obtain the raw acquisition data. Furthermore, the optical parameters include laser wavelength, output power, beam offset angle, and carrier rejection ratio; the operating parameters include operating temperature and vibration acceleration; and the electrical parameters include MZM bias voltage and laser drive current.

[0021] Specifically, various parameters are acquired synchronously through sensors. Generally, optical parameter sensors capture optical signal characteristics in real time, including laser wavelength, output power, beam offset angle, and carrier rejection ratio. Operating condition parameter sensors acquire operating temperature and vibration acceleration. Electrical parameter sensors acquire MZM bias voltage and laser drive current. All sensors acquire data synchronously according to a preset cycle, adding a timestamp to each data point. Acquired data is temporarily stored in a buffer to ensure temporal integrity and provide a data source for subsequent preprocessing. The data preprocessing module is used to perform time synchronization, noise reduction, and standardization on the raw collected data to obtain preprocessed data; Obtaining preprocessed data includes the following steps: Time synchronization of different types of data in the original collected data is performed based on time stamps to ensure data time sequence consistency and obtain synchronized data. The Kalman-Particle fusion filtering algorithm is used to denoise the synchronization data, removing high-frequency noise and low-frequency drift to obtain denoised data; The noise reduction data is standardized, and after unifying the data units, it is classified and organized according to the parameter type to form preprocessed data.

[0022] Specifically, the data preprocessing module receives the raw data from the parameter acquisition module and performs data purification through a three-step collaborative process. Generally, time synchronization is based on the timestamps of each data point, precisely aligning the acquisition times of optical, operating condition, and electrical parameters to ensure consistent data timing and synchronized data.

[0023] Noise reduction employs a Kalman-particle fusion filtering algorithm. The Kalman filter state equation is: The observation equation is: ,in for The state vector at any given time includes parameters such as laser wavelength, output power, beam offset angle, and operating temperature. The state transition matrix represents the temporal correlation of the parameters; This is process noise, reflecting the uncertainty of parameter changes; The noise observed is caused by sensor error; for Time sensor observations; Establish a mapping relationship between states and observations for the observation matrix.

[0024] Particle filtering initializes particles based on the output of a Kalman filter, retains effective particles through sequential importance resampling, and then fuses them to obtain denoised data. Normalization uses the z-score method, calculated as follows: in These are the parameter values ​​after noise reduction. This is the historical average of this type of parameter. The historical standard deviation of this type of parameter is provided. The data is categorized and organized by parameter type to ensure data structure and provide input for subsequent drift prediction.

[0025] The drift prediction module is used to predict parameter drift trends based on preprocessed data through time series modeling and multi-source data fusion, and obtain drift prediction results. Furthermore, obtaining the drift prediction results includes the following steps: Extract the time-series features of the target parameters from the preprocessed data and construct a three-dimensional feature vector, which includes the current value, historical period average, and rate of change. The Holt-Winters exponential smoothing method is used to model the three-dimensional feature vector, and the horizontal component, trend component and seasonal component are calculated to obtain the initial predicted value. By fusing preprocessed data through Bayesian filtering, the initial prediction values ​​are corrected to obtain the drift prediction results; The drift prediction result is compared with the dynamic threshold. If the drift prediction result exceeds the dynamic threshold, the drift prediction result is sent to the deviation tracing module.

[0026] Furthermore, setting the dynamic threshold includes the following steps: An initial threshold coefficient is set based on an ideal reference baseline of the optical functional module. Each preset duration is based on the cumulative usage time of the optical components and historical calibration data. The initial threshold coefficient is finely adjusted to form a threshold range that dynamically adapts to the aging of the components, thus obtaining a dynamic threshold.

[0027] Specifically, the drift prediction module receives preprocessed data from the data preprocessing module and performs proactive predictions through multi-step collaborative computation, providing a triggering basis for subsequent deviation tracing. Generally, the extraction of three-dimensional feature vectors is based on parameters such as laser wavelength and output power in the preprocessed data, including the current value, historical period average, and rate of change, providing basic input for time series modeling.

[0028] The Holt-Winters exponential smoothing method is used to model the three-dimensional eigenvectors, and the level equation is: The trend equation is The seasonal equation is The prediction equation is ,in for The horizontal component of the target parameter at any given time represents the current baseline state of the parameter; The trend component represents the trend of parameter change; The seasonal component characterizes the periodic fluctuations of the parameter; This is a horizontal smoothing coefficient that balances the weights of current and historical values. This is the trend smoothing coefficient, which controls the sensitivity to trend changes; It serves as a seasonal smoothing coefficient, adapting to periodic fluctuations; Seasonal cycles correspond to the fluctuation cycles of vehicle parameters. To predict the step size, the corresponding time duration must be predicted in advance; for The initial predicted value at time 1.

[0029] The initial predicted values ​​are corrected by fusing preprocessed data using Bayesian filtering, and the state equation is: The observation equation is ,in for The time-state vector contains the horizontal and trend components of the Holt-Winters output; The state transition matrix represents the temporal correlation of states; This represents process noise, reflecting the uncertainty of parameter fluctuations. for Observations of preprocessed data at any given time; The observation matrix correlates the state with the observation; The noise is caused by data error; the output is the corrected drift prediction result.

[0030] The dynamic threshold is set based on the ideal reference baseline of the optical functional module. The initial threshold coefficient is a preset value, and it is fine-tuned every preset duration based on the cumulative usage time of the optical component and historical calibration data. The adjustment formula is as follows: in for Threshold coefficient at time step; To adjust the coefficients and adapt to the aging rate of the components; This represents the cumulative usage time of the optical components; the output is a threshold range that dynamically adapts as the components age. The drift prediction result is compared with the dynamic threshold; if the drift exceeds the threshold, the drift prediction result is sent to the deviation tracing module to proactively trigger the tracing process.

[0031] The deviation tracing module is used to distinguish the sources of deviation based on the drift prediction results through causal inference and feature matching, and obtain the tracing results. Furthermore, to obtain the source tracing results, the following steps are included: Construct a directed acyclic causal graph with the correlation between operating conditions, parameters, and deviations; By fixing the confusion variables in the directed acyclic causal graph using the do-calculus operator, the correlation between the deviation and the operating parameters and the cumulative usage time of the optical components was observed. Based on the difference between the drift prediction result and the ideal reference baseline, the real-time deviation characteristics are calculated. Calculate the Mahalanobis distance between the real-time deviation characteristics and the preset environmental interference characteristic library and component aging characteristic library; By combining the causal inference results obtained through observations using the do-calculus operator with the Mahalanobis distance matching degree, the source of the deviation is determined to be environmental interference or component aging, and the source tracing results are output.

[0032] Specifically, the deviation tracing module receives the drift prediction results output by the drift prediction module, and uses causal analysis and feature matching to trace the source, providing a basis for subsequent difference decisions.

[0033] The construction of the directed acyclic causal graph (DAG) is based on the correlation between operating parameters and deviations in the preprocessed data, including nodes such as temperature, vibration, and usage time, clarifying the causal links between variables. The do-calculus operator is used to fix the confounding variables in the DAG, such as fixing the cumulative usage time of the optical components and the correlation between observation deviation and temperature / vibration, or fixing temperature / vibration and the correlation between observation deviation and usage time, thus initially distinguishing between environmental interference and component aging. Real-time deviation features are obtained based on the difference between the drift prediction results and the ideal reference baseline of the optical functional modules, providing input for subsequent feature matching.

[0034] Feature matching uses Mahalanobis distance calculation. The Mahalanobis distance formula for the environmental interference feature database is as follows: The Mahalanobis distance formula for the component aging feature library is: ,in This is a real-time deviation feature vector, which includes parameters such as wavelength drift and power attenuation. This is the mean vector of the environmental interference feature library; The covariance matrix of the environmental interference feature library; This is the mean vector of the component aging feature library; This is the covariance matrix of the component aging feature library; The Mahalanobis distance represents the similarity between real-time features and the feature library.

[0035] By combining the causal inference results obtained from the observations of the do-calculus operator with the Mahalanobis distance matching degree, if the Mahalanobis distance of the environmental interference feature library is smaller and the causal inference shows a strong correlation between the deviation and the operating condition, it is determined to be environmental interference; if the Mahalanobis distance of the component aging feature library is smaller and the causal inference shows a strong correlation between the deviation and the usage time, it is determined to be component aging. The source tracing results are output to the difference decision module to improve the targeting of compensation.

[0036] The differential decision-making module is used to customize differential compensation strategies and generate compensation control instructions based on the source tracing results. Furthermore, generating compensation control instructions includes the following steps: Based on the source tracing results, distinguish between environmental interference and component aging types; When there is environmental interference, the adjustment amounts of beam angle, optical power, bias voltage and drive current are calculated based on the correlation between temperature, vibration and parameter deviation. When components are aging, a step-by-step adjustment strategy is adopted to calculate the safe adjustment amount of each parameter; The upper limit of the adjustment constraint is adjusted by combining the preset parameters, and the adjustment amount is integrated to form a standardized compensation control command, which is then output to the compensation execution module.

[0037] Furthermore, a step-by-step adjustment strategy is adopted to calculate the safe adjustment amount for each parameter, including the following steps: Based on the cumulative usage time of optical components and historical degradation data of various parameters, aging stages are divided. For each aging stage, calculate the safe adjustment range of the corresponding parameter to ensure that the single adjustment does not exceed the preset proportion of the parameter's rated value. The adjustment amount is allocated in stages based on the current aging stage and the magnitude of the real-time deviation. Summarize the adjustments allocated at each stage to form the safety adjustment amount for each parameter.

[0038] Specifically, the difference decision module receives the source tracing results output by the deviation source tracing module, customizes compensation strategies for two types of deviation sources: environmental interference and component aging, and provides instructions to the compensation execution module to avoid the blindness of a single compensation method.

[0039] Generally, the specific type of deviation is first identified based on the source tracing results. Then, a compensation algorithm and parameter adjustment logic are selected accordingly to ensure a high degree of matching between the adjustment action and the cause of the deviation. When the source of the deviation is environmental interference, the adjustment amount is calculated based on the correlation between temperature vibration and parameter deviation. The formula for adjusting the drive current corresponding to laser wavelength drift is as follows: The formula for adjusting the MZM bias voltage according to temperature change is: ,in This refers to the adjustment amount of the laser drive current. This is the wavelength-current compensation coefficient. This is the laser wavelength shift. This is the bias voltage adjustment amount. This is the temperature-voltage compensation coefficient. For real-time temperature, The reference temperature is used. The amount of beam offset adjustment caused by vibration is determined based on the correlation between vibration acceleration and offset angle, ensuring that the optical path returns to the reference state.

[0040] When the deviation originates from component aging, a step-by-step adjustment strategy is used to calculate the safe adjustment amount. First, the aging stages are divided based on the cumulative usage time of the optical components and historical parameter decay data. The stage division formula is as follows: ,in For the current aging stage, The cumulative usage time of the optical components. To preset the stage duration threshold, the safe adjustment range of parameters is calculated for each aging stage. The formula for a single adjustment is: ,in This refers to the amount of parameter adjustment per transaction. To meet the total adjustment requirements for deviation compensation, To preset the single adjustment ratio, These are the parameter ratings.

[0041] The adjustment amount is allocated in stages according to the current aging stage and the real-time deviation. The adjustment amount of each stage is summarized to obtain the final safe adjustment amount. Then, combined with the preset parameter adjustment constraint upper limit, all adjustment amounts are integrated to form a standardized compensation control command, which is output to the compensation execution module to ensure that the adjustment action is safe and controllable.

[0042] The compensation execution module is used to adjust the optical and electrical parameters of the optical function module based on compensation control commands to obtain the calibrated parameter status.

[0043] Further, obtaining the calibrated parameter status includes the following steps: Based on the compensation control command, analyze the adjustment amount of each parameter and adjust the optical and electrical parameters of the autonomous driving system. The system collects and adjusts real-time parameters, generates a calibrated parameter status, and feeds it back to the drift prediction module.

[0044] Specifically, the compensation execution module receives standardized compensation control commands output by the difference decision module, performs parameter calibration by executing adjustment actions, and feeds back the calibration results to optimize closed-loop operation.

[0045] Generally, the compensation control command is first parsed to extract the adjustment amount and execution logic corresponding to each parameter, ensuring that the adjustment action is consistent with the command requirements. For beam angle adjustment, the optical path is finely adjusted by driving a piezoelectric ceramic micro-displacement stage. The adjustment amount is executed according to the parsed command value, and the relationship between displacement and voltage satisfies the formula. ,in This represents the micro-displacement of the piezoelectric ceramic, corresponding to the displacement required to adjust the beam angle. The voltage-displacement coefficient of piezoelectric ceramics. The driving voltage is used to achieve precise control of the beam angle through this relationship.

[0046] Optical power adjustment is achieved through a programmable electrically adjustable attenuator. Based on the optical power adjustment amount in the command, the attenuation coefficient is adjusted to match the target power value. MZM bias voltage adjustment uses a high-precision digital-to-analog converter to convert the digital adjustment amount into an analog voltage signal, driving the modulation unit parameters to reset. Laser drive current adjustment is achieved through a PWM drive module, with the duty cycle adjustment formula being... ,in To adjust the PWM duty cycle, As the baseline duty cycle, For the adjustment of the drive current, The rated drive current ensures smooth and controllable current adjustment.

[0047] After adjustment, real-time parameters are synchronously acquired through the parameter acquisition module. The deviation between the calibrated parameters and the ideal reference baseline of the optical functional module is calculated. The verification formula is as follows: ,in To calibrate the deviation, These are the real-time parameter values ​​after calibration. As an ideal reference baseline value, the calibrated parameter status is generated after successful verification. The calibrated parameter status is fed back to the drift prediction module to provide updated benchmark data for subsequent drift trend prediction, ensuring the accuracy of closed-loop prediction. At the same time, the adjustment process parameters are recorded to support the accumulation of historical attenuation data.

[0048] Example 2: In a second embodiment of the present invention, the present invention provides a novel application method for optical components in autonomous driving, such as... Figure 2 As shown, it includes the following steps: Optical functions: transmitting, modulating, and receiving optical signals in autonomous driving; Parameter acquisition: Acquire optical parameters, operating parameters, and electrical parameters during the optical function steps to obtain raw acquisition data; Data preprocessing: The raw collected data is time-synchronized, noise-reduced, and standardized to obtain preprocessed data; Drift prediction: Based on preprocessed data, the drift trend of parameters is predicted by time series modeling and multi-source data fusion, and the drift prediction result is obtained. Deviation source tracing: Based on the drift prediction results, the sources of deviation are distinguished through causal inference and feature matching to obtain the source tracing results; Differential decision-making: Based on the source tracing results, customize differentiated compensation strategies and generate compensation control instructions; Compensation execution: Based on the compensation control command, adjust the optical and electrical parameters in the optical function steps to obtain the calibrated parameter status.

[0049] In autonomous driving on urban roads, onboard environmental perception optical components need to be adapted to complex operating conditions such as extreme temperatures and vibrations over long periods, and are prone to component aging after prolonged use. Traditional optical components use threshold-triggered passive compensation, which cannot predict parameter drift trends in advance, nor can it accurately distinguish whether the deviation is caused by environmental interference or component aging. The compensation is not targeted enough and is prone to causing laser wavelength drift and beam offset, thereby affecting the accuracy of environmental perception and posing safety hazards to autonomous driving decisions. To solve the above problems, a novel application method of optical components in autonomous driving provided by this invention is adopted, the process of which is as follows: Figure 2 As shown. The specific implementation process of this method is as follows: First, the optical function steps are performed by emitting a laser of a preset wavelength, which is then transmitted, modulated, and received to provide basic optical signal support for environmental perception.

[0050] The parameter acquisition step is then carried out, synchronously collecting optical parameters, operating parameters, and electrical parameters during the operation of the optical structure. A time stamp is added to each piece of data, and the data is integrated in chronological order to form the original data, ensuring the integrity and relevance of the data.

[0051] Next, a data preprocessing step is performed. Based on the time stamp, the original data is synchronized in multiple data types. Interference signals are removed by a fusion filtering algorithm. Then, the data is standardized to unify the data units and output structured preprocessed data.

[0052] The next step is to perform a drift prediction step, extract the temporal features of the target parameters from the preprocessed data, construct a multi-dimensional feature vector, and predict the parameter drift trend through a temporal modeling and multi-source data fusion algorithm to obtain the drift prediction result. After comparing it with the dynamic threshold, if it exceeds the threshold, the deviation tracing step is triggered.

[0053] In the deviation tracing step, a causal graph of the relationship between operating conditions, parameters, and deviations is first constructed. Confusion variables are fixed through causal inference, and the similarity between real-time deviation features and the preset feature library is calculated by combining feature matching to distinguish whether the source of deviation is environmental interference or component aging.

[0054] Next, the difference decision-making step is executed, and differentiated compensation strategies are customized for different sources of deviation. The adjustment amount of each parameter is calculated, and standardized compensation control instructions are generated in combination with the preset constraint upper limit.

[0055] Finally, the compensation execution step is performed, and relevant parameters are adjusted according to the instructions. The adjusted real-time parameters are collected to form the calibrated parameter status, which is fed back to the drift prediction step to form a complete closed loop and continuously ensure the stability of the environmental perception accuracy of the optical component.

[0056] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A new application of a light assembly for unmanned vehicles, characterized by, include: Optical functional modules are used for the transmission, modulation, and reception of optical signals in autonomous driving. The parameter acquisition module is used to acquire the optical parameters, operating parameters, and electrical parameters of the optical functional module to obtain the raw acquisition data. The data preprocessing module is used to perform time synchronization, noise reduction, and standardization on the raw collected data to obtain preprocessed data. The drift prediction module is used to predict the parameter drift trend based on the preprocessed data by time series modeling and multi-source data fusion, and obtain the drift prediction result. The deviation tracing module is used to distinguish the sources of deviation based on the drift prediction results through causal inference and feature matching, and obtain the tracing results. The differential decision-making module is used to customize differential compensation strategies and generate compensation control instructions based on the source tracing results. The compensation execution module is used to adjust the optical and electrical parameters of the optical function module based on the compensation control command to obtain the calibrated parameter state.

2. The novel optical component for autonomous driving according to claim 1, characterized in that: The optical parameters include laser wavelength, output power, beam offset angle, and carrier rejection ratio; the operating parameters include operating temperature and vibration acceleration; and the electrical parameters include MZM bias voltage and laser drive current. Obtaining the preprocessed data includes the following steps: Time synchronization of different types of data in the original collected data is performed based on time stamps to ensure data time sequence consistency and obtain synchronized data. The Kalman-Particle fusion filtering algorithm is used to denoise the synchronization data, removing high-frequency noise and low-frequency drift to obtain denoised data; The noise reduction data is standardized, and after unifying the data units, it is classified and organized according to the parameter type to form preprocessed data.

3. A novel optical component for autonomous driving according to claim 1, characterized in that: Obtaining the drift prediction result includes the following steps: Extract the time-series features of the target parameters from the preprocessed data and construct a three-dimensional feature vector, which includes the current value, historical period average, and rate of change. The Holt-Winters exponential smoothing method is used to model the three-dimensional feature vector, and the horizontal component, trend component and seasonal component are calculated to obtain the initial predicted value. By fusing preprocessed data through Bayesian filtering, the initial prediction values ​​are corrected to obtain the drift prediction results; The drift prediction result is compared with the dynamic threshold. If the drift prediction result exceeds the dynamic threshold, the drift prediction result is sent to the deviation tracing module.

4. A novel optical component for autonomous driving according to claim 3, characterized in that: Setting the dynamic threshold includes the following steps: An initial threshold coefficient is set based on an ideal reference baseline of the optical functional module. Each preset duration is based on the cumulative usage time of the optical components and historical calibration data. The initial threshold coefficient is finely adjusted to form a threshold range that dynamically adapts to the aging of the components, thus obtaining a dynamic threshold.

5. A novel optical component for autonomous driving according to claim 1, characterized in that: Obtaining the source tracing results includes the following steps: Construct a directed acyclic causal graph with the correlation between operating conditions, parameters, and deviations; By fixing the confusion variables in the directed acyclic causal graph using the do-calculus operator, the correlation between the deviation and the operating parameters and the cumulative usage time of the optical components was observed. Based on the difference between the drift prediction result and the ideal reference baseline, the real-time deviation characteristics are calculated. Calculate the Mahalanobis distance between the real-time deviation characteristics and the preset environmental interference characteristic library and component aging characteristic library; By combining the causal inference results obtained through observations using the do-calculus operator with the Mahalanobis distance matching degree, the source of the deviation is determined to be environmental interference or component aging, and the source tracing results are output.

6. A novel optical component for autonomous driving according to claim 1, characterized in that: The generation of compensation control instructions includes the following steps: Based on the source tracing results, distinguish between environmental interference and component aging types; When there is environmental interference, the adjustment amounts of beam angle, optical power, bias voltage and drive current are calculated based on the correlation between temperature, vibration and parameter deviation. When components are aging, a step-by-step adjustment strategy is adopted to calculate the safe adjustment amount of each parameter; The upper limit of the adjustment constraint is adjusted by combining the preset parameters, and the adjustment amount is integrated to form a standardized compensation control command, which is then output to the compensation execution module.

7. A novel optical component for autonomous driving according to claim 6, characterized in that: The step-by-step adjustment strategy is adopted to calculate the safe adjustment amount of each parameter, including the following steps: Based on the cumulative usage time of optical components and historical degradation data of various parameters, aging stages are divided. For each aging stage, calculate the safe adjustment range of the corresponding parameter to ensure that the single adjustment does not exceed the preset proportion of the parameter's rated value. The adjustment amount is allocated in stages based on the current aging stage and the magnitude of the real-time deviation. Summarize the adjustments allocated at each stage to form the safety adjustment amount for each parameter.

8. A novel optical component for autonomous driving according to claim 1, characterized in that: Obtaining the calibrated parameter status includes the following steps: Based on the compensation control command, analyze the adjustment amount of each parameter and adjust the optical and electrical parameters of the autonomous driving system. The system collects and adjusts real-time parameters, generates a calibrated parameter status, and feeds it back to the drift prediction module.

Images