Inertial reference based road reflectivity correlation laser speed error determination method

Abstract

The application discloses a kind of inertial reference-based road reflectivity correlation laser speed measurement error determination method, it is related to laser speed measurement accuracy calibration field, the method includes according to signal-to-noise ratio extraction road reflectivity feature vector;According to original measurement speed, three-axis angular velocity and three-axis specific force, short-time speed is calculated based on inertial navigation mechanical arrangement and combined filtering, and consistency test is carried out by dynamics constraint, generates inertial reference speed;According to inertial reference speed and original measurement speed, determine instantaneous speed measurement error, and according to instantaneous speed measurement error and the road reflectivity feature vector of corresponding moment generates error training sample pair;According to error training sample pair, on-line learning and parameter dynamic adjustment are carried out to adaptive mapping model, and the error dynamic compensation model of laser speed measuring instrument based on road reflectivity function is obtained.The application can improve the precision and environmental adaptability of laser speed measurement.

Description

Technical Field

[0001] This application relates to the field of laser velocity measurement accuracy calibration, and in particular to a method for determining laser velocity measurement error based on road surface reflectivity correlation using an inertial reference. Background Technology

[0002] Laser speedometers, with their advantages of non-contact measurement, fast response speed, and wide speed measurement range, have become the core speed measurement component of vehicle-mounted inertial navigation systems. They are widely used in intelligent driving, vehicle positioning and navigation, and engineering vehicle speed measurement. Especially in environments where Global Navigation Satellite System (GNSS) signals are denied, such as tunnels and underground parking garages, laser speedometers are key devices for compensating for the accumulation of errors in the Inertial Navigation System (INS) and ensuring navigation continuity.

[0003] Currently, the speed measurement accuracy of laser speedometers is highly dependent on offline laboratory calibration before leaving the factory. They generally adopt a working mode with fixed scale factor and fixed error compensation parameters, which has many technical defects in actual vehicle operation.

[0004] Laser speed measurement relies on road surface echo signals to achieve Doppler speed measurement. However, the reflectivity of different road materials and road conditions varies greatly, which directly causes fluctuations in the laser echo signal-to-noise ratio and leads to errors such as zero bias and scale factor drift in speed measurement.

[0005] Existing online calibration methods rely on external ground-based equipment such as high-precision GNSS and ground-based velocity benchmarks to obtain reference speeds. This not only results in high calibration costs and cumbersome deployment processes, but also makes it impossible to carry out calibration operations in scenarios such as GNSS rejection, lack of ground benchmarks, and tunnels. As a result, laser velocimeters are in a long-term uncalibrated state, leading to a continuous deterioration in speed measurement accuracy.

[0006] Traditional calibration models are offline and fixed structures, which cannot update error parameters according to real-time road conditions and vehicle motion status. When the vehicle suddenly changes road conditions, the speed measurement error will change instantly. Existing methods do not have the ability to respond and compensate quickly, which further aggravates the drift of integrated navigation positioning.

[0007] Uncompensated laser velocity measurement errors will continue to be transmitted to the INS / laser integrated navigation filter, causing distortion of the filter observation residuals and accelerating the accumulation of errors in inertial devices. In scenarios without satellite signals, such as long-distance tunnels and underground road networks, positioning errors will diverge rapidly, seriously affecting the stability and availability of the vehicle navigation system.

[0008] In summary, a speed measurement error calibration method needs to be proposed to solve the above problems. Summary of the Invention

[0009] The purpose of this application is to provide a method for determining the error of laser velocity measurement based on road surface reflectivity correlation with inertial reference, which can improve the accuracy and environmental adaptability of laser velocity measurement.

[0010] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides a method for determining laser velocity measurement error based on road surface reflectivity using an inertial reference, comprising: acquiring raw data from a laser velocimeter, a signal-to-noise ratio (SNR), and raw data from an inertial measurement unit (IMU); the raw data from the laser velocimeter is the raw measured velocity; the raw data from the IMU includes triaxial angular velocity and triaxial specific force; extracting a road surface reflectivity feature vector based on the SNR; performing short-time velocity estimation based on the raw measured velocity, the triaxial angular velocity, and the triaxial specific force using inertial navigation mechanical arrangement and combined filtering, and performing consistency verification through dynamic constraints to generate an inertial reference velocity; determining the instantaneous velocity measurement error based on the inertial reference velocity and the raw measured velocity, and generating error training sample pairs based on the instantaneous velocity measurement error and the road surface reflectivity feature vector at the corresponding time; performing online learning and dynamic parameter adjustment on an adaptive mapping model based on the error training sample pairs to obtain a dynamic error compensation model for the laser velocimeter based on the road surface reflectivity function; obtaining a dynamic error compensation value for the raw measured velocity based on the road surface reflectivity feature vector and the dynamic error compensation model; and using the dynamic error compensation value to compensate for the raw data from the laser velocimeter.

[0011] In one embodiment, acquiring the raw data of the laser velocimeter, the signal-to-noise ratio, and the raw data of the inertial measurement unit specifically includes: synchronizing the laser velocimeter and the inertial measurement unit in time based on a second pulse synchronization signal from a unified clock source; and acquiring the raw data of the laser velocimeter, the signal-to-noise ratio, and the raw data of the inertial measurement unit.

[0012] In one embodiment, acquiring the original data of the laser velocimeter, the signal-to-noise ratio, and the original data of the inertial measurement unit specifically includes: acquiring the original data of the laser velocimeter, the signal-to-noise ratio, and the inertial measurement unit data; and sampling the original data of the laser velocimeter, the signal-to-noise ratio, and the inertial measurement unit data at the same frequency to ensure that the timestamps of each frame of data are aligned.

[0013] In one embodiment, extracting the road surface reflectivity feature vector based on the signal-to-noise ratio specifically includes: applying a sliding window average filter to the signal-to-noise ratio to obtain a smoothed signal-to-noise ratio sequence; extracting road surface reflectivity features based on the smoothed signal-to-noise ratio sequence; the road surface reflectivity features include the mean signal-to-noise ratio, the standard deviation of the signal-to-noise ratio, and the rate of change of the signal-to-noise ratio; and normalizing the road surface reflectivity features to obtain a road surface reflectivity feature vector.

[0014] In one embodiment, based on the original measured velocity, the three-axis angular velocity, and the three-axis specific force, short-time velocity estimation is performed using inertial navigation mechanical orchestration and combined filtering, and consistency verification is performed through dynamic constraints to generate an inertial reference velocity. Specifically, this includes: updating the carrier attitude matrix based on the three-axis angular velocity; performing coordinate system transformation on the three-axis specific force using the carrier attitude matrix to obtain the transformed three-axis specific force; obtaining the motion acceleration based on the transformed three-axis specific force and gravitational acceleration; integrating the motion acceleration and determining the current pure inertial calculated velocity based on the inertial reference velocity of the previous moment; using Kalman filtering to reduce noise based on the current pure inertial calculated velocity and the corrected velocity to obtain a denoised reference velocity; the corrected velocity is the velocity obtained after feedforward compensation of the original measured velocity using an adaptive mapping model; and determining the inertial reference velocity based on the denoised reference velocity and the motion acceleration using a reference velocity reliability assessment.

[0015] In one embodiment, determining the inertial reference velocity based on the denoised reference velocity and the motion acceleration using a reference velocity reliability assessment specifically includes: calculating the acceleration amplitude, rate of change of acceleration, and velocity change trend based on the motion acceleration; calculating the correlation coefficient between the first-order difference sequence of the denoised reference velocity and the original measured velocity; determining whether the acceleration amplitude is continuously lower than an amplitude threshold, the rate of change of acceleration is lower than an acceleration rate of change threshold, and the velocity change trend is lower than a velocity change threshold, and whether the correlation coefficient is higher than a set threshold; if so, then the denoising is determined. If the denoising reference speed is not the inertial reference speed, then the denoising reference speed is deemed unreliable. The current denoising reference speed is discarded, and no error training sample pairs are generated for the corresponding time. The online learning and parameter dynamic adjustment of the adaptive mapping model at the current time are suspended. The model parameters of the previous time are maintained to perform instantaneous error compensation on the original measured speed of the laser velocimeter that is currently being collected normally. At the same time, the observation noise covariance matrix of the combined navigation filter is adaptively adjusted, and the data fusion of the combined navigation filter continues to be performed until the denoising reference speed calculated at subsequent time points meets the reference speed reliability assessment conditions again.

[0016] In one embodiment, the adaptive mapping model is an adaptive lookup table model, a recursive least squares regression model, or a combined model; the combined model is a combination of the adaptive lookup table model and the recursive least squares regression model.

[0017] In one embodiment, when the adaptive mapping model is an adaptive lookup table model, the adaptive mapping model is learned online and its parameters are dynamically adjusted based on the error training samples to obtain a dynamic error compensation model for the laser speed measuring instrument based on the road surface reflectivity function. Specifically, this includes: spatially discretizing the road surface reflectivity feature vector of the error training samples to obtain multiple intervals; storing the error estimate and confidence weight in each interval; determining the interval and error compensation value based on the road surface reflectivity feature vector of the error training samples using nearest neighbor or interpolation algorithms; and updating the error estimate and confidence weight of the intervals using exponentially decaying average based on the error training samples to obtain the dynamic error compensation model for the laser speed measuring instrument based on the road surface reflectivity function.

[0018] In one embodiment, when the adaptive mapping model is a recursive least squares regression model, the adaptive mapping model is learned online and its parameters are dynamically adjusted according to the error training samples to obtain a dynamic error compensation model for the laser speed measuring instrument based on the road surface reflectivity function. Specifically, this includes: dynamically adjusting the parameters of the adaptive mapping model using the recursive least squares algorithm according to the error training samples to obtain a dynamic error compensation model for the laser speed measuring instrument based on the road surface reflectivity function.

[0019] In one embodiment, the method for determining laser velocity measurement error based on road surface reflectivity correlation according to inertial reference further includes: performing feedforward compensation on the original data of the laser velocity meter according to the dynamic error compensation value to obtain the corrected velocity; inputting the corrected velocity into the integrated navigation filter for data fusion to obtain the fused velocity; and feeding the fused velocity back to the short-time velocity estimation step as the initial velocity reference for estimating the inertial reference velocity at subsequent times, thus forming a closed-loop optimization of error calibration and integrated navigation.

[0020] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides a method for determining laser velocity measurement error based on road surface reflectivity using an inertial reference. It extracts the road surface reflectivity feature vector through the signal-to-noise ratio (SNR) of the laser velocimeter, overcoming the problem of traditional calibration neglecting road surface reflectivity interference. It transforms the SNR into a quantitative feature of road surface reflectivity, converting implicit error sources into modelable and compensable explicit parameters, thus fundamentally solving the velocity measurement deviation problem caused by reflectivity. Relying on the high short-term motion accuracy of inertial navigation systems, it selects stable motion segments through dynamic constraints, generating a high-confidence internal dynamic reference scale. This eliminates dependence on external equipment such as high-precision GNSS and ground velocity references, reducing calibration costs and expanding applicable scenarios. Based on error training samples, the adaptive mapping model is learned and its parameters are updated online. This allows it to adapt to known road conditions and quickly respond to unfamiliar roads and sudden road changes. The model converges quickly and has high compensation accuracy, further improving environmental adaptability. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart of a method for determining laser velocity measurement error based on road surface reflectivity in one embodiment of this application.

[0023] Figure 2 This is a schematic diagram of a method for determining laser velocity measurement error based on road surface reflectivity in accordance with an embodiment of this application. Detailed Implementation

[0024] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0025] Traditional calibration methods do not use road surface reflectivity as an error modeling factor and only use fixed parameters for compensation. This makes them unsuitable for changing road conditions, and the speed measurement error increases significantly with road surface changes, making it difficult to meet the requirements of high-precision navigation.

[0026] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0027] In one exemplary embodiment, such as Figure 1 As shown, a method for determining laser velocity measurement error based on road surface reflectivity correlation using an inertial reference is provided. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method includes the following steps: Step 101: Acquire the raw data of the laser velocimeter, the signal-to-noise ratio, and the raw data of the inertial measurement unit; the raw data of the laser velocimeter is the raw measured velocity; the raw data of the inertial measurement unit includes triaxial angular velocity and triaxial specific force.

[0028] Step 102: Extract the road surface reflectivity feature vector based on the signal-to-noise ratio.

[0029] Step 103: Based on the original measured velocity, the three-axis angular velocity, and the three-axis specific force, short-time velocity is calculated using inertial navigation mechanical arrangement and combined filtering, and consistency is checked through dynamic constraints to generate an inertial reference velocity.

[0030] Step 104: Determine the instantaneous speed measurement error based on the inertial reference speed and the original measured speed, and generate error training sample pairs based on the instantaneous speed measurement error and the road surface reflectivity feature vector at the corresponding time.

[0031] Step 105: Based on the error training samples, perform online learning and dynamic parameter adjustment on the adaptive mapping model to obtain the error dynamic compensation model of the laser speedometer based on the road surface reflectivity function.

[0032] Step 106: Based on the road surface reflectivity feature vector and the error dynamic compensation model, obtain the dynamic error compensation value of the original measured speed; the dynamic error compensation value is used to compensate the original data of the laser velocimeter.

[0033] In an exemplary embodiment, acquiring the raw data of the laser velocimeter, the signal-to-noise ratio, and the raw data of the inertial measurement unit specifically includes: synchronizing the laser velocimeter and the inertial measurement unit in time based on a second pulse synchronization signal from a unified clock source; and acquiring the raw data of the laser velocimeter, the signal-to-noise ratio, and the raw data of the inertial measurement unit.

[0034] In another exemplary embodiment, acquiring the raw data of the laser velocimeter, the signal-to-noise ratio, and the raw data of the inertial measurement unit specifically includes: acquiring the raw data of the laser velocimeter, the signal-to-noise ratio, and the inertial measurement unit data; and sampling the raw data of the laser velocimeter, the signal-to-noise ratio, and the inertial measurement unit data at the same frequency to ensure that the timestamps of each frame of data are aligned.

[0035] In an exemplary embodiment, extracting the road surface reflectivity feature vector based on the signal-to-noise ratio specifically includes: applying a sliding window average filter to the signal-to-noise ratio to obtain a smoothed signal-to-noise ratio sequence; extracting road surface reflectivity features based on the smoothed signal-to-noise ratio sequence; the road surface reflectivity features include the mean signal-to-noise ratio, the standard deviation of the signal-to-noise ratio, and the rate of change of the signal-to-noise ratio; and normalizing the road surface reflectivity features to obtain a road surface reflectivity feature vector.

[0036] In an exemplary embodiment, based on the original measured velocity, the three-axis angular velocity, and the three-axis specific force, short-time velocity estimation is performed using inertial navigation mechanical orchestration and combined filtering, and consistency verification is performed through dynamic constraints to generate an inertial reference velocity. Specifically, this includes: updating the carrier attitude matrix based on the three-axis angular velocity; performing coordinate system transformation on the three-axis specific force using the carrier attitude matrix to obtain the transformed three-axis specific force; obtaining the motion acceleration based on the transformed three-axis specific force and gravitational acceleration; integrating the motion acceleration and determining the current pure inertial calculated velocity based on the inertial reference velocity of the previous moment; using Kalman filtering to reduce noise based on the current pure inertial calculated velocity and the corrected velocity to obtain a denoised reference velocity; the corrected velocity is the velocity obtained after feedforward compensation of the original measured velocity using an adaptive mapping model; and determining the inertial reference velocity using a reference velocity reliability assessment based on the denoised reference velocity and the motion acceleration.

[0037] In practical applications, the inertial reference velocity is determined based on the denoised reference velocity and the motion acceleration using a reference velocity reliability assessment. Specifically, this includes: calculating the acceleration amplitude, rate of change of acceleration, and velocity change trend based on the motion acceleration; calculating the correlation coefficient between the first-order difference sequence of the denoised reference velocity and the original measured velocity; determining whether the acceleration amplitude remains below a threshold value within a time window, whether the rate of change of acceleration is below a threshold value, whether the velocity change trend is below a threshold value, and whether the correlation coefficient is above a set threshold; if so, the denoising is determined. If the denoising reference speed is not the inertial reference speed, then the denoising reference speed is deemed unreliable. The current denoising reference speed is discarded, and no error training sample pairs are generated for the corresponding time. The online learning and parameter dynamic adjustment of the adaptive mapping model at the current time are suspended. The model parameters of the previous time are maintained to perform instantaneous error compensation on the original measured speed of the laser velocimeter that is currently being collected normally. At the same time, the observation noise covariance matrix of the combined navigation filter is adaptively adjusted, and the data fusion of the combined navigation filter continues to be performed until the denoising reference speed calculated at subsequent time points meets the reference speed reliability assessment conditions again.

[0038] In an exemplary embodiment, the adaptive mapping model is an adaptive lookup table model, a recursive least squares regression model, or a combined model; the combined model is a combination of the adaptive lookup table model and the recursive least squares regression model.

[0039] In an exemplary embodiment, when the adaptive mapping model is an adaptive lookup table model, the adaptive mapping model is subjected to online learning and dynamic parameter adjustment based on the error training samples to obtain a dynamic error compensation model for the laser speed measuring instrument based on the road surface reflectivity function. Specifically, this includes: spatially discretizing the road surface reflectivity feature vector of the error training samples to obtain multiple intervals; storing the error estimate and confidence weight in each interval; determining the interval and error compensation value based on the road surface reflectivity feature vector of the error training samples using nearest neighbor or interpolation algorithms; and updating the error estimate and confidence weight of the intervals using exponentially decaying average based on the error training samples to obtain the dynamic error compensation model for the laser speed measuring instrument based on the road surface reflectivity function.

[0040] In an exemplary embodiment, when the adaptive mapping model is a recursive least squares regression model, the adaptive mapping model is learned online and its parameters are dynamically adjusted according to the error training samples to obtain a dynamic error compensation model for the laser speed measuring instrument based on the road surface reflectivity function. Specifically, this includes: dynamically adjusting the parameters of the adaptive mapping model using the recursive least squares algorithm according to the error training samples to obtain a dynamic error compensation model for the laser speed measuring instrument based on the road surface reflectivity function.

[0041] In an exemplary embodiment, the method for determining laser velocity measurement error based on road surface reflectivity correlation according to an inertial reference further includes: performing feedforward compensation on the original data of the laser velocity meter according to the dynamic error compensation value to obtain the corrected velocity; inputting the corrected velocity into the integrated navigation filter for data fusion to obtain the fused velocity; and feeding the fused velocity back to the short-time velocity estimation step as the initial velocity reference for estimating the inertial reference velocity at subsequent times, thus forming a closed-loop optimization of error calibration and integrated navigation.

[0042] In another exemplary embodiment, this application also provides a specific process for determining the error of laser velocity measurement based on road surface reflectivity using an inertial reference in practical applications, such as... Figure 2 As shown, it includes the following steps: S1. Build an onboard hardware data acquisition platform to simultaneously acquire raw data and signal-to-noise ratio from the laser speedometer. The raw data from the inertial measurement unit is processed, and the timestamps of the multiple sensors are aligned to obtain a standardized velocities and inertial measurement dataset.

[0043] The implementation process of step S1 includes the following: An onboard hardware platform consisting of a laser velocimeter, an inertial measurement unit, and a navigation computer was built.

[0044] Achieve time synchronization of multiple sensors: Hardware-triggered synchronization of the laser velocimeter and inertial measurement unit is achieved based on the pulse per second (PPS) synchronization signal from a unified clock source.

[0045] Acquire raw data from the laser velocimeter: including raw measured speed And signal-to-noise ratio (SNR); if the laser velocimeter's underlying hardware only outputs the raw Doppler frequency and echo signal receiving power Then it is necessary to calculate the scaling factor formula in advance ( ,in The original Doppler frequency (for laser wavelength) is converted into the original measurement velocity, and the echo signal received power is converted into signal-to-noise ratio (SNR).

[0046] Acquire raw data from the inertial measurement unit: including triaxial angular velocities. Compared to triaxial force .

[0047] The original data, signal-to-noise ratio, and inertial measurement unit data of the laser velocimeter are sampled at the same frequency to ensure that the timestamps of each frame of data are aligned.

[0048] S2. Preprocess and feature engineering the signal-to-noise ratio of the echo signal from the laser velocimeter, extract the quantized feature vector that can characterize the road surface reflectivity, and complete the feature normalization process.

[0049] The implementation process of step S2 includes the following: For the acquired raw signal-to-noise ratio sequence Sliding window averaging filtering is performed to smooth instantaneous fluctuations and obtain a smoothed signal-to-noise ratio sequence. The filtering formula is: .

[0050] In the formula, For window length, For half the length of the window, t For a moment, i This is the index (or sampling point number) of the discrete events traversed within the sliding window. Indicates at index i The raw signal-to-noise ratio value collected at each moment.

[0051] From the smoothed signal-to-noise ratio sequence The surface reflectance feature vector is extracted, and the core features include: mean feature, standard deviation feature and rate of change feature.

[0052] Mean characteristics The expression is: .

[0053] The feature extraction window is half the length.

[0054] Standard deviation characteristics The expression is: .

[0055] Change rate characteristics The expression is: .

[0056] The extracted multi-dimensional features are normalized using either Z-score normalization or Min-Max normalization, with the following formulas: Z-score standardization: .

[0057] Min-Max normalization: .

[0058] In the formula, The characteristic mean, The characteristic standard deviation, The minimum value of the characteristic. It is the maximum value of the characteristic; The original signal-to-noise ratio before normalization. The signal-to-noise ratio characteristic value after normalization.

[0059] After normalization, we get Road surface reflectivity eigenvector matrix at time 1 .

[0060] . The first eigenvector in the signal-to-noise ratio eigenvector n Each feature component This is the transpose of the matrix.

[0061] S3. Short-time velocity estimation is performed based on inertial navigation mechanical arrangement and combined filtering, and the reliability of the reference velocity is evaluated through dynamic constraint consistency verification, thus generating an inertial reference velocity.

[0062] The implementation process of step S3 includes the following: Perform short-time velocity estimation based on inertial navigation.

[0063] Using gyroscope data, i.e., three-axis angular velocity The carrier attitude matrix is ​​updated using the DCM attitude update algorithm based on quaternions or direction cosine matrices. The initial attitude is inherited from the navigation result of the previous moment or obtained through coarse alignment.

[0064] The triaxial force ratio of the inertial measurement unit Transform from the carrier coordinate system (b) to the navigation coordinate system (n) to obtain the transformed triaxial specific force. .

[0065] .

[0066] Compensating for gravitational acceleration in the navigation system Ignoring terms related to Earth's rotation, we obtain the acceleration due to motion. .

[0067] .

[0068] In short window Integrating the internal acceleration with respect to motion, combined with the inertial reference velocity from the previous moment. Calculate the pure inertial solution speed The formula is: .

[0069] For integration time variable, The starting time of integration, for The acceleration of motion under the constant navigation system. This is the time when integration ends.

[0070] Based on the current velocity calculated purely by inertia and the velocity obtained after feedforward compensation of the original measured velocity using an adaptive mapping model, a Kalman filter is used for noise reduction to obtain the denoised reference velocity. .

[0071] Construct the measurement equations for the Kalman filter and solve for the velocity using pure inertia. Speed ​​after error compensation from laser velocimeter The difference is used as a measurement of the system. .

[0072] .

[0073] The Kalman filter algorithm is used for state estimation to obtain the optimal estimate of the system velocity error at the current moment. The pure inertial solution velocity is then corrected by feedforward or feedback to obtain the noise-reduced reference velocity. .

[0074] .

[0075] In the formula, For filtering observation residuals, The state variables are estimated for the velocity error output of the Kalman filter.

[0076] Perform a reference speed reliability assessment.

[0077] Acceleration amplitude check: Calculation If it remains below the amplitude threshold within the window Then the condition is met.

[0078] Acceleration check: Calculate the rate of change of acceleration .

[0079] .

[0080] If it is below the acceleration rate of change threshold Then the condition is met.

[0081] Comparison of speed change trends: calculation and The correlation coefficient of the first-order difference sequence is considered to meet the condition if the correlation coefficient is higher than a set threshold. for t The original measured velocity at that moment.

[0082] When all the above tests are passed, the judgment is made. Inertial reference velocity .

[0083] S4. Calculate the instantaneous speed measurement error by subtracting the inertial reference speed from the original measured speed of the laser velocimeter, and bind the error value to the road surface reflectivity feature vector at the corresponding moment to generate error training sample pairs.

[0084] The implementation process of step S4 includes the following: Based on inertial reference velocity Calculate the instantaneous velocity measurement error of the laser velocimeter. The formula is: .

[0085] In the formula, Includes scale factor error and zero bias caused by road surface reflectivity; for t The original measured speed of the laser velocimeter.

[0086] Instantaneous speed measurement error Corresponding Road surface reflectivity feature vector at time 1 One-to-one binding is performed to form standardized error training sample pairs. The sample pairs need to carry timestamps and road surface environmental labels.

[0087] S5. Construct an adaptive mapping model, use error training samples to learn the model online and dynamically adjust the parameters, establish a precise mapping relationship between road surface reflectivity characteristics and speed measurement error, and obtain a dynamic error compensation model for the laser speed measuring instrument based on the road surface reflectivity function.

[0088] In this embodiment, it should also be noted that step S5 provides two adaptive mapping model structures, which can be selected or combined: an adaptive lookup table model and a recursive least squares regression model. When the two are used in combination, they are in a parallel fusion relationship, that is, the final dynamic error compensation value is output through weighted fusion. ,in The compensation value is the output of the lookup table model. The compensation value output by the regression model. The dynamic fusion weights are determined based on confidence level assessment. The specific implementation processes for online learning and dynamic parameter adjustment are as follows: S51. Implementation process of the adaptive lookup table model.

[0089] Discretize the road surface reflectivity feature space as follows: Each interval Storage error estimate and confidence weight .

[0090] Given the current features Find its interval using nearest neighbor or interpolation algorithms. Output As an error compensation value.

[0091] Receive new samples At that time, determine the interval to which it belongs. The interval parameters are updated using exponentially decaying average (EMA). To distinguish the states before and after the update, an iteration step variable is introduced. k The formula is: .

[0092] or .

[0093] In the formula, The sample learning rate, Forgotten factor for old estimates; This represents the current iteration step. This is the number of the previous iteration step; and The intervals before and after the update are respectively.j The error estimate; and These are the confidence weights before and after the update, respectively. In the first The newly input instantaneous speed measurement error sample value; if If the value is below the threshold, the compensation value is 0 or interpolated from the adjacent interval.

[0094] S52. Implementation process of recursive least squares regression model.

[0095] Assume that the velocity measurement error and reflectivity characteristics satisfy a regression relationship: .

[0096] In the formula, Let be the vector of parameters to be estimated. For characteristic basis functions, It is noise.

[0097] Online parameter updates using the RLS algorithm The iterative formula is: .

[0098] In the formula, Here is the gain matrix. For parameter estimation, the error covariance matrix, Forgetting factor It is an identity matrix. k This represents the current iteration step. For the first The instantaneous speed measurement error sample value of the step input.

[0099] Based on the error training sample pairs, the parameters of the adaptive mapping model are dynamically adjusted using the recursive least squares algorithm to obtain the updated error dynamic compensation model.

[0100] S6. Based on the current road surface reflectivity feature vector query error dynamic compensation model, obtain the dynamic error compensation value, perform feedforward compensation on the original output of the laser speedometer, and integrate the corrected speed into the combined navigation filter to form a closed-loop optimization of calibration and navigation.

[0101] The implementation process of step S6 includes the following: Perform real-time error feedforward compensation: continuously collect data during navigation. The original measured velocity at time and road surface reflectivity eigenvector ,Will Input the online-updated error dynamic compensation model to obtain the dynamic error compensation value. The original measured speed is corrected using the following formula: .

[0102] In the formula, for The original speed of the laser velocimeter. for t Speed ​​after time correction.

[0103] Output high-precision speed information: Speed ​​after time error compensation The slackline combined navigation filter of the INS / laser velocimeter is input, and Kalman filtering is used to complete data fusion. The slackline combined observation equation is: .

[0104] In the formula, For filtering observation residuals, for The pure inertial calculation speed at time step 1; the filter uses this residual for state estimation, and the final output is... High-precision reference speed after real-time noise reduction .

[0105] Achieving closed-loop optimization: High-precision reference speed output from the combined navigation filter. Feedback is fed back to the short-term velocity estimation step, serving as the initial velocity benchmark for calculating the inertial reference velocity at subsequent times, thus forming a closed-loop optimization for error calibration and integrated navigation.

[0106] Performance verification.

[0107] This application establishes a full-condition real-world test scenario for performance verification, covering complex road environments and variable vehicle motion states. An external true value benchmark segment is also set up for accuracy calibration. The specific scenario parameters are as follows.

[0108] a. Typical road conditions: Five representative road types were selected, including dry asphalt pavement, wet asphalt pavement, cement concrete pavement, gravel unpaved pavement, and tunnel inner wall pavement, to fully cover test environments with different reflectivity characteristics and simulate road switching scenarios during actual vehicle driving.

[0109] b. Vehicle motion state: Four core motion modes were set up: acceleration, deceleration and braking, constant speed cruise and steering. The variation law and compensation effect of speed measurement error under dynamic conditions were tested respectively.

[0110] c. True Value Reference Setting: Select a dedicated test section with unobstructed GNSS signals and centimeter-level positioning accuracy, and use the output results of a high-precision GNSS / INS integrated navigation system as the true value of velocity measurement for subsequent error quantification evaluation; for GNSS denied scenarios such as tunnels, set up a separate long-distance tunnel test section to verify the calibration performance without external reference.

[0111] The comparative verification method includes the following steps: To objectively verify the technical advantages of the method in this application, three sets of control experiments were set up, using a unified test scenario and data acquisition parameters to ensure the fairness and credibility of the comparison results.

[0112] This application's embodiment employs a road surface reflectivity correlation online calibration and error compensation method to achieve adaptive modeling and closed-loop optimization throughout the entire process.

[0113] Control group 1: A laser velocimeter with fixed parameters was calibrated offline in the laboratory. No parameter correction or error compensation was performed throughout the process, simulating the traditional fixed calibration scheme.

[0114] Control group 2: The conventional online calibration method without road surface reflectance correlation was used, and error fitting was performed only based on motion data without introducing reflectance feature modeling.

[0115] The evaluation index system includes a quantitative evaluation system based on four dimensions: speed measurement accuracy, model stability, environmental adaptability, and integrated navigation performance, to comprehensively measure the practicality and advancement of the method.

[0116] a. Speed ​​accuracy index: The root mean square error (RMSE) is used as the core index to calculate the deviation between the output speed of the laser velocimeter and the true speed before and after compensation. The formula is as follows: .

[0117] The total number of samples participating in the error assessment. The first output of the laser velocimeter i One measurement speed, For the corresponding number i We obtain the true speed value; at the same time, we calculate the maximum absolute error to intuitively reflect the speed measurement deviation.

[0118] b. Model convergence index: Monitor the parameter update process of the adaptive lookup table and RLS regression model, record the convergence speed of model parameters with driving time and mileage, and use the time / mileage for error fluctuation to drop to a stable threshold as the convergence criterion.

[0119] c. Environmental adaptability indicators: Simulate sudden road surface changes (such as dry asphalt turning into waterlogged road surface, paved road turning into gravel road surface), statistically analyze the sudden change amplitude of speed measurement error and compensation response delay, and evaluate the adaptability of the method to sudden changes in road surface reflectivity.

[0120] d. Integrated navigation improvement indicators: In tunnel scenarios where GNSS is denied, the cumulative rate of positioning error of INS / laser integrated navigation is continuously tested within a fixed duration / mileage. The positioning drift of the three methods is compared to measure the gain effect of velocity compensation on integrated navigation performance.

[0121] Mathematical models and analytical methods include: a. The verification results are presented in their entirety using standardized mathematical analysis and visualization techniques.

[0122] b. Error Analysis Model: Quantitative error assessment is carried out based on root mean square error (RMSE) and maximum absolute error. Combined with statistical indicators such as mean and standard deviation, the error distribution pattern and dispersion are analyzed.

[0123] c. Data visualization methods: plot trend curves of speed measurement error over time and road surface type to intuitively show the difference in error fluctuation before and after compensation; plot a schematic diagram of the correspondence between reflectivity characteristics and error compensation values ​​to present the modeling effect of the adaptive mapping model.

[0124] In summary, the laser speed measuring error calibration method based on road surface reflectivity described in this application effectively solves the problems of traditional laser speed measuring instruments relying on external benchmarks and the inability to eliminate road surface reflectivity interference. In scenarios without external ground truth such as GNSS, it achieves online accurate calibration and dynamic compensation of laser speed measuring errors, ensuring speed measuring stability under all operating conditions.

[0125] Breaking through the limitations of traditional calibration that ignores road surface reflectivity interference, this method transforms the laser echo signal-to-noise ratio into a quantitative characteristic of road surface reflectivity, converting implicit error sources into explicit parameters that can be modeled and compensated, thus fundamentally solving the problem of speed measurement deviation caused by reflectivity.

[0126] Leveraging the high short-term motion accuracy of inertial navigation systems, stable motion segments are selected through dynamic constraints to generate a high-confidence internal dynamic reference scale. This eliminates the reliance on external equipment such as high-precision GNSS and ground velocity benchmarks, reducing calibration costs and expanding applicable scenarios.

[0127] It adopts a dual-model architecture of adaptive lookup table and recursive least squares regression, supports online incremental learning and real-time parameter updates, can adapt to known road conditions, and can quickly respond to unfamiliar road conditions and sudden road changes. The model has fast convergence speed and high compensation accuracy.

[0128] By integrating the error-compensated speed measurement data into the integrated navigation filter, a closed-loop system of calibration-compensation-fusion-recalibration is formed, which significantly reduces positioning drift in GNSS denial scenarios and significantly improves the reliability and robustness of the vehicle-mounted integrated navigation system.

[0129] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0130] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for determining laser velocity measurement error based on road surface reflectivity correlation using an inertial reference, characterized in that, include: Acquire raw data from the laser velocimeter, signal-to-noise ratio, and inertial measurement unit; the raw data from the laser velocimeter is the raw measured velocity; the raw data from the inertial measurement unit includes triaxial angular velocity and triaxial specific force. Extract the road surface reflectivity feature vector based on the signal-to-noise ratio; Based on the original measured velocity, the triaxial angular velocity, and the triaxial specific force, short-time velocity is calculated using inertial navigation mechanical arrangement and combined filtering, and consistency is verified through dynamic constraints to generate an inertial reference velocity. The instantaneous velocity measurement error is determined based on the inertial reference velocity and the original measured velocity, and error training sample pairs are generated based on the instantaneous velocity measurement error and the road surface reflectivity feature vector at the corresponding moment. Based on the error training samples, the adaptive mapping model is learned online and the parameters are dynamically adjusted to obtain the error dynamic compensation model of the laser speed measuring instrument based on the road surface reflectivity function. The dynamic error compensation value of the original measured speed is obtained based on the road surface reflectivity feature vector and the error dynamic compensation model; the dynamic error compensation value is used to compensate the original data of the laser velocimeter.

2. The method for determining the error of laser velocity measurement based on road surface reflectivity according to claim 1, characterized in that, Acquire raw data from the laser velocimeter, signal-to-noise ratio, and inertial measurement unit, specifically including: The laser velocimeter and inertial measurement unit are synchronized in time using a second pulse synchronization signal based on a unified clock source. Acquire raw data from the laser velocimeter, signal-to-noise ratio, and inertial measurement unit.

3. The method for determining the error of laser velocity measurement based on road surface reflectivity according to claim 1, characterized in that, Acquire raw data from the laser velocimeter, signal-to-noise ratio, and inertial measurement unit, specifically including: Acquire raw data, signal-to-noise ratio, and inertial measurement unit data from the laser velocimeter; The raw data, signal-to-noise ratio, and inertial measurement unit data of the laser velocimeter are sampled at the same frequency to ensure that the timestamps of each frame of data are aligned.

4. The method for determining the error of laser velocity measurement based on road surface reflectivity according to claim 1, characterized in that, Extracting the road surface reflectivity feature vector based on the signal-to-noise ratio specifically includes: The signal-to-noise ratio is then subjected to sliding window averaging filtering to obtain a smoothed signal-to-noise ratio sequence; Based on the smoothed signal-to-noise ratio sequence, road surface reflectance features are extracted; the road surface reflectance features include the mean signal-to-noise ratio, the standard deviation of the signal-to-noise ratio, and the rate of change of the signal-to-noise ratio. The road surface reflectivity features are normalized to obtain the road surface reflectivity feature vector.

5. The method for determining the error of laser velocity measurement based on road surface reflectivity according to claim 1, characterized in that, Based on the original measured velocity, the triaxial angular velocity, and the triaxial specific force, short-time velocity is calculated using inertial navigation mechanical arrangement and combined filtering, and consistency is checked through dynamic constraints to generate an inertial reference velocity, specifically including: Update the carrier attitude matrix based on the three-axis angular velocities; The coordinate system is transformed by the carrier attitude matrix to obtain the transformed triaxial specific force. The acceleration due to motion is obtained based on the converted triaxial specific force and gravitational acceleration. The acceleration is integrated, and the pure inertial velocity at the current moment is determined based on the inertial reference velocity at the previous moment. Based on the current pure inertial calculated velocity and the corrected velocity, Kalman filtering is used to reduce noise, resulting in a denoised reference velocity; the corrected velocity is the velocity obtained by the adaptive mapping model after feedforward compensation of the original measured velocity. The inertial reference velocity is determined by using the reference velocity reliability assessment based on the noise-reduced reference velocity and the motion acceleration.

6. The method for determining the error of laser velocity measurement based on road surface reflectivity according to claim 5, characterized in that, The inertial reference velocity is determined based on the noise-reduced reference velocity and the motion acceleration using a reference velocity reliability assessment, specifically including: Calculate the acceleration amplitude, rate of change of acceleration, and velocity change trend based on the described motion acceleration; Calculate the correlation coefficient between the first-order difference sequence of the denoised reference velocity and the original measured velocity; Determine whether the acceleration amplitude is continuously lower than the amplitude threshold within the time window, whether the acceleration rate of change is lower than the acceleration rate of change threshold, whether the velocity change trend is lower than the velocity change threshold, and whether the correlation coefficient is higher than the set threshold. If so, then the noise-reduced reference velocity is determined to be the inertial reference velocity; If not, the denoised reference speed is deemed unreliable. The current denoised reference speed is discarded, and no error training sample pairs are generated for the corresponding time. The online learning and parameter dynamic adjustment of the adaptive mapping model at the current time are suspended. The model parameters of the previous time are maintained to perform instantaneous error compensation on the original measured speed of the laser velocimeter that is currently being collected normally. At the same time, the observation noise covariance matrix of the integrated navigation filter is adaptively adjusted, and the data fusion of the integrated navigation filter continues to be performed until the denoised reference speed calculated in subsequent time steps meets the reference speed reliability assessment conditions again.

7. The method for determining the error of laser velocity measurement based on road surface reflectivity according to claim 1, characterized in that, The adaptive mapping model is an adaptive lookup table model, a recursive least squares regression model, or a combined model; the combined model is a combination of the adaptive lookup table model and the recursive least squares regression model.

8. The method for determining the error of laser velocity measurement based on road surface reflectivity according to claim 7, characterized in that, When the adaptive mapping model is an adaptive lookup table model, the adaptive mapping model is learned online and its parameters are dynamically adjusted based on the error training samples to obtain the dynamic error compensation model of the laser velocimeter based on the road surface reflectivity function, specifically including: Based on the surface reflectivity feature vector of the error training sample pair, spatial discretization is performed to obtain multiple intervals; each interval stores the error estimate and confidence weight. Based on the road surface reflectivity feature vector of the error training samples, the interval and error compensation value are determined by the nearest neighbor or interpolation algorithm; Based on the error training samples, the error estimate and confidence weight of the interval are updated using exponential decay averaging, resulting in a dynamic error compensation model for the laser speedometer based on the road surface reflectivity function.

9. The method for determining the error of laser velocity measurement based on road surface reflectivity according to claim 7, characterized in that, When the adaptive mapping model is a recursive least squares regression model, the adaptive mapping model is learned online and its parameters are dynamically adjusted based on the error training samples to obtain the dynamic error compensation model of the laser speedometer based on the road surface reflectivity function, which specifically includes: Based on the error training samples, the parameters of the adaptive mapping model are dynamically adjusted using the recursive least squares algorithm to obtain the error dynamic compensation model of the laser speedometer based on the road surface reflectivity function.

10. The method for determining the error of laser velocity measurement based on road surface reflectivity according to claim 1, characterized in that, Also includes: The original data of the laser velocimeter is fed forward to compensate for the dynamic error compensation value to obtain the corrected speed. The corrected speed is input into the combined navigation filter for data fusion to obtain the fused speed. The fused velocity is fed back to the short-time velocity estimation step as the initial velocity benchmark for calculating the inertial reference velocity at subsequent times, forming a closed-loop optimization of error calibration and integrated navigation.

Images