Radar position precision compensation method, system and device and storage medium

Abstract

The invention provides a radar position precision compensation method, system and device and a storage medium, and belongs to the technical field of ultrasonic anti-interference, and the method comprises the steps: building a maintenance short-time track sequence from the original detection data of an ultrasonic radar sensor; estimating an installation deviation angle of the ultrasonic radar sensor based on the short-time trajectory sequence; processing the installation deviation angle to generate a final compensation value; and performing real-time geometric correction on the original measurement coordinates of the ultrasonic radar sensor according to the final compensation value, and outputting corrected target position information. A short-time track sequence of a static reference object of a vehicle in the current state is obtained by establishing and maintaining a short-time track sequence, the installation deviation angle of an ultrasonic radar sensor is estimated through stability calculation of the real-time continuous short-time track sequence, and original measurement coordinates of the ultrasonic radar sensor are corrected in real time. Therefore, real-time data correction of the ultrasonic radar sensor in daily driving of the vehicle is realized.

Description

Technical Field

[0001] This invention belongs to the field of ultrasonic anti-interference technology, and particularly relates to a radar position accuracy compensation method, system, device, and storage medium. Background Technology

[0002] With the widespread adoption of autonomous driving and advanced driver assistance systems (ADAS), ultrasonic radar is widely used in functions such as automatic parking and blind spot monitoring due to its low cost and high short-range detection accuracy. However, ultrasonic radar is typically installed behind plastic bumpers with complex curved surfaces. These bumpers are prone to deformation and displacement during production, assembly, and vehicle use, causing a deviation between the radar's physical mounting axis and its theoretical axis. This error directly leads to a shift in the radar's ranging and positioning reference, resulting in false or missed obstacle detections, severely impacting system safety and reliability. Existing technologies mostly focus on offline, static calibration methods or rely on additional hardware verification systems, which are insufficient to cover real-time data correction for automotive radar during driving. Summary of the Invention

[0003] To address the problems existing in the prior art, in a first aspect, the present invention provides a radar position accuracy compensation method, comprising: Establish and maintain short-time trajectory sequences from the raw detection data of ultrasonic radar sensors; Based on the short-time trajectory sequence, the installation deviation angle of the ultrasonic radar sensor is estimated; The installation deviation angle is processed to generate a final compensation value; The original measurement coordinates of the ultrasonic radar sensor are geometrically corrected in real time based on the final compensation value, and the corrected target position information is output.

[0004] In some implementations, estimating the installation deviation angle of the ultrasonic radar sensor based on the short-time trajectory sequence includes: The detection area of ​​the ultrasonic radar sensor is divided into several angular intervals; For each trajectory point within an angle interval, an ideal trajectory model is constructed based on the vehicle's own motion patterns, and the deviation between the observed trajectory and the ideal trajectory is compared. An observation equation regarding the installation deviation angle is constructed based on the systematic offset; The observation equations for all angle intervals are collected and solved using a weighted least squares fitting algorithm to obtain a preliminary estimate of the installation deviation angle.

[0005] In some implementations, the preliminary estimate of the installation deviation angle is obtained by solving the observation equations that aggregate all angular intervals using a weighted least squares fitting algorithm, including: The weights are determined by the target distance, echo intensity, and trajectory stability. The closer the target is, the stronger the echo intensity and the more stable the trajectory, the greater the weight. Assign corresponding weights to the observation equations for each angle interval, and construct a weighted least squares optimization objective. Solving the weighted least squares optimization objective yields preliminary estimates of the yaw and pitch installation error angles.

[0006] In some implementations, processing the installation deviation angle to generate a final compensation value includes: The installation deviation angle is calculated using a two-state recursive filter, which includes a dynamic state filter and a robust state filter. Calculate the deviation between the installation deviation angle and the robust state, and determine whether the deviation is within a preset confidence interval; If the deviation is within a preset confidence interval and continues for a preset number of times, the robust state is slowly updated; otherwise, only the dynamic state is updated. Based on the principle of prioritizing robust states and supplementing with dynamic states, the final compensation value is generated.

[0007] In some implementations, the step of slowly updating the robust state if the deviation is within a preset confidence interval and continues for a preset number of times; otherwise, only updating the dynamic state includes: A preset confidence interval is determined based on the historical update data of the robust state and the system's allowable error range; If the deviation between the new preliminary estimate and the robust state exceeds the preset confidence interval, it is judged as a random error or interference, and only the dynamic state is updated; If the deviation is within a preset confidence interval and the number of consecutive occurrences exceeds a threshold, it is determined to be a valid change in the installation deviation, and a slow update of the robust state is initiated.

[0008] In some embodiments, the step of performing real-time geometric correction on the original measurement coordinates of the ultrasonic radar sensor based on the final compensation value, and outputting the corrected target position information, includes: Obtain the raw polar coordinate data of the ultrasonic radar sensor, the raw polar coordinate data including distance and raw azimuth angle; Calculate the corrected azimuth angle based on the final compensation value; The vertical distance projection is corrected based on the final compensation value; The corrected target location information is output to the upper-level application system.

[0009] In some implementations, establishing and maintaining short-time trajectory sequences from raw detection data of ultrasonic radar sensors includes: The vehicle speed and yaw rate information are obtained, and the target motion state in the original detection data is compensated for. Based on the compensated motion state, stationary or quasi-stationary targets are selected, and moving targets are eliminated to remove interference. Each selected stationary or quasi-stationary target is assigned a unique identifier, and its position data in the vehicle coordinate system is continuously recorded to form and maintain a short-term trajectory sequence.

[0010] Secondly, this application provides a radar position accuracy compensation system, the system comprising: An ultrasonic processing unit is used to establish and maintain short-time trajectory sequences from the raw detection data of ultrasonic radar sensors; The deviation angle calculation module is able to estimate the installation deviation angle of the ultrasonic radar sensor based on the short-time trajectory sequence; The dynamic compensation module is used to process the installation deviation angle and generate the final compensation value; The data correction module is used to perform real-time geometric correction on the original measurement coordinates of the ultrasonic radar sensor based on the final compensation value, and output the corrected target position information.

[0011] Thirdly, this application provides an electronic device including: a processor, a memory, a communication interface, and one or more programs, the one or more programs being stored in the memory and configured to be executed by the processor, the programs including steps for performing some or all of the steps described in the embodiments of the first aspect of this application.

[0012] Fourthly, this application provides a computer-readable storage medium that stores a computer program for electronic data interchange, the computer program causing a computer to perform some or all of the steps described in the embodiments of the first aspect of this application. Attached Figure Description

[0013] Figure 1 This is a basic flowchart illustrating a radar position accuracy compensation method provided in the embodiments of this application; Figure 2 This is a detailed flowchart illustrating the installation deviation angle estimation of a radar position accuracy compensation method provided in this application embodiment; Figure 3 This is a detailed flowchart illustrating the weighted least squares solution of the installation deviation angle for a radar position accuracy compensation method provided in this application embodiment; Figure 4 This is a schematic diagram of the dual-state recursive filter processing of the installation deviation angle in a radar position accuracy compensation method provided in this application embodiment; Figure 5This is a schematic diagram of the confidence interval judgment and robust state update process of a radar position accuracy compensation method provided in the embodiments of this application; Figure 6 This is a detailed flowchart illustrating the real-time geometric correction process of the original measurement coordinates for a radar position accuracy compensation method provided in this application embodiment; Figure 7 This is a schematic diagram illustrating the detailed process of establishing and maintaining a short-time trajectory sequence for a radar position accuracy compensation method provided in this application embodiment; Figure 8 This is a schematic diagram of the module composition of a radar position accuracy compensation system provided in the embodiments of this application.

[0014] Explanation of reference numerals in the attached diagram: 10, ultrasonic processing unit; 20, deviation angle calculation module; 30, dynamic compensation module; 40, data correction module. Detailed Implementation

[0015] The embodiments of the technical solution of this application will now be described in detail with reference to the accompanying drawings. These embodiments are only used to more clearly illustrate the technical solution of this application and are therefore merely examples, and should not be used to limit the scope of protection of this application.

[0016] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0017] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0018] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0019] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0020] In the description of the embodiments of this application, the term "multiple" refers to two or more (including two), similarly, "multiple sets" refers to two or more (including two sets), and "multiple pieces" refers to two or more (including two pieces).

[0021] In the description of the embodiments of this application, the technical terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the embodiments of this application and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of this application.

[0022] In the description of the embodiments of this application, unless otherwise expressly specified and limited, technical terms such as "installation," "connection," "joining," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. For those skilled in the art, the specific meaning of the above terms in the embodiments of this application can be understood according to the specific circumstances.

[0023] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0024] Reference Figure 1 A radar position accuracy compensation method includes: S101. Establish and maintain short-time trajectory sequences from the raw detection data of the ultrasonic radar sensor; This step can be understood as follows: from the raw detection data output by the ultrasonic radar sensor, target data that is effective for estimating the installation deviation angle is selected, and by constructing and maintaining short-time trajectory sequences, the position change characteristics of the target are preserved, providing a stable and reliable data foundation for the accurate estimation of the subsequent installation deviation angle.

[0025] The original detection data refers to the data directly output by the ultrasonic radar sensor, which, after preliminary analysis, includes key parameters such as target distance, azimuth, and echo intensity. Establishing and maintaining short-time trajectory sequences involves continuously tracking and recording the position information of selected targets within a preset time window, forming a sequence while eliminating invalid data to ensure sequence validity. By constructing short-time trajectory sequences, interference data from moving targets and environmental noise can be filtered out, and time-continuous target position data can be provided for subsequent installation deviation angle estimation. This avoids the impact of random errors in a single frame of data on the deviation estimation results, thereby improving the robustness and accuracy of the entire compensation method. Simultaneously, the short-time trajectory maintenance mechanism can promptly eliminate failed targets, reducing the computational resource consumption of invalid data and ensuring the real-time performance of the compensation method.

[0026] For example, the raw ultrasonic radar detection data is filtered and denoised, and stationary or quasi-stationary targets are selected by combining the vehicle's motion information; a unique identifier is assigned to the selected targets, and their position data is recorded within a preset time window to form a short-term trajectory sequence; and the trajectory of continuously undetected targets is eliminated.

[0027] S102. Estimate the installation deviation angle of the ultrasonic radar sensor based on the short-time trajectory sequence; Specifically, this step utilizes the positional change patterns of the target in a short-time trajectory sequence to infer the angular offset caused by the installation deviation of the ultrasonic radar sensor. Because the ultrasonic radar exhibits deviation due to deflection, the detected target trajectory will systematically deviate from the ideal trajectory constructed based on vehicle motion patterns. The ideal trajectory refers to the pre-defined observation trajectory of a stationary reference object with the ultrasonic radar sensor installed without deviation. The systematic offset refers to the fact that at a certain point in time, all observed targets have shifted in the same direction relative to the ideal trajectory. Therefore, the installation deviation angle can be estimated by analyzing this systematic offset.

[0028] Based on short-time trajectory sequences, using the effective short-time trajectory constructed by S101 as data foundation, the estimated installation deviation angle determines the angle between the current actual direction of the ultrasonic radar and the theoretically designed installation direction. This deviation angle typically includes yaw deviation in the horizontal direction and pitch deviation in the vertical direction, and is a core parameter for subsequent coordinate correction. This step leverages the continuity of the short-time trajectory sequence to extract the systematic offset caused by installation deviation, avoiding misjudging random noise as installation deviation, thereby improving the accuracy of deviation angle estimation. By constructing an ideal trajectory model based on vehicle motion patterns, the deviation angle estimation process closely matches the actual driving scenario, enhancing the practicality and adaptability of the estimation results. This provides core parameter support for subsequent precise compensation and is a key step in improving position accuracy. Those skilled in the art will understand that accurate estimation of the installation deviation angle is a crucial prerequisite for achieving position accuracy compensation; only by accurately obtaining the deviation angle can the measurement errors caused by installation deviation be effectively eliminated.

[0029] For example, short-time trajectory sequences of stationary targets in different angle intervals are selected to construct an ideal trajectory model based on the vehicle's motion law. The systematic offset between the calculated observation trajectory and the ideal trajectory within a predetermined time (e.g., the first 1 second) is taken, and the observation equation is constructed and solved accordingly to obtain a preliminary estimate of the installation deviation angle.

[0030] S103. Process the installation deviation angle and generate the final compensation value; Specifically, the steps involve optimizing the preliminary estimate of the installation deviation angle obtained in S102 to eliminate estimation fluctuations caused by random errors, environmental interference, and other factors, generating a stable and reliable final compensation value. Since the preliminary estimate is obtained based on the systematic offset between the observed trajectory and the ideal trajectory, and this offset may be influenced by random noise, sudden interference, and other non-installation deviation factors, directly using it for correction would lead to unstable position accuracy. Therefore, specific processing methods are needed to filter effective deviation information to ensure the robustness of the compensation value.

[0031] For example, a dual-state recursive filter is used to process the initial estimate of the installation deviation angle. The validity of the deviation is judged by the confidence interval. The robust state is slowly updated for the valid deviation. Finally, a stable final compensation value is generated based on the weighted sum of the robust state and the dynamic state.

[0032] S104. Perform real-time geometric correction on the original measurement coordinates of the ultrasonic radar sensor based on the final compensation value, and output the corrected target position information.

[0033] Specifically, this step utilizes the final compensation value generated by S103 to perform real-time geometric correction on the original measurement coordinates of the ultrasonic radar sensor. This eliminates systematic errors caused by installation deviations (this step occurs during production inspection, correcting for factory installation errors using preset benchmark data and standard sites) or deflections during use, resulting in the simultaneous directional shift of all targets at the same point in time (occurring during user use, where the radar deflects due to loosening or deformation of the mounting base caused by collisions or other issues, obtained through abrupt changes in the position information of continuously tracked static reference objects). The result is an accurate output of target position information. Real-time geometric correction refers to dynamically compensating and correcting the data from the current sampling period, thus achieving online correction. The original measurement coordinates are typically polar coordinates output by the ultrasonic radar, which need to be converted to Cartesian coordinates in the vehicle coordinate system before correction. The core of the correction is to offset the directional shift caused by installation deviations, ensuring that the vehicle system can still obtain correct Cartesian coordinates even after radar deflection. For example, the original polar coordinate data of the ultrasonic radar is converted into Cartesian coordinates in the vehicle coordinate system, and the Cartesian coordinates are geometrically corrected using the final compensation value to correct the vertical distance projection. Finally, the corrected target position information is output to the upper-level application system in real time.

[0034] This application provides a radar position accuracy compensation method, including: S101. Establish and maintain short-time trajectory sequences from the raw detection data of the ultrasonic radar sensor; S102. Estimate the installation deviation angle of the ultrasonic radar sensor based on the short-time trajectory sequence; S103. Process the installation deviation angle and generate the final compensation value; S104. Perform real-time geometric correction on the original measurement coordinates of the ultrasonic radar sensor based on the final compensation value, and output the corrected target position information.

[0035] By establishing and maintaining short-time trajectory sequences, the short-time trajectory sequence of stationary reference objects under the current state of the vehicle is obtained. The installation deviation angle of the ultrasonic radar sensor is estimated by calculating the stability of the real-time continuous short-time trajectory sequence, and the original measurement coordinates of the ultrasonic radar sensor are corrected in real time, thereby realizing real-time data correction of the ultrasonic radar sensor during daily vehicle operation.

[0036] In some implementations, refer to Figure 2 S102. Based on the short-time trajectory sequence, the installation deviation angle of the ultrasonic radar sensor is estimated, including: S201. Divide the detection area of ​​the ultrasonic radar sensor into several angular intervals; Specifically, in the process of estimating the installation deviation angle in S102, the effective detection range of the radar is divided into segments according to the azimuth angle. For example, the detection range of -60° to 60° is divided into multiple intervals with a span of 15°, and for the short-time trajectory points of stationary targets in each interval, pre-processing actions of trajectory analysis and offset calculation are carried out respectively.

[0037] The impact of installation deviation on targets in different angle intervals is regular. Under the same deviation angle, the target offsets at different angles may have a fixed correlation. After partitioning, the trajectory points in each interval can be analyzed separately, avoiding the systematic offset ambiguity caused by the mixing of target data from different angles, and more accurately capturing the offset patterns in each angle interval.

[0038] In addition, the detection performance of ultrasonic radar may vary in different angular ranges. For example, the echo intensity is weaker in the edge range and the measurement accuracy of near and far targets is different. Partition processing can perform targeted analysis on the characteristics of each range to reduce the interference of detection performance differences on offset estimation.

[0039] S202. For the trajectory points in each angle interval, construct an ideal trajectory model based on the vehicle's own motion law, and compare the deviation between the observed trajectory and the ideal trajectory. This step can be understood as having two main objectives: to construct an ideal trajectory model and to use that ideal trajectory model to evaluate whether the current observed trajectory has deflected.

[0040] Constructing an ideal trajectory model refers to establishing the theoretical trajectory that a stationary target should present within a certain range when the ultrasonic radar has no installation deviation, based on the vehicle's own motion law. For example, when the vehicle is traveling at a constant speed in a straight line, the ideal trajectory of a stationary roadside target within the 15°-30° angle range to the left of the radar detection area is a straight line trajectory parallel to the vehicle's direction of travel, and its distance relative to the vehicle is a trigonometric function relationship between the horizontal distance and the included angle.

[0041] The ideal trajectory model can be obtained in several ways. For example, during the vehicle factory calibration phase, a standard test site and a preset ideal trajectory model are used for pre-adjustment to correct installation errors. For example, during the vehicle's commissioning phase, the calculation starts from the time point after factory calibration. During this phase, if the observed trajectory does not undergo abrupt changes (such as uniform linear motion that meets expectations; conversely, if the position information of all observed targets deviates in the same trend at the same moment, it is considered an abnormal abrupt change), it is memorized as the ideal trajectory model for the current scenario.

[0042] S203. Construct observation equations for the installation deviation angle based on systematic offset; Specifically, this step involves transforming the systematic offset of all targets moving in the same direction at the same time point, along with the geometric relationship between the yaw installation deviation angle (Δψ) and the pitch installation deviation angle (Δθ), into a mathematical equation that can be quantitatively solved.

[0043] Based on the principles of vehicle kinematics and the spatial geometric characteristics of radar detection, linear observation equations are independently established for stationary targets within each angular interval, with the systematic offset as the dependent variable and Δψ and Δθ as unknowns. The quantitative relationship among the three is clarified, providing mathematical model support for the subsequent simultaneous solution of the two deviation angles.

[0044] For example, for the yaw deviation angle Δψ, an equation is constructed using the lateral offset of a stationary target. Ideally, the lateral velocity of a stationary target should be 0. Due to the existence of Δψ, the observed lateral velocity offset is... The relationship between Δψ and the coordinate transformation deviation (small angle approximation: sinΔψ≈Δψ) is amplified as the vehicle moves forward, resulting in a systematic increase or decrease in the observed y-coordinate. The equation is as follows: Where v is the longitudinal velocity of the vehicle, To observe the noise, it is simplified to a standard linear form. ( (For design matrix).

[0045] For the pitch deviation angle Δθ, select an angle range that is sensitive to Δθ, such as the box directly in front, and extract the longitudinal position x of the stationary target and the vertical observation position within it.

[0046] The longitudinal displacement of the vehicle is expressed as , where v is the longitudinal speed of the vehicle (from the CAN bus). Let be the time interval between the i-th trajectory point and the previous frame. This represents the longitudinal distance traveled by the vehicle within the corresponding time period.

[0047] The longitudinal position of the target in the vehicle coordinate system is represented as follows: ,in The initial longitudinal position of the target (first frame of the short-time trajectory sequence). Let be the vertical position of the i-th trajectory point.

[0048] The systematic offset in the vertical direction is represented as: ,in This represents the vertical observation position (radar detection value) of the i-th trajectory point. For the ideal vertical position (radar installation height, a known constant). This is the systematic offset caused by Δθ. The vertical position of a stationary target under ideal conditions. It is a fixed value, that is Compared to Longitudinal difference between It should remain unchanged; when the height difference between the selected point and the radar horizontal plane is 0, then... =0, but in actual observations, due to beam tilt, the observed value is... It will exhibit a regular deviation as the vehicle moves, at different times. A matrix can reflect continuous time. The trend of positional change can reflect Δθ.

[0049] Similarly, using the small angle approximation condition sinΔθ≈Δθ, we get ,in Therefore A matrix with variables.

[0050] S204. Collect the observation equations for all angle intervals and solve them using a weighted least squares fitting algorithm to obtain a preliminary estimate of the installation deviation angle.

[0051] This step integrates the jointly constructed weighted observation equations for Δψ and Δθ into an overdetermined system of equations. Using a weighted least squares fitting algorithm, the combination of yaw angle (Δψ) and pitch angle (Δθ) that minimizes the sum of squared weighted residuals of all equations is solved. The final output is a preliminary estimate of the installation deviation angle without filtering or optimization. Observation equations for single angle intervals are susceptible to interference from local noise and missing targets. By aggregating equations from all bins into an overdetermined system, local errors can be offset through data redundancy, making the preliminary estimate more stable and avoiding misjudgments caused by data from a single bin.

[0052] For example, import the observation equation data for all bins and set the initial weights. (Determined by target distance, echo intensity, and trajectory stability), upper limit of iteration number (e.g., 20 times), and convergence threshold (e.g., 1e-6).

[0053] Calculate the residual of the equation in the current iteration, update the weights according to the rule that the smaller the residual, the greater the weight, and normalize them; First, solve for Δψ using bins that are sensitive to Δψ, then solve for Δθ using bins that are sensitive to Δθ, and finally optimize by combining all equations. If the changes in Δψ and Δθ in adjacent iterations are less than the convergence threshold, the iteration stops.

[0054] The final converged values ​​of Δψ and Δθ are used as preliminary estimates and passed to the subsequent filtering module.

[0055] In some implementations, refer to Figure 3 S204. The observation equations for all angle intervals are collected and solved using a weighted least squares fitting algorithm to obtain preliminary estimates of the installation deviation angle, including: S301. The weight is determined by the target distance, echo intensity, and trajectory stability. The closer the target is, the stronger the echo intensity and the more stable the trajectory, the greater the weight. Specifically, independent weighting coefficients are assigned to the observation equations for each angle interval, or the observation equations corresponding to stationary target trajectory points. The magnitude of the weights is determined by three core indicators: the closer the target is to the radar, the greater the weight; the stronger the target echo intensity, the greater the weight; and the more stable the target trajectory, the greater the weight. Finally, a set of weighting coefficients corresponding one-to-one with the observation equations is output for subsequent weighted least squares fitting solutions, thereby strengthening the contribution of high-precision data in the calculation, weakening the influence of low-quality data in the calculation, and improving the accuracy of the data.

[0056] For example, distance weight The weight is set according to the closer the distance, the greater the weight. The weight peaks at 1.0 for distances of 0.3~2m, decreases linearly to 0.5 for distances of 2~5m, and is set to 0.1 for distances >5m. echo intensity weight Based on the normalization processing of echo intensity values ​​(such as amplitude and signal-to-noise ratio) in the original radar detection data, the higher the intensity, the closer the weight is to 1.0, and the weight of echoes below the set threshold is set to below 0.2. Trajectory stability weights Based on the degree of fluctuation of the short-term trajectory sequence, the more trajectory points there are and the smaller the position fluctuation between adjacent frames (e.g., variance < threshold), the greater the weight; for example, if there are ≥10 trajectory points and the position variance <0.01m², the weight is set to 1.0; if there are <3 trajectory points or the fluctuation is large, the weight is set to 0.3 or less.

[0057] For a single observation equation, the normalized weights of the three indicators are weighted and fused according to equal weights or a preset ratio, as shown in the formula: (a, b, and c are weighting coefficients, satisfying a+b+c=1).

[0058] The final comprehensive weight Substitute the values ​​into the weighted least squares algorithm to construct the diagonal weight matrix. It works together with the observation equation and the design matrix to solve for the deviation angle.

[0059] S302. Assign corresponding weights to the observation equations for each angle interval and construct a weighted least squares optimization objective. Specifically, the observation equations for each angle interval are bound one-to-one with the corresponding weights determined in S301. This leads to the construction of a mathematical optimization model centered on minimizing the weighted sum of squared residuals from all observation equations. This ensures that high-quality observation equations—those with close proximity, strong echoes, and stable trajectories—dominate the optimization process, while mitigating the interference from low-quality observation equations, thus guaranteeing that the solution closely matches the actual installation deviation. Furthermore, the independent observation equations from different angle intervals are unified under a single optimization objective, avoiding inconsistencies caused by solving each bin's data separately, and leveraging data redundancy to enhance the robustness of the solution.

[0060] For example, each angle interval corresponds to one set of observation equations, and each set of equations is assigned a comprehensive weight calculated by S301. For example, the weights of the three target points P1, P2, and P3 are respectively... =0.9、 =1.0、 =0.7.

[0061] The weights of all observation equations are integrated into a diagonal matrix. = 。

[0062] according to ( The observation equation is used to calculate the residual. This reflects the deviation between observed values ​​and model predictions.

[0063] The optimization objective is to minimize the weighted sum of squared residuals of all equations. The mathematical expression is: ,in (Y=[0.075,0.080,-0.020]^T is the observation vector, and H=[-1.5,-1.5,-1.5]^T is the global design matrix).

[0064] S303. Solve the weighted least squares optimization objective to obtain preliminary estimates of the yaw and pitch installation error angles.

[0065] Specifically, based on the weighted least squares optimization objective constructed by S302, the optimal solution of the overdetermined equation system is obtained by solving the matrix operation, and the preliminary estimates of the yaw installation error angle Δψ and pitch installation error angle Δθ without filter wave optimization are obtained.

[0066] An example of a closed-form solution based on weighted least squares. First calculate the weighted product of the design matrix. Then find its inverse matrix, and finally combine it with... Multiplying them together yields the parameter estimates, where, This is a preliminary estimate of the yaw installation error angle to be determined. W is the design matrix, W is the weighted diagonal matrix, and Y is the observation vector, which in this example represents the systematic offset per unit time caused by installation error, i.e., the observed lateral velocity.

[0067] In some implementations, refer to Figure 4 S103. Process the installation deviation angle to generate the final compensation value, including: S401. The installation deviation angle is calculated using a dual-state recursive filter, which includes a dynamic state filter and a robust state filter. Specifically, based on the preliminary estimates of the installation deviation angles (Δψ, Δθ) from the bin-weighted least squares output, two independent parallel Kalman filters—a dynamic state filter and a robust state filter—are used for state prediction, Kalman gain calculation, and state updating, respectively. This yields real-time tracking of the dynamic state's rapid response to potential abrupt changes, as well as a robust state representing long-term stable deviations. This provides dual-path state data for subsequent confidence assessment and final compensation value generation. By processing these dual states in parallel, long-term, slowly varying installation deviations, such as bumper aging deformation, are decoupled from instantaneous impact interference, preventing single-state filters from misjudging interference due to excessive sensitivity or failing to detect true deformation due to insensitivity.

[0068] The dual-state recursive filter is essentially two Kalman filters with different parameter configurations, one maintaining the dynamic state and the other the robust state, and the two independently perform the prediction and update processes.

[0069] The core of dynamic state filters is to capture changes and respond quickly to possible sudden changes in installation angle while maintaining real-time performance. The processing flow fully follows the Kalman filter prediction, gain, and update logic, and the parameter configuration is biased towards sensitivity.

[0070] For example, the input to the dynamic state filter includes the value updated from the previous frame's dynamic state. and the corresponding error covariance Process noise covariance and measurement noise covariance Similarly, the input to the robust state filter includes the value updated from the robust state of the previous frame. and the corresponding error covariance Process noise covariance and measurement noise covariance .

[0071] Dynamic state prediction: This means that the state of the previous frame is used.

[0072] Dynamic covariance prediction: Accumulated process noise amplifies uncertainty.

[0073] Similarly: Robust condition prediction: .

[0074] Robust covariance prediction: .

[0075] Calculate the weight of the new measurement in the state update (Kalman gain): The value corresponding to the dynamic state filter ,because Larger Moderate, gain value Typically, the value is too large; the value corresponding to the robust state filter is... ,because Smaller Larger (preset), in a robust state, it hardly responds to new measurements.

[0076] Status Update: Dynamic state pre-update: Dynamic covariance update: Stable state pre-update: Steady-state covariance update: in, This represents the current frame observation of the binned weighted least squares output, such as current frame Δψ or current frame Δθ.

[0077] S402. Calculate the deviation between the installation deviation angle and the robust state, and determine whether the deviation is within the preset confidence interval; Specifically, the current installation deviation angle estimate is obtained from the dynamic state output of the two-state Kalman filter. Long-term stable installation deviation benchmark under robust conditions For the object, calculate the normalized deviation between the two, using the formula: in, , These are the estimation error covariances for dynamic and robust states, respectively, used to quantify state uncertainty.

[0078] Preset confidence interval coefficient (like =3, based on the statistical 3σ principle), if norm_diff < If the deviation is within the preset confidence interval, then the judgment is considered to be within the range specified. This judgment is essentially a statistical test. =3 corresponds to approximately 99.7% confidence level, ensuring that the bias is a reasonable random fluctuation rather than invalid interference.

[0079] S403. If the deviation is within the preset confidence interval and continues for a preset number of times, slowly update the robust state; otherwise, only update the dynamic state. Specifically, when norm_diff < If the normalization deviation is within the confidence interval, the system proceeds to the second judgment: whether the deviation state has lasted for a preset number of times. If so, it is determined to be an effective permanent deformation of the radar installation angle. The robust state is then updated slowly in small steps, for example, only 5% of the new dynamic state information is adopted each time to ensure a smooth transition of the compensation value. If any condition is not met, it is determined to be a transient interference. Only the dynamic state is updated to track the change, while the robust state remains unchanged to avoid system output oscillation.

[0080] S404. Based on the principle of prioritizing robust states and supplementing with dynamic states, the final compensation value is generated.

[0081] Specifically, clearly define a stable state. Weighting coefficients With dynamic state Weighting coefficients The final compensation value is generated by weighted summation. .

[0082] In some implementations, refer to Figure 5 S403. If the deviation is within the preset confidence interval and continues for a preset number of times, slowly update the robust state; otherwise, only update the dynamic state, including: S501. Determine the preset confidence interval based on the historical update data of the robust state and the system's allowable error range; Specifically, in a stable state Historical estimates and corresponding error covariance Based on historical update data, combined with the system's allowable error range (and measurement noise variance) Related, for example, 3 Determine the preset information interval.

[0083] Using the statistical 3σ principle, the confidence interval formula is: in This corresponds to a 99.7% confidence interval. This represents the system's allowable error variance.

[0084] S502. If the deviation between the new preliminary estimate and the robust state exceeds the preset confidence interval, it is judged as a random error or interference, and only the dynamic state is updated. Specifically, the absolute deviation between the new preliminary estimate and the robust state is calculated. If the value exceeds the confidence interval determined by S501, it is determined to be a random error, such as measurement noise or transient interference. The dynamic state and the corresponding error covariance are updated only according to the Kalman filter formula. The robust state remains unchanged, avoiding system output oscillation.

[0085] S503. If the deviation is within the preset confidence interval and the number of consecutive occurrences exceeds the threshold, it is determined to be a valid change in the installation deviation, and a slow update of the robust state is initiated.

[0086] Specifically, when the deviation is within the preset information interval and the number of consecutive occurrences of the deviation exceeds the threshold (e.g., 30 times for a radar frame rate of 10Hz, corresponding to 3 seconds), it is determined to be a valid change in radar installation deviation (e.g., permanent deformation caused by a minor collision). Initiating a robust state for slow updates, using a small step weighted update, ensures a smooth transition of the compensation value, balancing accuracy and stability.

[0087] In some implementations, refer to Figure 6 S104. Based on the final compensation value, perform real-time geometric correction on the original measurement coordinates of the ultrasonic radar sensor, and output the corrected target position information, including: S601. Obtain the raw polar coordinate data of the ultrasonic radar sensor, which includes the distance and the raw azimuth angle. Specifically, this step extracts polar coordinate data from the raw detection data of the ultrasonic radar sensor. The raw polar coordinate data contains two core parameters: the straight-line distance between the target and the radar. and the original azimuth angle This provides the basic input for subsequent calibration.

[0088] S602. Calculate the corrected azimuth angle based on the final compensation value; Specifically, this step is based on the final yaw error compensation value output by the two-state Kalman filter. For the original azimuth angle Make corrections, specifically as follows: ,in This is the corrected azimuth data.

[0089] S603. Correct the vertical distance projection based on the final compensation value; Based on the final pitch error compensation value Combined with radar theory installation angle Compensation will be provided, including It is usually 0°, that is, horizontal installation.

[0090] The corrected formula is: Thus, the original slant distance It is converted into the true horizontal projection distance in the vehicle's horizontal plane to eliminate vertical distance measurement distortion caused by pitch deviation.

[0091] S604. Output the corrected target position information to the upper-level application system.

[0092] Specifically, the corrected horizontal projection distance and the corrected azimuth angle As the new polar coordinate system, the position information can be directly output to the vehicle's upper-level application system, or converted into position information in the Cartesian coordinate system. The output is then used for calls from upper-level applications, including but not limited to automatic parking and blind spot monitoring, for obstacle location and path planning.

[0093] In some implementations, refer to Figure 7 S101. Establishing and maintaining short-time trajectory sequences from the raw detection data of the ultrasonic radar sensor includes: S701. Obtain the vehicle speed and yaw rate information, and perform compensation calculations on the target motion state in the original detection data; Specifically, the vehicle's longitudinal velocity v and yaw rate ω are collected in real time via the vehicle's CAN / FlexRay bus. Based on the vehicle's kinematic discrete-time model, the relative motion of the target in the original radar detection data is decomposed into the apparent motion caused by the vehicle's motion and the target's actual motion.

[0094] If the target is completely stationary, its position in the world coordinate system (ground) remains unchanged at times t and t+Δt. However, due to the vehicle's own motion (forward movement and turning), the observed position of this stationary target in the vehicle coordinate system V will change systematically. This change needs to be predicted.

[0095] This change needs to be predicted.

[0096] The change in vehicle pose from t to t+Δt can be described as follows: Based on rigid body kinematics, it can be deduced that for a point P stationary in the world coordinate system, its coordinates in the vehicle coordinate system change from time t to ( , Transformed to time t+Δt , Its prediction formula is: This process can be understood as follows: first, the target's position in the vehicle coordinate system (X-axis forward, Y-axis left) at time t. , And vehicle motion parameters longitudinal displacement and changes in heading angle A change in the vehicle's heading angle (turning) will cause the vehicle's coordinate system to "rotate," so the target coordinates at time t need to be rotated and corrected first. right Rotation: right Rotation: This corrects the visual offset of the target coordinates caused by the vehicle turning, ensuring that the coordinates after rotation reflect the target position when the vehicle is only translating.

[0097] After this, the vehicle's movement along the X-axis will cause the target's relative longitudinal position in the vehicle coordinate system to shift backward. Therefore, the longitudinal displacement needs to be subtracted from the rotated x-coordinate, resulting in the final predicted x-direction value being the rotated x-coordinate minus the longitudinal displacement. .

[0098] S702. Based on the compensated motion state, select stationary or quasi-stationary targets and eliminate moving target interference. Specifically, calculate the absolute velocity of the target after compensation. , , For residual motion, the calculation method is as follows: This refers to the inter-frame interval.

[0099] Set absolute speed threshold For example, 0.5 m / s or 1.0 m / s can be adjusted according to the radar noise level. If < If the target is identified as a stationary target, its trajectory points are input into the subsequent bin-weighted least squares process to estimate the radar installation error.

[0100] S703. Assign a unique identifier to each selected stationary or quasi-stationary target, continuously record its position data in the vehicle coordinate system, and form and maintain a short-term trajectory sequence.

[0101] Specifically, each selected stationary / quasi-stationary target is assigned a unique ID to ensure the continuity of target tracking between frames.

[0102] The target's position data in the vehicle coordinate system is continuously recorded at the radar frame rate to construct a dynamic sliding short-time trajectory sequence. The sequence meets the constraints of 2-5 seconds duration and uniform azimuth sampling.

[0103] In some implementations, refer to Figure 8 This application provides a radar position accuracy compensation system, the system comprising: The ultrasonic processing unit 10 is used to establish and maintain short-time trajectory sequences from the raw detection data of the ultrasonic radar sensor. The deviation angle calculation module 20 is capable of estimating the installation deviation angle of the ultrasonic radar sensor based on a short-time trajectory sequence. The dynamic compensation module 30 is used to process the installation deviation angle and generate the final compensation value; The data correction module 40 is used to perform real-time geometric correction on the original measurement coordinates of the ultrasonic radar sensor based on the final compensation value, and output the corrected target position information.

[0104] The modules are interconnected to support data flow. The specific data processing procedures and decision-making logic have been explained in detail and will not be repeated here.

[0105] In some embodiments, an electronic device is also included, comprising a processor, a memory, a communication interface, and one or more programs, the one or more programs being stored in the memory and configured to be executed by the processor, the programs including instructions for performing all or part of any of the methods described in the above embodiments.

[0106] This application also provides a computer-readable storage medium that stores a computer program for electronic data interchange, the computer program causing a computer to perform some or all of the steps of any of the methods described in the above method embodiments.

[0107] All software, hardware, and combined software and hardware implementation schemes based on the technical concept of this invention shall fall within the protection scope of this invention.

[0108] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be covered within the scope of protection of the present invention.

[0109] It should be noted that this application is not limited to the above-described embodiments. The above embodiments are merely examples, and any embodiments with the same structure and effect as the technical concept within the scope of this application are included in the technical scope of this application. Furthermore, various modifications that can be conceived by those skilled in the art to the embodiments, and other ways of constructing by combining some of the constituent elements of the embodiments, without departing from the spirit of this application, are also included in the scope of this application.

Claims

1. A radar position accuracy compensation method, characterized in that, include: Establish and maintain short-time trajectory sequences from the raw detection data of ultrasonic radar sensors; Based on the short-time trajectory sequence, the installation deviation angle of the ultrasonic radar sensor is estimated; The installation deviation angle is processed to generate a final compensation value; The original measurement coordinates of the ultrasonic radar sensor are geometrically corrected in real time based on the final compensation value, and the corrected target position information is output.

2. The radar position accuracy compensation method according to claim 1, characterized in that, The estimation of the installation deviation angle of the ultrasonic radar sensor based on the short-time trajectory sequence includes: The detection area of ​​the ultrasonic radar sensor is divided into several angular intervals; For each trajectory point within an angle interval, an ideal trajectory model is constructed based on the vehicle's own motion patterns, and the deviation between the observed trajectory and the ideal trajectory is compared. An observation equation regarding the installation deviation angle is constructed based on the systematic offset; The observation equations for all angle intervals are collected and solved using a weighted least squares fitting algorithm to obtain a preliminary estimate of the installation deviation angle.

3. The radar position accuracy compensation method according to claim 2, characterized in that, The observation equations, which aggregate all angle intervals, are solved using a weighted least squares fitting algorithm to obtain preliminary estimates of the installation deviation angle, including: The weights are determined by the target distance, echo intensity, and trajectory stability. The closer the target is, the stronger the echo intensity and the more stable the trajectory, the greater the weight. Assign corresponding weights to the observation equations for each angle interval, and construct a weighted least squares optimization objective. Solving the weighted least squares optimization objective yields preliminary estimates of the yaw and pitch installation error angles.

4. The radar position accuracy compensation method according to claim 1, characterized in that, The process of processing the installation deviation angle to generate the final compensation value includes: The installation deviation angle is calculated using a two-state recursive filter, which includes a dynamic state filter and a robust state filter. Calculate the deviation between the installation deviation angle and the robust state, and determine whether the deviation is within a preset confidence interval; If the deviation is within a preset confidence interval and continues for a preset number of times, the robust state is slowly updated; otherwise, only the dynamic state is updated. Based on the principle of prioritizing robust states and supplementing with dynamic states, the final compensation value is generated.

5. The radar position accuracy compensation method according to claim 4, characterized in that, If the deviation is within a preset confidence interval and continues for a preset number of times, the robust state is slowly updated. Otherwise, only update dynamic states including: A preset confidence interval is determined based on the historical update data of the robust state and the system's allowable error range; If the deviation between the new preliminary estimate and the robust state exceeds the preset confidence interval, it is judged as a random error or interference, and only the dynamic state is updated; If the deviation is within a preset confidence interval and the number of consecutive occurrences exceeds a threshold, it is determined to be a valid change in the installation deviation, and a slow update of the robust state is initiated.

6. The radar position accuracy compensation method according to claim 1, characterized in that, The step of performing real-time geometric correction on the original measurement coordinates of the ultrasonic radar sensor based on the final compensation value, and outputting the corrected target position information includes: Obtain the raw polar coordinate data of the ultrasonic radar sensor, the raw polar coordinate data including distance and raw azimuth angle; Calculate the corrected azimuth angle based on the final compensation value; The vertical distance projection is corrected based on the final compensation value; The corrected target location information is output to the upper-level application system.

7. The radar position accuracy compensation method according to claim 1, characterized in that, Establishing and maintaining short-time trajectory sequences from raw detection data of ultrasonic radar sensors includes: The vehicle speed and yaw rate information are obtained, and the target motion state in the original detection data is compensated for. Based on the compensated motion state, stationary or quasi-stationary targets are selected, and moving targets are eliminated to remove interference. Each selected stationary or quasi-stationary target is assigned a unique identifier, and its position data in the vehicle coordinate system is continuously recorded to form and maintain a short-term trajectory sequence.

8. A position accuracy compensation system for automotive ultrasonic radar, characterized in that, The system includes: An ultrasonic processing unit is used to establish and maintain short-time trajectory sequences from the raw detection data of ultrasonic radar sensors; The deviation angle calculation module is able to estimate the installation deviation angle of the ultrasonic radar sensor based on the short-time trajectory sequence; The dynamic compensation module is used to process the installation deviation angle and generate the final compensation value; The data correction module is used to perform real-time geometric correction on the original measurement coordinates of the ultrasonic radar sensor based on the final compensation value, and output the corrected target position information.

9. An electronic device, characterized in that, include: The processor, memory, communication interface, and one or more programs, said one or more programs being stored in the memory and configured to be executed by the processor, said programs including instructions for performing the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program for electronic data interchange, the computer program causing a computer to perform the method according to any one of claims 1 to 7.

Images