Vehicle direction calibration method and device, vehicle and storage medium

Abstract

The invention discloses a vehicle direction calibration method and device, a vehicle and a storage medium, and relates to the technical field of vehicles. The method comprises the following steps: acquiring an actual vehicle direction angle in a current acquisition period and sensor data, acquired by a plurality of vehicle sensors, in the current acquisition period; determining a corresponding sensor weight coefficient based on the error information of the current acquisition period of each vehicle sensor; performing fusion processing on the sensor data of the current acquisition period based on the sensor weight coefficient to determine a fusion vehicle direction angle of the current acquisition period; and determining a vehicle direction angle correction value based on a vehicle direction angle difference value between the fused vehicle direction angle of the current acquisition period and the actual vehicle direction angle of the current acquisition period, and calibrating the fused vehicle direction angle according to the vehicle direction angle correction value. In this way, the robustness, precision and scene adaptability of vehicle direction calibration are remarkably improved.

Description

Technical Field

[0001] This application relates to the field of vehicle technology, and in particular to a vehicle direction calibration method, a vehicle direction calibration device, a computer-readable storage medium, and a vehicle. Background Technology

[0002] Vehicle orientation calibration is one of the core functions of intelligent driving systems, and its accuracy directly affects the vehicle's path tracking, obstacle avoidance, and overall driving safety. Currently, traditional vehicle orientation calibration solutions mainly rely on a single sensor, such as an inertial measurement unit (IMU), a visual camera, or a global navigation satellite system (GNSS). However, single-sensor methods have significant drawbacks: inertial measurement units suffer from zero-bias drift, random noise, and integration errors, which accumulate over long periods or in high-dynamic scenarios; visual cameras are susceptible to interference from environmental factors such as changes in lighting, rain, fog, and low-light conditions at night, leading to feature matching failures or misidentification; and GNSS is prone to losing positioning in signal-blocked scenarios such as tunnels and densely built-up areas, and its civilian-grade accuracy typically only reaches the meter level, which is insufficient to meet the centimeter-level accuracy requirements of autonomous driving.

[0003] Related technologies improve calibration performance through dual-sensor fusion (such as the combination of inertial measurement unit and GNSS), but these solutions generally fail to fully address issues such as dynamic weight allocation between sensors, time synchronization errors, and effective suppression of abnormal data in complex scenarios. This leads to a significant decrease in the accuracy of orientation calibration under extreme conditions such as partial sensor failure and multi-source data conflict, and even misjudgment of orientation, which seriously affects the reliability and safety of the system. Summary of the Invention

[0004] This application aims to at least partially address one of the technical problems in related technologies. To this end, the first objective of this application is to propose a vehicle orientation calibration method. The method includes: acquiring the actual vehicle orientation angle in the current acquisition period and sensor data from multiple vehicle sensors collected in the current acquisition period; determining corresponding sensor weight coefficients based on the error information of each vehicle sensor in the current acquisition period; fusing the sensor data in the current acquisition period based on the sensor weight coefficients to determine the fused vehicle orientation angle in the current acquisition period; determining a vehicle orientation angle correction value based on the difference between the fused vehicle orientation angle and the actual vehicle orientation angle in the current acquisition period; and calibrating the fused vehicle orientation angle based on the vehicle orientation angle correction value. This application achieves adaptive optimal fusion of multi-source data by dynamically allocating weights through real-time evaluation of the error state of each sensor, and performs closed-loop correction of the fusion result by combining an external high-precision orientation reference, thereby significantly improving the robustness, accuracy, and scene adaptability of vehicle orientation calibration.

[0005] The second objective of this application is to provide a vehicle orientation calibration device.

[0006] The third objective of this application is to provide a computer-readable storage medium.

[0007] The fourth objective of this application is to propose a vehicle.

[0008] To achieve the above objectives, a first aspect of this application proposes a vehicle direction calibration method, the method comprising: acquiring the actual vehicle direction angle in the current acquisition period and sensor data acquired by multiple vehicle sensors in the current acquisition period; determining a corresponding sensor weighting coefficient based on the error information of each vehicle sensor in the current acquisition period; performing fusion processing on the sensor data in the current acquisition period based on the sensor weighting coefficient to determine the fused vehicle direction angle in the current acquisition period; determining a vehicle direction angle correction value based on the vehicle direction angle difference between the fused vehicle direction angle in the current acquisition period and the actual vehicle direction angle in the current acquisition period, and calibrating the fused vehicle direction angle according to the vehicle direction angle correction value.

[0009] According to one embodiment of this application, determining a vehicle direction angle correction value based on the difference between the fused vehicle direction angle and the actual vehicle direction angle in the current acquisition period includes: when the difference between the fused vehicle direction angle and the actual vehicle direction angle is greater than a preset vehicle direction angle difference threshold, determining the vehicle direction angle correction value based on a preset PD adjustment strategy according to the vehicle direction angle difference.

[0010] According to one embodiment of this application, sensor data in the current acquisition period is fused based on sensor weighting coefficients to determine the fused vehicle heading angle for the current acquisition period. This includes: acquiring the vehicle control command for the current acquisition period and the vehicle state vector for the previous acquisition period; calculating the vehicle state vector for the current acquisition period using a preset fusion algorithm based on the vehicle control command for the current acquisition period, the vehicle state vector for the previous acquisition period, the sensor data for the current acquisition period, and the corresponding sensor weighting coefficients; and determining the fused vehicle heading angle based on the vehicle state vector for the current acquisition period.

[0011] According to one embodiment of this application, the preset fusion algorithm includes an extended Kalman filter algorithm.

[0012] According to one embodiment of this application, determining the corresponding sensor weight coefficient based on the error information of the current acquisition period of each vehicle sensor includes: determining the confidence value of the current acquisition period of each vehicle sensor based on the error information of the current acquisition period of each vehicle sensor; and determining the sensor weight coefficient of the current acquisition period of each vehicle sensor based on the confidence value of the current acquisition period of each vehicle sensor and a preset adjustment factor.

[0013] According to one embodiment of this application, the vehicle sensor includes at least two of the following: an inertial measurement unit (IMU), a camera, a radar, and a wheel speed sensor. The error information for the current acquisition cycle of the IMU includes inertial zero bias; the error information for the current acquisition cycle of the camera includes the number of false detection points and the total number of detected feature points; the error information for the current acquisition cycle of the radar includes the number of iterations required for registration of radar point cloud data with preset map data; and the error information for the current acquisition cycle of the wheel speed sensor includes the measured vehicle speed and the wheel speed integral vehicle speed. The confidence level for the current acquisition cycle of each vehicle sensor is determined based on the error information for the current acquisition cycle of each vehicle sensor. The confidence values ​​include: determining the integral value of the inertial zero bias in the current acquisition period as the confidence value of the inertial measurement unit in the current acquisition period; determining the false detection rate of camera features in the current acquisition period as the confidence value of the camera in the current acquisition period, wherein the false detection rate of camera features is determined based on the ratio between the number of false detection points of camera features and the total number of detection points of camera features; determining the number of iterations required to register the radar point cloud data and the preset map data in the current acquisition period as the confidence value of the radar in the current acquisition period; and determining the wheel radius error in the current acquisition period as the confidence value of the wheel speed sensor, wherein the wheel radius error is determined based on the measured vehicle speed and the wheel speed integral vehicle speed.

[0014] According to one embodiment of this application, after acquiring sensor data for the current acquisition period collected by multiple vehicle sensors, the method further includes: preprocessing the sensor data for the current acquisition period; wherein the preprocessing includes one or more of time synchronization, spatial calibration, and noise filtering.

[0015] To achieve the above objectives, a second aspect of this application provides a vehicle direction calibration device, comprising: an acquisition module for acquiring the actual vehicle direction angle in the current acquisition period and sensor data collected by multiple vehicle sensors in the current acquisition period; a first determination module for determining a corresponding sensor weight coefficient based on the error information of each vehicle sensor in the current acquisition period; a second determination module for performing fusion processing on the sensor data in the current acquisition period based on the sensor weight coefficient to determine the fused vehicle direction angle in the current acquisition period; a third determination module for determining a vehicle direction angle correction value based on the vehicle direction angle difference between the fused vehicle direction angle in the current acquisition period and the actual vehicle direction angle in the current acquisition period; and a calibration module for calibrating the fused vehicle direction angle according to the vehicle direction angle correction value.

[0016] To achieve the above objectives, a third aspect of this application provides a computer-readable storage medium storing a vehicle direction calibration program, which, when executed by a processor, implements the aforementioned vehicle direction calibration method.

[0017] To achieve the above objectives, a fourth aspect of this application provides a vehicle including a memory, a processor, and a vehicle direction calibration program stored in the memory and executable on the processor. When the processor executes the vehicle direction calibration program, it implements the aforementioned vehicle direction calibration method.

[0018] According to the vehicle direction calibration method, apparatus, vehicle, and storage medium of this application, the actual vehicle direction angle of the current acquisition period and sensor data of the current acquisition period collected by multiple vehicle sensors are acquired; a corresponding sensor weight coefficient is determined based on the error information of each vehicle sensor in the current acquisition period; the sensor data of the current acquisition period is fused based on the sensor weight coefficient to determine the fused vehicle direction angle of the current acquisition period; a vehicle direction angle correction value is determined based on the difference between the fused vehicle direction angle of the current acquisition period and the actual vehicle direction angle of the current acquisition period, and the fused vehicle direction angle is calibrated based on the vehicle direction angle correction value. This application achieves adaptive optimal fusion of multi-source data by dynamically allocating weights through real-time evaluation of the error state of each sensor, and performs closed-loop correction of the fusion result by combining an external high-precision direction reference, thereby significantly improving the robustness, accuracy, and scene adaptability of vehicle direction calibration. Attached Figure Description

[0019] Figure 1 Here is a flowchart of a vehicle direction calibration method according to some embodiments of this application; Figure 2 Here is a flowchart of a vehicle direction calibration method according to other embodiments of this application; Figure 3 This is a block diagram of a vehicle direction calibration device according to some embodiments of this application; Figure 4 This is a block diagram of a vehicle according to some embodiments of this application. Detailed Implementation

[0020] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0021] The vehicle direction calibration method, apparatus, vehicle, and storage medium of this application are described in detail below with reference to the accompanying drawings.

[0022] Figure 1 This is a flowchart of a vehicle direction calibration method according to some embodiments of this application. (Refer to...) Figure 1The vehicle direction calibration method in this application embodiment may include the following steps: S110, acquire the actual vehicle heading angle for the current acquisition period and the sensor data for the current acquisition period collected by multiple vehicle sensors.

[0023] Specifically, the actual vehicle heading angle for the current data acquisition period can be obtained through high-precision map matching or RSU (Roadside Unit) acquisition. Multiple vehicle sensors can be at least two of the following: inertial measurement unit (IMU), camera, radar, and wheel speed sensor. The IMU is used to acquire the vehicle's angular velocity and acceleration for the current data acquisition period; the camera is used to acquire video data for the current data acquisition period; the radar is used to acquire radar point cloud data for the current data acquisition period; and the wheel speed sensor is used to acquire the vehicle's wheel speed for the current data acquisition period.

[0024] S120 determines the corresponding sensor weight coefficient based on the error information of the current acquisition cycle of each vehicle sensor.

[0025] Specifically, after collecting data from multiple sensors in the current acquisition period, it is necessary to determine the corresponding sensor weighting coefficient based on the error information of each vehicle sensor in the current acquisition period. Then, based on these weighting coefficients, the data from the multiple sensors in the current acquisition period are fused to obtain fused data. Finally, the vehicle's heading angle is estimated based on the fused data. For example, the weighting coefficient of each sensor in the current acquisition period can be determined by looking up a preset mapping table between sensor error information and sensor weighting coefficients in the current acquisition period. This preset mapping table includes multiple sensor error information and the corresponding sensor weighting coefficient for each sensor error information.

[0026] In other words, if a sensor exhibits significant anomalies in the current period (such as a sharp increase in GNSS positioning variance or a sudden drop in visual feature matching rate), its weight will be reduced; conversely, if the sensor data is highly consistent with the multi-source trend and its own noise is stable, its weight will be increased accordingly.

[0027] In this way, online real-time evaluation and weight allocation of sensor reliability are realized, ensuring that the fusion system can still prioritize state estimation based on the most reliable data source even when some sensors are degraded or there is environmental interference, thereby significantly improving the environmental adaptability and fault tolerance of the vehicle direction calibration system.

[0028] To further illustrate the above embodiments, in this application embodiment, determining the corresponding sensor weight coefficient based on the error information of the current acquisition cycle of each vehicle sensor includes: determining the confidence value of the current acquisition cycle of each vehicle sensor based on the error information of the current acquisition cycle of each vehicle sensor; and determining the sensor weight coefficient of the current acquisition cycle of each vehicle sensor based on the confidence value of the current acquisition cycle of each vehicle sensor and a preset adjustment factor. The preset adjustment factor can be calibrated according to actual conditions, and no specific limitations are imposed here.

[0029] Specifically, the error information of the current acquisition cycle of each vehicle sensor can be input into the corresponding confidence calculation formula to output the confidence value of each vehicle sensor for the current acquisition cycle. For example, the error information of the current acquisition cycle of the inertial measurement unit can be input into the corresponding confidence calculation formula to calculate the confidence of the inertial measurement unit for the current acquisition cycle; the error information of the current acquisition cycle of the camera can be input into the corresponding confidence calculation formula to calculate the confidence of the camera for the current acquisition cycle; the error information of the current acquisition cycle of the radar can be input into the corresponding confidence calculation formula to calculate the confidence of the radar for the current acquisition cycle; and the error information of the current acquisition cycle of the wheel speed sensor can be input into the corresponding confidence calculation formula to calculate the confidence of the wheel speed sensor for the current acquisition cycle.

[0030] Then, input the confidence value of each vehicle sensor for the current acquisition period into the following formula to calculate the sensor weight coefficient for each vehicle sensor for the current acquisition period: wi ( t )=exp( λ Ci ( t )) / ∑exp( λ Ci ( t )); in, wi ( t ) represents the sensor weighting coefficient for the current acquisition period of the i-th vehicle sensor; Ci ( t ) represents the confidence level of the current acquisition period of the i-th vehicle sensor; λ This indicates the preset adjustment factor.

[0031] To further illustrate the above embodiments, in this application embodiment, the vehicle sensor includes at least two of the following: an inertial measurement unit, a camera, a radar, and a wheel speed sensor. The error information for the current acquisition cycle of the inertial measurement unit includes inertial zero bias; the error information for the current acquisition cycle of the camera includes the number of false detection points and the total number of detected feature points; the error information for the current acquisition cycle of the radar includes the number of iterations required for registration of radar point cloud data with preset map data; and the error information for the current acquisition cycle of the wheel speed sensor includes the measured vehicle speed and the wheel speed integral vehicle speed. The current acquisition cycle of each vehicle sensor is determined based on the error information for its current acquisition cycle. The confidence value of the period includes: determining the integral value of the inertial zero bias of the current acquisition period as the confidence value of the current acquisition period of the inertial measurement unit; determining the false detection rate of camera features in the current acquisition period as the confidence value of the current acquisition period of the camera, wherein the false detection rate of camera features is determined based on the ratio between the number of false detection points of camera features and the total number of detection points of camera features; determining the number of iterations required to register the radar point cloud data and the preset map data in the current acquisition period as the confidence value of the current acquisition period of the radar; and determining the wheel radius error of the current acquisition period as the confidence value of the wheel speed sensor, wherein the wheel radius error is determined based on the measured vehicle speed and the wheel speed integral vehicle speed.

[0032] For example, the error information of the current acquisition cycle of the inertial measurement unit (IMU) includes inertial bias. Inertial bias is an inherent error source of gyroscopes and accelerometers that drifts slowly. Calculating the integral value of the inertial bias in the current acquisition cycle yields an estimate of the cumulative error of the angle or velocity. The larger this cumulative error value, the more severe the drift of the IMU, and the lower the reliability of its estimation results, thus the lower the confidence level. Therefore, the integral value of the inertial bias in the current acquisition cycle can be determined as the confidence level value of the IMU in the current acquisition cycle.

[0033] The error information for the current acquisition cycle of the camera includes the camera feature false detection rate, which is determined based on the ratio between the number of falsely detected camera features and the total number of detected camera features. The camera feature false detection rate directly reflects the reliability of the visual algorithm in the current environment. In rain, fog, backlight, or when features are scarce, the false detection rate will spike. The higher the false detection rate, the less reliable the output pose estimation result, and the lower the confidence level. Therefore, the camera feature false detection rate can be used as the confidence value for the current acquisition cycle of the camera.

[0034] The error information for the current radar acquisition cycle includes the number of iterations required to register radar point cloud data with preset map data. The number of iterations reflects the difficulty of matching the current radar observation with the prior map. Fewer iterations indicate smooth matching, good data consistency, and high confidence; more iterations indicate high point cloud noise, significant scene changes, or inaccurate motion estimation, leading to matching difficulties and low confidence. Therefore, the number of iterations required to register radar point cloud data with preset map data can be determined as the confidence value for the current radar acquisition cycle.

[0035] The error information in the current acquisition cycle of the wheel speed sensor includes the wheel radius error. This wheel radius error is determined based on the measured vehicle speed and the wheel speed integral. Specifically, wheel speed error = |measured vehicle speed - wheel speed integral| / measured vehicle speed. Tire pressure, wear, and slippage all cause changes in the effective rolling radius of the wheel. The larger the wheel radius error, the greater the error in the speed and displacement calculated from the wheel speed. Therefore, the radius error value directly determines the reliability of the wheel speed sensor data.

[0036] S130: Based on the sensor weighting coefficients, the sensor data of the current acquisition period is fused to determine the fused vehicle direction angle of the current acquisition period.

[0037] Specifically, each sensor, through its specific observation model, maps its own measurements (such as GPS position, inertial measurement unit angular velocity, visual lane line angle, and wheel speedometer mileage) to a contribution to the overall state vector. This is then weighted to integrate the heading angle measurements or related state observations from different sensors. Higher weights correspond to lower observation noise settings, meaning the algorithm assigns higher confidence to the sensor data when calculating the Kalman gain, resulting in a larger correction of the state vector by its observation residuals. Conversely, lower weights correspond to higher observation noise settings, significantly suppressing state corrections caused by the data in the update step. This achieves optimal fusion in the state space, with observations from highly reliable sensors dominating the convergence direction of state estimation, particularly for the calibration of the crucial vehicle heading angle, thus mathematically guaranteeing the optimality and environmental adaptability of the output.

[0038] In some embodiments, sensor data in the current acquisition period is fused based on sensor weighting coefficients to determine the fused vehicle heading angle for the current acquisition period. This includes: acquiring the vehicle control command for the current acquisition period and the vehicle state vector for the previous acquisition period; calculating the vehicle state vector for the current acquisition period using a preset fusion algorithm based on the vehicle control command for the current acquisition period, the vehicle state vector for the previous acquisition period, the sensor data for the current acquisition period, and the corresponding sensor weighting coefficients; and determining the fused vehicle heading angle based on the vehicle state vector for the current acquisition period.

[0039] In some embodiments, the preset fusion algorithm includes the extended Kalman filter algorithm.

[0040] Specifically, the vehicle state vector of the previous acquisition cycle is X_{k-1}, where X_{k-1} includes the vehicle pose (x, y, θ) and velocity ( v x , v y , ω The vehicle control command for the current acquisition period is u_k, which includes the steering wheel angle and accelerator / brake pedal opening. The sensor data for the current acquisition period is z_k, which includes observations from various sensors (e.g., GPS, inertial measurement unit, camera, wheel speedometer). The sensor weight coefficient corresponding to the sensor data for the current acquisition period is w_i, which represents the vehicle's real-time trust in different sensors. In the preset fusion algorithm, the observation noise covariance matrix R_k can be calculated by weighting the confidence levels of each sensor.

[0041] Specifically, using a vehicle kinematics model (such as the Ackermann steering model) and control command u_k, the state X_{k-1} from the previous moment is predicted to the current moment, resulting in the prior state estimate X_{k|k-1}, as shown in the following formula: X_{k|k-1}=f(X_{k-1},u_k)+w_k; Where f() is the state transition function (based on the vehicle kinematics model, such as the Ackermann steering model); w_k is the process noise.

[0042] Then, the predicted value X_{k|k-1} is corrected using the current sensor data z_k. The algorithm calculates the difference between the predicted and actual observed values, and then, based on the weights of each sensor, i.e., the R_k matrix, feeds the correction amount back into the state estimate in the optimal proportion (Kalman gain K_k), to obtain the posterior state estimate X_{k|k}, which is as follows: X_{k|k}=X_{k|k-1}+K_k*[z_k-h(X_{k|k-1})]; Where h() is the observation model.

[0043] It should be noted that the specific role of the weights is that if a sensor has a high weight (high confidence level), the corresponding R_k element value in the algorithm will be small, and the Kalman gain K_k will adopt the observation value of that sensor for correction to a greater extent.

[0044] The vehicle state vector for the current acquisition period is X_k, which is the final output of the fusion algorithm. It is the optimal and smoothest estimate of the vehicle's current state, combining model predictions and all weighted sensor observations. Furthermore, the heading angle component directly extracted from the current state vector X_k is determined as the fused vehicle heading angle.

[0045] S140, determine the vehicle heading angle correction value based on the difference between the fused vehicle heading angle in the current acquisition period and the actual vehicle heading angle in the current acquisition period, and calibrate the fused vehicle heading angle according to the vehicle heading angle correction value.

[0046] Specifically, after determining the fused vehicle heading angle for the current acquisition period, the vehicle heading angle difference between the fused vehicle heading angle for the current acquisition period and the actual vehicle heading angle for the current acquisition period is calculated. The vehicle heading angle correction value corresponding to the vehicle heading angle difference for the current acquisition period can be determined by looking up a preset relationship mapping table between the vehicle heading angle difference and the vehicle heading angle correction value. The preset relationship mapping table includes multiple vehicle heading angle differences and the vehicle heading angle correction value corresponding to each vehicle heading angle difference. Finally, the fused vehicle heading angle is calibrated based on the vehicle heading angle correction value.

[0047] To further illustrate the above embodiments, in this application embodiment, determining a vehicle direction angle correction value based on the difference between the fused vehicle direction angle and the actual vehicle direction angle in the current acquisition period includes: when the difference between the fused vehicle direction angle and the actual vehicle direction angle is greater than a preset vehicle direction angle difference threshold, determining the vehicle direction angle correction value based on a preset PD adjustment strategy. The preset vehicle direction angle difference threshold can be calibrated according to actual conditions; for example, the preset vehicle direction angle difference threshold can be 0.5°, and no specific limitation is made here.

[0048] Specifically, the vehicle steering angle difference can be compared with a preset vehicle steering angle difference threshold to determine whether a vehicle steering angle correction value needs to be determined based on a preset PD adjustment strategy. For example, if the vehicle steering angle difference is less than or equal to the preset vehicle steering angle difference threshold, then no calibration of the fused vehicle steering angle is required; if the vehicle steering angle difference is greater than the preset vehicle steering angle difference threshold, then the vehicle steering angle correction value is determined based on the preset PD adjustment strategy. Specifically, the vehicle steering angle correction value is determined based on the preset PD adjustment strategy as follows: δ = Kp Δ θ + Kd Δ θ ; in, δ Indicates the vehicle steering angle correction value; Kp This represents the proportional adjustment coefficient; Kd Denotes the differential adjustment coefficient; Δ θ This represents the difference in vehicle steering angle. Kp and Kd All can be determined based on the actual situation; no specific restrictions are imposed here.

[0049] This application achieves adaptive optimal fusion of multi-source data by dynamically allocating weights based on real-time evaluation of the error states of each sensor, and performs closed-loop correction on the fusion result using an external high-precision orientation reference, thereby significantly improving the robustness, accuracy, and scene adaptability of vehicle orientation calibration. It overcomes the inherent limitations of single sensors (such as zero-bias drift of inertial measurement units, environmental sensitivity of vision, and signal obstruction by GNSS), and solves the problem of failing to achieve dynamic adaptive allocation of sensor weights during the fusion process, making it difficult to effectively suppress orientation calibration errors under abnormal operating conditions such as multi-source data conflicts and partial sensor failures.

[0050] In some embodiments, after acquiring sensor data for the current acquisition period from multiple vehicle sensors, the sensor data for the current acquisition period is preprocessed; wherein, the preprocessing includes one or more of time synchronization, spatial calibration, and noise filtering.

[0051] Specifically, due to inherent differences in sampling mechanisms, physical characteristics, and data features among different sensors, the accuracy of vehicle direction calibration is affected. Therefore, it is necessary to acquire sensor data from multiple vehicle sensors for the current acquisition period and then preprocess the sensor data for the current acquisition period. This preprocessing includes one or more of the following: time synchronization, spatial calibration, and noise filtering.

[0052] Time synchronization: Based on a combination of hardware triggering (such as GPS PPS signal) and software interpolation (linear interpolation method), sensor data with different sampling frequencies are aligned to the same time axis (error ≤ 1ms).

[0053] Spatial calibration: The extrinsic parameters (rotation matrix R and translation vector T) of each sensor relative to the vehicle coordinate system are pre-calibrated using a calibration board (such as a checkerboard + mirror array), with the error controlled within ±0.1° rotation angle and ±0.5cm translation amount.

[0054] Noise filtering: Adaptive Kalman filtering is used to filter the inertial measurement unit data; visual data is filtered by bilateral filtering and optical flow method to suppress image noise, and sub-pixel positioning of feature points (such as lane lines and curbs) is combined to improve accuracy; voxel filtering is used to reduce point cloud density (from 100,000 points / frame to 20,000 points / frame), and radar data is filtered by outlier removal algorithm to remove dynamic interference (such as birds and falling objects).

[0055] As a concrete example, refer to Figure 2 The vehicle direction calibration method in this application embodiment may further include the following steps: Step 1: The hardware device layer includes an inertial measurement unit (IMU), a binocular vision camera, a radar (16-line LiDAR), and a Hall effect wheel speed sensor. The IMU is used to collect the vehicle's angular velocity and acceleration for the current acquisition period; the camera is used to collect video data for the current acquisition period; the radar is used to collect radar point cloud data for the current acquisition period; and the wheel speed sensor is used to collect the vehicle's wheel speed for the current acquisition period.

[0056] Step 2: Preprocess the sensor data for the current acquisition period. The preprocessing includes one or more of the following: time synchronization, spatial calibration, and noise filtering.

[0057] Step 3: Determine the confidence value of the current acquisition cycle for each vehicle sensor based on the error information of the current acquisition cycle for each vehicle sensor.

[0058] Step 4: Determine the sensor weight coefficient for each vehicle sensor in the current acquisition period based on the confidence value and preset adjustment factor. A schematic diagram of the dynamic weight allocation algorithm mathematical model is shown below. Figure 3 As shown.

[0059] Step 5: Based on the sensor weighting coefficients, the sensor data of the current acquisition period is fused using the extended Kalman filter algorithm to determine the fused vehicle heading angle of the current acquisition period.

[0060] Step 6: If the difference between the fused vehicle steering angle and the actual vehicle steering angle is greater than the preset vehicle steering angle difference threshold, the vehicle steering angle correction value is determined based on the preset PD adjustment strategy and the EPS electric power steering system is used to calibrate the fused vehicle steering angle based on the vehicle steering angle difference.

[0061] As another specific example, the orientation calibration process in a high-speed, sunny day scenario is as follows: Scenario Description: A vehicle is traveling at 100 km / h on a highway with ample, unobstructed lighting. The directional error caused by the zero-bias drift of the inertial measurement unit needs to be calibrated.

[0062] Implementation steps: Step 1 Sensor data acquisition: The inertial measurement unit outputs angular rate (0.1° / s) and acceleration (0.5g); the visual camera detects a clear lane line (lateral offset 0cm); the matching error between the LiDAR point cloud and the high-precision map is 0.8cm; the wheel speed sensor measures the vehicle speed as 100.2km / h (integral error 0.1km / h).

[0063] Step 2: Data preprocessing: After AKF filtering, the angular rate noise of the inertial measurement unit data is reduced from ±0.5° / s to ±0.1° / s; the visual data is filtered by both sides to remove specular interference, and the lane line detection confidence is 95%; the LiDAR point cloud is filtered by voxels and retains 20,000 points with a matching error of 0.8cm; the wheel speed data is compared with the integrated vehicle speed of the inertial measurement unit to correct the wheel speed error.

[0064] Step 3: Error Modeling and Weight Allocation: The cumulative error of the inertial measurement unit (IMU) zero-bias drift is 0.05° (CIMU = 0.05), the visual false detection rate is 0% (CVision = 0), the LiDAR matching error is 0.8 cm (CLiDAR = 0.2), and the wheel speed integral error is 0.1 km / h (CWheel Speed = 0.1). Dynamic weight calculation yields: wIMU = 0.35, wVision = 0.4, wLiDAR = 0.2, wWheel Speed = 0.05.

[0065] Step 4: Multi-source fusion: The extended Kalman filter fuses the four types of data, and the output fusion direction angle θ = 120.05° (true value is 120.00°), with an error of 0.05°.

[0066] Step 5 Orientation calibration: If the error is less than the threshold (0.5°), no correction is needed for the steering actuator; update the zero bias parameter of the inertial measurement unit (correct the inertial zero bias from 0.01° / s to 0.005° / s).

[0067] As another concrete example, the orientation calibration process in a rainy tunnel scenario is as follows: Scenario description: The vehicle enters the tunnel at 60km / h, loses GPS signal, and the visual camera is blurred due to rain and fog (feature point matching success rate is 70%). It is necessary to rely on LiDAR and inertial measurement unit to calibrate the direction.

[0068] Implementation steps: Step 1: Sensor data acquisition: The inertial measurement unit outputs angular rate (0.3° / s) and acceleration (0.8g); the visual camera detects a blurred lane line (lateral offset ±10cm); the LiDAR point cloud and tunnel wall map matching error is 1.2cm; the wheel speed sensor measures the vehicle speed as 58.5km / h (integral error 1.5km / h).

[0069] Step 2 Data Preprocessing: After AKF filtering, the angular rate noise of the inertial measurement unit data is reduced to ±0.2° / s; the visual data is filtered by bilateral filtering to remove raindrop interference, and the lane line detection confidence is 60%; the LiDAR point cloud retains 18,000 points after outlier removal, with a matching error of 1.2cm; the wheel speed data has increased error due to water accumulation, and is corrected by integrating the vehicle speed through the inertial measurement unit (corrected vehicle speed is 60.2km / h).

[0070] Step 3: Error Modeling and Weight Allocation: The cumulative error of the inertial measurement unit (IMU) zero-bias drift is 0.3° (CIMU = 0.3), the visual false detection rate is 30% (CVision = 0.7), the LiDAR matching error is 1.2 cm (CLiDAR = 0.6), and the wheel speed integral error is 1.5 km / h (CWheel Speed = 0.9). Dynamic weight calculation yields: wIMU = 0.25, wVision = 0.1, wLiDAR = 0.4, wWheel Speed = 0.25 (LiDAR weight is increased to compensate for visual failure).

[0071] Step 4 Multi-source fusion: The extended Kalman filter is mainly composed of LiDAR (40%) and inertial measurement unit (25%), and the output fusion direction angle θ fusion = 45.12° (true value is 45.00°) with an error of 0.12°.

[0072] Step 5: Orientation calibration: The error is less than the threshold, no correction is required; update the LiDAR environmental compensation model (temperature 25℃→26℃, ranging error correction +0.02%).

[0073] Corresponding to the above embodiments, this application also proposes a vehicle direction calibration device.

[0074] Reference Figure 3 The vehicle direction calibration device 200 includes: an acquisition module 210, a first determination module 220, a second determination module 230, a third determination module 240, and a calibration module 250.

[0075] The acquisition module 210 acquires the actual vehicle heading angle for the current acquisition period and sensor data from multiple vehicle sensors for the current acquisition period. The first determination module 220 determines the corresponding sensor weighting coefficient based on the error information of each vehicle sensor for the current acquisition period. The second determination module 230 performs fusion processing on the sensor data for the current acquisition period based on the sensor weighting coefficients to determine the fused vehicle heading angle for the current acquisition period. The third determination module 240 determines a vehicle heading angle correction value based on the difference between the fused vehicle heading angle and the actual vehicle heading angle for the current acquisition period. The calibration module 250 calibrates the fused vehicle heading angle according to the vehicle heading angle correction value.

[0076] According to one embodiment of this application, the third determining module 240 is specifically used to determine a vehicle steering angle correction value based on a preset PD adjustment strategy when the difference between the fused vehicle steering angle and the actual vehicle steering angle is greater than a preset vehicle steering angle difference threshold.

[0077] According to one embodiment of this application, the second determining module 230 is specifically used to: acquire the vehicle control command of the current acquisition cycle and the vehicle state vector of the previous acquisition cycle; calculate the vehicle state vector of the current acquisition cycle using a preset fusion algorithm based on the vehicle control command of the current acquisition cycle, the vehicle state vector of the previous acquisition cycle, the sensor data of the current acquisition cycle and the corresponding sensor weight coefficients; and determine the fused vehicle direction angle based on the vehicle state vector of the current acquisition cycle.

[0078] According to one embodiment of this application, the preset fusion algorithm includes an extended Kalman filter algorithm.

[0079] According to one embodiment of this application, the first determining module 220 is specifically used to: determine the confidence value of the current acquisition period of each vehicle sensor based on the error information of the current acquisition period of each vehicle sensor; and determine the sensor weight coefficient of the current acquisition period of each vehicle sensor based on the confidence value of the current acquisition period of each vehicle sensor and a preset adjustment factor.

[0080] According to one embodiment of this application, the vehicle sensor includes at least two of an inertial measurement unit, a camera, a radar, and a wheel speed sensor. The error information for the current acquisition cycle of the inertial measurement unit includes inertial zero bias; the error information for the current acquisition cycle of the camera includes the number of false detection points of camera features and the total number of detected camera features; the error information for the current acquisition cycle of the radar includes the number of iterations required to register radar point cloud data with preset map data; and the error information for the current acquisition cycle of the wheel speed sensor includes the measured vehicle speed and the wheel speed integral vehicle speed. The first determining module 220 is specifically used to determine the inertial bias of the current acquisition cycle. The integral value of zero bias is determined as the confidence value of the current acquisition cycle of the inertial measurement unit; the false detection rate of camera features in the current acquisition cycle is determined as the confidence value of the current acquisition cycle of the camera, wherein the false detection rate of camera features is determined based on the ratio between the number of false detection points of camera features and the total number of detection points of camera features; the number of iterations required to register the radar point cloud data and the preset map data in the current acquisition cycle is determined as the confidence value of the current acquisition cycle of the radar; the wheel radius error in the current acquisition cycle is determined as the confidence value of the wheel speed sensor, wherein the wheel radius error is determined based on the measured vehicle speed and the wheel speed integral vehicle speed.

[0081] According to one embodiment of this application, after acquiring sensor data for the current acquisition period from multiple vehicle sensors, the sensor data for the current acquisition period is preprocessed; wherein, the preprocessing includes one or more of time synchronization, spatial calibration and noise filtering.

[0082] Corresponding to the above embodiments, this application also proposes a computer-readable storage medium.

[0083] The present application provides a computer-readable storage medium storing a vehicle direction calibration program, which, when executed by a processor, implements the aforementioned vehicle direction calibration method.

[0084] It should be noted that the above explanation of the embodiments and beneficial effects of the vehicle direction calibration method also applies to the computer-readable storage medium of the embodiments of this application. To avoid redundancy, it will not be elaborated in detail here.

[0085] Corresponding to the above embodiments, this application also proposes a vehicle.

[0086] See Figure 4 As shown, the vehicle 300 of this application includes a memory 310, a processor 320, and a vehicle direction calibration program stored in the memory 310 and executable on the processor 320. When the processor executes the vehicle direction calibration program, it implements the aforementioned vehicle direction calibration method.

[0087] It should be noted that the above explanation of the embodiments and beneficial effects of the vehicle direction calibration method also applies to the vehicles in the embodiments of this application, and will not be elaborated in detail here to avoid redundancy.

[0088] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0089] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0090] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0091] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0092] In this application, unless otherwise expressly specified and limited, the terms "installation," "connection," "joining," and "fixing," etc., 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, unless otherwise expressly limited. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0093] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A vehicle direction calibration method, characterized in that, The method includes: Acquire the actual vehicle heading angle for the current acquisition period and the sensor data for the current acquisition period collected by multiple vehicle sensors; The corresponding sensor weighting coefficient is determined based on the error information of the current acquisition cycle of each of the vehicle sensors. The sensor data of the current acquisition period are fused based on the sensor weighting coefficients to determine the fused vehicle direction angle of the current acquisition period. A vehicle heading angle correction value is determined based on the difference between the fused vehicle heading angle in the current acquisition period and the actual vehicle heading angle in the current acquisition period, and the fused vehicle heading angle is calibrated based on the vehicle heading angle correction value.

2. The vehicle direction calibration method according to claim 1, characterized in that, The determination of the vehicle heading angle correction value based on the difference between the fused vehicle heading angle of the current acquisition period and the actual vehicle heading angle of the current acquisition period includes: If the difference between the fused vehicle heading angle and the actual vehicle heading angle is greater than a preset vehicle heading angle difference threshold, a vehicle heading angle correction value is determined based on the vehicle heading angle difference according to a preset PD adjustment strategy.

3. The vehicle direction calibration method according to claim 1, characterized in that, The process of fusing sensor data for the current acquisition period based on the sensor weighting coefficients to determine the fused vehicle direction angle for the current acquisition period includes: Obtain the vehicle control command for the current acquisition cycle and the vehicle state vector for the previous acquisition cycle; Based on the vehicle control command of the current acquisition cycle, the vehicle state vector of the previous acquisition cycle, the sensor data of the current acquisition cycle, and the corresponding sensor weight coefficients, a preset fusion algorithm is used to calculate the vehicle state vector of the current acquisition cycle. The fused vehicle orientation angle is determined based on the vehicle state vector of the current acquisition period.

4. The vehicle direction calibration method according to claim 3, characterized in that, The preset fusion algorithm includes the extended Kalman filter algorithm.

5. The vehicle direction calibration method according to claim 1, characterized in that, The step of determining the corresponding sensor weight coefficient based on the error information of the current acquisition period of each vehicle sensor includes: The confidence value of the current acquisition period of each vehicle sensor is determined based on the error information of the current acquisition period of each vehicle sensor. The sensor weight coefficient for the current acquisition period of each vehicle sensor is determined based on the confidence value of the current acquisition period of each vehicle sensor and a preset adjustment factor.

6. The vehicle direction calibration method according to claim 5, characterized in that, The vehicle sensors include at least two of the following: an inertial measurement unit (IMU), a camera, a radar, and a wheel speed sensor. The error information for the current acquisition cycle of the IMU includes inertial zero bias; the error information for the current acquisition cycle of the camera includes the number of false detection points and the total number of detected feature points; the error information for the current acquisition cycle of the radar includes the number of iterations required for registration of radar point cloud data with preset map data; and the error information for the current acquisition cycle of the wheel speed sensor includes the measured vehicle speed and the wheel speed integral vehicle speed. Determining the confidence value for the current acquisition cycle of each vehicle sensor based on the error information for the current acquisition cycle of each vehicle sensor includes: The integral value of the inertial zero bias in the current acquisition cycle is determined as the confidence value of the inertial measurement unit in the current acquisition cycle. The false detection rate of camera features in the current acquisition period is determined as the confidence value of the camera in the current acquisition period, wherein the false detection rate of camera features is determined based on the ratio between the number of false detection points of camera features and the total number of detection points of camera features; The number of iterations required to register the radar point cloud data of the current acquisition period with the preset map data is determined as the confidence value of the radar for the current acquisition period. The wheel radius error of the current acquisition period is determined as the confidence value of the wheel speed sensor, wherein the wheel radius error is determined based on the measured vehicle speed and the wheel speed integral vehicle speed.

7. The vehicle direction calibration method according to claim 1, characterized in that, After acquiring sensor data from multiple vehicle sensors for the current acquisition period, the method further includes: Preprocess the sensor data for the current acquisition period; The preprocessing includes one or more of time synchronization, spatial calibration, and noise filtering.

8. A vehicle direction calibration device, characterized in that, The device includes: The acquisition module is used to acquire the actual vehicle heading angle in the current acquisition period and the sensor data collected by multiple vehicle sensors in the current acquisition period. The first determining module is used to determine the corresponding sensor weighting coefficient based on the error information of the current acquisition cycle of each vehicle sensor; The second determining module is used to perform fusion processing on the sensor data of the current acquisition period based on the sensor weighting coefficient, so as to determine the fused vehicle direction angle of the current acquisition period; The third determining module is used to determine a vehicle heading angle correction value based on the vehicle heading angle difference between the fused vehicle heading angle of the current acquisition period and the actual vehicle heading angle of the current acquisition period. The calibration module is used to calibrate the fused vehicle heading angle based on the vehicle heading angle correction value.

9. A computer-readable storage medium, characterized in that, It stores a vehicle direction calibration program, which, when executed by a processor, implements the vehicle direction calibration method according to any one of claims 1-7.

10. A vehicle, characterized in that, The system includes a memory, a processor, and a vehicle direction calibration program stored in the memory and executable on the processor. When the processor executes the vehicle direction calibration program, it implements the vehicle direction calibration method according to any one of claims 1-7.

Images