High-precision perception-driven automatic driving positioning method and system

Abstract

The invention provides a high-precision perception-driven automatic driving positioning method and system, and relates to the technical field of automatic driving, and the method comprises the steps: collecting vehicle environment data in real time through a multi-mode sensor array, and generating a multi-source perception data set; performing time-space synchronization processing; performing motion distortion correction on the fused sensing data stream to generate dynamic compensation sensing data; inputting a multi-source fusion positioning model, and performing three-dimensional space matching in combination with high-precision map data to obtain real-time positioning coordinates; and performing iterative optimization on the real-time positioning coordinates through an adaptive error correction module to generate a vehicle positioning result. According to the method and the device, the technical problem that the positioning precision is affected due to data distortion caused by the influence of the vehicle motion on the sensor data is solved, the distortion caused by the vehicle motion is eliminated through motion distortion correction and compensation, and the positioning precision and stability are improved.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, and particularly to an autonomous driving positioning method and system driven by high-precision perception. Background Art

[0002] Autonomous driving technology is one of the core technologies for realizing autonomous vehicle driving, and high-precision positioning is a key link to ensure the safe and efficient operation of the autonomous driving system. Autonomous driving positioning technology needs to accurately determine the position of the vehicle based on the perception information of the vehicle's surrounding environment, combined with maps and motion models. The level of positioning accuracy directly affects the decision-making and control of the autonomous driving system, and thus determines the safety and reliability of autonomous driving. During the driving process of the vehicle, especially in situations such as high-speed driving, sudden braking, and sharp turning, the data collected by the sensors may be distorted due to the vehicle's motion state, resulting in data distortion. Due to the data distortion problem, the autonomous driving system may deviate during positioning. Especially during a long driving process, the errors will accumulate continuously, ultimately affecting the overall positioning accuracy and stability, and thus affecting driving safety.

[0003] In summary, there is a technical problem in the prior art that the sensor data is affected by the vehicle's motion, resulting in data distortion, which affects the positioning accuracy. Summary of the Invention

[0004] The purpose of this application is to provide an autonomous driving positioning method and system driven by high-precision perception, so as to solve the technical problem in the prior art that the sensor data is affected by the vehicle's motion, resulting in data distortion, which affects the positioning accuracy.

[0005] In view of the above problems, this application provides an autonomous driving positioning method and system driven by high-precision perception.

[0006] In the first aspect, this application provides an autonomous driving positioning method driven by high-precision perception. The autonomous driving positioning method driven by high-precision perception is implemented through an autonomous driving positioning system driven by high-precision perception. Among them, the autonomous driving positioning method driven by high-precision perception includes: real-time collecting vehicle environment data through a multi-modal sensor array to generate a multi-source perception data set; performing spatio-temporal synchronization processing on the multi-source perception data set to establish a fusion perception data stream with a unified spatio-temporal reference; performing motion distortion correction on the fusion perception data stream to generate dynamically compensated perception data; inputting the dynamically compensated perception data into a multi-source fusion positioning model, and performing three-dimensional space matching in combination with high-precision map data to obtain real-time positioning coordinates; and iteratively optimizing the real-time positioning coordinates through an adaptive error correction module to generate a vehicle positioning result.

[0007] Optionally, configure the lidar to scan the target area at a preset scanning frequency to obtain three-dimensional point cloud data; trigger the vision sensor to collect images according to the three-dimensional point cloud data to obtain multiple regional image data; control the millimeter-wave radar to sense the target area in a multi-target tracking mode to obtain the velocity vector information of moving objects; synchronize and align the time stamps of the three-dimensional point cloud data, the multiple regional image data, and the velocity vector information to generate the multi-source perception dataset.

[0008] Optionally, establish a transformation matrix between the coordinate systems of the multi-modal sensor array, map the lidar coordinate system to the vehicle body coordinate system according to the transformation matrix to obtain the first perception dataset; perform dynamic reprojection on the multiple regional image data of the vision sensor through a sliding window mechanism to obtain the second perception dataset; perform interpolation compensation on the asynchronous sampling data extracted based on the millimeter-wave radar to obtain the third perception dataset; construct spatio-temporal consistency constraint conditions to verify the geometric consistency of the first perception dataset, the second perception dataset, and the third perception dataset in three-dimensional space. If the geometric data in three-dimensional space is consistent, generate the multi-source perception dataset.

[0009] Optionally, obtain the angular velocity data and linear acceleration data of the target vehicle through inertial measurement of the target vehicle; perform motion compensation based on the angular velocity data and the linear acceleration data to obtain the vehicle pose change amount; perform inverse motion compensation on the three-dimensional point cloud data combined with the vehicle pose change amount, and denoise the point cloud trajectory of the dynamic obstacle to generate the dynamically compensated perception data.

[0010] Optionally, extract lidar geometric features based on the three-dimensional point cloud data, and extract visual semantic features based on the multiple regional image data; construct a deep neural network model, and perform cross-modal feature fusion on the visual semantic features and the lidar geometric features through the deep neural network model to obtain visual-lidar data; perform motion state estimation on the velocity vector information of the moving object according to the extended Kalman filter algorithm to obtain the motion state estimation result; introduce an attention mechanism to dynamically adjust the weight distribution of the visual-lidar data and the motion state estimation result to construct the multi-source fusion positioning model.

[0011] Optionally, extract the road topology in the high-precision map and evaluate the matching degree with the real-time positioning coordinates, and obtain the matching degree evaluation result; perform particle filtering according to the matching degree evaluation result to obtain the particle filtering result, dynamically adjust the real-time positioning coordinates according to the particle filtering result to obtain the particle distribution density; obtain the historical positioning trajectory of the target vehicle, and perform batch non-linear optimization on the historical positioning trajectory through a sliding window to obtain the positioning optimization trajectory data; calculate the error by combining the positioning optimization trajectory data with the particle distribution density to obtain the positioning error data. When the positioning error exceeds the preset threshold, trigger the positioning mode to reset the error and determine the vehicle positioning result.

[0012] Optionally, monitor the matching degree between the real-time positioning coordinates output by the multi-source fusion positioning model and the high-precision map in real time, determine the positioning error data according to the matching degree, and judge whether the positioning error data exceeds the preset threshold; when the positioning error data exceeds the preset threshold, generate an error trigger signal, start the multi-source data consistency verification based on the error trigger signal, and select the trigger positioning mode according to the verification result. The positioning mode includes a tightly coupled positioning mode or a redundant positioning mode; perform error correction based on the tightly coupled positioning mode or the redundant positioning mode, and generate the vehicle positioning result according to the correction result in combination with the spatial feature constraint relationship between the real-time positioning coordinates and the high-precision map.

[0013] In a second aspect, the present application also provides a high-precision perception-driven autonomous driving positioning system for performing the high-precision perception-driven autonomous driving positioning method as described in the first aspect. The high-precision perception-driven autonomous driving positioning system includes: a data acquisition module for real-time collecting vehicle environment data through a multi-modal sensor array to generate a multi-source perception data set; a spatio-temporal synchronization processing module for performing spatio-temporal synchronization processing on the multi-source perception data set to establish a fusion perception data stream with a unified spatio-temporal reference; a motion distortion correction module for performing motion distortion correction on the fusion perception data stream to generate dynamically compensated perception data; a spatial matching module for inputting the dynamically compensated perception data into a multi-source fusion positioning model and performing three-dimensional spatial matching in combination with high-precision map data to obtain real-time positioning coordinates; a vehicle positioning module for iteratively optimizing the real-time positioning coordinates through an adaptive error correction module to generate a vehicle positioning result.

[0014] One or more technical solutions provided in the present application have at least the following beneficial effects:

[0015] Collect vehicle environment data in real time through a multi-modal sensor array to generate a multi-source perception data set; perform spatio-temporal synchronization processing on the multi-source perception data set to establish a fused perception data stream with a unified spatio-temporal reference; perform motion distortion correction on the fused perception data stream to generate dynamically compensated perception data; input the dynamically compensated perception data into a multi-source fusion positioning model, and perform three-dimensional space matching in combination with high-precision map data to obtain real-time positioning coordinates; perform iterative optimization on the real-time positioning coordinates through an adaptive error correction module to generate a vehicle positioning result. That is to say, by performing spatio-temporal synchronization processing on all sensor data, then performing motion distortion correction, eliminating the distortion caused by vehicle motion, determining the positioning coordinates in combination with a high-precision map, and continuously optimizing the positioning result using an adaptive error correction module, the accuracy and stability of positioning are improved, thereby improving the safety and reliability of autonomous driving.

[0016] The above description is only an overview of the technical solution of the present application. In order to be able to understand the technical means of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present application more obvious and understandable, the following specifically illustrates the specific embodiments of the present application. It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present application, nor is it used to limit the scope of the present application. Other features of the present application will become easily understood through the following description. Brief Description of the Drawings

[0017] In order to more clearly illustrate the technical solutions in the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings described below are only exemplary, and for those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0018] Figure 1 It is a flowchart of the autonomous driving positioning method driven by high-precision perception of the present application;

[0019] Figure 2 It is a structural diagram of the autonomous driving positioning system driven by high-precision perception of the present application.

[0020] Explanation of the reference numerals in the drawings: data acquisition module 11, spatio-temporal synchronization processing module 12, motion distortion correction module 13, space matching module 14, vehicle positioning module 15. Detailed Description of the Embodiments

[0021] By providing a high-precision perception-driven autonomous driving positioning method and system, the present application solves the technical problem in the prior art that sensor data is affected by vehicle movement, resulting in data distortion and thus affecting positioning accuracy. By performing spatio-temporal synchronization processing on all sensor data, then performing motion distortion correction to eliminate the distortion caused by vehicle movement, determining the positioning coordinates in combination with a high-precision map, and using an adaptive error correction module to continuously optimize the positioning result, the positioning accuracy and stability are improved, thereby enhancing the safety and reliability of autonomous driving.

[0022] Next, the technical solutions in the present application will be described clearly and completely with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application. Additionally, it should be noted that for the sake of description, only the parts related to the present application are shown in the accompanying drawings rather than all of them.

[0023] Embodiment 1. Please refer to the attached Figure 1 , the present application provides a high-precision perception-driven autonomous driving positioning method. Among them, the high-precision perception-driven autonomous driving positioning method is executed by a high-precision perception-driven autonomous driving positioning system. The high-precision perception-driven autonomous driving positioning method specifically includes the following steps:

[0024] S100: Real-time collect vehicle environment data through a multi-modal sensor array to generate a multi-source perception data set.

[0025] Specifically, a vehicle is equipped with a multi-modal sensor array composed of a lidar, a vision sensor, and a millimeter-wave radar. These sensors have their own focuses. The lidar is used to collect accurate three-dimensional point cloud data, the vision sensor captures high-resolution image information, and the millimeter-wave radar performs stably in bad weather such as rain, snow, and fog. During the movement of the vehicle, the multi-modal sensor array continuously collects vehicle environment data. While collecting data, each sensor attaches a timestamp to the data it collects. The vehicle environment data includes information such as the roads, obstacles, pedestrians, and traffic signs around the vehicle.

[0026] Due to the different operating frequencies and response times of different sensors, it is necessary to align the data collected by each sensor through timestamp synchronization. By means of a unified clock signal, it is ensured that all sensor data have a common time reference. Software algorithms are used to interpolate and align data with different clock sampling rates to ensure that there are data from lidar, cameras, and millimeter-wave radars at each moment. Through the multi-modal sensor array and precise timestamp synchronization and alignment, a high-quality multi-source perception dataset is effectively generated, improving the accuracy and robustness of environmental perception.

[0027] Further, the S100 of this application includes:

[0028] Configure the lidar to scan the target area at a preset scanning frequency to obtain three-dimensional point cloud data; trigger the vision sensor to perform image acquisition according to the three-dimensional point cloud data to obtain multiple regional image data; control the millimeter-wave radar to sense the target area in a multi-target tracking mode to obtain the velocity vector information of moving objects; perform timestamp synchronization and alignment on the three-dimensional point cloud data, the multiple regional image data, and the velocity vector information to generate the multi-source perception dataset.

[0029] Establish a transformation matrix between the coordinate systems of the multi-modal sensor array, map the lidar coordinate system to the vehicle body coordinate system according to the transformation matrix to obtain the first perception dataset; perform dynamic reprojection on the multiple regional image data of the vision sensor through a sliding window mechanism to obtain the second perception dataset; perform interpolation compensation on the asynchronous sampling data extracted based on the millimeter-wave radar to obtain the third perception dataset; construct spatio-temporal consistency constraint conditions to verify the geometric consistency of the first perception dataset, the second perception dataset, and the third perception dataset in three-dimensional space. If the geometric data in three-dimensional space are consistent, then generate the multi-source perception dataset.

[0030] Specifically, first configure the lidar, set the scanning frequency (such as 10 scans per second). The lidar emits laser pulses at the set frequency and records the time difference of the reflected light, thereby obtaining continuous three-dimensional point cloud data. The three-dimensional point cloud data is a set of discrete data points reflecting the spatial structure of the target area obtained by the lidar. Each point has specific three-dimensional coordinate information and is used to construct a three-dimensional model of the environment. For example, in a 30-second experiment, the lidar will collect 300 frames of point cloud data, with each frame containing approximately 100,000 points on average, which details the geometric structure of the vehicle's surrounding environment.

[0031] Based on the three-dimensional point cloud data of the lidar, the vision sensor (such as a camera) is triggered to collect images. For example, when the lidar detects an object in a certain area, the camera will be triggered to capture an image of that area, obtaining multiple regional image data corresponding to the three-dimensional point cloud data. The camera captures images at a set resolution and frame rate (such as 1920x1080 resolution, 30 frames per second). The vision sensor usually refers to a camera, which can capture high-definition images, is triggered according to the point cloud data generated by the lidar, and collects image data of a specific area to supplement the texture and color information of the point cloud data.

[0032] At the same time, the millimeter-wave radar is configured in a multi-target tracking mode to continuously perceive the movement of multiple moving objects in the target area and obtain the velocity vector information of each target. The millimeter-wave radar is a sensor that uses electromagnetic waves in the millimeter-wave band for target detection, mainly used for multi-target tracking and obtaining the velocity vector information of moving objects, and performs excellently especially in bad weather such as rain and snow. Assuming that the millimeter-wave radar operates at a rate of 50 Hz, about 1500 velocity vector data can be obtained during the same test period, and these data detail the movement state of the target object.

[0033] To ensure the spatio-temporal consistency of various sensor data, a high-precision timestamp synchronization technology is adopted. According to the relative positions and orientations of the multi-modal sensor array, a transformation matrix between the coordinate systems of the multi-modal sensor array is established, which is used to transform the data of one sensor (such as a lidar) from its own coordinate system to the vehicle body coordinate system. The transformation matrix is usually obtained through external parameter calibration to ensure that the data of each sensor is aligned in the same reference frame. Taking the lidar as an example, by using a laser calibration plate or a calibration target, the data of the lidar and other sensors on the vehicle (such as a camera) are collected to achieve the geometric relationship calibration between the sensors. Using calibration software (such as the camera calibration toolbox of MATLAB or special lidar calibration software), the transformation matrix from the lidar coordinate system to the vehicle body coordinate system is calculated. This matrix is usually a 4×4 homogeneous transformation matrix, which contains a rotation matrix and a translation vector. The three-dimensional point cloud data collected by the lidar is subjected to coordinate transformation according to the homogeneous transformation matrix. Through matrix operations, each point is mapped into the vehicle body coordinate system to form the first perception data set.

[0034] For visual sensor data, a sliding window mechanism is adopted to perform dynamic reprojection on continuously captured image data. Taking 5 frames as a window, using the real-time motion data of the vehicle (such as vehicle speed and angular velocity obtained through IMU), each frame of image is geometrically corrected according to its time position within the window to eliminate image distortion caused by dynamic factors such as vehicle acceleration and turning. For example, the original image has an average blur error of 0.1 meters caused by vehicle speed changes, while after dynamic reprojection processing, this error is reduced by 80% to 0.02 meters, ensuring that the image can accurately reflect the environment. The second perception data set is a set of visual image data that is free of motion distortion and accurately reflects environmental characteristics after being processed by the sliding window mechanism and dynamic reprojection.

[0035] For millimeter-wave radar, due to its asynchronous sampling problem, an interpolation compensation technique is adopted to perform temporal continuity compensation on the collected data. Mathematical compensation (such as linear interpolation or spline interpolation) is performed on the time interval data generated by the asynchronous sampling of the millimeter-wave radar, making the data more continuous and consistent in the time dimension. Assuming that the original sampling frequency of the millimeter-wave radar is 45Hz, but the data has intermittent inconsistencies, through the linear interpolation method, the time error of the interpolated data is reduced to within 2 milliseconds, thereby generating the third perception data set, which contains accurate motion target speed vector information. The third perception data set is a data set containing motion object speed vector information obtained from the millimeter-wave radar after interpolation compensation processing.

[0036] To ensure the spatial and temporal consistency of the first, second, and third perception data sets, spatio-temporal consistency constraint conditions are constructed. By establishing spatial and temporal matching conditions, it is ensured that data from different sensors maintain geometric consistency in three-dimensional space. Using known static features in the environment (such as building boundaries and road contours) and vehicle dynamic parameters, the geometric positions of each data set in three-dimensional space are compared and verified. The geometric consistency of the first perception data set, the second perception data set, and the third perception data set in three-dimensional space is verified. If the geometric data of the first, second, and third perception data sets are consistent in three-dimensional space, a multi-source perception data set is obtained. Through the above steps, problems such as image distortion caused by coordinate system differences and vehicle motion between different sensors and time deviation caused by sampling asynchrony are eliminated, and the finally fused multi-source perception data set greatly improves the positioning accuracy and environmental perception stability of the autonomous driving system.

[0037] S200: Perform spatio-temporal synchronization processing on the multi-source perception data set to establish a fused perception data stream with a unified spatio-temporal reference.

[0038] Specifically, perform spatio-temporal synchronization processing on the aforementioned multi-source perception dataset, calibrate and align the multi-source data in terms of time and space to ensure that the data of each sensor reflects the environmental information under the same moment and the same spatial reference. Each type of data in lidar, vision sensors, and millimeter-wave radar carries its own timestamp and spatial information. Due to the different sampling frequencies and data formats of these sensors, there will be inconsistencies in time and space when directly used, and spatio-temporal synchronization processing is required.

[0039] Adopt high-precision clock synchronization technology, such as the PTP protocol, to uniformly calibrate the time of all sensors, so that the sampling moments of each sensor can be mapped to the same global clock. At the same time, through external parameter calibration and coordinate transformation, convert the data of each sensor into the vehicle body coordinate system to form a unified spatial reference. For example, in an experiment, the timestamp error of the lidar is controlled within 5 milliseconds, and after coordinate transformation, the geometric error between the lidar point cloud and the vehicle body coordinate system is less than 0.02 meters. Through algorithms such as interpolation and nearest neighbor matching, correct the data collected at different times so that the lidar point cloud, image data, and velocity vector information of the millimeter-wave radar at each moment can be accurately corresponded.

[0040] The specific process of time synchronization processing includes: selecting a sensor as the master clock, usually the sensor with the highest time resolution and stability, and aligning the timestamps of all sensors to the time reference of the master clock. For sensors that cannot be fully synchronized, record and compensate for the time delay between them and the master clock. The spatial synchronization processing includes: selecting the vehicle body coordinate system as the unified spatial reference system and converting the data of the sensors into the vehicle body coordinate system.

[0041] Fuse the data after time synchronization and spatial coordinate transformation to form a coherent and unified data stream. This data stream contains the information of all sensors and is arranged according to the unified time and spatial reference. The fused perception data stream refers to the continuous data stream formed by fusing the data from different sensors under the unified spatio-temporal reference, which reflects the comprehensive perception information of the vehicle's surrounding environment. Through spatio-temporal alignment processing, strictly align the data of each sensor to the unified spatio-temporal reference, and solve the error problem caused by the inconsistency of sampling time and spatial reference of multi-sensor data.

[0042] S300: Perform motion distortion correction on the fused perception data stream to generate dynamically compensated perception data.

[0043] Furthermore, S300 of this application includes:

[0044] By performing inertial measurement on the target vehicle, angular velocity data and linear acceleration data of the target vehicle are obtained; based on the angular velocity data and the linear acceleration data, motion compensation is performed to obtain the vehicle pose change amount; the three-dimensional point cloud data is combined with the vehicle pose change amount for inverse motion compensation, and the point cloud trajectory of the dynamic obstacle is denoised to generate the dynamic compensation perception data.

[0045] Specifically, motion distortion correction refers to correcting the data distortion caused by the displacement and rotation of the vehicle during the scanning process, so that the data collected by each sensor can reflect the real and static environment. The vehicle uses the built-in inertial measurement unit (IMU) to collect angular velocity and linear acceleration data in real time, which reflects the motion state of the vehicle in three-dimensional space. For example, at time t, the IMU records the angular velocity of the vehicle as (0.1 rad / s, 0.2 rad / s, 0.05 rad / s) and the linear acceleration as (0.5 m / s 2 , -0.3 m / s 2 , 0 m / s 2 ). The angular velocity data is the data collected by the inertial measurement unit (IMU) that describes the rotation rate of the vehicle around each axis, and the unit is usually radians per second. The linear acceleration data is the change rate that describes the vehicle accelerating or decelerating in a straight line direction, and the unit is usually meters per second squared.

[0046] Using the angular velocity and linear acceleration data, the pose (i.e., pose) change amount of the vehicle during the data collection period is calculated by mathematical methods such as integration, so as to correct the sensor data distortion caused by the vehicle motion. For example, in a 0.1-second acquisition period, the average angular velocity recorded by the IMU is 0.03 rad / s, and the average linear acceleration is 0.5 m / s 2 , and the initial velocity of the vehicle during this period is 0.5 m / s. To obtain the pose change amount of the vehicle during this acquisition period, it is necessary to perform integral calculations on the angular velocity and acceleration data respectively. For the angular velocity data, △θ = ω × △t = 0.003 rad, where △θ is the angular velocity, ω is the average angular velocity, and △t is the duration of 0.1 s. For the linear acceleration data, if the initial velocity v0 is considered, the displacement is calculated by the formula △s = v0 × △t + 1 / 2a × (△t) 2 = 0.0525 m. Through integral operation, the pose change amount of the vehicle translating about 0.05 meters and rotating 0.003 rad in this period is obtained. Based on this, the vehicle pose change amount is obtained, that is, the comprehensive amount of translation and rotation changes that the vehicle undergoes in a certain period of time, usually expressed in the form of displacement and rotation angle.

[0047] Construct a motion transformation matrix using the vehicle pose change amount, then take its inverse matrix to perform reverse motion compensation. Multiply each point cloud data point of the three-dimensional point cloud data by the reverse motion compensation matrix to obtain the corrected points, freeze the vehicle's motion during scanning, and convert the data to a stationary reference state, thereby eliminating the geometric distortion caused by vehicle displacement to the point cloud data. For example, in an experiment, the trajectory of the dynamic obstacle point cloud before compensation showed an average position error of about 0.15 meters, while after reverse motion compensation, the error was reduced to about 0.05 meters.

[0048] After completing the reverse motion compensation, since the point cloud data of the dynamic obstacle may still be affected by environmental noise or sensor errors, its trajectory will show certain fluctuations. Apply denoising processing to these trajectory data, usually using median filtering or Kalman filtering algorithms to smooth the point cloud trajectory and further remove outliers. After reverse motion compensation and denoising processing, dynamic compensation perception data is generated. This data set can more realistically and stably reflect the environmental characteristics and the motion state of dynamic obstacles after removing the scanning errors and noise interference caused by vehicle motion.

[0049] Motion distortion correction refers to calculating and compensating for the distortion of sensor data caused by vehicle motion. During vehicle driving, due to the influence of motion, the data collected by the sensor may have deviations in position and direction, resulting in distortion of point cloud data, image data, etc. Motion distortion correction is to compensate for these distortions through algorithms to restore the original environmental data. After motion distortion correction, the error of the perception data caused by vehicle motion is corrected to obtain dynamic compensation perception data. By performing motion distortion correction on the fused perception data stream, the vehicle can eliminate the data distortion caused by motion, and the denoising processing further improves the stability and accuracy of the point cloud data, ensuring high-precision environmental modeling and dynamic target detection.

[0050] S400: Input the dynamic compensation perception data into a multi-source fusion positioning model, perform three-dimensional space matching in combination with high-precision map data, and obtain real-time positioning coordinates.

[0051] Specifically, the dynamically compensated perception data after motion distortion correction and denoising processing is input into the multi-source fusion positioning model. The algorithm in the multi-source fusion positioning model is used to match the input perception data with the high-precision map data. The static environmental features such as road networks, building boundaries, and road signs contained in the high-precision map data provide a positioning reference for the vehicle. The input dynamically compensated perception data (such as the three-dimensional point cloud of lidar) is compared with the static environmental features in the high-precision map data. For example, the lidar point cloud can help identify the boundaries of roads and the locations of buildings, and then perform matching on the map. Through the three-dimensional space matching of the multi-source fusion positioning model and the high-precision map data, the precise position of the current vehicle is calculated. This positioning coordinate is real-time and continuously updated as the vehicle moves. For example, assume that the vehicle is driving on a known road. The lidar senses obstacles (such as street lights) on both sides of the road, and the vision sensor identifies a stop sign. The multi-source fusion positioning model matches these perception data with the road information in the high-precision map, thereby accurately calculating the position coordinates of the current vehicle. By combining the dynamically compensated perception data and the high-precision map data for three-dimensional space matching, high-precision real-time positioning is achieved, accurately positioning to the reference points on the map, ensuring the accuracy of navigation and path planning.

[0052] Furthermore, step S400 of the present application includes:

[0053] Extracting lidar geometric features based on the three-dimensional point cloud data, and extracting visual semantic features based on the multiple regional image data; constructing a deep neural network model, and performing cross-modal feature fusion on the visual semantic features and the lidar geometric features through the deep neural network model to obtain visual-lidar data; performing motion state estimation on the velocity vector information of the moving object according to the extended Kalman filter algorithm to obtain a motion state estimation result; introducing an attention mechanism to dynamically adjust the weight distribution of the visual-lidar data and the motion state estimation result, and constructing the multi-source fusion positioning model.

[0054] Specifically, by processing the three-dimensional point cloud data collected by the lidar, geometric features are extracted. The lidar geometric features refer to the environmental geometric information extracted by processing the three-dimensional point cloud data, including the shape, size, edge features, etc. of obstacles. For example, an algorithm based on normal estimation is used to extract information such as the surface shape and edges in the point cloud. Taking roads and obstacles as examples, the point cloud data of the lidar can accurately reflect the geometric shapes of obstacles, such as lane lines and three-dimensional buildings.

[0055] The image data collected by the vision sensor uses existing deep learning methods such as convolutional neural networks (CNNs) to extract semantic features in the image. The vision sensor can identify and label objects in the image, such as pedestrians, vehicles, traffic signs, etc. Visual semantic features include the category of the object (such as pedestrians, vehicles, road signs, etc.), the color of the object, texture, etc.

[0056] Construct a deep neural network model to fuse visual semantic features with lidar geometric features, which is achieved through a multi-channel neural network. The geometric information of the lidar and the semantic information of the image are input respectively, and the network automatically learns the association between the two. For example, the model can learn how to combine the pedestrian category in vision with the obstacle geometric information in lidar to obtain a more accurate obstacle position and type. The deep neural network model usually contains multiple input channels, one of which is used to receive visual semantic features, and the other input channel is used to receive lidar geometric features. By fusing the features of these two channels, the network can learn the association between them. The design of the deep neural network requires adding a fusion layer in the middle layer or the last few layers of the network, which is specifically used to fuse features from different modalities, including weighted sum fusion, concatenation fusion, and attention mechanism fusion. For example, in the fusion layer, visual features and lidar features are concatenated and mapped to a unified feature space through a fully connected layer.

[0057] After being fused by the deep neural network, a unified dataset is obtained, which contains comprehensive information of vision and lidar. Vision-lidar data is a dataset fused by the deep neural network model, which contains semantic information from the vision sensor (image) and geometric information from the lidar (point cloud).

[0058] For moving objects (such as other vehicles or pedestrians), the extended Kalman filter algorithm is used to estimate their motion states. Based on the current velocity vector information and the prediction model, the future position and velocity of the object are estimated. For example, when the velocity vector of another vehicle is detected, the EKF can estimate the future motion trajectory of this vehicle and provide an accurate state estimate. The extended Kalman filter (EKF) algorithm is a non-linear filtering algorithm, which is commonly used for state estimation of dynamic systems, especially in the presence of noise. Through the EKF, the motion state (position, velocity, etc.) of the object can be estimated. The motion state estimation result obtained by estimating through the extended Kalman filter algorithm is the current state of the object, including its position, velocity, and acceleration, etc.

[0059] Combine visual-lidar data and motion state estimation results, and dynamically adjust the weights of different data sources through an attention mechanism. When a moving object with a relatively high speed is detected, more attention may be paid to the motion state estimation of the moving object, while in a static environment, more reliance is placed on the fused features of vision and lidar. The attention mechanism is used to dynamically adjust the importance of different input data, automatically adjusting the attention to different features according to the context information or the requirements of the current task, thereby improving the performance of the model.

[0060] Integrate the feature fusion and motion state estimation results in the above steps through a multi-source fusion positioning model, which can provide high-precision positioning information based on the input data of sensors. For example, the system may combine visual and lidar data to provide the relative positions of the current vehicle and other objects (such as obstacles, pedestrians, etc.) in the environment. The multi-source fusion positioning model is a model that performs precise positioning and state estimation based on the fusion of data from multiple sensors (such as vision, lidar, IMU, etc.). By leveraging the complementary characteristics of multiple sensors, the positioning accuracy and robustness are improved.

[0061] By constructing a multi-source fusion positioning model, combining the geometric features of lidar and the semantic features of vision, and dynamically adjusting the data weights through an attention mechanism, not only the accuracy of environmental perception is improved, but also the tracking ability of moving objects is enhanced, enabling more accurate prediction of the states of moving objects, thereby improving the positioning and navigation performance of autonomous vehicles in complex environments.

[0062] S500: Iteratively optimize the real-time positioning coordinates through an adaptive error correction module to generate a vehicle positioning result.

[0063] Further, S500 of this application includes:

[0064] Extract the road topology structure in the high-precision map and evaluate the matching degree with the real-time positioning coordinates to obtain the matching degree evaluation result; perform particle filtering according to the matching degree evaluation result to obtain a particle filtering result, and dynamically adjust the real-time positioning coordinates according to the particle filtering result to obtain a particle distribution density; obtain the historical positioning trajectory of the target vehicle, and perform batch nonlinear optimization on the historical positioning trajectory through a sliding window to obtain positioning optimization trajectory data; combine the positioning optimization trajectory data with the particle distribution density to calculate the positioning error data. When the positioning error exceeds a preset threshold, trigger a positioning mode to reset the error and determine the vehicle positioning result.

[0065] Monitor the matching degree between the real-time positioning coordinates output by the multi-source fusion positioning model and the high-precision map in real time, determine the positioning error data according to the matching degree, and judge whether the positioning error data exceeds a preset threshold; when the positioning error data exceeds the preset threshold, generate an error trigger signal, start the multi-source data consistency verification based on the error trigger signal, select a trigger positioning mode according to the verification result, and the positioning mode includes a tightly coupled positioning mode or a redundant positioning mode; perform error correction based on the tightly coupled positioning mode or the redundant positioning mode, and generate the vehicle positioning result according to the correction result in combination with the spatial feature constraint relationship between the real-time positioning coordinates and the high-precision map.

[0066] Specifically, the road topology structure in the high-precision map is the spatial relationship and connection method between road elements in the map, such as road intersections, turns, section connections, etc. Extracting the road topology structure from the high-precision map includes elements such as the geometric information of the road, intersections, turns, and traffic signs. Match the real-time positioning coordinates with the road topology structure in these maps to evaluate the matching degree between the real-time positioning and the map data. A high matching degree indicates that the relationship between the current position of the vehicle and the road on the map is highly consistent, while a low matching degree may indicate errors in positioning. For example, if the current positioning coordinates of the vehicle are (100.5, 200.3) and the corresponding position in the high-precision map is (100.55, 200.35), the matching degree is evaluated by calculating the distance between the two points.

[0067] According to the matching degree evaluation result, use the particle filter method for dynamic adjustment. The particle filter generates multiple particles to represent the possible poses of the vehicle and updates the weights and positions of each particle through a filtering algorithm. That is to say, according to the result of the particle filter, the positioning coordinates of the vehicle are corrected. During the particle filter process, the distribution of particles represents the possibilities of different positions in the state space. The particle distribution density refers to the concentration of particles distributed in space, and a higher density area indicates that this area is more likely to be the real position of the vehicle. For example, assuming that the matching degree between the high-precision map and the real-time positioning coordinates is 0.95, then the particle filter will adjust the positions and distributions of the particles according to this evaluation, making the distribution of particles closer to the real position. The particle filter can continuously adjust the position estimation of the vehicle in a dynamic environment (such as in the presence of obstacles or other changes).

[0068] Obtain the historical positioning trajectory of the target vehicle, that is, the trajectory record of the vehicle over a period of time in the past, including the historical position data of the vehicle. Use the sliding window technique to divide the historical positioning trajectory into multiple data blocks. Each sliding window contains a fixed number of consecutive position data. As time goes by, the window slides in the historical data, adding new positioning data at each update and removing the oldest data. In each window, a non-linear optimization algorithm (such as Levenberg-Marquardt, least squares method, etc.) is used to optimize the trajectory data. The goal of optimization is to minimize the error between trajectory points. After the optimization process ends, the optimized trajectory data within each sliding window is obtained and combined into the final positioning optimized trajectory.

[0069] Batch non-linear optimization refers to the overall optimization of a series of data points in the historical positioning trajectory. Different from traditional linear optimization, non-linear optimization deals with problems with non-linear relationships and is very effective for complex motion trajectories in practice (such as turning and accelerating that may occur during vehicle driving). Batch optimization can obtain more comprehensive and accurate trajectory corrections by processing multiple data points simultaneously. The positioning optimized trajectory data is the optimized trajectory data, representing the precise path that the vehicle has passed through during the historical time period, closer to the actual path, and removing the noise and errors in the original trajectory.

[0070] In the particle filter algorithm, particles represent multiple possible positions of the vehicle, and the particle distribution density represents the concentration of particles at different positions in the state space. The area with a higher particle density indicates a higher probability of that position and may be the real position. The particle filter dynamically estimates the state of the vehicle by updating the weights and positions of the particles. Combine the optimized trajectory data with the particle filter result, and evaluate the gap between the two by comparing the particle positions generated by the particle filter and the positions of the optimized trajectory. Quantify the gap between the positioning optimized trajectory and the actual vehicle path. By calculating the difference between the optimized trajectory and the particle distribution density, positioning error data can be obtained, which is used to evaluate the accuracy of the positioning system. The smaller the error, the more accurate the positioning system. By calculating the error for each trajectory point and summarizing all the error data into an overall error data set, the deviation between the optimized positioning data and the actual path is represented.

[0071] Preset an error bound to judge whether the positioning result is accurate. When the positioning error exceeds the preset threshold, trigger the positioning mode to reset the error. Real-time monitor the output of the multi-source fusion positioning model, that is, the real-time positioning coordinates of the vehicle, and use the preset position and spatial information in the high-precision map to calculate the matching degree between the real-time positioning coordinates and the high-precision map. According to the matching degree, determine the positioning error data, that is, the error between the real-time positioning coordinates and the map position, and judge whether the positioning error data exceeds the preset threshold, that is, the maximum acceptable error range.

[0072] When the positioning error exceeds the threshold, it is considered that the positioning accuracy is insufficient. Error correction needs to be initiated to generate an error trigger signal and start the multi-source data consistency verification. Check whether the data from different sensors (such as GNSS, IMU, vision, lidar, etc.) is consistent and make adjustments according to the verification results. According to the results of the data consistency verification, select the trigger positioning mode, including using the tightly coupled positioning mode or the redundant positioning mode. The tightly coupled positioning mode is to tightly fuse the data from multiple sensors through a factor graph optimization framework to improve the positioning accuracy. For example, combine GNSS pseudorange and IMU acceleration data to reduce the error impact of a single sensor. The redundant positioning mode means that when the data of some sensors is inaccurate, other sensors are enabled, such as vision-lidar odometry, and positioning is performed through a loop detection mechanism to avoid relying on a single sensor.

[0073] If the tightly coupled positioning mode is selected, the factor graph optimization framework will be used to fuse multi-band GNSS pseudorange / carrier phase and IMU data to correct the positioning error. If the redundant positioning mode is selected, the loop detection mechanism of the vision-lidar odometry will be activated to correct the positioning error by detecting repeated features in the environment. After selecting the positioning mode, the real-time positioning coordinates will be error-corrected according to the selected mode. The corrected positioning data will be spatially feature-matched with the high-precision map to further optimize the positioning result and ensure that the positioning accuracy meets the preset requirements.

[0074] By dynamically monitoring the matching degree between the real-time positioning coordinates and the high-precision map, errors can be detected and corrected in a timely manner, which can effectively improve the positioning accuracy. According to the size of the error, different positioning modes (tightly coupled or redundant modes) are selected for correction to ensure that the positioning result maintains a high accuracy in any environment. The error trigger mechanism ensures that when the positioning accuracy deviates from the predetermined range, it can respond in a timely manner and initiate the necessary correction steps to maintain high-precision real-time positioning in complex environments and ensure the safety and reliability of the vehicle during the autonomous driving process.

[0075] In summary, the high-precision perception-driven autonomous driving positioning method provided by this application has the following beneficial effects:

[0076] Real-time collect vehicle environment data through a multi-modal sensor array to generate a multi-source perception data set; perform spatio-temporal synchronization processing on the multi-source perception data set to establish a fused perception data stream with a unified spatio-temporal reference; perform motion distortion correction on the fused perception data stream to generate dynamically compensated perception data; input the dynamically compensated perception data into a multi-source fusion positioning model, and perform three-dimensional space matching in combination with high-precision map data to obtain real-time positioning coordinates; perform iterative optimization on the real-time positioning coordinates through an adaptive error correction module to generate a vehicle positioning result. That is to say, by performing spatio-temporal synchronization processing on all sensor data, then performing motion distortion correction to eliminate the distortion caused by vehicle motion, determining the positioning coordinates in combination with a high-precision map, and continuously optimizing the positioning result using an adaptive error correction module, the accuracy and stability of the positioning are improved, thereby enhancing the safety and reliability of autonomous driving.

[0077] Embodiment 2, based on the same inventive concept as the high-precision perception-driven autonomous driving positioning method in the foregoing Embodiment 1, the present application also provides a high-precision perception-driven autonomous driving positioning system. Please refer to the attached Figure 2 , the high-precision perception-driven autonomous driving positioning system includes:

[0078] A data acquisition module 11, configured to collect vehicle environment data in real time through a multi-modal sensor array to generate a multi-source perception data set; a spatio-temporal synchronization processing module 12, configured to perform spatio-temporal synchronization processing on the multi-source perception data set to establish a fused perception data stream with a unified spatio-temporal reference; a motion distortion correction module 13, configured to perform motion distortion correction on the fused perception data stream to generate dynamically compensated perception data; a space matching module 14, configured to input the dynamically compensated perception data into a multi-source fusion positioning model, and perform three-dimensional space matching in combination with high-precision map data to obtain real-time positioning coordinates; a vehicle positioning module 15, configured to perform iterative optimization on the real-time positioning coordinates through an adaptive error correction module to generate a vehicle positioning result.

[0079] Furthermore, the data acquisition module 11 in the high-precision perception-driven autonomous driving positioning system is further configured to:

[0080] Configure a lidar to scan a target area at a preset scanning frequency to obtain three-dimensional point cloud data; trigger a vision sensor to perform image acquisition according to the three-dimensional point cloud data to obtain multiple regional image data; control a millimeter-wave radar to perceive the target area in a multi-target tracking mode to obtain speed vector information of moving objects; synchronize and align the time stamps of the three-dimensional point cloud data, the multiple regional image data, and the speed vector information to generate the multi-source perception data set.

[0081] Furthermore, the data acquisition module 11 in the high-precision perception-driven autonomous driving positioning system is further configured to:

[0082] Establish a transformation matrix between the coordinate systems of the multi-modal sensor array, map the lidar coordinate system to the vehicle body coordinate system according to the transformation matrix to obtain the first perception data set; perform dynamic reprojection on the multi-region image data of the vision sensor through a sliding window mechanism to obtain the second perception data set; perform interpolation compensation on the asynchronous sampling data extracted based on the millimeter-wave radar to obtain the third perception data set; construct a spatio-temporal consistency constraint condition to verify the geometric consistency of the first perception data set, the second perception data set, and the third perception data set in three-dimensional space. If the geometric data in the three-dimensional space is consistent, generate the multi-source perception data set.

[0083] Furthermore, the motion distortion correction module 13 in the high-precision perception-driven autonomous driving positioning system is further configured to:

[0084] Obtain the angular velocity data and linear acceleration data of the target vehicle through inertial measurement of the target vehicle; perform motion compensation based on the angular velocity data and the linear acceleration data to obtain the vehicle pose change amount; perform reverse motion compensation on the three-dimensional point cloud data combined with the vehicle pose change amount, and denoise the point cloud trajectory of the dynamic obstacle to generate the dynamically compensated perception data.

[0085] Furthermore, the spatial matching module 14 in the high-precision perception-driven autonomous driving positioning system is further configured to:

[0086] Extract lidar geometric features based on the three-dimensional point cloud data, and extract visual semantic features based on the multi-region image data; construct a deep neural network model, and perform cross-modal feature fusion on the visual semantic features and the lidar geometric features through the deep neural network model to obtain visual-lidar data; perform motion state estimation on the velocity vector information of the moving object according to the extended Kalman filter algorithm to obtain the motion state estimation result; introduce an attention mechanism to dynamically adjust the weight distribution of the visual-lidar data and the motion state estimation result, and construct the multi-source fusion positioning model.

[0087] Furthermore, the vehicle positioning module 15 in the high-precision perception-driven autonomous driving positioning system is further configured to:

[0088] Extract the road topology structure in the high-precision map and evaluate the matching degree with the real-time positioning coordinates, and the matching degree evaluation result; perform particle filtering according to the matching degree evaluation result to obtain a particle filtering result, and dynamically adjust the real-time positioning coordinates according to the particle filtering result to obtain a particle distribution density; obtain the historical positioning trajectory of the target vehicle, and perform batch nonlinear optimization on the historical positioning trajectory through a sliding window to obtain positioning optimization trajectory data; calculate the error by combining the positioning optimization trajectory data with the particle distribution density to obtain positioning error data. When the positioning error exceeds a preset threshold, trigger a positioning mode to reset the error and determine the vehicle positioning result.

[0089] Further, the vehicle positioning module 15 in the high-precision perception-driven autonomous driving positioning system is further configured to:

[0090] Monitor the matching degree between the real-time positioning coordinates output by the multi-source fusion positioning model and the high-precision map in real time, determine the positioning error data according to the matching degree, and judge whether the positioning error data exceeds a preset threshold; when the positioning error data exceeds the preset threshold, generate an error trigger signal, start multi-source data consistency verification based on the error trigger signal, select a trigger positioning mode according to the verification result, and the positioning mode includes a tightly coupled positioning mode or a redundant positioning mode; perform error correction based on the tightly coupled positioning mode or the redundant positioning mode, and generate the vehicle positioning result according to the correction result in combination with the spatial feature constraint relationship between the real-time positioning coordinates and the high-precision map.

[0091] Each embodiment in this specification is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The high-precision perception-driven autonomous driving positioning method and specific examples in the foregoing Figure 1 The high-precision perception-driven autonomous driving positioning method and specific examples in Embodiment 1 are equally applicable to the high-precision perception-driven autonomous driving positioning system in this embodiment. Through the foregoing detailed description of the high-precision perception-driven autonomous driving positioning method, those skilled in the art can clearly know the high-precision perception-driven autonomous driving positioning system in this embodiment. Therefore, for the sake of brevity of the specification, it will not be elaborated here. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description in the method part.

[0092] The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features disclosed herein.

[0093] Obviously, those skilled in the art can make several improvements and modifications to the present application without departing from the principles of the present application, and these improvements and modifications also fall within the protection scope of the present application.

Claims

1. A high-precision perception-driven autonomous driving positioning method, characterized in that: include: Collect vehicle environment data in real time through a multi-modal sensor array to generate a multi-source perception dataset; Performing spatiotemporal synchronization processing on the multi-source perception data set to establish a fused perception data stream with a unified spatiotemporal reference; Performing motion distortion correction on the fused perception data stream to generate dynamic compensation perception data; The dynamic compensation perception data is input into a multi-source fusion positioning model, combined with high-precision map data for three-dimensional space matching, to obtain real-time positioning coordinates; The real-time positioning coordinates are iteratively optimized through an adaptive error correction module to generate a vehicle positioning result.

2. The high-precision perception-driven autonomous driving positioning method according to claim 1, characterized in that: The vehicle environment data is collected in real time through a multi-modal sensor array to generate a multi-source perception dataset, including: Configure the laser radar to scan the target area at a preset scanning frequency to obtain three-dimensional point cloud data; Triggering a visual sensor to collect images according to the three-dimensional point cloud data to obtain image data of multiple regions; Control the millimeter-wave radar to sense the target area in a multi-target tracking mode and obtain the velocity vector information of the moving object; The three-dimensional point cloud data, the multiple regional image data and the velocity vector information are time stamped and synchronously aligned to generate the multi-source perception data set.

3. The high-precision perception-driven autonomous driving positioning method according to claim 2, characterized in that: The three-dimensional point cloud data, the plurality of regional image data and the velocity vector information are synchronized with each other by time stamps to generate a multi-source perception data set, including: Establishing a transformation matrix between the multimodal sensor array coordinate systems, mapping the laser radar coordinate system to the vehicle body coordinate system according to the transformation matrix, and obtaining a first perception data set; Dynamically reprojecting the image data of multiple regions of the visual sensor through a sliding window mechanism to obtain a second perception data set; Extract asynchronous sampling data based on millimeter wave radar for interpolation compensation to obtain a third perception data set; Construct spatiotemporal consistency constraints to verify the geometric consistency of the first perception data set, the second perception data set, and the third perception data set in the three-dimensional space. If the geometric data of the three-dimensional space are consistent, generate the multi-source perception data set.

4. The high-precision perception-driven autonomous driving positioning method according to claim 2, characterized in that: Performing motion distortion correction on the fused perception data stream to generate dynamic compensation perception data includes: By performing inertial measurement on the target vehicle, the angular velocity data and linear acceleration data of the target vehicle are obtained; Performing motion compensation based on the angular velocity data and the linear acceleration data to obtain a change in vehicle posture; Inverse motion compensation is performed on the three-dimensional point cloud data in combination with the vehicle posture change, and the point cloud trajectory of the dynamic obstacle is denoised to generate the dynamic compensation perception data.

5. The high-precision perception-driven autonomous driving positioning method according to claim 2, characterized in that: The construction process of the multi-source fusion positioning model includes: Extracting laser radar geometric features based on the three-dimensional point cloud data, and extracting visual semantic features based on the multiple regional image data; Constructing a deep neural network model, and performing cross-modal feature fusion of the visual semantic features and the laser radar geometric features through the deep neural network model to obtain visual-lidar data; The velocity vector information of the moving object is used to estimate the motion state according to the extended Kalman filter algorithm to obtain a motion state estimation result; An attention mechanism is introduced to dynamically adjust the visual-lidar data and the motion state estimation results for weight distribution, so as to construct the multi-source fusion positioning model.

6. The high-precision perception-driven autonomous driving positioning method according to claim 1, characterized in that: The real-time positioning coordinates are iteratively optimized through an adaptive error correction module to generate a vehicle positioning result, including: Extracting the road topology structure in the high-precision map and performing matching evaluation on the real-time positioning coordinates, and obtaining a matching evaluation result; Perform particle filtering according to the matching evaluation result to obtain a particle filtering result, and dynamically adjust the real-time positioning coordinates according to the particle filtering result to obtain a particle distribution density; Obtaining the historical positioning trajectory of the target vehicle, performing batch nonlinear optimization on the historical positioning trajectory through a sliding window, and obtaining positioning optimization trajectory data; The positioning optimization trajectory data is combined with the particle distribution density to perform error calculation to obtain positioning error data. When the positioning error exceeds a preset threshold, the positioning mode is triggered to reset the error and determine the vehicle positioning result.

7. The high-precision perception-driven automatic driving positioning method according to claim 6, characterized in that: When the positioning error exceeds a preset threshold, the positioning mode is triggered to reset the error and determine the vehicle positioning result, including: Monitor in real time the matching degree between the real-time positioning coordinates output by the multi-source fusion positioning model and the high-precision map, determine the positioning error data according to the matching degree, and judge whether the positioning error data exceeds a preset threshold; When the positioning error data exceeds the preset threshold, an error trigger signal is generated, multi-source data consistency verification is started based on the error trigger signal, and a trigger positioning mode is selected according to the verification result, and the positioning mode includes a tight combination positioning mode or a redundant positioning mode; Error correction is performed based on the tightly combined positioning mode or the redundant positioning mode, and the vehicle positioning result is generated according to the correction result combined with the real-time positioning coordinates and the spatial feature constraint relationship of the high-precision map.

8. High-precision perception-driven autonomous driving positioning system, characterized by: The steps for implementing the high-precision perception-driven autonomous driving positioning method according to any one of claims 1 to 7, wherein the high-precision perception-driven autonomous driving positioning system comprises: A data acquisition module, used to collect vehicle environment data in real time through a multi-modal sensor array to generate a multi-source perception data set; A spatiotemporal synchronization processing module, used to perform spatiotemporal synchronization processing on the multi-source perception data set to establish a fused perception data stream with a unified spatiotemporal reference; A motion distortion correction module, used to perform motion distortion correction on the fused perception data stream to generate dynamic compensation perception data; A spatial matching module is used to input the dynamic compensation perception data into a multi-source fusion positioning model, and perform three-dimensional spatial matching in combination with high-precision map data to obtain real-time positioning coordinates; The vehicle positioning module is used to iteratively optimize the real-time positioning coordinates through the adaptive error correction module to generate a vehicle positioning result.

Images