A high-speed multi-scene vehicle-road cooperative driving simulation system

Abstract

The application relates to the road condition simulation technical field, in particular to a high-speed multi-scene vehicle-road cooperative driving simulation system, which comprises a scene simulation and data collection module, a noise analysis module, a point cloud filtering module and a vehicle-road cooperation module; the scene simulation and data collection module collects point cloud data around a simulated vehicle in real time through a virtual simulation scene; the noise analysis module analyzes local and global outlier features of all point cloud data in each voxel, and calculates noise interference prominence of each point cloud data; the point cloud filtering module identifies and removes abnormal point cloud based on the noise interference prominence; voxel filtering processing is performed on the remaining point cloud data; the vehicle-road cooperation module uses the filtered point cloud data to update a three-dimensional point cloud map of the simulated vehicle, and realizes vehicle-road cooperative simulation testing; the application aims to improve the filtering effect of the voxel filtering algorithm on the point cloud data, so as to guarantee the quality of real-time updating of the three-dimensional point cloud map.

Description

Technical Field

[0001] This application relates to the field of road condition simulation technology, specifically to a high-speed multi-scenario vehicle-road cooperative driving simulation system. Background Technology

[0002] Cooperative vehicle-infrastructure system (CVIS), a crucial technology for intelligent transportation development, utilizes wireless communication and next-generation internet technologies to comprehensively implement dynamic information interaction between vehicles, roads, and pedestrians. It fully realizes effective coordination among people, vehicles, and roads, providing a reliable means to improve traffic efficiency and enhance traffic safety, enabling efficient, orderly, and safe traffic flow. Currently, a CVIS-based cooperative driving simulation system can be built on a six-degree-of-freedom motion platform, providing three degrees of freedom in translation and three degrees of freedom in rotational motion. This allows for simulation testing of CVIS in various traffic scenarios on highways, verifying the traffic safety and smoothness of CVIS, and further improving the coordination efficiency between vehicles and road infrastructure.

[0003] In existing technologies, a highway simulation environment and virtual sensors are constructed using simulation software. These virtual sensors are then used in vehicle-to-infrastructure (V2I) driving simulations to collect point cloud data around the simulated vehicle in real time. By employing coordinate system transformation, the discrete coordinates of the point cloud data in the map coordinate system are obtained, thereby updating the 3D point cloud map of the simulated vehicle in real time. This helps improve traffic safety and flow in V2I driving. However, because the real-time collected point cloud data contains redundant or noisy data, existing technologies typically use voxel filtering algorithms to filter it. Traditional voxel filtering algorithms generally directly calculate the centroid of all point cloud data within a voxel as the representative point. However, abnormal point clouds caused by noise within a voxel can affect the selection of the representative point. Existing technologies do not accurately identify and remove abnormal point clouds caused by noise within a voxel, resulting in poor accuracy in selecting the representative point. This affects the quality of the real-time updated 3D point cloud map and is detrimental to traffic safety and flow in V2I driving. Summary of the Invention

[0004] To address the aforementioned technical problems, the purpose of this application is to provide a high-speed, multi-scenario vehicle-road cooperative driving simulation system, the specific technical solution of which is as follows: This application proposes a high-speed, multi-scenario vehicle-road cooperative driving simulation system, the system comprising: Scene simulation and data acquisition module: Real-time acquisition of point cloud data around the simulated vehicle through virtual simulation scene; Noise Analysis Module: Divide the point cloud data into voxel grids, analyze the local and global outlier features of all point cloud data within each voxel and form two-dimensional feature points, use the difference distance between the two-dimensional feature points to obtain the neighborhood point cloud set of each point cloud data within each voxel, and combine the difference distance between the two-dimensional feature points and the two-dimensional feature origin to calculate the noise interference prominence of each point cloud data. Point cloud filtering module: Calculates the interference confidence of each point cloud data based on the salience of noise interference and the proportion of the part exceeding the segmentation threshold, so as to identify and remove abnormal point clouds; and performs voxel filtering on the remaining point cloud data. Vehicle-road cooperative module: The filtered point cloud data is used to update the 3D point cloud map of the simulated vehicle to realize vehicle-road cooperative simulation test.

[0005] Preferably, the virtual simulation scene is built using PreScan simulation software.

[0006] Preferably, the local outlier feature is calculated by using the LOF algorithm to calculate the local outlier factor for each point cloud data within each voxel.

[0007] Preferably, the global outlier feature is determined by calculating the average of the absolute differences between the local outlier factors of each point cloud data and all point cloud data.

[0008] Preferably, the two-dimensional feature points consist of local outlier features after range normalization and global outlier features.

[0009] Preferably, obtaining the neighborhood point cloud set of each point cloud data within each voxel includes: using the two-dimensional feature points of all point cloud data within all voxels as input to the OPTICS clustering algorithm, and outputting the neighborhood point cloud set of each point cloud data within each voxel.

[0010] Preferably, the noise interference salience is positively correlated with the difference distance between the two-dimensional feature points and the two-dimensional feature origin of each point cloud data, and negatively correlated with the number of point clouds in the neighborhood point cloud set of each point cloud data.

[0011] Preferably, the segmentation threshold is obtained by adaptively calculating the noise interference salience of all point cloud data in the current point cloud space using the maximum inter-class variance algorithm.

[0012] Preferably, the method for identifying and removing abnormal point clouds is as follows: taking the interference confidence level of all point cloud data as input, calculating its upper quartile, and marking point cloud data higher than the upper quartile as abnormal point clouds and removing them.

[0013] Preferably, during the voxel filtering process, all remaining point cloud data within each voxel are used as input to the voxel filtering algorithm. The voxel filtering algorithm calculates the centroid of all remaining point cloud data within each voxel as the representative point of each voxel. The voxel filtering algorithm outputs the filtered point cloud dataset in the current point cloud space.

[0014] This application has at least the following beneficial effects: (1) This application accurately measures the global deviation of outliers between each point cloud data and all point cloud data within its voxel, which more clearly reflects the global abnormal point cloud features within the voxel. It also constructs two-dimensional feature points for each point cloud data by combining local outliers and global outlier features, which is beneficial for more accurately distinguishing normal point cloud data and abnormal point cloud data within each voxel in the future.

[0015] (2) This application takes into account the clustering results of two-dimensional feature points of all point cloud data and the prominence of local outliers and global outliers of point cloud data, and accurately measures and analyzes the prominence of noise interference in point cloud data, so as to highlight the abnormal point cloud features caused by noise more significantly and improve the accuracy of subsequent selection of representative points within voxels.

[0016] (3) This application accurately measures the credibility of point cloud data affected by noise by using a feature measurement method, and uses a boxplot to accurately identify and remove abnormal point clouds caused by noise within voxels based on the credibility of interference of all point cloud data, so as to avoid affecting the selection of representative points within voxels in the current point cloud space, improve the effect of voxel filtering algorithm on point cloud data filtering, and thus ensure the quality of real-time updated three-dimensional point cloud map. Attached Figure Description

[0017] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a high-speed multi-scenario vehicle-road cooperative driving simulation system provided in one embodiment of this application. Detailed Implementation

[0019] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a high-speed multi-scenario vehicle-road cooperative driving simulation system proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0021] The following description, in conjunction with the accompanying drawings, details a specific scheme for a high-speed, multi-scenario vehicle-road cooperative driving simulation system provided in this application.

[0022] Please see Figure 1 The diagram illustrates a flowchart of a high-speed multi-scenario vehicle-road cooperative driving simulation system provided in one embodiment of this application. The system includes: a scenario simulation and data acquisition module, a noise analysis module, a point cloud filtering module, and a vehicle-road cooperative module.

[0023] Scene simulation and data acquisition module: Real-time acquisition of point cloud data around the simulated vehicle through virtual simulation scene.

[0024] The PreScan simulation software is used to build virtual simulation scenarios and vehicle terminal systems, and provides a rich set of sensor models, such as cameras, lidar, millimeter-wave radar, and roadside units (RSUs). The virtual simulation scenario refers to a highway scenario, and the road conditions in the virtual simulation scenario include intersections, tunnels, and severe weather.

[0025] Vehicle-road cooperative driving simulation tests were conducted in a virtual simulation scenario. The hardware of the simulation system mainly consists of a six-degree-of-freedom motion platform, a simulated vehicle, and a VR display device. The simulated vehicle is installed on the six-degree-of-freedom motion platform, which is a load-bearing platform supported by six electric cylinders. The load-bearing platform and the piston rod end of the electric cylinders are connected by six sets of Hooke hinges. The cylinder body end of the electric cylinder is also connected to the lower platform by six sets of Hooke hinges. The six electric cylinders are driven by servo motors.

[0026] In the vehicle-road cooperative driving simulation test, virtual sensors acquire and collect real-time LiDAR data, roadside perception data, and connected vehicle data around the simulated vehicle. The LiDAR data includes a point cloud dataset of the surrounding environment, used to update the 3D point cloud map of the simulated vehicle in real time. The roadside perception data includes data collected and broadcast by RSU roadside units and intelligent transportation infrastructure (such as traffic lights, variable message signs, etc.), such as traffic light status data and traffic sign data. The connected vehicle data includes data broadcast by other connected vehicles through V2X (Vehicle-to-Everything) communication, such as the position, speed, driving direction, braking intention, and steering intention of other connected vehicles.

[0027] Noise Analysis Module: The point cloud data is divided into voxel grids, the local and global outlier features of all point cloud data within each voxel are analyzed and formed into two-dimensional feature points, the difference distance between the two-dimensional feature points is used to obtain the neighborhood point cloud set of each point cloud data within each voxel, and the noise interference salience of each point cloud data is calculated by combining the difference distance between the two-dimensional feature points and the two-dimensional feature origin.

[0028] Because real-time point cloud data contains redundant or noisy point cloud data, voxel filtering algorithms are needed to filter the real-time point cloud dataset. However, abnormal point clouds caused by noise within voxels can affect the selection of representative points for voxels, thus affecting the quality of real-time updated 3D point cloud maps and hindering traffic safety and flow in vehicle-to-infrastructure (V2I) driving. Therefore, it is necessary to accurately identify and remove abnormal point clouds caused by noise within voxels to improve the accuracy of voxel representative point selection.

[0029] Therefore, the real-time acquired point cloud dataset is divided into voxel grids. The point cloud dataset corresponding to the current acquisition time is used as the voxel filtering algorithm. The preset voxel parameter voxel_size in the algorithm is (0.1m, 0.1m, 0.1m). The voxel filtering algorithm is used to divide the real-time acquired point cloud dataset into voxel grids to obtain each voxel in the current point cloud space. Each voxel contains multiple point cloud data. The voxel grid division is a well-known technique, and the specific process will not be described in detail.

[0030] Furthermore, the abnormal point clouds generated by noise within voxels are analyzed. All point cloud data within all voxels in the current point cloud space are used as input to the LOF (Local Outlier Factor) algorithm. The preset neighborhood parameter in the algorithm is 10. The LOF outlier detection algorithm is used to obtain the local outlier factor of each point cloud data within each voxel. The LOF outlier detection algorithm is a well-known technology, and the specific process will not be described in detail.

[0031] Since the LOF outlier detection algorithm only identifies anomalous point clouds through local density differences, it does not fully consider the global deviation of outliers between each point cloud data and all point cloud data within its voxel, thus failing to accurately identify global anomalous point cloud features within each voxel.

[0032] Therefore, for each voxel, the global outlier feature of each point cloud data within the voxel is calculated: In the formula, Let j be the global outlier feature of the j-th point cloud data. The number of point cloud data within a voxel. and These are the local outlier factors for the j-th and k-th point cloud data points, respectively.

[0033] Global outlier features reflect the global deviation of each point cloud data from all point cloud data within the voxel. The greater the global deviation of outlier, the greater the global deviation of the point cloud data from all point cloud data within the voxel, the higher the degree of anomaly prominence of the point cloud data within the voxel, and the more likely it is to belong to noisy point cloud data within the voxel.

[0034] To more accurately identify anomalous point clouds caused by noise within voxels, the local outlier factors and global outlier features of all point cloud data within all voxels are normalized by range, so that the quantification results of local outliers and global outlier features are in the same range [0,1]. The local outlier factors and global outlier features of each point cloud data after range normalization are used to form two-dimensional feature points of each point cloud data. , ),in, and These are the local outlier factor and global outlier feature after range normalization, which are helpful for more accurately distinguishing normal point cloud data and abnormal point cloud data within each voxel.

[0035] Furthermore, the two-dimensional feature points of all point cloud data within all voxels are used as input to the OPTICS clustering algorithm (Ordering Points To Identify the Clustering Structure). The algorithm has a preset neighborhood radius parameter of 0.2 and a preset minimum number of neighbors of 8. The OPTICS clustering algorithm is used to obtain the neighborhood point cloud set of each point cloud data within each voxel. The OPTICS clustering algorithm is a well-known technique, and its specific process will not be described in detail.

[0036] Generally, the greater the difference between the two-dimensional feature points of each point cloud data and the two-dimensional feature origin (0,0), the more prominent the local and global outlier characteristics of the point cloud data. At the same time, the smaller the number of points in the neighborhood point cloud set of the point cloud data, the greater the difference between the two-dimensional feature points of the point cloud data and the two-dimensional feature points of most point cloud data in the voxel, which can better highlight the abnormal point cloud features caused by noise.

[0037] Therefore, for each voxel, the saliency of noise interference in each point cloud data within the voxel is calculated.

[0038] The noise interference salience is positively correlated with the difference distance between the two-dimensional feature points and the two-dimensional feature origin of each point cloud data, and negatively correlated with the number of point clouds in the neighborhood point cloud set of each point cloud data.

[0039] It is understandable that a positive correlation means that the dependent variable increases as the independent variable increases, and decreases as the independent variable decreases; a negative correlation means that the dependent variable decreases as the independent variable increases, and increases as the independent variable decreases. This is determined by the actual application, and this application does not impose any special restrictions.

[0040] Specifically, for each voxel, calculate the noise saliency of the j-th point cloud data: In the formula, Let the noise saliency of the j-th point cloud data be denoted as . Let be the distance between the two-dimensional feature point of the j-th point cloud data and the two-dimensional feature origin (0,0). The number of point clouds in the neighborhood point cloud set for each point cloud data. This is a preset error parameter, designed to avoid a denominator of 0. Its value ranges from (0,1), and in this embodiment, it is set to 0.5. The difference distance can be measured using Euclidean distance or Mahalanobis distance; this embodiment uses Euclidean distance to measure the difference distance.

[0041] Noise interference salience reflects the degree to which abnormal point cloud features are prominent due to noise in each point cloud data within each voxel. The greater the noise interference salience, the more prominent the local and global outlier characteristics of the point cloud data are, and the more significantly the abnormal point cloud features caused by noise are highlighted, which will result in poorer accuracy in selecting representative points within the voxel.

[0042] Point cloud filtering module: Calculates the interference confidence of each point cloud data based on the salience of noise interference and the proportion of the part exceeding the segmentation threshold, so as to identify and remove abnormal point clouds; and performs voxel filtering on the remaining point cloud data.

[0043] Furthermore, to more reliably and accurately identify and remove anomalous point clouds caused by noise within voxels, the salience of noise interference in all point cloud data within all voxels in the current point cloud space is used as the input to the maximum inter-class variance (MIL) algorithm. The MIL algorithm is then used to obtain the segmentation threshold of the current point cloud space. The MIL algorithm is a well-known technique, and its specific process will not be elaborated further.

[0044] Generally, if the salience of noise interference in a point cloud data is higher than the segmentation threshold of the current point cloud space, it means that the point cloud data is more likely to be affected by noise interference. When measuring the abnormal point cloud features of noisy point cloud data within voxels, a larger gain should be given to more reliably and accurately identify abnormal point clouds caused by noise within voxels.

[0045] Therefore, the difference between the noise saliency of each point cloud data within each voxel and the segmentation threshold of the current point cloud space is calculated, and all differences are normalized by range. The normalized difference is recorded as the normalized difference of each point cloud data within each voxel. The larger the normalized difference, the more likely the point cloud data is to be affected by noise interference. Therefore, the greater the gain is assigned when measuring the abnormal point cloud features of the point cloud data.

[0046] Based on the above analysis, for each voxel, calculate the perturbation confidence level of each point cloud data within the voxel: In the formula, Let j be the confidence level of the j-th point cloud data being disturbed. Let be the normalized difference of the j-th point cloud data.

[0047] The perturbation confidence level reflects the degree to which each point cloud data within each voxel is affected by noise. The higher the perturbation confidence level, the higher the confidence level of the abnormal point cloud features generated by the noise. In this case, the point cloud data is more likely to affect the selection of representative points within the voxel. Therefore, it is necessary to accurately identify and remove abnormal point clouds generated by noise within the voxel to improve the accuracy of the selection of representative points within the voxel and avoid affecting the quality of real-time updated 3D point cloud maps.

[0048] Therefore, a boxplot is used to accurately identify anomalous point clouds caused by noise within voxels. The perturbation confidence of all point cloud data within all voxels in the current point cloud space is used as the input of the boxplot. The upper quartile of the perturbation confidence of all point cloud data is calculated using the boxplot. Point cloud data with perturbation confidence higher than the upper quartile are marked as anomalous point clouds in the current point cloud space. The boxplot is a well-known technique, and the specific process will not be described in detail.

[0049] Furthermore, the real-time acquired point cloud data is accurately filtered. In each voxel of the current point cloud space, marked abnormal point clouds are removed. All remaining point cloud data in all voxels after the removal of abnormal point clouds are used as input to the voxel filtering algorithm. The voxel filtering algorithm calculates the centroid of all remaining point cloud data in each voxel as the representative point of each voxel. The voxel filtering algorithm outputs the filtered point cloud dataset in the current point cloud space. The voxel filtering algorithm is a well-known technique, and the specific process will not be described in detail.

[0050] Vehicle-road cooperative module: The filtered point cloud data is used to update the 3D point cloud map of the simulated vehicle to realize vehicle-road cooperative simulation test.

[0051] Furthermore, to update the 3D point cloud map in real time and ensure traffic safety and smoothness in vehicle-road cooperative driving, a map coordinate system transformation is performed on the filtered point cloud dataset in the current point cloud space. This yields all discrete coordinates of the current point cloud dataset in the 3D point cloud map. These discrete coordinates are then sent to the vehicle-to-vehicle terminal system of the simulated vehicle, which updates the 3D point cloud map in real time. The map coordinate system conversion utilizes well-known techniques, and the specific process is not detailed here.

[0052] Simultaneously, the real-time collected roadside perception data and connected vehicle data are sent to the vehicle-to-vehicle terminal system of the simulated vehicle. The vehicle-to-vehicle terminal system performs time alignment of the roadside perception data and connected vehicle data by updating the 3D point cloud map in real time. Then, the final perception data after time alignment is input into the PreScan simulation software, which can realize closed-loop operation between vehicle perception data and vehicle simulation driving software. The PreScan simulation software performs fusion processing on the input final perception data to realize the testing of various vehicle-road cooperative functions, such as adaptive cruise control, lane keeping, and automatic emergency braking, thereby realizing the vehicle-road cooperative driving simulation test of the simulated vehicle.

[0053] Thus, the invention of a high-speed, multi-scenario vehicle-road cooperative driving simulation system was completed.

[0054] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0055] It should be noted that, unless otherwise specified and limited, terms such as “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a circuit structure, article, or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such article or device. Without further limitations, an element defined by the phrase “comprising one…” does not exclude the presence of other identical elements in the article or device that includes said element. Furthermore, the term “and / or” as used herein includes any and all combinations of one or more of the associated listed items.

[0056] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not invented in this application.

[0057] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A high-speed, multi-scenario vehicle-road cooperative driving simulation system, characterized in that, The system includes: Scene simulation and data acquisition module: Real-time acquisition of point cloud data around the simulated vehicle through virtual simulation scene; Noise Analysis Module: The module divides the point cloud data into voxel grids, analyzes the local and global outlier features of all point cloud data within each voxel, and forms two-dimensional feature points. It uses the difference distance between these two-dimensional feature points to obtain the neighborhood point cloud set for each point cloud data within each voxel. Combining the difference distance between the two-dimensional feature points and the origin of the two-dimensional feature points, it calculates the noise saliency of each point cloud data. Local outlier features are calculated using the LOF algorithm to determine the local outlier factor for each point cloud data within each voxel. Global outlier features are determined by averaging the absolute differences between the local outlier factors of each point cloud data and all point cloud data. The two-dimensional feature points consist of local outlier features normalized to range and global outlier features. The noise saliency is positively correlated with the difference distance between the two-dimensional feature points and the origin of the two-dimensional feature points for each point cloud data, and negatively correlated with the number of points in the neighborhood point cloud set for each point cloud data. Point cloud filtering module: Calculates the interference confidence of each point cloud data based on the salience of noise interference and the proportion of the part exceeding the segmentation threshold, so as to identify and remove abnormal point clouds; and performs voxel filtering on the remaining point cloud data. For each voxel, calculate the perturbation confidence level of each point cloud data within the voxel: In the formula, Let j be the confidence level of the j-th point cloud data being disturbed. Let be the normalized difference of the j-th point cloud data; where the method for determining the normalized difference of the point cloud data is as follows: calculate the difference between the noise interference salience of each point cloud data in each voxel and the segmentation threshold of the current point cloud space, and perform range normalization processing on all differences, and record the data after range normalization processing as the normalized difference of each point cloud data in each voxel. The noise saliency of the j-th point cloud data; The method for identifying and removing abnormal point clouds is as follows: taking the interference confidence of all point cloud data as input, calculating its upper quartile, and marking point cloud data higher than the upper quartile as abnormal point clouds and removing them. Vehicle-road cooperative module: The filtered point cloud data is used to update the 3D point cloud map of the simulated vehicle to realize vehicle-road cooperative simulation test.

2. The high-speed multi-scenario vehicle-road cooperative driving simulation system as described in claim 1, characterized in that, The virtual simulation scene was built using PreScan simulation software.

3. The high-speed multi-scenario vehicle-road cooperative driving simulation system as described in claim 1, characterized in that, The step of obtaining the neighborhood point cloud set of each point cloud data within each voxel includes: using the two-dimensional feature points of all point cloud data within all voxels as input to the OPTICS clustering algorithm, and outputting the neighborhood point cloud set of each point cloud data within each voxel.

4. The high-speed multi-scenario vehicle-road cooperative driving simulation system as described in claim 1, characterized in that, The segmentation threshold is obtained by adaptively calculating the salience of noise interference in all point cloud data in the current point cloud space using the maximum inter-class variance algorithm.

5. The high-speed multi-scenario vehicle-road cooperative driving simulation system as described in claim 1, characterized in that, During the voxel filtering process, all remaining point cloud data within each voxel are used as input to the voxel filtering algorithm. The voxel filtering algorithm calculates the centroid of all remaining point cloud data within each voxel as the representative point of each voxel. The voxel filtering algorithm outputs the filtered point cloud dataset in the current point cloud space.

Images