Intelligent automobile road curvilinear abstraction method and vectorized path

Abstract

The invention discloses an intelligent automobile road curvilinear abstraction method and a vectorized path, and belongs to the technical field of automatic driving. The scheme comprises the steps of road data acquisition; performing rolling airspace processing, dynamically caching data in a set space-time window, and performing coordinate conversion; road curvilinear abstraction: abstracting road segments into straight line segments and curve segments, and generating a feature point set; vectorizing the path, connecting the straight line segments, and connecting the straight line segments and the circular curve segments by adopting an easement curve; vectorization path integration is carried out, and a multi-section path function with the driving distance as an independent variable is generated; global vectorization path integration is carried out, and coordinate system unification is carried out. According to the method, aiming at the problems that the prior art depends on a high-cost sensor, the computing power demand is relatively large, the path smoothness is insufficient and the like, the vectorization path is generated based on a low-cost multi-source sensor acquisition vehicle and a dynamic rolling airspace processing method, the cost can be reduced, the computing speed can be optimized, and a high-comfort multi-section road curve function and a vectorization path function can be obtained.

Description

Technical Field

[0001] The present invention relates to the field of driverless cars, and in particular to a method for abstracting curves in roads of intelligent cars and a vectorized path. Background Art

[0002] For intelligent vehicles, the abstraction and vectorization of their driving paths are crucial foundations for implementing intelligent driving technology. By collecting and processing vehicle path data, the real road is abstracted into geometric curves. These curves are then converted into computable and predictable formulas, providing a crucial basis for subsequent autonomous vehicle path tracking. This is also a crucial step in transitioning autonomous driving technology from the laboratory to actual road testing and commercial application, laying the technical foundation for the future of large-scale autonomous driving.

[0003] Current methods for collecting autonomous driving paths typically require deploying multiple sensors, such as lidar or cameras, on vehicles to collect point cloud data and geographic data on the road. These vehicles and equipment are expensive, making them unsuitable for large-scale deployment. Furthermore, the collection process requires high computing power and is computationally complex. Furthermore, for autonomous vehicles, generating a good path function is crucial for both tracking and driving comfort.

[0004] The invention with application number CN118587375A is entitled "A method for generating a reference path for an unmanned mining vehicle based on cubic spline curve sampling." This method is based on a road boundary map, annotates reference path anchor points in the boundary map, performs lateral offset sampling on the reference path anchor points, obtains a sampling point set, and constructs a cost function based on the sampling point set based on the cubic spline curve; selects the optimal reference path from the candidate reference path set as the reference path generation result. Although this method can obtain a suitable reference path, in its specific implementation, it first obtains a road boundary map based on aerial images. This step places high demands on the clarity and accuracy of the road map. General aerial images are difficult to ensure the required accuracy. Furthermore, the process of marking reference path anchor points in this method requires manual annotation, which requires a large amount of work and is difficult to achieve efficiency. In addition, the accuracy of manual annotation is difficult to unify, which has an adverse impact on subsequent work.

[0005] The invention with application number CN117111061A is titled "A Lane-Level Mapping Method Using Point Cloud Data for Traffic Radar." This method uses traffic radars installed at the edge of the road to obtain point cloud data of each vehicle on the road. The point cloud data of each vehicle is clustered and tracked to obtain trajectory data for each vehicle. Each segment of trajectory data is fitted according to a preset fitting function to obtain the trajectory curve equation for the vehicle on each road. The multiple trajectory curve equations are divided into several trajectory clusters, which are then clustered and analyzed to extract and integrate the road curves to obtain the complete road curve. In its implementation, this method requires the deployment of a large number of traffic radars along an entire route. These radars can only detect a specific route, resulting in high equipment costs and limited widespread use in practice.

[0006] The invention with application number CN117874866A is entitled "A method for generating and connecting plane roads based on control points". This method first determines the curved line shape of the generated road, and then sets up several control points at the corners of the road route to determine road drawing parameters such as the route angle, turning radius and tangent length. During use, this method requires manual determination of a large number of control points. The rationality of the control point selection will directly affect the subsequent route generation work. In addition, during the use of this method, the required line shape function formulas need to be collected in advance and organized into a formula library. During use, due to the diversity of road curves, the formula library data is difficult to meet the needs of all road curves.

[0007] The invention with application number CN105740505A is named "A method for restoring the spatial linear shape of a road based on GPS-RTK technology". This method first collects data on the road linear shape, then processes and analyzes the collected data, and finally obtains the spatial linear shape of the entire path. During the use of this method, the method used is to first collect data and then perform computer processing, which is not real-time. In addition, when collecting data for the entire path, if the road is long and the data is large, the data storage requirements and subsequent computing and processing capabilities will also be greater.

[0008] In summary, adopting a low-cost acquisition and path abstraction method and using good path line types to represent the path is of great significance for the research and further use of autonomous vehicles. Summary of the Invention

[0009] In response to the above-mentioned problems, the purpose of the present invention is to provide a method for abstracting curved roads and vectoring paths for intelligent automobiles, using low-cost sensors to solve the problems of high cost and high computing power requirements of existing equipment, and adopting gentle curves to generate paths to improve the comfort of drivers and passengers.

[0010] In order to achieve the above object, the present invention adopts the following technical solutions:

[0011] Step 1: Road data collection (1): obtain vehicle position, front wheel angle, wheel speed and other data in real time through on-board sensors;

[0012] Step 2, rolling spatial domain processing (2), dynamically caching data within a set spatiotemporal window, and converting the original first ground coordinate (211) into a third static coordinate (231) with the starting point of the driving trajectory as the origin;

[0013] Step 3: Road curve abstraction (3): Based on the front wheel angle threshold (301) and the wheel speed difference threshold (302), the road segments are abstracted into straight segments and curved segments and a feature point set is generated (303). The feature angle of the curved segment is extracted (304).

[0014] Step 4, path vectorization (4), generating a straight line segment function through a straight line equation based on the feature point set (303), calculating the circular curve radius based on the feature turning angle, and connecting the straight line segment and the circular curve segment using a transition curve (401); Step 5, local multi-segment path generation (5), generating a local multi-segment path function (501) with the travel distance as the independent variable;

[0015] Step 6: Global vectorized path integration (6), integrating all local multi-segment path functions (501), generating a global vectorized path (601), and obtaining a vectorized path function P total (S)(602), and perform coordinate system one (7).

[0016] Furthermore, the conversion of the original first ground coordinate (211) into a local third static coordinate (231) with the starting point of the driving trajectory as the origin specifically includes:

[0017] Step 2-1, obtaining the first coordinate (211) O0 (M0, N0) of the starting point of the vehicle's driving trajectory;

[0018] Step 2-2, based on the earth ellipsoid model parameters, convert the first ground coordinate (211) of the starting point into the second plane coordinate (221) J1 (x0, y0) in the second plane coordinate system (22) on the meridian section projection plane according to formula (1.1);

[0019]

[0020] Wherein, a(222) is the length of the semi-major axis of the Earth ellipsoid, b(223) is the length of the semi-minor axis of the Earth ellipsoid, and N0(212) is the latitude coordinate of the starting point O0;

[0021] Step 2-3, taking the starting point O0 as the origin, establishing a local plane third static coordinate system (23), wherein the longitude direction is the third x-axis and the latitude direction is the third y-axis;

[0022] Step 2-4: Calculate the intermediate auxiliary point (232) A0(M0,N according to formula (1.2) k )'s second face coordinate (221)A A (x A ,y A ):

[0023]

[0024] Step 2-5: Determine the remaining collection points O using formula (1.3) k The third static coordinate (231)(x k ,y k );

[0025]

[0026] Where, ΔM(213) is O k The longitude difference between the starting point O0 and the starting point is ΔM=M k -M0.

[0027] Furthermore, the process of abstracting the road segments into straight line segments and curve segments and generating a feature point set (303) includes:

[0028] Step 3-1, setting a front wheel angle threshold (301) and a wheel speed difference threshold (302) for determining whether a road curve is a straight segment or a curved segment;

[0029] Step 3-2: Perform sliding window filtering on the front wheel angle and wheel speed difference, and set an outlier removal strategy to ensure data stability;

[0030] Step 3-3: Use the sliding window statistical strategy to determine whether the current segment is a curve segment or a straight line segment. If more than 60% of the frames in the window meet the threshold judgment condition, it is marked as a curve segment;

[0031] Step 3-4: record the boundary points between the curve segment and the straight line segment in the airspace. The set of all boundary points in the path is the feature point set (303);

[0032] Step 3-5: In the curve section, if the front wheel angle is stable at α f If ±1° lasts for ≥t seconds, record α f (304) is the characteristic corner.

[0033] The path vectorization (4) specifically includes:

[0034] Step 4-1: Connect the feature point set (303) with straight-line function equations into spaced straight-line segments in the third static coordinate system (23);

[0035] Step 4-2: Establish a fourth moving coordinate system (24) with the third static coordinate (231) of the first straight-line end point as the origin and the first straight-line direction as the positive direction;

[0036] Step 4-3: Establish a first transition curve:

[0037]

[0038] where, A (401) is the transition curve parameter; A = L s ×R; L s (402) is the total arc length of the transition curve; L (403) is the arc length of the transition curve from the starting point to a point on the curve;

[0039] Step 4-4: Calculate the radius R (404) of the bend center circle based on the feature rotation angle (304),

[0040]

[0041] where, L c (405) is the distance between the front and rear axles of the acquisition vehicle; α f (304) is the feature rotation angle;

[0042] According to the radius R (404) of the center circle, establish a circular curve formula in the fourth moving coordinate system (24):

[0043] <000014​​​​​​​​​​​​​​​​​​​​​​​​

[0050]

[0051] Among them, s(502) is the vehicle driving distance length, and s1(503) is the length of the first straight line segment;

[0052] (2) When s1 ≤ s < s2, the fourth moving coordinate (241) of the path point is:

[0053]

[0054] Among them, s2(504) is the length of the first transition curve, and s3(505) is the length of the circular curve segment,

[0055] (3) When s2 ≤ s < s3, the fourth moving coordinate (241) of the path point is:

[0056]

[0057] Among them, (m0, n0) is the center coordinate of the circular curve segment, L s (401) is the length of the transition curve segment, and s4(506) is the length of the second transition curve;

[0058] (4) When s3 ≤ s < s4, the fourth moving coordinate (241) of the path point is:

[0059]

[0060] Among them,

[0061]

[0062] Among them, r(507) is the radius of curvature of a point on the transition curve;

[0063] (5) When s ≥ s4

[0064] The fourth moving coordinate (241) of the path point is:

[0065]

[0066] Furthermore, according to the global vectorized path integration (6) described in claim 1, it is characterized in that the vectorized path (601) is represented as a concatenation of multiple local multi-segment path functions (501), and the global vectorized path function P total (S)(602) is the integration of all local multi-segment path functions (501), as shown in formula (6.1);

[0067]

[0068] Among them, L k (602) is the total length of the kth local multi-segment path; P k (S)(603) is the parametric equation of the kth local multi-segment path;

[0069] Furthermore, the fourth moving coordinate (241) in the local multi-segment path function (501) is converted into the third static coordinate (231). The specific conversion process is as follows:

[0070]

[0071] Wherein, T0 (701) represents the angle between the positive direction of the fourth moving coordinate system (24) and the positive direction of the third static coordinate system (23);

[0072] mD (702) is the m-direction coefficient, which takes a value of 1 when the m-axis in the fourth moving coordinate system (24) is in the forward direction of the tangent line, and takes a value of -1 when the m-axis in the fourth moving coordinate system (24) is in the backward direction of the tangent line;

[0073] nD(703) is the n-direction coefficient, which takes the value of 1 when the n-axis is on the same side as the m-axis and takes the value of -1 when the n-axis is on the opposite side of the m-axis;

[0074] is the fourth moving coordinate (241);

[0075] represents the coordinates of the origin of the kth fourth moving coordinate system (24) in the third static coordinate system (23).

[0076] In summary, the present invention has the following advantages:

[0077] 1. The present invention provides a method for abstracting curved roads and vectoring paths for intelligent automobiles. During the coordinate conversion process, the Earth's meridian section coordinate system is introduced to compensate for the distortion of traditional projected coordinate conversion, thereby improving the accuracy of the local coordinate system.

[0078] 2. The present invention provides a method for abstracting curved roads and vectoring paths for intelligent automobiles. By collecting the latitude and longitude coordinate information of the road and vehicle-related parameters, and through multiple low-cost sensor data collection and relatively simple algorithms, the present invention requires lower costs and lower computing power requirements than using lidar to collect routes, thus reducing computing time.

[0079] 3. The present invention provides a method for abstracting curved roads and vectoring paths for intelligent automobiles. By setting a dynamic window method, dynamic data is scrolled and processed, and characteristic data is extracted and recorded. Compared with recording data for the entire road, this method only needs to record the characteristic data of the road, which occupies less storage during the collection and vectorization process, facilitating the subsequent calculation and recording of path functions. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Table 1 is a comparison table of the meanings of the serial numbers in the present invention;

[0081] Figure 1 This is a specific flow chart of a method for abstracting curves and vectoring paths for intelligent automobile roads in the present invention;

[0082] Figure 2 A diagram showing a process of converting a first ground coordinate into a third static coordinate in an embodiment of the present invention;

[0083] Figure 3 This is a diagram of a multi-segment curve structure in an embodiment of the present invention;

[0084] Table 1: Comparison table of serial number meanings

[0085]

[0086]

[0087] DETAILED DESCRIPTION

[0088] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. In addition, the present invention can also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited by the specific embodiments disclosed below.

[0089] The present invention provides a method for abstracting curves of intelligent automobile roads and a vectorized path, which specifically includes the following steps:

[0090] Step 1: Road data collection (1), the specific process includes:

[0091] The vehicle position, front wheel angle, wheel speed and other data are acquired in real time through the vehicle-mounted sensors; in this embodiment, the acquisition vehicle needs to use an RTK system with a differential positioning function, as well as left and right wheel speed sensors and front wheel angle acquisition equipment. During the data acquisition process, the vehicle needs to travel on the road where the data needs to be collected, and continuously acquire the current latitude and longitude coordinate position of the vehicle, i.e., the first ground coordinate (211), through the RTK, and acquire the front wheel angle of the vehicle at the current position and the rotation speed of the left and right vehicles through the left and right wheel speed sensors and front wheel angle sensors on the acquisition vehicle, and record the data in real time.

[0092] During the data collection process, in view of the possibility of signal loss of the RTK positioning system, the IMU inertial navigation system can be used to compensate for the RTK signal, predict the current position based on historical data, and improve the path continuity maintenance rate in the signal loss scenario.

[0093] Step 2, rolling airspace processing (2), specifically includes: setting a rolling airspace for the collected data stream, monitoring the first ground coordinate (211) of the collected vehicle position data, the front wheel angle and the left and right wheel speeds within the airspace for a period of time, and converting the original first ground coordinate (211) into a local third static coordinate (231) with the starting point of the driving trajectory as the origin, and dynamically updating the airspace content;

[0094] In rolling spatial domain processing (2), dynamic window parameters can be adjusted in real time for different scenarios, and the window size, data sampling frequency, etc. can be changed according to the actual scenario;

[0095] Furthermore, converting the original first ground coordinate (211) into a local third static coordinate (231) with the starting point of the driving trajectory as the origin within the window specifically includes:

[0096] Step 2-1, obtaining the first coordinate (211) O0(M0, N0) of the starting point of the vehicle's driving trajectory, where M0 is longitude and N0 is latitude;

[0097] Step 2-2, based on the earth ellipsoid model parameters, convert the first ground coordinate (211) of the starting point into the second plane coordinate (221) J1 (x0, y0) in the meridian section projection coordinate system (22) according to formula (1.1);

[0098]

[0099] Where a(222) is the length of the semi-major axis of the Earth ellipsoid, b(223) is the length of the semi-minor axis of the Earth ellipsoid, and N o (212) is the latitude coordinate of the starting point O0;

[0100] Step 2-3, taking the starting point O0 as the origin, establishing a local plane third static coordinate system (23), wherein the longitude direction is the third x-axis and the latitude direction is the third y-axis;

[0101] Step 2-4: Calculate the intermediate auxiliary point (232)A(M0,N) according to formula (1.2) k )'s second face coordinate (221)A A (x A ,y A ),

[0102]

[0103] Step 2-5: Determine the remaining collection points O using formula (1.3) k The third static coordinate (231)(x k ,y k );

[0104]

[0105] Where, ΔM(213) is O k The longitude difference between the starting point O0 and the starting point is ΔM=M k -M0.

[0106] During the coordinate conversion process in the airspace, a dynamic switching mechanism of multiple coordinate systems can be introduced. When the vehicle is at low speed, the plane Cartesian coordinate system is used to simplify the calculation. At medium and high speeds, it is switched to the coordinate conversion method based on the earth ellipsoid model in this step to further improve the calculation speed. In addition, when dealing with complex terrain and large altitude change rates such as mountain roads, a high-precision digital elevation model can be introduced to perform Z-axis correction on the ellipsoid projection to eliminate the horizontal projection distortion caused by the undulating terrain, so as to reduce the path error of mountain roads.

[0107] Step 3: Road curve abstraction (3): Based on the front wheel angle threshold (301) and the wheel speed difference threshold (302), the road segments are abstracted into straight segments and curved segments and a feature point set is generated (303). The feature angle of the curved segment is extracted (304).

[0108] Step 3-1, setting a front wheel turning angle threshold (301) and a wheel speed difference threshold (302) for determining whether a road curve is a straight segment or a curved segment, and collecting the front wheel turning angle and wheel speed difference of the current position of the vehicle in real time when the vehicle is traveling on the road;

[0109] Step 3-2: Perform sliding window filtering on the front wheel angle and wheel speed difference, and set an outlier removal strategy to ensure data stability;

[0110] Step 3-3: Use the sliding window statistical strategy to determine whether the current segment is a curve segment or a straight line segment. If more than 60% of the frames in the window meet the threshold judgment condition, it is marked as a curve segment;

[0111] Step 3-4: The determined vehicle trajectory is recorded in the rolling airspace by using the determined straight line and curve segments and the third static coordinate of the vehicle. The intersection coordinates of the straight line segment and the curve segment are recorded as the feature point coordinates of the vehicle trajectory point. The set of all feature points in the trajectory constitutes a feature point set (303);

[0112] Step 3-5: In the curve section, if the front wheel angle is stable at α f If ±1° lasts for ≥t seconds, record α f (304) is the characteristic corner.

[0113] In step 3, in view of the different effects of the collected values ​​on the segmentation abstraction effect caused by different vehicle speeds, the reference threshold can be dynamically set according to the vehicle speed or actual road conditions to improve the accuracy of segmentation abstraction. The sliding filter can remove the mutation points and small jitters in the front wheel angle data, making the segmentation judgment more stable and helping to smooth the wheel speed difference data to avoid short-term wheel speed changes from misleading the path segmentation. In specific implementation, when the vehicle is driving on a flat road, the weighted sliding average method is used for filtering processing, which can more accurately track rapidly changing steering. When the vehicle is in poor road conditions, the sensor signal is prone to mutation. At this time, the median filter method is used to filter out instantaneous abnormal fluctuations.

[0114] Step 4, path vectorization (4), connecting the straight line segments by straight line equations according to the feature point set (303), calculating the circular curve radius based on the feature turning angle, and connecting the straight line segments and the circular curve segments by using a transition curve (401); the feature point set (303) is used to describe the straight line segment area of ​​the road, and the third static coordinate (311) data in the feature point set (303) is divided into a group of two coordinates, and each group of data can represent a straight line. By recording the feature point set (303), the required storage space can be reduced;

[0115] Step 4: Path vectorization (4) includes the following steps:

[0116] Step 4-1, connecting the feature point set (303) into spaced straight line segments in the third static coordinate system (23) using a straight line function equation;

[0117] Step 4-2: Establish a fourth moving coordinate system (24)mon at the end point of the first straight line along the direction of the first straight line, and establish the first transition curve:

[0118]

[0119] Where A(401) is the parameter of the transition curve; A=L s ×R;L s (402) is the total arc length of the transition curve; L(403) is the arc length of the transition curve from the starting point to a point on the curve; L s ≥v 2 / (3.6 2 a max R), a max =2.5m / s 2 ;

[0120] Step 4-3: Calculate the curve center circle radius R (404) based on the characteristic turning angle (304).

[0121]

[0122] Among them, L c (405) is the distance between the front and rear axles of the acquisition vehicle, and α f (304) is the characteristic rotation angle;

[0123] According to the radius R (403) of the center circle, a circular curve formula is established in the fourth moving coordinate system (24):

[0124]

[0125] The circular curve path formula is obtained.

[0126] Among them, (m0, n0) is the center of the circular curve;

[0127] β (407) is the tangential angle of any point on the circular curve, β0 (408) is the deflection angle of the first transition curve, R (403) is the radius of the circular curve, and L r (406) is the arc length from a certain point on the arc line to the starting point of the circular curve;

[0128] [[ID=Z30]]Step 4-4: Establish the second transition curve:

[0129]

[0130] Among them, θ (409) is the included angle between the second straight line and the first straight line;

[0131] To achieve the continuity of the path function in the geometric structure, the system executes a continuity constraint mechanism when splicing each path segment. Specifically, it includes: the position continuity constraint ensures that there are no breakpoints in the path, the tangent continuity constraint guarantees the natural transition of the vehicle driving direction, and in some scenarios, the curvature continuity constraint is further implemented to achieve smooth and controllable path turning. Each segment of the curve uses the end point and direction of the previous segment as the initial boundary conditions, and continuous transition is achieved by adjusting the parameters A (401) and the arc length L (403) in the construction of the transition curve and the circular curve. At the same time, a numerical verification function is provided for error detection and path segment screening in engineering implementation.

[0132] Step 5: Vectorize path integration (5) to generate a multi-segment path function (501) with the driving distance as the independent variable

[0133] Furthermore, the multi-segment path function (501) specifically includes:

[0134] (1) When 0 ≤ s < s1, the fourth moving coordinates (241) of the path points are:

[0135]

[0136] Among them, s(502) is the vehicle driving distance length, and s1(503) is the length of the first straight line segment;

[0137] (2) When s1 ≤ s < s2, the fourth dynamic coordinate (241) of the path point is:

[0138]

[0139] Among them, s2(504) is the length of the first transition curve, s3(505) is the length of the circular curve segment,

[0140] (3) When s2 ≤ s < s3, the fourth dynamic coordinate (241) of the path point is:

[0141]

[0142] Among them, (m0, n0) is the center coordinate of the circular curve segment, L s (401) is the length of the transition curve segment, s4(506) is the length of the second transition curve;

[0143] (4) When s3 ≤ s < s4, the fourth dynamic coordinate (241) of the path point is:

[0144]

[0145] Among them,

[0146]

[0147] Among them, r(507) is the radius of curvature of a point on the transition curve;

[0148] (5) When s ≥ s4

[0149] The fourth dynamic coordinate (241) of the path point is: <​​​​​​​​​​​​​​​(S)(603) is the parametric equation of the kth local multi-segment path;

[0154] Step 7: Convert the fourth moving coordinate (241) in the local multi-segment path function (501) into the third static coordinate (231). The specific conversion process is as follows:

[0155]

[0156] Wherein, T0 (701) represents the angle between the positive direction of the fourth moving coordinate system (24) and the positive direction of the third static coordinate system (23);

[0157] mD (702) is the m-direction coefficient, which takes a value of 1 when the m-axis in the fourth moving coordinate system (24) is in the forward direction of the tangent line, and takes a value of -1 when the m-axis in the fourth moving coordinate system (24) is in the backward direction of the tangent line;

[0158] nD(703) is the n-direction coefficient, which takes the value of 1 when the n-axis is on the same side as the m-axis and takes the value of -1 when the n-axis is on the opposite side of the m-axis;

[0159] is the fourth moving coordinate (241);

[0160] represents the coordinates of the origin of the kth fourth moving coordinate system (24) in the third static coordinate system (23).

[0161] In summary, the present invention has the following advantages:

[0162] 1. The present invention provides a method for abstracting curved roads and vectoring paths for intelligent automobiles. During the coordinate conversion process, the Earth's meridian section coordinate system is introduced to compensate for the distortion of traditional projected coordinate conversion, thereby improving the accuracy of the local coordinate system.

[0163] 2. The present invention provides a method for abstracting curved roads and vectoring paths for intelligent automobiles. By collecting the latitude and longitude coordinate information of the road and vehicle-related parameters, and through multiple low-cost sensor data collection and relatively simple algorithms, the present invention requires lower costs and lower computing power requirements than using lidar to collect routes, thus reducing computing time.

[0164] 3. The present invention provides a method for abstracting curved roads and vectoring paths for intelligent automobiles. By setting a dynamic window method, dynamic data is scrolled and processed, and characteristic data is extracted and recorded. Compared with recording data for the entire road, this method only needs to record the characteristic data of the road, which occupies less storage during the collection and vectorization process, facilitating the subsequent calculation and recording of path functions.

Claims

1. A method for abstracting curves of intelligent automobile roads and a vectorized path, characterized in that: The following steps are involved: Step 1: Road data collection (1): obtain vehicle position, front wheel angle, wheel speed and other data in real time through on-board sensors; Step 2, rolling spatial domain processing (2), dynamically caching data within a set spatiotemporal window, and converting the original first ground coordinate (211) into a third static coordinate (231) with the starting point of the driving trajectory as the origin; Step 3: Road curve abstraction (3): Based on the front wheel angle threshold (301) and the wheel speed difference threshold (302), the road segments are abstracted into straight segments and curved segments and a feature point set is generated (303). The feature angle of the curved segment is extracted (304). Step 4, path vectorization (4), generating a straight line segment function through a straight line equation according to the feature point set (303), calculating the circular curve radius based on the feature angle, and connecting the straight line segment and the circular curve segment using a transition curve (401); Step 5: Generate a local multi-segment path (5), generating a local multi-segment path function (501) with the travel distance as an independent variable; Step 6: Global vectorized path integration (6), integrating all local multi-segment path functions (501), generating a global vectorized path (601), and obtaining a vectorized path function P total (S)(602), and perform coordinate system one (7).

2. The rolling airspace processing (2) according to claim 1, characterized in that The conversion of the original first ground coordinate (211) into a third static coordinate (231) with the starting point of the driving track as the origin specifically includes: step 2-1, obtaining the first ground coordinate (211) O0 (M0, N0) of the starting point of the vehicle driving track; Step 2-2, based on the earth ellipsoid model parameters, convert the first ground coordinate (211) of the starting point into the second plane coordinate (221) J1 (x0, y0) in the second plane coordinate system (22) on the meridian section projection plane according to formula (1.1); Wherein, a(222) is the length of the semi-major axis of the Earth ellipsoid, b(223) is the length of the semi-minor axis of the Earth ellipsoid, and N0(212) is the latitude coordinate of the starting point O0; Step 2-3, taking the starting point O0 as the origin, establishing a local plane third static coordinate system (23), wherein the longitude direction is the third x-axis and the latitude direction is the third y-axis; Step 2-4: Calculate the intermediate auxiliary point (232) A0(M0,N according to formula (1.2) k )'s second face coordinate (221)A A (x A ,y A ): Step 2-5: Determine the remaining collection points O using formula (1.3) k The third static coordinate (231)(xk,yk); Where, ΔM(213) is O k The longitude difference between the starting point O0 and the starting point is ΔM=M k -M0.

3. The road curve abstraction (3) according to claim 1, characterized in that: The process of abstracting the road segments into straight line segments and curve segments and generating a feature point set (303) includes: Step 3-1, setting a front wheel angle threshold (301) and a wheel speed difference threshold (302) for determining whether a road curve is a straight segment or a curved segment; Step 3-2: Perform sliding window filtering on the front wheel angle and wheel speed difference, and set an outlier rejection strategy. Step 3-3: Use a sliding window statistical strategy to determine whether the current road is a curved segment or a straight segment. If more than 60% of the frames in the window meet the threshold judgment condition, it is marked as a curved segment. Step 3-4: record the boundary points of the curve segment and the straight line segment in the airspace, and the set of the third static coordinates (231) of all the boundary points in the path is the feature point set (303); Step 3-5: In the curve section, if the front wheel angle is stable at α f If ±1° lasts for ≥t seconds, record α f (304) is the characteristic corner.

4. The path vectoring (4) according to claim 1, characterized in that The curve segment path formulation specifically includes: Step 4-1, connecting the feature point set (303) into spaced straight line segments in the third static coordinate system (23) using a straight line function equation; Step 4-2, establishing a fourth moving coordinate system (24) with the third static coordinate (231) of the end point of the first straight line as the origin and the direction of the first straight line as the positive direction; Step 4-3: Create the first transition curve: Where A(401) is the parameter of the transition curve; A=L s ×R;L s (402) is the total arc length of the transition curve; L(403) is the arc length of the transition curve from the starting point to a point on the curve; Step 4-4: Calculate the curve center circle radius R (404) based on the characteristic turning angle (304). Among them, L c (405) is the distance between the front and rear axles of the collection vehicle; α f (304) is the characteristic corner; According to the radius R (404) of the central circle; a circular curve is established in the fourth moving coordinate system (24); Step 4-5: Establish the second transition curve: where θ (409) is the included angle between the second straight line and the first straight line.

5. The local multi-segment path generation (5) according to claim 1, characterized in that The local multi-segment path function (501) specifically includes: (1) When 0 ≤ s < s1, the fourth moving coordinate (241) of the path point is: where s (502) is the vehicle driving distance length, and s1 (503) is the length of the first straight line segment; (2) When s1 ≤ s < s2, the fourth moving coordinate (241) of the path point is: where s2 (504) is the length of the first transition curve, and s3 (505) is the length of the circular curve segment; (3) When s2 ≤ s < s3, the fourth moving coordinate (241) of the path point is: Where (m0, n0) is the coordinate of the center of the circular curve segment, L s (401) is the length of the transition curve segment, s4 (506) is the length of the second transition curve; (4) When s3 ≤ s < s4, the fourth moving coordinate (241) of the path point is: where, where r (507) is the radius of curvature of a point on the transition curve; (5) When s ≥ s4 the fourth moving coordinate (241) of the path point is:

6. The global vectorized path integration (6) according to claim 1, characterized in that The vectorized path (601) is represented as a series connection of multiple local multi-segment path functions (501), defining a global vectorized path function P total (S)(602) is the integration of all local multi-segment path functions (501), as shown in formula (6.1); Among them, L k (603) is the total length of the kth local multi-segment path; P k (S)(604) is the parametric equation of the kth local multi-segment path.

7. The coordinate system (7) according to claim 1, characterized in that Convert the fourth moving coordinate (241) in the local multi-segment path function (501) to the third static coordinate (231), and the specific conversion process is as follows: where, T0 (701) represents the included angle between the positive direction of the fourth moving coordinate system (24) and the positive direction of the third static coordinate system (23); mD (702) is the m-direction coefficient, which takes the value of 1 when the m-axis in the fourth moving coordinate system (24) extends in the forward direction of the tangent, and takes the value of -1 when the m-axis in the fourth moving coordinate system (24) extends in the backward direction of the tangent; nD (703) is the n-direction coefficient, which takes the value of 1 when the n-axis is on the same side of the m-axis, and takes the value of -1 when the n-axis is on the opposite side of the m-axis; is the fourth moving coordinate (241); represents the coordinates of the origin of the kth fourth moving coordinate system (24) in the third static coordinate system (23).

Images