Generating an HD map

Abstract

Procedure for generating an HD map using the following steps: - Providing (S1) a trajectory of an area to be mapped; - Discretizing (S2) the trajectory according to a predefined distance with trajectory points; - Determining (S3) geometric features of the trajectory at the grid points; - Clustering (S4) of adjacent trajectory points whose geometric features meet one or more predefined criteria into feature-based trajectory segments of a feature-based segmentation of the trajectory.

Description

AREA OF INVENTION

[0001] The present invention relates to a method for generating an HD map. SUMMARY OF THE INVENTION

[0002] Accordingly, the following is planned: - a method for generating an HD map with the following steps: Providing a trajectory of an area to be mapped; discretizing the trajectory according to a predefined distance with trajectory points; determining geometric features of the trajectory at the trajectory points; clustering adjacent trajectory points whose geometric features meet one or more predefined criteria into feature-based trajectory segments of a feature-based segmentation of the trajectory.

[0003] A high-definition map (HD map) is a high-resolution digital map that provides a precise and detailed representation of road infrastructure. It contains not only geometric information such as road layout and intersections, but also semantic data such as lane markings, traffic signs, and traffic lights. HD maps are used particularly for autonomous driving systems to enable precise navigation and decision-making, and operate with centimeter-level accuracy.

[0004] A trajectory describes the recorded path that an object – such as a vehicle – travels in space. It consists of a sequence of points determined by GPS data along the object's path and serves as the basis for creating high-resolution maps (HD maps).

[0005] Clustering, or grouping, means grouping similar points of a trajectory together based on predetermined characteristics. Equidistant clusters are groups of trajectory points evenly spaced at fixed intervals along the trajectory. This method divides the trajectory into segments, regardless of its geometric properties, with each segment having a constant length or distance between points. This method is simple and regular, but it ignores potential variations in road structure or movements along the trajectory.

[0006] Feature-based clusters, on the other hand, are formed by analyzing specific geometric features such as direction, curvature, or other properties of the trajectory. Points with similar features are grouped into clusters, enabling adaptive segmentation. This method adapts to the actual shape and structure of the trajectory, allowing curves, straight sections, or special features like intersections to be detected and segmented more precisely.

[0007] The cross product of two vectors is calculated by taking the product of specific components of the two original vectors for each component of the resulting vector and finding their difference. To calculate the x-component, the y-component of the first vector is multiplied by the z-component of the second vector, and the product of the z-component of the first vector and the y-component of the second vector is subtracted. For the y-component, the product of the z-component of the first vector and the x-component of the second vector is taken, and the product of the x-component of the first vector and the z-component of the second vector is subtracted. For the z-component, the product of the x-component of the first vector and the y-component of the second vector is taken, and the product of the y-component of the first vector and the x-component of the second vector is subtracted.

[0008] In two-dimensional space, the cross product between two vectors is a scalar and corresponds to the z-component of two three-dimensional vectors, since two-dimensional vectors can also be described as three-dimensional vectors with a z-component z=0.

[0009] Straight trajectory segments are sections of a path where an object moves in a straight line without changing direction. In these sections, the direction remains constant, and there are no bends or curves. In contrast, curved trajectory segments exhibit a change in direction, causing the path to be bent or winding. These segments may contain slight curves or more pronounced turns and are characterized by a noticeable curvature that defines the shape of the path.

[0010] Computer program products typically comprise a sequence of instructions that, when the program is loaded, cause the hardware to perform a specific procedure that leads to a particular result.

[0011] This application addresses the problem of generating high-resolution maps in connection with trajectory segmentation. Conventional methods for equidistant trajectory segmentation are often inaccurate and lead to unsatisfactory results because the trajectories used to generate the maps often have complex geometric features that complicate precise segmentation. The present invention solves this problem by proposing a feature-based trajectory segmentation method.

[0012] The process comprises several steps, including the discretization of the trajectory according to a predefined spacing with trajectory points and the determination of geometric features at the trajectory points. By clustering adjacent trajectory points that meet certain predefined criteria, feature-based trajectory segments are generated.

[0013] Discretizing the trajectory enables efficient processing and analysis of the data by converting the continuous trajectory into a manageable format suitable for computer-aided processing.

[0014] By determining geometric features at the trajectory points, a detailed characterization of the trajectory can be achieved.

[0015] Clustering trajectory points with similar features leads to feature-based segmentation, which enables intuitive and relevant structuring of the trajectory data and forms the basis for further analyses and applications.

[0016] Advantageous designs and further developments result from the further sub-claims as well as from the description with reference to the figures in the drawing.

[0017] In one embodiment, several feature-based segmentations are determined for a trajectory, and the several feature-based segmentations are aggregated into a fused segmentation.

[0018] Determining multiple feature-based segmentations increases the robustness and accuracy of HD mapping by taking into account different aspects of the trajectory.

[0019] Aggregation into a fused segmentation enables a comprehensive representation of the trajectory, combining the advantages of individual segmentations and compensating for potential weaknesses of individual approaches.

[0020] In one embodiment, a feature-based segmentation is determined with respect to at least one geometric feature from the following list: direction at a grid point, cosine of the direction at a grid point, sine of the direction at a grid point, curvature at a grid point, tangent of the direction at a grid point, radius of a circle through a predetermined selection of trajectory points, a product of direction and curvature at a grid point.

[0021] According to a preferred embodiment of the invention, the segmentation is based on the following features: direction at a grid point, cosine of the direction at a grid point, sine of the direction at a grid point, curvature at a grid point.

[0022] It goes without saying that equivalent results can be achieved with a different selection of features.

[0023] In one embodiment, intersections of trajectory segments are identified by means of a cross product of trajectory segments.

[0024] Accurate localization of intersections improves the safety and reliability of driver assistance systems by providing an important source of information for decision-making.

[0025] Two line segments can either intersect (form a crossing) or not intersect. To find out if two segments intersect, you calculate the cross product of the vectors that form these segments.

[0026] If two vectors lie on a particular side of a line segment, their cross products have the same sign. If they lie on opposite sides, their cross products have different signs.

[0027] First, vectors are formed from the endpoints of the two line segments, and then their cross products are calculated to determine the relative orientation of the vectors to each other.

[0028] Two line segments are defined by two points each. A first segment is formed by points P1 and P2, a second segment by points P3 and P4. First, a vector P1P2 is calculated from point P1 to point P2. Similarly, the vectors P1P3 and P1P4 are calculated between point P1 and points P3 and P4, respectively.

[0029] To determine whether the two segments intersect, the cross product of these vectors, P1P2 × P1P3 and P1P2 × P1P4, is used.

[0030] The cross products indicate whether points P3 and P4 lie on the same or opposite sides of segment P1P2. The sign of the cross products is evaluated: if the signs are different, it is determined that points P3 and P4 lie on opposite sides of the first segment.

[0031] This is repeated for the second segment P3P4. To do this, the cross products P3P4 × P3P1 and P3P4 × P3P2 are calculated. If the signs of the cross products are different here as well, then points P1 and P2 lie on opposite sides of the second segment.

[0032] An intersection of the two segments is therefore determined when both sign comparisons show different signs. This is how it is determined whether the two line segments intersect.

[0033] In one embodiment, the feature-based trajectory segments cluster straight and curved trajectory segments.

[0034] Clustering into straight and curved trajectory segments allows for differentiation between various road alignments. This can increase data processing efficiency by enabling a simplified representation of complex trajectories.

[0035] In one embodiment, segmentation into feature-based trajectory segments is based on sine and cosine values ​​of the direction.

[0036] It is advantageous if the direction at a grid point dictates whether the cosine value or the sine value is used for clustering a grid point.

[0037] It is conceivable that the sine function is used for directions from 0° to 45°, as it allows for a more stable and refined differentiation of directional changes in this range. Between 45° and 90°, the cosine function is used instead, since it exhibits greater changes in this angular range and thus allows for a more precise description of the directions. The use of the cosine function continues in the range of 90° to 135°, as the sine function has almost reached its maximum value in this section and exhibits only minor changes, while the cosine function offers greater sensitivity and differentiability. This method continues in further angular ranges, always choosing the function that shows greater changes in the respective range and thus enables a better differentiation of directions.In this way, the stability and precision of the calculations are optimized by making optimal use of the properties of sine and cosine in different direction ranges.

[0038] Switching between sine and cosine in different angular ranges makes sense because the values ​​of these trigonometric functions are particularly sensitive to changes in direction in certain ranges. This also helps to avoid numerical singularities.

[0039] A computer program product according to a method of an embodiment of the invention performs the steps of a method according to the preceding description when the computer program product runs on a computer, in particular an in-vehicle computer. When the program in question is used on a computer, the computer program product produces an effect, namely the improvement of accuracy and efficiency in the creation of high-resolution maps (HD maps). CONTENT OF THE DRAWINGS

[0040] The present invention will be explained in more detail below with reference to the exemplary embodiments shown in the schematic figures of the drawings. These show: Fig. 1 A schematic block diagram of an embodiment of the invention.

[0041] The accompanying drawings are intended to provide a further understanding of the embodiments of the invention. They illustrate embodiments and, in conjunction with the description, serve to explain the principles and concepts of the invention. Other embodiments and many of the advantages mentioned will become apparent with reference to the drawings. The elements of the drawings are not necessarily shown to scale.

[0042] In the figures of the drawings, identical, functionally equivalent and equally effective elements, features and components - unless otherwise stated - are each provided with the same reference symbols. DESCRIPTION OF EXAMPLES OF EXECUTION

[0043] Fig. Figure 1 shows a schematic block diagram of a procedure for generating an HD map with steps S1-S4.

[0044] In step S1, a trajectory of an area to be mapped is provided. In step S2, the trajectory is discretized with trajectory points at a predefined interval. In step S3, geometric features of the trajectory are determined at the trajectory points. In step S4, adjacent grid points whose geometric features meet one or more predefined criteria are clustered into feature-based trajectory segments of a feature-based segmentation of the trajectory. Reference sign S1-S4 process steps

Claims

[1] Method for generating an HD map using the following steps: - Providing (S1) a trajectory of an area to be mapped; - Discretizing (S2) the trajectory according to a predefined distance with trajectory points; - Determining (S3) geometric features of the trajectory at the trajectory points; - Clustering (S4) of adjacent trajectory points whose geometric features meet one or more predefined criteria into feature-based trajectory segments of a feature-based segmentation of the trajectory. [2] Method according to claim 1, wherein several feature-based segmentations are determined for a trajectory and the several feature-based segmentations are aggregated to a fused segmentation. [3] Method according to one of the preceding claims, wherein a feature-based segmentation is determined with respect to at least one geometric feature of the following list: Direction at a grid point, cosine of direction at a grid point, sine of direction at a grid point, curvature at a grid point, tangent of direction at a grid point, radius of a circle through a predetermined selection of trajectory points, a product of direction and curvature at a grid point. [4] Method according to one of the preceding claims, wherein intersections of trajectory segments are identified by means of a cross product of trajectory segments. [5] Method according to any of the preceding claims, wherein the feature-based trajectory segments cluster straight and curved trajectory segments. [6] Method according to one of the preceding claims, wherein segmentation into feature-based trajectory segments is based on sine and cosine values ​​of the direction. [7] Method according to claim 6, wherein the direction in a grid point determines whether the cosine value or the sine value of a grid point is used for clustering. [8] Computer program product comprising instructions that cause a hardware component of a computer to execute the method according to any of the preceding claims when the computer program is loaded onto / from the hardware component or executed by it.

Images